bank efficiency and lending propensity: evidence … · bank efficiency and lending propensity:...
TRANSCRIPT
I
BANK EFFICIENCY AND LENDING PROPENSITY:
EVIDENCE FROM COMMERCIAL BANKS IN INDONESIA
Thesis submitted for the degree of
Doctor of Philosophy
at the University of Leicester
by:
Mokhamad Anwar
School of Management
University of Leicester
December, 2014
II
BANK EFFICIENCY AND LENDING PROPENSITY:
EVIDENCE FROM COMMERCIAL BANKS IN INDONESIA
Abstract
Indonesia is one of the emerging economies1, which has been adopting a bank-based system
in the economy. It is very important to investigate the Indonesian commercial banks‟
performance given their substantial contribution to the development of the country. This
thesis aims to measure and analyse the performance of Indonesian banks in terms of their
efficiency and lending propensity over the period 2002-2010. The period testifies the
recovery phase after the turmoil caused by the Asian Financial Crisis of 1997-98 as well as
the revocation of the regulation of minimum threshold on commercial banks‟ small business
loans (finance) in 2001. This thesis employs frontier methods in estimating bank efficiency
where both parametric and non-parametric linear programming approaches are used, namely
Data Envelopment Analysis (DEA) and Stochastic Frontier Analysis (SFA). The former is
used to estimate the technical efficiencies and the latter is used to estimate the cost
efficiencies of Indonesian banks over the period 2002-2010. This thesis also examines the
lending propensity of Indonesian banks reflected by the tendency of their total loans and
small business loans over the study period. The findings suggest that the technical efficiency
of Indonesian banks tends to decrease whilst the opposite tendency is associated with their
cost efficiency during the period. The downward trend of their technical efficiency stems
from the fact that the period was the post-crisis of 1997-98 where banks were still unsteady
to maintain high level of outputs over inputs. While the upward trend of their cost efficiency
reflects their intensity to operate more economically in employing their resources during the
period. The latter result testifies that Indonesian bank management took lessons learnt from
the failure of their previous operations during the crisis. In addition, their total lending
propensity is prone to increase over time during the period albeit they have not reached yet
the optimum proportions. In contrast, their lending propensity to small businesses witnessed
a diminishing pattern over the period. A regulatory change in 2001 seems to discourage
Indonesian banks to lend to small businesses. Loans to deposits or lending proportions
emerges to be the most important factor enhancing bank efficiency, whilst bank size and
bank deposits are the foremost factors influencing the lending propensity of Indonesian
banks over the study period.
Mokhamad Anwar
1World Economic Outlook (International Monetary Fund). April 2012 Edition.
III
This thesis is dedicated to my family
IV
Acknowledgment
First of all, I give thanks to Allah for all his special gifts with endless enjoyment and
happiness in life. I would like to thank Prof. Meryem Duygun and Dr. Mohamed Shaban
who have always given a lot of inspiration and guidance with kindness and sincerity
during my study at the PhD programme, School of Management, University of
Leicester.
I also thank the people of the University of Leicester, School of Management (ULSM)
especially Prof. Jo Brewis, Dr. Dimitris Papadopoulus, and Dr. Sarah Robinson (PhD
programmes Directors) and also Mrs. Teresa for all their help and kindness in guiding
me during my PhD studies at ULSM.
I would also like to thank all the members of my family; especially my mother
(H.Aminah), my father-in-law (H.E.Marwana), my wife (Aam Salamah), my children
(Fakhri & Fadli) and all my brothers and sisters as well as all my relatives who always
gave me their support and prayed for the success of my study with sincerity. Without
their support I would not have been able to finish my work.
My gratitude also goes to my friends: Saleem, Ting and Carol for their kind support
from the very beginning stages of my PhD journey until I finally finished this thesis.
Thanks also to the PhD Cohort at the ULSM. Thanks are also addressed to all my
friends from Masleics (Malaysian & Indonesian Societies in Leicester).
I would also like to thank Prof. Emmanuel Haven, Dr. Alexandra Dias, Dr Tomasz P.
Wisniewski, Dr. Christophe Schinkus, Dr Zsuzsanna Vargha as well as Prof. Meryem
Duygun and Dr. Mohamed Shaban over their trusts on me to work as an Associate
Tutor for the courses they foster at the School of Management.
My thanks are also addressed to Prof. Subal Kumbhakar and Prof. Ragnar Tveterås for
giving me the opportunity to participate in Summer School 2011 so that I could learn
SFA efficiency techniques at Stavenger Universitetet, Norway.
I am really thankful to the Directorate General of Higher Education (DIKTI),
Government of Indonesia, which has provided me a scholarship to study towards a PhD
at the ULSM. Many thanks also to all my colleagues at the Faculty of Economics and
Business, Universitas Padjadjaran (FEB UNPAD) Indonesia especially Dr. Nury Effendi
(Dean of FEB), Prof. Ernie T. Sule, Prof. Armida Alisjahbana, Dr. Sulaeman Rachman
and Dr. Poppy Rufaidah, who always encouraged me to study further and provided me
recommendations. Thanks also to my colleagues at the Center for Management Studies
(LMFE UNPAD) which is currently led by Dr. Aldrin Herwany) and Center for
Business Incubation, PT. Telkom-LMFE UNPAD.
Many thanks are also addressed to Prof. Jonathan Williams (Bangor Business School)
and Prof. Peter Jackson (ULSM) for their very valuable comments on this thesis.
Finally I am really grateful to all the academics of the University of Leicester and the
employees of the School of Management at the University of Leicester.
V
Table of Contents
Abstract.....................................................................................................................................II
Acknowledgment .................................................................................................................... IV
Abbreviations.........................................................................................................................XII
1. Chapter one: Introduction ................................................................................................... 1
1.1. Motivation............................................................................................................................2
1.2. Objectives and Contribution ............................................................................................ 5
1.3. The Choice of Methodology .............................................................................................. 6
1.4. Thesis Organisation ........................................................................................................... 8
2. Chapter Two: The Indonesian Banking Sector and Small Business Finance ............... 11
2.1. Introduction......................................................................................................................11
2.2. Industry Analysis of the Indonesian Banking Sector ................................................... 13
2.2.1. The Origin of Indonesian Banks ............................................................................ 13
2.2.2. Financial Deregulation Package (1983-1993) ........................................................ 18
2.2.3. The 1997-98 Asian Economic Crisis ...................................................................... 21
2.2.4. After the 1997-98 Crisis .......................................................................................... 23
2.2.5. The 2008 World Economic Crisis .......................................................................... 25
2.2.6. After the 2008 Crisis................................................................................................ 26
2.3 Small Businesses Finance and Total Finance ................................................................. 27
2.3.1 Definition of a Small Business (SB) and Small Business Finance (SBF) ............. 27
2.3.2. Definition of Small Businesses Finance in Indonesia ........................................... 28
2.3.3. The Regulation of minimum SBF Threshold in the Indonesian Commercial
Banks .................................................................................................................................. 30
2.3.4. The Development of Small Business Finance in Indonesia over the period 2002-
2009 ..................................................................................................................................... 31
2.3.5. The Development of Total Loans of Indonesian Banks during 2002-2010......... 34
2.4. Conclusions.......................................................................................................................38
3. Chapter three: Literature Review ..................................................................................... 40
3.1. Introduction......................................................................................................................40
3.2. Theoretical and Empirical Literature ........................................................................... 41
3.2.1. The Theory of Efficiency ........................................................................................ 41
3.2.1.1. The Brief Theory of Production Efficiency ........................................................ 41
3.2.1.2. Techniques of Measuring Efficiency ................................................................... 43
3.2.1.2.1. Non-parametric Techniques ............................................................................. 43
3.2.1.2.2. Parametric Techniques ..................................................................................... 44
3.2.2. The Empirical Literature on the Efficiency and Productivity Studies ............... 45
VI
3.2.2.1. Efficiency and Productivity Studies on Indonesian banks................................ 49
3.2.2.2. Efficiency and Productivity Studies in other Emerging Economies Banking . 52
3.2.2.2.1. South East Asia (SEA) ...................................................................................... 52
3.2.2.2. East Asia (EA) ....................................................................................................... 55
3.2.3. A Brief Theory of Loan Management in Commercial Banks ............................. 58
3.2.4. The Empirical Literature on the Bank Lending Propensities ............................. 61
3.3. Conclusion.........................................................................................................................66
4. Chapter Four: Methodology .............................................................................................. 69
4.1. Introduction.......................................................................................................................69
4.2. A Comparison of the DEA and SFA Approaches ......................................................... 70
4.3. Non-parametric Methods: Data Envelopment Analysis (DEA) .................................. 74
4.3.1. Technical Efficiency (DEA) .................................................................................... 75
4.3.2. The CCR and BCC Models .................................................................................... 75
4.3.3. Nonparametric Productivity Measurement: DEA-based Malmquist
Productivity Index ............................................................................................................. 78
4.3.3.1. The Malmquist Index of Productivity ................................................................ 79
4.4. Parametric Methods: Stochastic Frontier Analysis (SFA) .......................................... 81
4.4.1. The Stochastic Production Frontier....................................................................... 81
4.4.2. Estimating the parameters ..................................................................................... 82
4.4.2.1. Ordinary Least Squares (OLS) Estimation ....................................................... 82
4.4.2.2. Maximum Likelihood (ML) Estimation ............................................................. 83
4.4.2.3. Estimating Efficiency ........................................................................................... 84
4.4.2.4. The Stochastic Cost Frontier ............................................................................... 84
4.4.2.5. Stochastic Frontier Models for Panel Data. ....................................................... 86
4.4.2.5.1. Time-invariant Inefficiency Models ................................................................. 88
4.4.2.5.2. Time-varying Inefficiency Models ................................................................... 89
4.4.2.5.3. Environmental effects model ............................................................................ 90
4.4.2.5.4. Battese and Coelli 1995 Model ......................................................................... 92
4.5. Multiple Regression Estimation ..................................................................................... 93
4.5.1. Multiple Regressions for Panel Data ..................................................................... 93
4.5.2. Random Effect Model ............................................................................................. 95
4.5.3. Fixed Effect Model .................................................................................................. 96
4.5.4. TOBIT Regression ................................................................................................... 98
4.5.5. Two-Stage Least Squares (TSLS) Regression ..................................................... 100
4.6. Data and Variables ........................................................................................................ 100
4.6.1. The Variables Specification and Definition ........................................................ 101
VII
4.6.1.1. The Inputs and Outputs Specification .............................................................. 102
4.6.1.2. Inputs and Outputs Definition .......................................................................... 103
4.6.2. Environmental and Bank specific variables ........................................................ 104
4.6.2.1. Bank specific variables ....................................................................................... 104
4.6.2.2. Environmental variables .................................................................................... 104
4.7. Software:.........................................................................................................................104
5. Chapter Five: Small Business Finance and Indonesian Commercial Banks’ Technical
Efficiency: DEA Approach .................................................................................................. 105
5.1. Introduction.....................................................................................................................105
5.2. Descriptive Statistics ..................................................................................................... 106
5.2.1. Data of Inputs and Outputs .................................................................................. 106
5.2.2. Data of Bank-Specific Variables and Macroeconomic Variables ..................... 109
5.3. Efficiency and Productivity of Indonesian Commercial Banks ................................ 112
5.3.1. Technical efficiency of Indonesian commercial banks ....................................... 112
5.3.1.1. Overal performance of Indonesian banks ........................................................ 112
5.3.1.2. Ownership and Banks’ efficiency ..................................................................... 114
5.3.1.2. Size and Banks’ efficiency ................................................................................. 119
5.3.2. The Productivity of Indonesian commercial banks ............................................ 121
5.4. Determinants of Efficiency ........................................................................................... 123
5.5. Conclusions..................................................................................................................... 128
6. Chapter Six: Small Business Finance and Indonesian Commercial Banks’ Cost
Efficiency: SFA Approach ................................................................................................... 130
6.1. Introduction.................................................................................................................... 130
6.2. Descriptive Statistics ..................................................................................................... 131
6.2.1. Data for Estimating Cost Efficiency .................................................................... 131
6.2.2. Small Business Finance of Indonesian commercial banks ................................. 137
6.2.3. Data of Bank-Specific Variables and Macroeconomic Variables ..................... 139
6.3. Cost Efficiency of Indonesian Commercial Banks ..................................................... 141
6.3.1. Average Cost Efficiency of Indonesian Commercial Banks .............................. 141
6.3.1.1. Overall Performance of Indonesian Banks’ Cost Efficiency .......................... 142
6.3.1.2. Ownership and Banks’ Cost Efficiency ............................................................ 145
6.3.1.3. Size and Banks’ Cost Efficiency ........................................................................ 145
6.4. Determinants of Cost Efficiency ................................................................................... 152
6.5. Conclusions..................................................................................................................... 164
7. Chapter seven: The Lending Propensities of Indonesian Commercial Banks over the
Period 2002-2010 .................................................................................................................. 166
VIII
7.1. Introduction....................................................................................................................166
7.2. Previous studies ............................................................................................................. 167
7.3. The Importance of Banks’ Lending Propensities ....................................................... 172
7.4. Methodology ................................................................................................................... 173
7.4.1. Demand and Supply Model of Small Business Finance ..................................... 173
7.4.2. Two-Stage Least Square (TSLS) Technique ....................................................... 175
7.5. Data..................................................................................................................................177
7.6. Empirical results ............................................................................................................ 177
7.6.1. Descriptive Statistics ............................................................................................. 177
7.6.1.1. Descriptive Statistics of the key variables. ....................................................... 178
7.6.1.2. Distribution by Peer group ................................................................................ 179
7.6.2. Regression Results ................................................................................................. 186
7.7. Conclusions..................................................................................................................... 190
8. Chapter eight: Conclusion ............................................................................................... 195
8.1. Empirical Findings ........................................................................................................ 195
8.2. Policy implications ......................................................................................................... 201
8.3. Limitations...................................................................................................................... 204
8.4. Future Research ............................................................................................................. 205
Bibliography.......................................................................................................................... 206
APPENDICES: ..................................................................................................................... 218
Appendix 1. Figure 1.1. Thesis Objectives ......................................................................... 218
Appendix 2. Table 2.6. Indonesian Government Policy Responses in 1998 and 2008 .... 218
Appendix 3. Figure 5.4. Average technical efficiency of Indonesian banks (model 1) ... 219
Appendix 4. Figure 5.5. Average technical efficiency of Indonesian banks (model 2) ... 219
Appendix 5. Figure 5.6. Average technical efficiency of Indonesian banks (model 3) .. 219
Appendix 6. Figure 5.8. Average Technical Efficiency By Bank Peer Group (Model 2)
................................................................................................................... 220
Appendix 7. Figure 5.9. Average Technical Efficiency By Bank Peer Group (Model 3)
................................................................................................................... 220
Appendix 8. Table 5.8. Malmquist Index Summary of Annual Means (Model 2) ......... 220
Appendix 9. Table 5.9. Malmquist Index Summary of Annual Means (Model 3) ......... 221
Appendix 10. Figure 6.7. Evolution of Indonesian Banks’ Cost Efficiency – Pooled -
Model 2 ................................................................................................................... 221
Appendix 11. Figure 6.8. The Evolution of Indonesian Banks’ Cost Efficiency – BC92 -
Model 2 ................................................................................................................... 221
IX
Appendix 12. Figure 6.9. Evolution of Indonesian Banks’ Cost Efficiency – Pooled -
Model 3 ................................................................................................................... 222
Appendix 13. Figure 6.10. The Evolution of Indonesian Banks’ Cost Efficiency – BC92 -
Model 3 ................................................................................................................... 222
Appendix 14. Figure 6.13. Cost Efficiency – Pooled - Model 2 (by Bank Peer Group) . 223
Appendix 15. Figure 6.14. Cost Efficiency – BC92 - Model 2 (by Bank Peer Group) ... 223
Appendix 16. Figure 6.15. Cost Efficiency – Pooled - Model 3 (by Bank Peer Group) . 224
Appendix 17. Figure 6.16. Cost Efficiency – BC92 - Model 3 (by Bank Peer Group) ... 224
Appendix 18. Table 6.17. Coefficient of Cost Efficiency Estimation Variables – Pooled
Model 1 ....................................................................................................................... 225
Appendix 19. Table 6.18. Coefficient of Cost Efficiency Estimation Variables – BC92
Model 1 ....................................................................................................................... 226
Appendix 20. Table 6.19. Coefficient of Cost Efficiency Estimation Variables – Pooled
Model 2 ....................................................................................................................... 227
Appendix 21. Table 6.20. Coefficient of Cost Efficiency Estimation Variables – BC92
Model 2 ....................................................................................................................... 228
Appendix 22. Table 6.21. Coefficient of Cost Efficiency Estimation Variables – Pooled
Model 3 ....................................................................................................................... 229
Appendix 23. Table 6.22. Coefficient of Cost Efficiency Estimation Variables – BC92
Model 3 ....................................................................................................................... 230
Appendix 24. First Regressions in completion to Table 7.3 (The Regression Results 1)
...................................................................................................................................... 230
Appendix 25. First Regressions in completion to Table 7.4 (The Regression Results 2)
...................................................................................................................................... 230
Appendix 26. Correlation Test........................................................................................... 230
List of Tables:
Table 2.1. Major Economic Reforms (1966-1982) ............................................................... 17
Table 2.2. Income and Inflation Indicators .......................................................................... 21
Table 2.3.Total Indonesian commercial banks, Pre- and Post-Crisis ................................ 22
Table 2.4. Indonesian Economic Indicator, 2006-2010 ....................................................... 25
Table 2.5. Indonesian Macroeconomic & Financial Indicator, 2005-2010 ........................ 26
Table 2.7. The Highlights of SBF and MSMEs Finance ..................................................... 33
Table 2.8. The Highlights of Total Loans & the Profitability of Indonesian Banks ......... 34
Table 2.9. Productive Loans & Total Loans of Indonesian Banks ..................................... 35
Table 3.1. A commercial bank’s balance sheet .................................................................... 58
Table 4.1.The Comparison between SFA and DEA methods ............................................. 73
X
Table 5.1. Inputs and Outputs ............................................................................................. 106
Table 5.2. The Proportion of Small Business Finance ....................................................... 107
Table 5.3. Bank-Specific and Macroeconomic Variables .................................................. 109
Table 5.4. Ranking of Indonesian Banks Efficiency (2002-2010) ..................................... 116
Table 5.5. Distribution of Technical Efficiency By Bank Peer Group .. Error! Bookmark not
defined.
Table 5.6. Malmquist Index Summary over the period 2002-2010 .................................. 121
Table 5.7. Malmquist Index Summary of Annual Means (Model 1) ............................... 122
Table 5.10. Correlation Matrix of Independent Variables-1 ............................................ 125
Table 5.11. Correlation Matrix of Independent Variables-2 ............................................ 125
Tabel 5.12. Determinants of Indonesian banks’ technical efficiency (2002-2010) .......... 126
Table 6.1. Output Components ........................................................................................... 131
Table 6.2. Total Cost and Price of Inputs ........................................................................... 132
Table 6.3. Total Costs to Total Assets ................................................................................. 132
Table 6.4. Expenses of Indonesian banks (2002-2010) ...................................................... 134
Table 6.5. Price of Funds (2002-2010) ................................................................................. 135
Table 6.6. Price of Labour (2002-2010) .............................................................................. 136
Table 6.7. Price of Capital (2002-2010) ............................................................................... 136
Table 6.8. Banks’ Small Business Finance by Different Ownership & Operation (2002-
2010) ...................................................................................................................................... 137
Table 6.9. Bank-Specific and Macroeconomic Variables .................................................. 140
Table 6.10. Ranking of Indonesian Banks Cost Efficiency (2002-2010) .......................... 146
Table 6.11. Distribution of Average Cost Efficiency By Bank Peer Group .................... 146
Table 6.12. Correlation Matrix of Independent Variables-1 ............................................ 153
Table 6.13. Correlation Matrix of Independent variables-2 ............................................. 154
Tabel 6.14. Determinants of banks cost efficiency – Model 1 ........................................... 154
Tabel 6.15. Determinants of banks cost efficiency – Model 2 ........................................... 159
Tabel 6.16. Determinants of banks cost efficiency – Model 3 ........................................... 162
Table 7.1. Descriptive statistics of the key variables ......................................................... 178
Table 7.2. Distribution by Peer group ................................................................................ 180
Table 7.3. TSLS Regression Results 1................................................................................. 187
Table 7.4. TSLS Regression Results 2................................................................................. 189
List of Figures:
Figure 2.1. The Indonesian Banking Architecture (IBA) ................................................... 24
Figure 2.2. The Trend of Loans of Indonesian Banks (2002-2010) .................................... 36
XI
Figure 2.3. The Trend of Loans Proportions of Indonesian Banks (2002-2010) ............... 36
Figure 2.4. The Trend of Loans Growth and GDP Growth (2002-2010) .......................... 37
Figure 3.1. Technical and Allocative Efficiency ................................................................... 42
Figure 5.1. Average technical efficiency of the Indonesian banks (Model 1,2 and 3) ..... 112
Figure 5.2. Numbers of sharia banks and Efficiencies Model 2 ....................................... 117
Figure 5.3. SBF percentage and Efficiencies Model 2 ....................................................... 118
Figure 5.7. Average Technical Efficiency By Bank Peer Group (Model 1) ..................... 120
Figure 6.1. Banks’ Small Business Finance Proportions (2002-2010) .............................. 138
Figure 6.2. Average Cost Efficiency of Indonesian Banks (Pooled-Model 1,2,3) ............ 142
Figure 6.3. Average Cost Efficiency of Indonesian Banks (BC92-Model 1,2,3) .............. 143
Figure 6.4. Cost Efficiency of the Indonesian Banks: By Different Ownership .............. 145
Figure 6.5. The Evolution of Indonesian Banks’ Cost Efficiency – Pooled - Model 1 .... 148
Figure 6.6. The Evolution of Indonesian Banks’ Cost Efficiency – BC92 - Model 1 ...... 149
Figure 6.11. Cost Efficiency – Pooled - Model 1 (by Bank Peer Group) ......................... 151
Figure 6.12. Cost Efficiency – BC92 - Model 1 (by Bank Peer Group) ......................... 1511
Figure 7.1. Average TLTA across different groups ........................................................... 182
Figure 7.2. Average TLTD across different groups ........................................................... 183
Figure 7.3. Average TL across different groups ................................................................ 184
Figure 7.4. Average SBLTL across different groups......................................................... 184
Figure 7.5. Average SBLTA across different groups ........................................................ 185
Figure 7.6. Average SBL across different groups .............................................................. 186
XII
Abbreviations
ADB Asian Development Bank
BC Battese and Coelli
BCC Banker, Charnes and Cooper
BI Bank Indonesia
BILC Bank Indonesia Liquidity Credit
BNI Bank Negara Indonesia
BRI Bank Rakyat Indonesia
BTN Bank Tabungan Negara
CAR Capital Adequacy Ratio
CCR Charnes, Cooper and Rhodes
CDO Collateral Debt Obligation
CE Cost Efficiency
CMAD Corrected Mean Absolute Deviation
CNY Chinese Yuan
COLS Corrected Ordinary Least Squares
CONV Conventional Banks
CPI Consumer Price Index
CRS Constant Returns to Scale
DEA Data Envelopment Analysis
DFA Distribution Free Approach
DMU Decision Making Unit
EA East Asia
EC Efficiency Change
EFFCH Efficiency Change
FA Fixed Assets
FB Foreign Banks
FDH Free Distribution Hull
FDIC Federal Deposit Insurance Corporation
FE Foreign Exchange
FEB Foreign Exchange Banks
FEM Fixed-Effects Model
GA General and Administrative expenses
GDP Gross Domestic Products
GDPGR Gross Domestic Products Growth
GOI Government of Indonesia
IBA Indonesian Bank Architecture
IBRA Indonesian Bank Restructuring Agency
ICT Information and Communication Technology
IDIA Indonesian Deposit Insurance Agency
IDR Indonesian Rupiahs
IF Instrumental Variable
IFC International Finance Corporation
IMF International Monetary Fund
INFL Inflation rate
ISE Indonesia Stock Exchange
JV Joint Venture Banks
KUR Kredit Usaha Rakyat (People‟s Business Credit)
LDR Loans to Deposits Ratio
XIII
LGOB Local Government Banks
LNTA Natural logarithm of total assets
LNTL Natural Logarithm of Total Loans
LOI Letter of Intent
LP Linear Programming
LR Likelihood Ratio
MKTINDEX Market Index
ML Maximum Likelihood
MPI Malmquist Productivity Index
MSME Micro, Small, and Medium Enterprises
MTFPI Malmquist Total Factor Productivity Index
NFEB Non-Foreign Exchange Banks
NICA Netherlands-Indies Civil Administration
NOR New Order Regime
NPL Non-Performing Loans
OF Other Finance
OI Other Income
OL Other Loans
OLS Ordinary Least Square
OOR Old Order Regime
PAKFEB the package of February 1991
PAKJAN the package of January 1990
PAKJUN the Package of June 1983
PAKMEI the Package of May 1993
PAKTO the package of October 1988
PB Private Banks
PC Price of Capital
PEFFCH Pure Technical Efficiency Change
PF Price of Fund
PL Price of Labour
REM Random-Effects Model
ROA Return on assets
SB Small Business
SBF Small Business Finance
SBL Small Business Lending
SBLTA Small Business Loans to Total Assets
SBLTD Small Business Loans to Total Deposits
SBLTL Small Business Loans to Total Loans
SEA South East Asia
SECH Scale Efficiency Change
SFA Stochastic Frontier Analysis
SHARIA Sharia (Islamic) Banks
SI Securities and Investment
SOB State-Owned Banks
TCI Technological Change Index
TCTA Total Costs per Total Assets
TD Total Deposits
TECHCH Technological Change
TF Total Finance
TFA Thick Frontier Approach
XIV
TFP Total Factor Productivity
TFPCH Total Factor Productivity Change
TL Total Loans
TLTA Total Loans to Total Assets
TLTD Total Loans to Total Deposits
TSLS Two-Stage Least Squares
UNEMP Unemployment rate
USD United States Dollars
VRS Variable Returns to Scale
WB World Bank
WTO World Trade Organization
1
1. Chapter one: Introduction
The role of commercial banks is very important in an economy. They serve customers‟
needs either by providing saving and deposit accounts as well as providing lending or
financing facilities. For some countries, the contribution of banks is very dominant as a
major source of funds for supporting the development of public sector as well as private
entities.
Considering their essential contributions, it is very crucial to assess their performance.
The performance of banks reflects their ability and quality of their management in
carrying out their operations efficiently. Mishkin (2010) and Saunders and Cornett
(2012) indicate that commercial banks mainly create profit by issuing loans or financing.
The contribution of loans in their revenues is regarded more than fifty percent of the
operations.
Efficiency performance is required to reveal the efficiency level of banks in terms of
managing their inputs to create outputs. In addition, the lending or financing profile of
banks in terms of their lending propensity is also useful to identify their loan
management. Macerinskiene and Ivaskeviciute (2008) suggest the importance of loan
management considering that improper loan management in turn has a negative impact
on the bank performance. Hence, the investigation of bank performance is very vital to
be carried out either in terms of bank efficiency as well as its lending propensity.
Indonesia is known as a developing country and a member of the Group of 20
economies (G20). The country continues to develop and support all of the sectors‟
economic activities by stimulating the role of commercial banks in the provision of
funds. Past experience of the 1997-98 financial crises has led Indonesian policy makers
and the management of commercial banks in Indonesia to create a more supportive
environment for Indonesian banks to exceed the last decade of the 2000s in which the
world economic crisis of 2008 occurred.
2
This thesis presents insight analysis of Indonesian bank performance in terms of bank
efficiency as well as bank lending propensity during the period 2002-2010. The period
is considered to be a recovery period post the 1997-98 crises and after the government
of Indonesia (GOI) undertook some significant policies during the period 1997-2000.
The period is also regarded as the implementation period after GOI issued the special
regulation to relax the commercial banks in providing small businesses finance (SBF) in
2001. The thesis examines the technical efficiency, cost efficiency as well as the lending
propensity of Indonesian banks during the period.
1.1. Motivation
During the late 1990‟s, Southeast Asian countries were engulfed by a financial turmoil
which later on was known as the Asian Financial Crisis 1997-98. Among other Asian
countries, Indonesia was regarded the most suffered country in the region reflected by
the sharp currency depreciation over the period [Enoch et al (2001), Fane and McLeod
(2002), Harada and Ito (2005), Sato (2005)].
This crisis has left the Indonesian banking sector in a fragile state leading to the possible
reasons that the International Monetary Fund (IMF) intervened the Government of
Indonesia (GOI) by proposing several adjusted programmes started in1997. The GOI
then carried out several actions as policy responses to the recommendations of the IMF
to overcome the problems of bank liquidity and solvency over the subsequent years. The
period of 1997-2001 testifies the era of banking collapse and restructuring in Indonesia
(Fane and McLeod, 2002).
The GOI commenced some restructuring programmes i.e. the liquidation of 16 banks on
November 1997; the establishment of the Indonesian Bank Restructuring Agency
(IBRA) on January 1998 to restructure and manage the assets of some particular banks;
the recapitalisation program2 to inject the commercial banks by government funds to
2 The recapitalisation program valued at IDR 658 trillion was equal to 52 per cent of GDP in 2000 (Sato,
2005).
3
empower banks‟ assets and capital; and the programme of mergers and acquisitions over
a number of banks as well as the liquidation of some particular banks3.
All of these programmes had some impact on the structure of Indonesian banks during
the period 1996-2001: 1) the total number of banks was reduced from 239 banks at the
end of 1996 to 145 banks at the end of 2001. 2) the total assets of banks to nominal
GDP declined from 72.8 per cent in 1996 to 70.9 per cent in 2001. 3) the loans to
deposits ratio dropped from 104.0 per cent in 1996 to only 38.0 per cent in 2001and 4)
gross non-performing loans increased from 9.3 per cent in 1996 to 58.7 percent in 1998
and 12.1 per cent in 2001 (Sato, 2005).
In terms of lending behaviour, before the crisis of 1997-98, banks were more focused on
corporate and commercial sectors in their loan portfolios. It is believed that these
segments of loans are more risky compared to loans to the small business sector. Small
business loans tend to be less risky because they are diversified over more borrowers.
During the Asian crisis period, in Indonesian commercial banks the non-performing
loans for the small business portfolio sector were much lower than those of other sectors
(Timberg, 1999). This indicates that loans to small businesses in Indonesia were more
resilient over the crisis period.
In fact, GOI through the Ministry of Finance and Central Bank of Indonesia (BI) always
give considerable contributions for the development of small businesses through the
issuances of some policies. One of the most influential policies is the implementation of
the obligations of commercial banks to provide small business loans with a certain
minimum percentage (minimum threshold) of their total loans. GOI through the policy
package of January 1990 (PAKJAN 90) imposed a compulsory clause for each
commercial bank operating in Indonesia to have a minimum of 20 per cent of their
portfolio had to be devoted to small business loans (finance). The package was then
revised by The Directors of BI Decree No. 30/4/KEP/DIR on April 1997 which stated
that the threshold had changed to 22.5 per cent of a bank‟ net loan expansion.
3 The liquidation was applied to the banks included in the Category C which had a capital adequacy ratio
(CAR) of less than minus 25 per cent (Fane and McLeod, 2002).
4
However, BI subsequently launched a controversial policy in 2001 when it abolished
the provision of minimum threshold of small business finance through the BI regulation
No. 3/2/PBI/2001. These regulations eradicate the compulsory clause of commercial
banks to fulfil a minimum threshold of their financing devoted to small businesses. The
portion of a bank‟s small business finance was then adjusted by the condition of each
bank. Banks were only required to include the projections of their small business
finance in their business plans and convey the achievement of their loans on their
published annual reports. The policy of 2001was considered unfavourable in terms of
discouraging commercial banks in Indonesia to provide financing for small businesses
whereas the loan quality of this scheme was much better than those of loans for other
sector.
The thesis focuses only on the period 2002 to 20104 considering that in the year 2012
the Government of Indonesia turned back to issue the regulation (BI Regulation
No.14/22/PBI/2012) to reapply the minimum threshold of 20% bank loan portfolios
should be devoted to micro, small and medium enterprises (MSMEs). The difference
from the latter regulation is that the obligation for commercial banks to fulfil the
minimum financing portfolio by 20% is not only intended to Small businesses (SBs) but
also to micro, small and medium enterprises (MSMEs). Indeed, SBs is a part of
MSMEs.
Some policies undertaken over the period 1997-2001 in terms of bank restructuring as
well as the abolishment of the minimum threshold for small business finance may have
had any impact on Indonesian banks performance in terms of bank efficiency as well as
lending propensity during 2002-2010. Hence, this has been one of the major motivations
of this thesis.
Another motivation of this thesis is that studies about the performance of Indonesian
banks particularly in terms of „bank efficiency and lending propensity‟ are quite rare to
date. There are few studies in the literature that discussed these issues on Indonesian
bank efficiency, but they mainly use one method [Margono et al (2010) and Besar
(2011) use the SFA method, whilst Harada and Ito (2005), Hadad et al (2011a, b),
4 The data of the year 2011 was not completely available during the process of writing the thesis.
5
Suzuki and Sastrosuwito (2011) employ the DEA method, in estimating Indonesian
bank efficiency, respectively]. This circumstance has motivated the thesis to include
both the DEA and SFA techniques in estimating Indonesian bank efficiency over the
period 2002-2010. In addition, to the best of my knowledge, there has not been any
research completed yet on the lending propensity of Indonesian banks to date. Hence,
this also motivates this thesis to reveal Indonesian bank lending propensity especially in
terms of total loans as well as small business loans over the period 2002-2010.
1.2. Objectives and Contribution
This thesis aims to measure and analyse the performance of Indonesian banks in terms
of efficiency and productivity. The study employs non-parametric Data Envelopment
Analysis (DEA) and parametric Stochastic Frontier Analysis (SFA) techniques in order
to estimate technical and cost efficiency respectively during the period 2002 to 2010.
This period can be described as a recovery period hence it is after the Government of
Indonesia (GOI) implemented some restructuring programmes in banking. The thesis
also investigates whether bank specific characteristics and the environmental variables
determine the efficiency level of Indonesian banks. Finally, the thesis also provides an
empirical analysis by examining the lending propensities of Indonesian banks as well as
their determinants over the study period. The study expands its analysis of Indonesian
bank performance by measuring bank productivity using the DEA Malmquist total
factor productivity index. The main objectives of this thesis are reflected in the figure
1.1 (see Appendix 1).
In summary, the research questions of this thesis are:
1. What was the performance of Indonesian banks in terms of DEA technical
efficiency and productivity during the recovery period (2002 to 2010)?
2. Did environmental and bank-specific variables affect the efficiency levels of
Indonesian banks in terms of DEA technical efficiency over the recovery
period?
3. What was the performance of Indonesian banks in terms of SFA cost efficiency
during the recovery period?
6
4. Did environmental and bank-specific variables affect the efficiency levels of
Indonesian banks in terms of SFA cost efficiency over the recovery period?
5. What were the lending propensities of Indonesian banks over the recovery
period and after GOI launched the regulatory change in 2001 on the abolishment
of minimum threshold of bank‟s small business finance?
6. What were the determinants of bank lending propensity of Indonesian banks
over the recovery period?
The major contribution of this thesis comes in three folds: i) it is the first empirical
study which measures the efficiency of Indonesian banks using both DEA and SFA over
the recovery period (2002-2010) with a sample that represents 93% of the banking
sector. ii) It is the first study on Indonesian banks that examines the impact of
environmental and bank-specific variables on the efficiency levels of Indonesian banks
particularly in terms of DEA technical efficiency and SFA cost efficiency. iii) Finally,
this is the first study to investigate the lending propensity of Indonesian banks and their
determinants over the recovery period (2002 to 2010).
1.3. The Choice of Methodology
In order to answer the objectives of this thesis in terms of estimating Indonesian bank
efficiency performance, this thesis employs a frontier approach. The approach measures
the performance of a bank by relatively comparing it as a distance from the best practice
frontier (the „best practice‟ bank(s)). This approach provides two prominent methods,
namely the DEA non-parametric method and the SFA parametric method. The DEA
non-parametric method estimates a firm‟s efficiency through mathematical linear
programming while the SFA parametric method estimates a firm‟s efficiency through
econometric principles.
This thesis adopts both methodologies to obtain efficiency estimates of the Indonesian
commercial banks during the period 2002-2010. In estimating the technical efficiency of
Indonesian banks, DEA technical efficiency is employed because of the possibility that
this measure decomposes the technical efficiency into two parts: “pure technical
efficiency” and “scale efficiency”. In addition, the first empirical chapter of this thesis
7
(chapter five) examines the efficiency levels of Indonesian banks by only accounting for
the conversion of physical inputs (general and administrative expenses, fixed assets, and
total deposits) into outputs (total finance, small business finance, other finance, and
securities and investments) relative to best practice. This efficiency measurement is
based on only the technical process without regarding price and cost.
While SFA cost efficiency is used in the second empirical chapter (chapter six),
considering that in the measurement of economic or cost efficiency of a bank, it
involved input prices as well as output quantities in the model. Cost efficiency is
composed of two components: allocative efficiency and technical efficiency. The former
computes the ability of a firm to employ inputs in optimal proportions given their prices
while the later measures the firm's ability to obtain the maximum output from given
inputs (Coelli et al, 2005). Unlike the DEA approach it is cumbersome to disentangle
allocative efficiency from scale efficiency given the cost efficiency measure obtained
from the SFA model.
Regarding the estimation of the SFA cost efficiency, the thesis uses Battese and Coelli
1992 (BC92). BC92 (the error correction model) is employed considering the focuses of
this study to use two-step method as it is also applied to the first empirical chapter and
this practice was also adopted by other authors [(Pitt and Lee, 1981), (Andries, 2011),
(Assaf and Josiassen, 2012), etc]. There are two alternatives using two step methods,
those are Kumbakar 1990 and Battese and Coelli 1992. The advantage of the Battese
and Coelli 1992 approach over that of Kumbakhar 1990 is that their model involves
only one unknown parameter whilst the latter involves two unknown parameters (Assaf
and Josiassen, 2012).
Indeed, the BC 1992 is less flexible compared to the BC 1995 particularly in terms of
concerning the yearly rank order over time. Since this study focuses on the average
basis of the efficiency estimation (achieved by commercial banks in Indonesia) over the
study period, not yearly basis, it is considered sufficient to estimate cost efficiency using
the BC1992. In addition, the study also employs the standard pooled method to enrich
the analysis. The findings of both methods have provided us in understanding the cost
efficiency attained by Indonesian commercial banks during the period.
8
Actually there are some alternative models could be employed instead of cost efficiency
estimation (i.e. technical efficiency, profit efficiency, etc) in terms of SFA technique.
This thesis uses the SFA cost efficiency estimation concerning the fact that during the
recovery period Indonesian banks embraced the consequences from the restructuring
programme (mergers, acquisitions, etc) undertaken by the Government of Indonesia i.e.
employee layoff, bank branches closures, etc. These could be regarded as the cost
efficiency programme initiated by Government of Indonesia during the period 1997-
2001. These circumstances are assumed to have any impacts on the cost efficiency
performance of Indonesian banks over 2002 to 2010. These become the reasons why the
cost efficiency model is employed in this thesis. Other authors [(Margono et al (2010)
and Besar (2011)] also used the SFA cost efficiency models in their research.
In revealing the factors affecting both efficiency levels of Indonesian banks, this thesis
uses TOBIT regression given that the efficiency scores (dependent variables) have the
features of censored value between 0 and 1. In addition, this thesis adopts two-stage
least square regression in estimating the determinants of Indonesian banks‟ lending
propensity over the study period. This regression accounts for the simultaneity problem
between the supply and the demand function for loan provisions. This method was also
adopted by McNulty et al (2013).
1.4. Thesis Organisation
This thesis is composed of eight chapters including the current introduction chapter.
Chapter two provides information about the circumstances of the Indonesian banking
sector and small business finance in details. Chapter three reviews previous studies and
chapter four presents the methodologies employed to obtain the empirical results.
Chapter five, six and seven provide analyses of the empirical findings covering the DEA
technical efficiency and the SFA cost efficiency of Indonesian commercial banks as well
as Indonesian bank lending propensity over the study period. Finally, chapter eight
provides conclusions for the empirical results of the thesis.
The contents of each chapter are summarised as follows:
9
Chapter two describes the history and the evolution of the Indonesian banking industry
since the first commercial bank was established in Indonesia up to date. The chapter
also explains the political and economic aspects through the policies imposed by the
Government of Indonesia (GOI) in managing the banking Industry for different eras i.e.
The post-Independence period, the deregulation era, until the post-Asian financial crisis
1997-98 that hit the Indonesian banking industry. The last two sections of this chapter
reveal the efforts of GOI (through Bank Indonesia) to reconstitute the new architecture
of Indonesian banking. The last section explains the conditions of small business
finance covering the definition and the legislation that prevails in Indonesia particularly
within the study period.
The third chapter reviews the literature examining the study of bank efficiency as well
as bank lending propensity. The chapter reviews the theory of production efficiency
revealing the Farrel‟s (1957) propositions on technical efficiency, allocative efficiency,
and economic efficiency as well as describing the techniques used to estimate
efficiency. The chapter then exposes some empirical evidence on efficiency and
productivity studies in Indonesia as well as in some emerging economies in South East
Asia and the East Asia region. The chapter also describe the brief theory of loan
management in commercial banks and then provides some empirical evidence on the
lending propensity of commercial banks.
Chapter four initiates with a brief summary on the comparison between the non-
parametric DEA approach and the parametric SFA approach. The chapter then
elucidates in detail the DEA approach in estimating efficiency as well as the DEA-based
Malmquist productivity index computation especially for the production function on
technical efficiency. The following section describes the parametric SFA method with
some underlying estimation models specifically for the cost frontier. This chapter also
presents the multiple regression techniques including TOBIT regression for examining
the determinants of efficiency and the two-stage least square (TSLS) regression for
determining the lending propensities of commercial banks.
Chapter five analyses the tendency of the technical efficiency of Indonesian commercial
banks as well as their determinants over the study period. The chapter comprises two
major sections. The first section discusses the empirical findings of the technical
10
efficiencies of Indonesian banks in terms of the non-parametric DEA approach as well
as the DEA-Malmquist productivity index. It then compares the performance of bank
efficiency across different groups based on their ownership and operation. The second
section analyses the determinants of Indonesian bank technical efficiency over the study
period. The section reveals some significant factors either from bank specific factors
and macroeconomic variables that significantly impact on the Indonesian bank
efficiency over the study period.
Chapter six provides an empirical analysis of Indonesian bank efficiency particularly in
terms of cost efficiency resulting from the estimation of the SFA parametric method
over the study period. The chapter also presents a comparison of cost efficiency
performance across different groups of banks by ownership and operation. The last
section of this thesis then analyses the determinants of Indonesian bank cost efficiency
during the period.
Chapter seven discusses the empirical evidence of Indonesian commercial banks
regarding their lending propensity over the period 2002 to 2010. This chapter follows
McNulty et al (2013) in revealing the lending propensity of Indonesian banks across
different groups of banks and examines the determinants of the lending propensity of
Indonesian banks.
The thesis provides concluding remarks in the last chapter (chapter eight). The chapter
begins with a summary of the empirical findings followed by policy implications on
Indonesian bank efficiency as well as lending propensity. Finally, the chapter concludes
by summarizing the limitations of the thesis and suggesting new opportunities for future
research.
11
2. Chapter Two: The Indonesian Banking Sector and Small Business Finance
2.1. Introduction
The banking sector in Indonesia has been present since the pre-independence period.
The Dutch colony authority holder5 in Indonesia established some commercial banks to
operate in several cities of Indonesia in the 18th century6. The Indonesian banking has
experienced various periods since then (i.e. the pre- and post-independent period, the
deregulation period, the 1997-1998 Asian financial crises period and the post-Asian
financial crises period).
The period between 1980s-1990s can be described as the deregulation7 era of the
Indonesian banking sector. The aim of the deregulation policies is to improve the
effectiveness and efficiency of the Indonesian banking sector. This in turn aims to have
a positive impact on the real sector in terms of stimulating the development of the
Indonesian economy. The deregulation process consists of some packages launched by
the Government of Indonesia (GOI) during the period 1983-1993. These packages8 [i.e.
the package of June 1983 (PAKJUN 83), the package of October 1988 (PAKTO 88), the
package of January 1990 (PAKJAN 1990), the package of February1991 (PAKFEB 91),
and the package of May 1993 (PAKMEI 93)] have brought the Indonesian banking
sector to grow very rapidly.
The rapid development of Indonesian banks, especially in terms of increasing the
number of banks has not been accompanied by the sound quality of bank management.
Thus, when the financial crisis occurred in neighbouring countries (e.g. Thailand,
Malaysia), the effects easily spread to Indonesia, starting in mid-1997.
During the crises period (1997-1998), the Indonesian banking sector was at the worst
condition. Many commercial banks were forced into liquidation9or merged with other
5 As a Dutch colony, Indonesia was under the authority of a Governor General.
6 See more details in the section 2.2.1.
7 Deregulation means the process of eliminating or reducing state regulations. In this case, the
deregulation of the banking sector in Indonesia. 8 See in detail at the Section 2.2.2 in this chapter.
9 It began with the liquidation of 16 banks in November 1997, followed by the closure of other 38 banks
in 1999 as the impact of the financial and banking crisis in 1997-1998 in Indonesia.
12
banks (arranged by the Central Bank of Indonesia (BI)) due to their poor performance.
The government of Indonesia (GOI) then established the Indonesian Bank Restructuring
Agency (IBRA)10
on January 26, 1998, with the authorities to manage and restructure
non-performing banks as well as their non-performing assets.
The following period (2000-2010) was a period of recovery and re-establishment of the
Indonesian banking sector after suffering from the crisis. In this period, the GOI through
BI launched a program known as the "Indonesian Banking Architecture (IBA)" in 2004.
IBA was designed with the aim of achieving a sound, strong, and efficient banking
sector to promote the stability of the Indonesian financial system as well as enhance the
Indonesian economic growth (Indonesian Banking Report, 2004).
In terms of bank efficiency and bank lending propensity, the performance of Indonesian
banks over the decade of 2000s is very important to examine particularly after the
regulation11
launched in 2001. The regulation states that all commercial banks are not
obliged to maintain the minimum threshold of small business finance (SBF). The SBF
portion of each bank is then adjusted to its own ability and willingness. These
circumstances along with the previous programme undertaken by the GOI during the
period 1997-2001 (liquidation, recapitalisation, mergers and acquisitions) seem to have
impacted on the Indonesian bank efficiency as well as bank lending propensity over the
period 2002-2010.
This chapter is organised as follows. The first section discusses the historical
background of the Indonesian banking sector (e.g. the pre-independence period, the
post-independence period, the financial deregulations period, the 1997-98 Asian
economic crises period, the post-1998 Asian crises period, 2007-2008, the world
economic crisis period, and the post-2008 crises period). The second section explains
the evolution of small business finance in Indonesia in terms of the regulatory
framework and its highlights over the period 2002-2010. Finally, the last section
concludes.
10
IBRA was established as one item in a series of Letter of Intent (LOI) between the GOI and IMF with
the first LOI was signed on October 31, 1997. 11
BI Regulation No. 3/2/PBI/2001 about Small Business Finance on January 4, 2001. It states that the
provision of small business finance is readjusted based on the ability of each bank respectively.
13
2.2. Industry Analysis of the Indonesian Banking Sector
2.2.1. The Origin of Indonesian Banks
The evolution of Indonesian banking started in 1746 when Bank Van Leening was
firstly established as a commercial bank operated in Indonesia. It was the first bank to
serve banking transactions of people in Indonesia during the time. Following Bank Van
Leening other commercial banks were established to meet the needs of the economy
(Nederlandsche Handel Maatschapij in 1824, De Javasche Bank in 1828,
Escomptobank in 1857, and Nederlandsche Indische Handelsbank in 1864). All these
banks were affiliated with or under the authority of the Dutch Governor General12
in
Indonesia. In order to expand the scope of banking transactions in Indonesia, several
banks affiliated with other countries were also established [the Chartered Bank of India,
Australia and China were founded in 1859, Hongkong and Shanghai Banking
Corporation (HSBC) in 1884, Bank of China in 1915, Yokohama Specie Bank in 1919
and Mitsui Bank in 1925].
As the development of economic activities increased rapidly in different provinces
during the early 1900s, some local commercial banks were then established (Bank
Vereeniging Oey Tiong Ham was established in Semarang (1906), Chung Hwa
Shangieh Maatschapij in Medan (1913), Batavia Bank in Batavia/Jakarta (1918) and
Bank Nasional Indonesia in Surabaya (1928)). The aim of the local bank
establishments was to serve banking transactions of commercial activities in several
major cities in Indonesia during the early 1900s. These cities are four big cities in
Indonesia located in Java and the Sumatra islands.
Indonesia proclaimed the independence on August 17, 1945. Since then, Indonesia's
banking sector has entered a new era characterized by the formation of state-owned
banks. Bank Negara Indonesia (BNI) and Bank Rakyat Indonesia (BRI) were established
in 1946 as the first two state-owned banks operated in Indonesia. The aims of
establishing BNI were to facilitate new business entities and entrepreneurs through
several services including foreign exchange transactions, commercial loans and
12
During that time Indonesia was a Dutch colony, governed or managed by the Dutch Governor General.
14
government subsidized loans. While BRI was established to support the development of
rural banks spread over thousands of villages across Indonesia. Both banks have been
continuously contributing to serve their customers through an extensive network of
branches in Indonesia and several branches overseas up to the present.
Bank Tabungan Pos was then reactivated in 1950 as a legacy of the Dutch authorities
(NICA13
). Bank Tabungan Pos was operated to focus on providing loans to local
governments in terms of developing public infrastructures: market development,
electricity distribution facilities, and bus terminals. It is important to note that public
infrastructures were very crucial for many provinces and districts to stimulate economic
activities in their regions after five years of declaring the Independence.
In addition, Bank Industri Negara and Bank Tani Nelayan were also established in 1955
and 1957 respectively. Bank Industri Negara was assigned to enhance the development
of the plantation, mines and industry sectors. These sectors required the assistance of
commercial banking mainly to finance their activities. Bank Tani Nelayan was
appointed to assist farmers and fishermen in enhancing their production and marketing
activities (Emergency Law Republic of Indonesia No. 18/1957).
The period (1959-1960), witnessed the nationalisation of several Dutch banks
[Nationale Handels Bank NV was nationalised to be Bank Umum Negara, whilst
Escomptobank and Nederlandsche Handels Maatschappij were nationalised to be Bank
Dagang Negara and Bank Expor-Impor Indonesia respectively]. These last two banks14
then became very popular as state-owned banks which support commercial and trade
activities in Indonesia. Bank Dagang Negara was assigned to provide banking
transactions for customers concerning commercial and trade affairs while Bank Expor-
Impor Indonesia was in charge of supporting the export and import transactions
between its customers and their counterparts.
13
Netherlands-Indies Civil Administration (NICA) was in charge of controlling the Allied Forces Netherlands East
Indies region after Japan surrendered unconditionally to the Allies in World War II in mid-August 14th 1945. Area
located in the Dutch East Indies now Indonesia. NICA rode ally when it came to Indonesia after the end of World
War II. 14
These banks, along with the other two banks (Bank Pembangunan Indonesia and Bank Bumi Daya) are
then merged in 1998 to be PT. Bank Mandiri (Persero), a state-owned bank with the biggest assets in
Indonesia today.
15
The history of Indonesian banking arrived at a new era when the GOI established Bank
Indonesia (BI) as a central bank on July 1, 1953. BI had a duty to oversee the
operations of the entire system of commercial banks in Indonesia. This duty is
reinforced by Law no. 11/1953 on the Central Bank. Since then, the operations of all
commercial banks operating in Indonesia were under the supervision of BI.
In 1965 the GOI decided to merge BI with other state-owned banks to mark the launch
of the “single bank15
” concept. With the concept of the single bank in place, four
commercial banks, namely Bank Negara Indonesia (BNI), Bank Koperasi Tani dan
Nelayan, Bank Umum Negara and Bank Tabungan Negara (BTN) merged and were
named as Bank Negara Indonesia (BNI). The problem was laid in the context of
unification between the function of a banking supervision as well as a commercial bank
into one bank. This seemed to be unusual practice since there must be a conflict of
interests as a dual function bank, although the bank operation was divided into several
units (from unit 1 until unit 5). BI turned into BNI Unit 1 and kept its position as the
central bank. The remaining units were established based on the different focus of
financing as of a commercial bank16
. The concept finally ended and was no longer valid
after the tragedy of Black September in 196517
. The change of national leadership
regime from the old order to the new order in 1967 also sparked a desire to convert this
entity into its original state.
The period (1967-1983) witnessed the rise of the New Order Regime (NOR) under the
command of President Suharto. Under NOR the functions of Bank Indonesia was
formally restored under the provision of the Decree No. 600 of 1968 by the Minister of
Finance. In the early period of the new regime, BI began to play a strategic role in
economic development. BI has three main roles as the holder of the monetary policy
authority, the regulator and supervisor of commercial banks, as well as the regulator for
the payment system. BI was very important in this period because Indonesia was
15
The single bank concept was conducted during the Sukarno‟s regime (Old Order Regime) with the aim
of operating the monetary and banking affairs efficiently in one bank. 16
Unit 2 to unit 5 has its own function respectively as a saving bank, a trade and commercial bank, a
cooperative bank and a farmers‟ bank. 17
The tragedy was named G30S (the 30th
of September movement, 1965). It was a self-proclaimed
organization of Indonesian national armed forces members who, in the early hours of 1 October 1965,
assassinated six Indonesian army generals in an abortive coup (coup d'état). Finally, by the end of the
day, the coup attempt had failed in Jakarta and the city can be controlled under the command of Major
General Suharto.
16
situated in the economic recovery after a downturn in the economy at the end of the Old
Order regime (OOR). Economic performance during the NOR was quite well marked
by the achievement of economic growth of around 6-7 per cent a year, and the inflation
rate was reduced to under 10 per cent a year.
In addition, during the NOR period BI also played a significant role in ensuring that the
financial system including commercial banks operated efficiently. Moreover, in this
regime, State-owned banks (SOB) carried out the greater duty in the process of bank
lending channel to finance various development projects proposed by the GOI. The
program of financing was mainly focused on the development of infrastructure as well
as the agricultural and industrial sectors.
Based on Law no. 14/1967 on banking and Law no. 13/1968 on Central bank, banks are
classified based on their functions. There are 5 types of banks: bank central,
commercial banks, savings banks, development (construction) banks, and rural banks.
The central bank is the monetary authority as well as the bank regulatory and
supervisory body. Commercial banks engage in providing short-term loans, while
development-banks provide medium and long term financing, especially in the
construction purposes. A savings-bank is a bank that concentrates on customer savings
accounts while rural-banks focus on financing the agricultural sector in rural areas.
The NOR has implemented several economic reforms within the period of 1966-1982.
These reforms consist of some economic and financial reforms dealing with the
economic activities and financial transactions during the time. Those transactions are
mostly held either by the Central Bank of Indonesia and commercial banks (Ariff and
Khalid, 2005).
The completed major economic reforms 1966-1982 are shown in the following table:
17
Table 2.1. Major Economic Reforms (1966-1982)
First reform (1966-69): Economic stabilization
1967 Legalized private trade in foreign exchange through Foreign Exchange Bourse.
New Bank Act permits foreign exchange by foreign banks and joint venture banks.
1969 Approved local banks permitted to trade in foreign exchange.
Second reform (1970-78): Non-oil exporting capacity
1970 Foreign investment laws to ease entry of foreign capital, technology and skill, especially
for primary sector, capital intensive activities.
1971 Peg rupiah to USD at 415 rupiah after 10% devaluation; exchange rates unified to two,
one for all transactions and 10% lower for essential items.
Central bank takes over foreign exchange transactions; Foreign exchange bourse almost
abolished;
Major policy of directed credits to preferred firms to create industrial capacity.
1970-76 PT.Danareksa18 established to distribute national investment certificate in unit trusts.
1977 First local firm listed on the Jakarta Exchange.
1978 When the aim of making Indonesia competitive for non-oil exporting, the rupiah devalued
by 50% in November 1978.
Third reform (1978-82): Towards non-oil export capacity and foreign investments
1978 Exchange liberalization; basket peg to trading partners‟ currencies.
Special exchange rate for foreign capital and central bank takes a swap to cover the risk.
Directed credit over 1970-82 failure; reduced directed.
Source: Ariff and Khalid (2005)
The first economic reform occurred in 1966-1969 which was recognized as the
Economic stabilization. The reform was intended to manage the economy after the
worst civil unrest of 1965 with the goal of creating many new jobs to reduce
unemployment. Of those reforms, there was also a new policy that allowed foreign
banks and joint venture banks to participate in trading currencies in the foreign
exchange market. Furthermore, as can be seen from Table 2.1, local banks were also
allowed to conduct foreign exchange transactions. It seemed that the GOI induced the
participation of various groups of banks to stimulate Indonesian economic activity
during this period.
The second economic reform wave of 1971-1978 was intended to expand the non-oil
exporting capacity. These reforms were undertaken to enhance indigenous
manufacturing capacity away from the heavy dependence on manufacturing imports that
had drained foreign exchange (Ariff and Khalid, 2005). The reform also extended to the
banking sector by means of allocating a directed credit (with funds originated from GOI
channelled to State-owned banks) mostly towards companies that could create and
develop the capacity of labour intensive industries. In addition, the GOI began to further
develop the capital market by providing an opportunity for the public to invest their
18
PT Danareksa is an Indonesian State-Owned Enterprise engaged in the field of financial services. The
company was established in 1976 with the main activity in the field of capital markets and money markets
as a finance company, securities broker, underwriter, as well as investment management and mutual
funds.
18
funds in the form of unit trusts in PT. Danareksa (Mutual funds Company) and allowing
local companies to list their shares on the Jakarta Stock Exchange.
The third reform was still aimed at increasing non-oil production capacity as well as
foreign investment. In support of these objectives, the central bank seems to be more
involved with providing special rates for foreign capital and take foreign currency
swaps to cover the risk.
2.2.2. Financial Deregulation Package (1983-1993)
One of the most important stages in the evolution of Indonesian banking sector is the
period of financial deregulation during 1983-1995. In this period, the GOI launched
some deregulation packages that had brought substantial implications for the banking
industry in Indonesia then. The objectives of these packages are to develop a sound
banking system and to support a wider involvement of privately-owned banks in the
provision of funds for the development of Indonesia as a whole. One of the visible
impacts of the deregulation is the establishment of large number of private banks and
the support of government institutions given to them by utilising their functions in terms
of current accounts, saving accounts, and credit facilities.
The first deregulation package was launched on June 1983, known as PAKJUN 83 (the
policy package of June 1983). This package has several objectives: First; abolishing the
credit ceiling for commercial banks. Second; commercial banks were given the authority
to freely set their interest rates for both loans and deposits. Third; Bank Indonesia
Liquidity Credit (BILC) through commercial banks was terminated except for the
certain loans related to cooperative development and exports. Fourth; the reserve
requirement was reduced from 15% to 2% in order for banks to expand their loan
facilities to customers. All these policies succeeded in increasing the involvement of the
private sector and private banks in supporting the Indonesian economy in the era of
competition and the on-going globalization.
The second package was named PAKTO 88 (the policy package of October 1988). The
package was a trigger for the establishment of a large number of new private
19
commercial banks since the minimum paid-in capital required for establishing a bank
was very low at that time (IDR 10 billion or USD 5 million). The package also consists
of the ease of the requirements for licensing to be a foreign-exchange bank19
as well as
joint-venture banks. BI also set up a secondary banking market represented in
traditional-market banks and rural banks in all regions of Indonesia. The aim of this
program is to expand the reach of financial aid, especially in rural areas, in order to
increase their involvement in the development of the Indonesian economy.
Through these two comprehensive reform packages, the structure of banking deposits
has dramatically changed in 1989. As Ariff and Khalid (2005) elaborate there was a
movement of deposits towards private banks at large-scale. The market share of
deposits had a severe shift from 80% of the deposits at state-owned banks in 1983 to
80% of the deposits moved to private banks by the end of 1989. This shift in market
power was supported by an increase in the quality of private banking services and the
enormous branch network across the country (Ariff and Khalid, 2005).
The third package is PAKJAN 90 (the policy package of January 1990). Through this
policy, banks and financial institutions are encouraged to be more independent and are
able to carry out funding as well as lending activities effectively while gradually
reducing dependence on Bank Indonesia Liquidity Credit (BILC). The package aimed
for the following: First, Lending rates for some activities or sectors which previously
provided subsidies were transferred to the market mechanism. Second, the BILC was
granted only in limited quantities to support self-sufficiency, cooperatives development
and increased investment. In addition, to support the development of small businesses
(SBs) and in an effort to equitable development, banks are obliged to provide at least 20
per cent of the loan portfolios to finance the small business sector (small business loans
or small business finance).
Another package is PAKFEB 91 (the policy package of February 1991). This package
contains the prudential banking principles that commercial banks should comply to.
Through this package, the capital adequacy requirement of commercial banks as well as
their liquidity and asset quality were strictly monitored by the central bank. Similarly,
19
A foreign exchange bank is a commercial bank which also provides transactions in other currencies
besides Indonesian Rupiahs.
20
loans to individual borrowers or groups’ affiliated with the bank20
as well as net
foreign-exchange positions were also strictly monitored. This package has restrained the
excessive growth of the bank at time created a concern of inducing inflationary state for
the economy.
In the period 1992-1993, Indonesian banks began to experience a crucial issue because
of the growing number of bad loans (non-performing loans) that have the potential to
reduce the profitability of banks. Due to the fact that non-performing loans (NPL)
increased, some commercial banks were reluctant to expand their financing. This has
further augmented the restrictive effect of PAKFEB 1991 on bank loan expansion. To
overcome the problems encountered by the commercial banks, the GOI then launched
PAKMEI 93 (the policy package of May 1993). This package aimed to relax some of
the strict rules put in place by the previous policy so that commercial banks can operate
more freely in expanding their lending activities to customers.
The Indonesian economy started to grow rapidly starting from the year 1994 with the
real-estate sector being the top choice for investors. PAKMEI 93 has been successful in
encouraging the credit growth, but it might have passed through the level that could put
a heavy pressure on the monetary control. In other words, the amount of credit given to
a variety of business sectors, especially the real-estate sector, has led Indonesia to face
an overheated economy characterized by a high inflation rate during that period. The
fallout from the overheated economy coupled with the weak Indonesian economy paved
the way to the financial and economic crisis in 1997-1998 to occur. The crisis in turn
triggered a succession of national leadership with the fall of the New Order regime and
replaced by the Reformation Order in 1998.
In general, the condition of the Indonesian economy was very manageable before the
crisis occurred. A comparison of fundamental macroeconomic indicators across the
several Association of South East Asian Nations (ASEAN) countries during the period
1985-1997 is as follows:
20
Provisions of legal lending limit (LLL) for an individual or a group affiliated with a bank is 10 per cent,
while for those who are not affiliated with the bank is 20 per cent.
21
Table 2.2. Income and Inflation Indicators
Indonesia Philippines Malaysia Thailand
Real GDP Growth 1985-97 6.6% 7.6% 3.8% 8.5%
GDP per head, 1986 $480 $530 $1700 $820
GDP per head, 1996 $1100 $1200 $4600 $3100
Annual Increase in GDP per
head
12.92% 12.64% 17.06% 27.80%
Inflation, 1996 6.5% 8.4% 3.5% 5.8%
Source: Heffernan (2009)
The above table shows that the average real GDP growth of Indonesia during the period
1985-1997 was 6.6 per cent a year which was above the average real GDP growth of
Malaysia (3.8 per cent) in the same period. The data also show that Indonesia's GDP per
capita has increased more sharply compared to the Philippines (12.92 per cent compared
to 12.64 per cent). Indonesia's inflation rate in 1996 was also well below that of the
Philippines (6.5 per cent compared to 8.4 per cent, respectively). Although the
performance of the Indonesian economy was still below that of Malaysia and Thailand,
Indonesia had emerged as one of the emerging economies in the region during the
period.
2.2.3. The 1997-98 Asian Economic Crisis
The crises of 1997-98 were the most vulnerable era in the history of Indonesian banks.
Having enjoyed a boom in the period of liberalisation, the crisis occurred without any
official forecast of a sudden down turn (Heffernan, 2009). The crisis originated in
Thailand starting by a stock market crash in February 1997 that had led to a sharp
devaluation of the Thai currency (The Bath decreased by up to 30 per cent from its
value) by the end of the year. This incident then quickly turned into a currency crisis
that spread over the financial sector in Thailand. That currency and financial crisis in
Thailand had a contagion effect and transmitted quickly to other Asian economies i.e.
Indonesia, South Korea, Malaysia and the Philippines. The wave of the crisis has forced
all these countries to take rescuing actions through various monetary policies such a
devaluation policy, currency pegging, or the implementing of a currency board system.
22
The currency crisis has led the Indonesian Rupiah (IDR) to float in August 1997. On
July 1997, one USD was equal to 2,400 IDR, but after that date it started to devaluate
sharply afterward, and by February 1998, one USD had exceeded 10,000 IDR. The GOI
then took action to address the situation through several policies: restructuring the
banking sector, cancelling some government projects, and providing liquidity credit for
some insolvent banks. These policies were implemented to rescue the banking system
during the period and to save the Indonesian economy as a whole.
The GOI was under an agreement with the international monetary fund (IMF) to start
restructuring the banking system through implementing a series of actions that both
parties agreed upon. The restructuring package included 50 banks, of which 16 banks
were closed, and the remaining 34 banks underwent different treatment depending on
their financial condition: the six largest private banks were under more intensive
supervision; the ten insolvent banks were instituted to a rehabilitation program; and at
least one merger programme was implemented (Heffernan, 2009). In addition to the
restructuring programme other policies were also implemented for example: providing
liquidity assistance (BILC) for some commercial banks; introducing a blanket guarantee
scheme for all the liabilities of commercial banks in terms of rebuilding confidence in
the Indonesian banking system; and the establishment of the Indonesian Bank
Restructuring Agency (IBRA) which had the authority to take over, manage and
restructure the assets of non-performing (insolvent) banks.
The dramatic changes in the Indonesian banking sector as a result of the restructuring
programme within the two-year period can be observed in the table 2.3 as follows:
Table 2.3.Total Indonesian commercial banks, Pre- and Post-Crisis
Source: Heffernan (2009)
Bank by ownership Pre-crisis, July 1997 Post-crisis, August 1999
No. Of banks Market share No. Of
banks
Market
share
Private domestic banks 160 50% 82 17%
State domestic banks 34 42% 31 73%
Joint ventures/foreign banks 44 8% 40 10%
23
From the table above, it can be seen that the number of private domestic banks has
declined sharply to almost half, due to the closure, merger and acquisition programmes.
The market share of state domestic banks has increased from 42 per cent in 1997 to 73
per cent in 1999. The sharp rise of the state-owned banks‟ market share was caused by
the lack of confidence of private domestic bank customers that moved their funds into
state-owned banks as a safer resort.
2.2.4. After the 1997-98 Crisis
The period after the crisis is considered an important phase in the history of the
Indonesian banking sector particularly within the framework of the improvement of the
banking quality in the country. There were two milestones in this period: the
restructuring and the reconstruction programs. Both programs were very beneficial for
the survival of the Indonesian banking sector in the present and the future.
The restructuring program is characterized by the formation of IBRA which was formed
on January 26, 1998. IBRA had a duty to restructure some particular banks requiring
special assistance and take over some of the many non-performing loans of commercial
banks through an asset transfer scheme. IBRA also managed and took over some poor
performing banks by merging several banks and arranging the acquisition of some weak
banks by the healthy banks.
According to the Indonesia‟s Economic Report (2004), IBRA had successfully managed
several non-performing assets transferred by commercial banks and in return, they were
given government bonds as compensation. Regarding the debt restructuring program,
until September 2004, the total volume of bank restructuring funds in the 15 largest
banks was about IDR 38.4 trillion.
The Indonesian Banking Architecture (IBA) was launched by Bank Indonesia on 9th
January, 2004 (Indonesia‟s Economic Report, 2004). The vision of IBA is to realise the
sound, strong, and efficient banking system in order to create the stability of financial
system as well as stimulate national economic growth (see the figure 2.1).
24
The IBA vision is based on six pillars that are considered essential to support the
achievement of IBA‟s objectives:
1. Creating a healthy domestic banking structure which is able to meet the needs of
the society and encourage sustainable national economic development.
2. Creating an effective bank regulation and supervision system which conducts to
the international standards.
3. Constructing a strong Indonesian banking industry which is very high
competitive as well as having resilient at risk.
4. Creating good corporate governance in order to strengthen the internal condition
of the national banking system.
5. Realising a complete infrastructure to support the creation of a healthy banking
industry.
6. Realising an empowerment of consumer protection for banking services.
Figure 2.1. The Indonesian Banking Architecture (IBA)
Source: Bank Indonesia (www.bi.go.id)
Through these programs, it is expected that by 2019, the structure of the Indonesian
banking system will be improved. The structure would be envisaged as follows (Besar,
2011:22-23):
2 or 3 banks are likely to emerge as international banks. These banks possess the
capacity and ability to operate on an international scale and having total capital
exceeding IDR 50 trillion (USD 5 billion).
Up to 5 national banks. These banks have a broad scope of business and
operating nationwide with total capital between IDR 10 trillion (USD 1 billion)
and IDR 50 trillion (USD 5 billion).
25
30 to 50 specialized banks with operations focused on particular business
segments according to the capability and competence of each bank. These banks
will have capital of IDR 100 billion (USD 10 million) up to IDR 10 trillion
(USD 1 billion).
Rural Banks are the banks operating in rural area, and banks with limited scope
of business, having capital of less than IDR100 billion (USD 10 million).
In addition, to improve the quality of the banking infrastructure and maintain the
Indonesian banking consolidation process, GOI has also established the Indonesian
Deposit Insurance Agency (IDIA) on December 22, 2005 (Indonesian Economic
Report, 2005). The short program of this agency was the reduction of the guaranteed
bank deposit accounts. For example, until March 2006, the maximum deposit insurance
of IDR 5 billion was reduced to IDR 1 billion in September 2006, and became only IDR
100 million in March 2007. The purpose behind the policy was to direct commercial
banks in Indonesia to undertake their businesses more prudently (Indonesian Economic
Report, 2005).
2.2.5. The 2008 World Economic Crisis
The world economic crisis in 2008 had some impact on the Indonesian economy by the
weakening in the macroeconomic indicators (See Table 2.4). The BI was forced to
increase interest to maintain liquidity in the banks as a defective policy to face the
negative consequences of the crisis on the Indonesian banking sector.
Table 2.4. Indonesian Economic Indicator, 2006-2010
Indicators 2006 2007 2008 2009 2010
GDP growth (%) 5.5 6.3 6.0 4.6 6.1
Time deposits rate, 12months (%) 10.7 7.7 10.3 7.9 7.2
Time deposits rate, 6months (%) 11.6 8.2 10.4 9.6 7.9
IDR/USD (end of period) 9020 9419 10950 9400 8991
IDR/USD (Average of period) 9159 9141 9699 10390 9090
Source: www.adb.org/statistics
The effect of the crisis on the Indonesian economy seemed to continue until 2009
although there have been improvements in some economic indicators. Nonetheless, the
26
GDP growth dropped slightly to 4.6 per cent from 6 per cent in the previous year and an
Indonesian Rupiah decreased from IDR 9,699 to IDR 10,390 per one USD. These
incidents had its toll on slowing down the Indonesian economic activities throughout the
year. The Central Bank's discretionary policy taken has obtained good results. The
decision to reduce the BI rate (from 9.25 per cent in 2008 to 6.5 per cent in 2009),
managed to suppress the inflation rate as well as the bank deposit rate and stimulated
the Indonesian stock exchange index.
In addition, the crisis reduced trust among the commercial banks. The inter-bank
lending and borrowing declined by 59.3 per cent from IDR 206.0 trillion on December
2007 to IDR 83.8 trillion on December 2008. The decline in confidence among
commercial banks was almost similar to the 1997-98 crises at which they experienced a
shortage of liquidity so that they were reluctant to lend each other in the interbank
market (Basri and Siregar, 2009)
2.2.6. After the 2008 Crisis
The period after the crisis of 2008 is considered as a recovery phase for the Indonesian
economy after a slight decline occurred in 2008. In general, table 2.5 shows the progress
of Indonesian economic performance in terms of lower inflation, lower interest rates,
lower unemployment rates and rising stock market indices in 2009-2010.
Table 2.5. Indonesian Macroeconomic & Financial Indicator, 2005-2010
Indicator 2005 2006 2007 2008 2009 2010
GDP Growth (%) 5.7 5.5 6.3 6 4.6 6.1
Inflation rate (%) 17.1 6.6 6.6 11.1 2.8 7.0
Un-employment rate
(%) 11.2 10.3 9.1 8.4 7.9 7.1
ISE composite index 1163 1806 2746 1355 2534 3704
Interest rate/BI rate (%) 12.75 9.75 8 9.25 6.5 6.5
Source: BAPPENAS RI (Ministry of National Planning Republic of Indonesia)
Table 2.5 shows that the economic condition in Indonesia has significantly improved in
2010. The inflation rate, the unemployment rate, the Indonesia Stock Exchange (ISE)
composite index, and the interest rate all have improved in 2009 compared to 2008
27
levels. The adopted policy by BI to lower interest rates in 2009 and 2010 had attained
satisfactory results: stimulated economic growth; reduced unemployment and increased
the ISE composite index in 2010. Given the above statistics one can argue that
Indonesia has successfully managed the crisis of 2008 compared to the 1997-1998 one.
Basri (2011) provides a comparison table (See Table 2.6., Appendix 2) that elaborates
how the GOI dealt with the 2008 crisis differently compared to the 1998 crisis from
different aspects: monetary policy, fiscal policy, banking health, response towards
banking sector, and structural reform policies.
2.3 Small Businesses Finance and Total Finance
This section presents overview knowledge on commercial banks‟ overall lending
(financing) activities and to small business lending (financing) in Indonesia. The content
of this chapter is essential for understanding the effects of small business finance as a
proportion of total finance on the Indonesian bank performance.
2.3.1 Definition of a Small Business (SB) and Small Business Finance (SBF)
In general, the definition of a small business and small business finance across the globe
is different depending on the provisions prevailing in the country. Every country
decides on their criteria for a small business and small businesses finance.
International Finance Corporation (IFC)21
has compiled the definition of micro small
medium enterprises (MSMEs) across 120 countries. Several criteria for SBs in these
countries are as follows:
In the UK, a small business is defined as follows: a company that has turnover of
not more than £5.6 million, balance sheet total, not more than £2.8 million of
total assets, and not more than 50 employees.
In the US, a small business is a company that has no more than 500 employees
21 International Finance Corporation (IFC) is an organization and a member of the World Bank Group
that has two primary goals i.e. to end poverty of the world by 2030 and to boost shared prosperity in
every developing country in the world. See in details in www.ifc.org.
28
(for most manufacturing and mining industries) and has average annual receipts
of $7 million (for most non-manufacturing industries).
In Australia, a small business is a business employing less than 20 people. The
small business also covers non-employing businesses i.e. sole proprietorships
and partnerships without employees, micro businesses which employ less than 5
people, and other small businesses which employ 5 or more people but less than
20 people..
In China, a small business is a business which has total assets of less than ¥40
million and average business revenue of less than ¥30 million. It is also
classified as a small business a company that has number of employees less than
300 (for the Industry sector), 600 (for construction sector), 100 (for wholesale
and retail industries), 500 (for the transportation industry) and 400 (for hotel and
restaurant industries).
The criteria for small business finance also vary across countries in the world. Some
countries defined small business finance (SBF) as follows:
In the US, small business finance is all commercial and industrial (C&I) loans
with original amounts under to $1 million [Strahan and Weston (1998), Devaney
and Weber (2002)].
In Argentina, the Central bank of Argentina has defined small business loans as
the total debts in between $50,000-$2.5 million (Clarke et al, 2005).
The Chilean Banking Superintendence has defined small business loans as the
total debts of less than $1.5 million (Clarke et al, 2005).
The Banking Superintendence of Peru has defined small business finance as the
loans amounted in the range between $20,000 and $500,000 (Clarke et al, 2005).
2.3.2. Definition of Small Businesses Finance in Indonesia
Small business finance (SBF) is one of the commercial banks‟ financing schemes in
addition to corporate and consumer financing. Commercial banks provide various
financing with the aim of diversifying their financing portfolios as well as reaching
numerous customer segmentations. Through diversified portfolios, bank managers
29
expect that the risk associated with financing will be reduced.
The Indonesian government defines a small business as well as small business finance
through the Indonesian government laws notified by parliament and the regulations
issued by Bank Indonesia. Accordingly, commercial banks then adjust their financing
schemes to small businesses in accordance with prevailing regulations.
The Government of Indonesia (GOI) has defined a small business as follows:
“A small business (SB) is an economic and productive entity, carried out by an
individual or a business entity that is not a subsidiary or a branch of a company and is
not owned and controlled or a part, directly or indirectly, of a medium or large
enterprise that meets the following criteria: 1) It has a net worth of more than IDR 50
million up to IDR 500 million, not including its land and buildings. 2) It has the annual
sales of more than IDR 300 million up to IDR 2,500 million ” (The Law No. 20/2008
about micro, small, and medium enterprises, Ch. 1 article 1 page 2 and Ch. 4 article 6
page 5-6) .
The definition of small business finance in Indonesia is as follows:
“Small business finance (SBF) is loans or financing facility provided by commercial
banks for investment or working capital purposes to small businesses with a maximum
amount of IDR 500 million” (Bank Indonesia Regulation No. 3/2/PBI/200122
).
Based on those criteria and the definition of a small business above, commercial banks
provide financing to small businesses along with other financings i.e. corporate
financing and consumer financing. Then, commercial banks regularly report the
exposures of their finance portfolios to the Central Bank of Indonesia on a monthly and
yearly basis. Accordingly, the Central Bank of Indonesia announces the level of total
small business finance maintained by all commercial banks in Indonesia in its reports.
22
BI Regulation No. 3/2/PBI/2001 on Small business loans (finance) on January 4, 2001.
30
2.3.3. The Regulation of minimum SBF Threshold in the Indonesian Commercial
Banks
Basic rules concerning the obligation of commercial banks to allocate 20 per cent of
their loan portfolios was first stipulated in one of the banking deregulation packages
which was published in 1990 (PAKJAN 90)23
. The package policy states that banks are
obliged to provide at least 20 per cent of their loan portfolio towards small business
finance. The policy was later revised by the Decree of BI Directors No. 30/4/KEP/DIR
on April 1997 which changed the minimum threshold to 22.5 percent of the net loan
(finance) expansion. The latter regulation was applied onwards until 2001.
The regulations about the minimum SBF requirements for Indonesian commercial banks
were then revoked by BI Regulation Number 3/2/PBI/2001 about Small Business
Finance on January 4, 2001. The regulation states that the provision of small business
finance is readjusted based on the ability of each bank. In other words, commercial
banks were no longer compelled to provide a minimum percentage of their finance
portfolios to small businesses.
Although the government provides flexibility for commercial banks to manage their
small business finance without giving the minimum limit, the central bank keeps
requiring all commercial banks to stipulate the distribution plan of SBF on their annual
business plans. In addition, banks are required to report the disbursement of SBF to BI
at the end of each month. For transparency purposes, the commercial banks are required
to inform the public about their achievements of their SBF through the publication of
financial statements/annual report.
In 2012 the GOI turned back to issue new regulations24
regarding the obligation of
commercial banks to distribute at least 20 per cent loan portfolios in this sector. The
difference is, the obligation to fulfil the minimum financing portfolio by 20 per cent is
23
On January 29, 1990 the GOI issued a policy package known as PAKJAN (the package of January
1990). The package is a continuation of the previous packages in the banking sector with the aim of
mobilizing funds, improving efficiency and reducing subsidies from Bank Indonesia. The most important
issue of this package is the improvement of the provision of Small businesses (SBs) finance with the
provision of at least 20 percent of the bank financing shall be distributed to SBs. 24
BI Regulation No.14/22/PBI/2012 on 21 December, 2012 regarding the commercial bank loans or
finance for MSMEs and the technical assistance for the development of MSMEs.
31
currently not only intended to Small businesses (SBs) but it is aimed at financing Micro
Small and Medium Enterprises (MSMEs). Indeed, SBs is a part of MSMEs.
In addition, banks are allowed to achieve the ratio up to the following 6 years with the
stages as follows:
Until the end of 2013 and 2014, the ratio of MSMEs finance is tailored to the
capabilities of commercial banks, which must be included in their business plan.
Until the end of 2015, the ratio of the total MSMEs finance should achieve at
least five percent.
Until the end of 2016, the ratio of the total MSMEs finance should achieve a
minimum of ten percent.
Until the end of 2017, the ratio of the total MSMEs finance should achieve a
minimum of fifteen percent.
Until the end of 2018 onwards, the ratio of the total MSMEs finance should
achieve a minimum of twenty percent.
Sanctions against banks that do not comply with the above are as follows:
The banks are required to organize training for Micro, Small and Medium
Enterprises which are not or have never received loans or financing of MSMEs.
The amount of training funds is calculated based on the percentage difference
between the ratio of MSMEs financing that must be met and the realization of
the ratio achievement at the end of each year, with a maximum of IDR 10
billion.
The launch of this new regulation could be interpreted that the GOI has considered the
importance of MSMEs as an important driver for the Indonesian economy. This thesis
aims to investigate the performance of Indonesian banks in terms of bank efficiency and
bank lending propensity over the period 2002 to 2010. The study period witnessed the
period of time when there were no government regulations that forced Indonesian banks
to provide a minimum particular percentage (minimum threshold) of their financings to
small businesses. Therefore, the findings of this thesis are very useful as an evidence of
whether the decision of the GOI to re-enact the regulation of small business finance (in
this case, MSMEs finance) in 2012 was right or not.
32
2.3.4. The Development of Small Business Finance in Indonesia over the period
2002-2009
Indonesian commercial banks maintained small business finance as one of their loan
portfolios during the decade of the 2000s. During the period, there was only one scheme
launched by the government of Indonesia which was classified as SBs finance. The
scheme was named Kredit Usaha Rakyat (KUR) in 2008. The purpose of the scheme is
to assist commercial banks in accelerating the SBF as well as the development of small
businesses in Indonesia.
KUR is one of the loan schemes provided by the commercial banks to micro, small and
medium enterprises and cooperatives with the guarantee pattern, cooperating with the
insurance agency established by the Government. KUR can be used to finance all
productive enterprises including those in the agriculture sector which are viable
(feasible) but not yet bankable25
of additional collateral aspects (Technical Guidelines
People's Business Credit, Ministry of Agriculture, GOI, 2012).
For commercial banks, this scheme should have increased the amount of SBF exposure
owned by the banks. With this scheme, commercial banks transfer the credit risk to the
insurance company. The role of commercial banks is only to assess feasibility and
decide on granting loans or financing. The guarantee fees on the loans are paid by the
government to the credit insurance company. Hence, the role of the GOI in this scheme
is laid on the payment of the guarantee fees only and not providing the interest subsidy
as the previous policies in the decade of 1990s.
The table below presents the comparison between Small Business (SBs) Finance, Micro
Small and Medium Enterprises (MSMEs) Finance, and the Total Finance of Indonesian
commercial banks for the period of 2002-2009:
25
Bankable means that the project or the company has met all the requirements set by commercial banks
in terms of the project feasibility and collateral availability.
33
Table 2.7. The Highlights of SBF and MSMEs Finance IDR Billion
No Type of finance 2002 2003 2004 2005 2006 2007 2008 2009
1 SBs Finance 62,266 72,647 85,191 96,580 102,028 112,575 132,115 153,553
2 MSMEs Finance 160,977 207,088 271,092 354,908 410,442 502,798 633,944 737,386
3 Total Finance 371,058 440,505 559,470 695,648 792,297 1,002,012 1,307,688 1,437,930
No %Type of finance 2002 2003 2004 2005 2006 2007 2008 2009
1 SBs Finance 16.78% 16.49% 15.23% 13.88% 12.88% 11.23% 10.10% 10.68%
2 MSMEs Finance 43.38% 47.01% 48.46% 51.02% 51.80% 50.18% 48.48% 51.28%
3 Total Finance 100.00% 100.00% 100.00% 100.00% 100.00% 100.00% 100.00% 100.00%
Source: Indonesian Bank Statistics, 2002-2009
Table 2.7 reveals that SBF has increased in the Indonesian Rupiahs (IDR) absolute
value from IDR 62.27 trillion in 2002 to IDR 153.55 trillion in 2009. The average
increased rate of SBF over the period is 13.84 per cent a year. However in terms of the
portfolio percentage, SBF of Indonesian commercial banks tended to decrease from
16.78 per cent in 2002 to the only 10.68 per cent in 2009. The regulation to revoke the
obligation for commercial banks to maintain 20 per cent of their finance portfolios to
small businesses26
has impacted to the major decline in the SBF portions over time
during the period 2002-2009.
On the other hand, MSMEs finance is likely to increase either in nominal value or in
percentage terms. MSMEs finance of Indonesian commercial banks has increased from
IDR 160.98 trillion in 2002 to IDR 737.39 trillion in 2009. The average increase rate of
MSMs finance over the period is 24.43 per cent which is much higher than that of the
average increase rate of SBF which achieved 13.84 per cent. In terms of the percentage
of finance portfolio, MSMs finance also experienced a considerable increase from 43.38
percent in 2002 to 51.28 percent in 2009. Because MSMEs actually consists of micro,
small and medium enterprises, then it can be concluded that the high increase rate of
MSMEs finance is caused by a high increase in the financing portfolios of micro and
medium enterprises.
The upward tendency of Indonesian commercial banks to provide MSMEs finance also
indicates that there is a change in the orientation of commercial bank financing in
Indonesia. They are now more likely to focus on financing micro, small and medium
26
BI Regulation Number 3/2/PBI/2001 about Small Business Finance on January 4, 2001.
34
enterprises (MSMEs) rather than financing on small businesses (SBs) only. This
tendency is in line with the new regulation issued by the GOI, namely the Law Number
20/200827
regarding micro, small and medium enterprises. Under the new rules,
commercial banks should have adjusted their progress reports on the micro, small and
medium enterprises (MSMEs) finance rather than on SBs finance only. These reports
are submitted regularly at the end of the month. These policies are in line with the new
regulations as stipulated in the BI Regulation No.14/22/PBI/2012 on 21 December,
2012 regarding commercial bank loans or finance for MSMEs and technical assistance
for the development of MSMEs.
2.3.5. The Development of Total Loans of Indonesian Banks during 2002-2010
In terms of total lending facilities (total loans) of Indonesian banks over the period 2002
to 2010, some following tables and figures are presented as follows:
Table 2.8. The Highlights of Total Loans & the Profitability of Indonesian Banks
Main Indicators 2002 2003 2004 2005 2006 2007 2008 2009 2010
Total Loans (IDR Trillion) 371.1 440.5 559.5 695.6 792.3 1,002 1,307.7 1,437.9 1,765.8
Loans to Deposits ratio (%) 44.4 49.6 58.1 61.7 61.6 66.3 74.6 72.9 75.5
Gross Non-Performing Loans (%) 7.5 6.8 4.5 7.6 6.1 4.1 3.2 3.3 2.6
Return on Assets (%) 2 2.6 3.5 2.6 2.6 2.8 2.3 2.6 2.9
Net Interest Margin (%) 4.1 4.6 6.4 5.6 5.8 5.7 5.7 5.6 5.7
Source: Indonesian Economic Report 2011, Bank Indonesia
Table 2.8 testifies the increasing tendency of the total loans of Indonesian commercial
banks over the period 2002 to 2010 in terms of total absolute value of Rupiahs. It shows
the upward tendency from IDR 371.1 trillion in 2002 to IDR 1.765.8 in 2010. In terms
of loans to deposits ratio (LDR), it is also likely to increase over time from 44.4 per cent
in 2002 to 75.5 per cent in 2010. These two indicators imply the upward propensity of
all Indonesian commercial banks over the period 2002 to 2010. These also show that the
function of banks as financial intermediary institutions has increased over the period
particularly in the role of the provision of loans to lending customers. The increased
27
As mandated by the Law No. 20/2008 regarding the micro, small and medium enterprises (MSMEs), the
GOI has changed the focus of attention not only for small businesses but also for micro and medium
enterprises.
35
trend of the banks‟ loans also reflects the revival phase of the Indonesian banking
industry after suffering from the Asian crisis of 1997-98.
In terms of the loan quality, the table witnesses the increasing loan quality of Indonesian
banks evidenced by the downward trend of the gross non-performing loans from 7.4 per
cent in 2002 to the only 2.6 in 2010. In addition, in terms of bank profitability, the table
shows good performance of banks in terms of return on assets and net interest margin
which obtained an average performance of 2.7 per cent and 5.5 per cent respectively.
In the assessment of the contribution of the commercial loans on the economy, it can be
observed by their productive loans (working capital loans and investment loans)
compared to the total loans (see table 2.9).
Table 2.9. Productive Loans & Total Loans of Indonesian Banks
IDR Billion
Year Working Capital Loans Investment Loans Total Productive Loans TOTAL Loans
2002 282,486 82,924 365,410 371,058
2003 343,626 94,316 437,942 440,505
2004 436,684 116,864 553,548 559,470
2005 557,207 132,463 689,670 695,648
2006 405,551 148,770 554,321 792,297
2007 518,339 183,694 702,033 1,002,012
2008 668,007 254,373 922,380 1,307,688
2009 686,983 295,914 982,897 1,437,930
2010 868,356 345,700 1,214,056 1,765,845
Average 529,693 183,891 713,584 930,273
Source: Indonesian Banking Statistics, 2002-2010, Bank Indonesia
On average, of the total loans provided yearly over the period 2002 to 2010 by
commercial banks which are accounted for IDR 930.27 trillion, an average of IDR
713.58 was bestowed to productive loans, and the rests are for consumptive purposes
loans. Of all productive loans, commercial banks in Indonesia tend to finance
predominantly to working capital purposes which accounted for an average of IDR
529.69 trillion a year over the period. While an average loan of IDR 183.89 was
attributed to investment purposes.
36
Figure 2.2. The Trend of Loans of Indonesian Banks (2002-2010)
Source: Indonesian Banking Statistics, 2002-2010, Bank Indonesia
Meanwhile, the trend of productive loans and total loans which are presented by the
above figure demonstrate that all productive loans (working capital and investment
loans) have an upward trend over the period 2002-2010. These achievements have
supported the trend of the total loans in similar pattern over time across the period.
Figure 2.3. The Trend of Loans Proportions of Indonesian Banks (2002-2010)
Source: Indonesian Banking Statistics, 2002-2010, Bank Indonesia
In terms of loan proportion to the total loans, the figure 2.3 shows that the percentage of
productive loans has tended to decrease since 2006. This means that since that year, the
contribution of consumptive loans increased over time until 2010. The decreasing trend
of productive loans seems to be affected by the decreasing pattern of working capital
loans during 2006 to 2010. While investment loans appear to be constant at an average
of about 20 per cent. The total loans provided by commercial banks are still very
important in the bank-based system in the vein of Indonesia. Banks‟ loans fuel the
-
500,000
1,000,000
1,500,000
2,000,000
2002 2003 2004 2005 2006 2007 2008 2009 2010
IDR
Bill
ion
Year
The Trends of Loans (Indonesian banks, 2002-2010)
Working Capital
Investment
Total Productive Loans
TOTAL Loans
0.0%
20.0%
40.0%
60.0%
80.0%
100.0%
120.0%
2002 2003 2004 2005 2006 2007 2008 2009 2010
Year
The trend of Loans Proportions (Indonesian banks, 2002-2010)
Working Capital
Investment
Total Productive Loans
TOTAL Loans
37
engine of the economy to produce goods and services in the country. The figure below
can be used to briefly observe the association between the growth of loans and the
growth of gross domestic products (GDP).
Figure 2.4. The Trend of Loans Growth and GDP Growth (2002-2010)
Source: Indonesian Banking Statistics, 2002-2010, Bank Indonesia
Figure 2.4 represents loan growth in terms of working capital loans, investment loans
and the total loans in supporting the growth of the GDP. The growth of working capital
loans accounted for 17.30 per cent on average. It seems that the negative growth in 2006
and the small growth in 2009 affected the average growth of the loans over the period.
The decline in working-capital loans growth in 2006 and small growth achievement in
2009 seem to be affected by the mini-crisis of 2005 and the world crisis of 2008.
Meanwhile, the growth of the investment loans accounted for 19.03 per cent on average
yearly over the period. The mini-crisis of 2005 and the world crisis of 2008 only slightly
affected the achievement of investment loan growth in 2006 and 2009 but the growth
was still positive in two digits.
Overall, the growth of the total loans which accounted for the average of 21.23 percent
has contributed to the positive growth of the GDP of Indonesia which accounted for
5.39 per cent growth on the average. The figure reflects the contributions of the total
loans as well as the productive loans of commercial banks in enhancing economic
activities in Indonesia over the period 2002 to 2010.
2002 2003 2004 2005 2006 2007 2008 2009 2010
Working Capital Loans Growth 20.65% 21.64% 27.08% 27.60% -27.22% 27.81% 28.87% 2.84% 26.40%
Investment Loans Growth 12.87% 13.74% 23.91% 13.35% 12.31% 23.48% 38.48% 16.33% 16.82%
TOTAL Loans Growth 17.40% 18.72% 27.01% 24.34% 13.89% 26.47% 30.51% 9.96% 22.80%
GDP Growth 4.50% 4.80% 5.00% 5.70% 5.50% 6.30% 6.00% 4.60% 6.10%
-40.00%-30.00%-20.00%-10.00%
0.00%10.00%20.00%30.00%40.00%50.00%
Loans Growth & GDP Growth
38
2.4. Conclusions
Indonesia's banking industry has a long history since the colonial era when Indonesia
was a Dutch colony. Some commercial banks were established in the period, and they
were owned either by colonial powers or by local communities in several cities in
Indonesia.
After Indonesia proclaimed its independence in 1945, some significant progress was
achieved by the Indonesian government through the establishment of several new
commercial banks that concentrated on particular sectors in accordance with the
requirements in the economy. Several periods took place with tight regulations set by
Bank Indonesia as a central bank as well as commercial banks regulatory and oversight
agency. Regulations were applied to commercial banks to serve the public in accordance
with the objectives of the government.
The 1980s-1996 periods is considered to be a very dynamic one since some
deregulation packages were launched for the purpose of accelerating the evolution of
Indonesian banking effectively. However, success in augmenting the number of banks
was actually not accompanied good quality bank management. Thus, when the crisis
occurred in the neighbouring countries in 1997, it easily spread to Indonesia.
The crisis of 1997-98 provided some very important lessons for the Indonesian
government and banking authorities to launch their new architectural Indonesian bank
in 2004. The new architecture was in keeping with the policy of Indonesian banking
prudence and sound supervision. This programme successfully led Indonesian banking
to pass the global crisis of 2008 where Indonesian banks are less affected compared to
other Asian countries.
The performance of Indonesian banks witnessed improvement during the decade of
2000s. Many banks rebounded to make profits after suffering from losses during the
crisis of 1997-98. The portfolio composition of earning assets particularly in lending or
finance facilities is very interesting to explore. In this regard, the thesis provides
efficiency estimates of Indonesian commercial banks, deriving the involvement of small
business finance or small business loans in the estimation of the bank efficiency for both
DEA and SFA techniques.
39
In addition, the lending propensity of Indonesian banks is also interesting to explore.
The propensity of Indonesian banks in terms of small business loans and total loans over
the period 2002-2010 is very important to be examined since the period was considered
to be a revival period after the Asian financial crisis of 1997-98 as well as a period of
implementing the government policy28
to relax the minimum threshold of SBF exposure
that should be maintained by Indonesian commercial banks.
28
BI Regulation Number 3/2/PBI/2001 about Small Business Finance on January 4, 2001.
40
3. Chapter three: Literature Review
3.1. Introduction
Measuring bank performance has been a crucial issue in banking research for decades.
There are two mainstream approaches in assessing the performance of banks, namely
the financial ratios approach and the frontier approach. The financial ratios approach
explores the financial condition of each bank in terms of liquidity, profitability,
leverage, and activity ratios and compares the ratios with the average industry ratios as
the benchmarks. Meanwhile, the frontier approach measures bank performance by
comparing the performance of each bank relative to a best practice frontier. In the
frontier approach, there are two prominent methods employed, namely the non-
parametric method and the parametric method. This thesis investigates the efficiency of
Indonesian banks by both the DEA non parametric method and SFA parametric method,
presented in the two empirical chapters (chapters five and six). These chapters also
expand their analysis by revealing the determinants of DEA technical efficiency as well
as SFA cost efficiency of Indonesian banks during the period 2002-2010.
On the other hand, studies on banks‟ lending propensity have been growing in the
academic literature [see for example Strahan and Wetson (1998), Peek and Rosengren
(1998), Goldberg and White (1998), Berger et al. (2001, 2007), Akhavein et al. (2004),
Laderman (2008), Shen et al. (2009), McNulty (2013)]. These studies mostly discuss the
propensity of commercial banks to provide small business loans as the main point of
view. This is applied due to the fact that not all commercial banks have the willingness
to provide loans for small businesses. Thus, it has attracted many authors to examine the
lending propensities of banks including the propensity of banks to provide lending to
SBs. This thesis also investigates the propensities of Indonesian commercial banks in
providing their lending facilities in terms of total loans and small business loans
(chapter 7). The chapter also provides an extensive analysis by examining the
determinants of Indonesian bank lending propensity over the period 2002 to 2010.
41
This chapter provides a brief theoretical background and empirical literature on two
main issues in this thesis: bank efficiency and bank lending propensity. The chapter is
organised as follows. The first section reviews a brief theory of production efficiency as
well as the empirical evidence on bank efficiency in some countries. The focus of the
empirical evidence presented is the efficiency and productivity studies on the
Indonesian commercial banks as well the efficiency and productivity studies on the
emerging economies particularly in areas adjacent to Indonesia i.e. the South East Asia
and East Asia countries. The second section also provides a brief theory of loan
management and the previous literature on bank lending propensity. The final section
provides conclusions.
3.2. Theoretical and Empirical Literature
3.2.1. The Theory of Efficiency
3.2.1.1. The Brief Theory of Production Efficiency
Efficiency measurement was firstly discussed by Farrel (1957) based on the previous
work of Koopmans (1951) and Debreu (1951). He accounts for multiple inputs for the
measurement of efficiency. Farrel (1957) identified that firm efficiency covers two
components: 1) technical efficiency, which denotes the ability of a firm to obtain
maximal output from a given sets of inputs and 2) allocative efficiency, which indicates
the ability of a firm to employ the inputs in the optimal allocations given the price and
the technology of production. The combination of the two measures (technical
efficiency and allocative efficiency), are then recognized as an economic efficiency
measurement.
Figure 3.1 illustrates the original ideas of Farrel. He simplifies his explanation through
the usage of the two factor of productions, where, in this case ( and as the two
inputs to obtain an output (q) under the constant-return to scale assumption. denotes
the unit isoquant of fully efficient firms, which reflects the various combination of the
two inputs that can be employed by a perfectly efficient firm to produce unit output.
42
is the isocost-line (slope) which reflects the ratio of the price of the two inputs.
Figure 3.1. Technical and Allocative Efficiency
S
X2/q
P
Q
A
R
Q’
0
X1/q
Source: Coelli et.al (2005), Greene (2013)
If a firm produces an output by using a quantity of inputs at point P, the distance of QP
is the technical inefficiency of the firm. That distance reflects the amount of all inputs
that could be reduced without influencing the output. The technical inefficiency of the
firm would value QP/0P.
The point Q is technically efficient since it lies on the efficient isoquant. Thus, the
technical efficiency (TE) of the firm would be as follows (Coelli et al., 2005):
(3.1)
In general, the technical efficiency of a firm would always be between 0 and 1 (0 ≤ TE
≤1) and the score reflects the technical efficiency level of the firm. 0 means least
efficient while 1 denotes the most efficient one.
However, a firm can reallocate their inputs with the new combination along the isoquant
curve to obtain the technical efficient as well as the allocative efficient. At the
intersection of the isocost line and isoquant curve (at the point ) the firm can achieve
the overall efficiency or economic efficiency since it can experience the lowest cost for
the optimal output. If a firm produces at , the distance RQ would be the amount of
43
reduction in the cost of production that could be gained by the firm. Hence, the firm that
operates at would be technical efficient since it is on the isoquant curve (producing
maximal output) and also allocative efficient because it undergoes the optimal
allocation of production inputs. In other words, the firm operates in economic (overall)
efficient.
In the case of cost efficiency, we can assume w as the vector of input prices, and x as
the observed vector of inputs employed (related with P), while is the input vector
employed in the technically efficient point Q and denotes the input vector in the cost
minimizing condition at . Then, we can state that cost efficiency (CE) of the firm is
measured as follows (Coelli et al, 2005):
(3.2)
3.2.1.2. Techniques of Measuring Efficiency
There are two mainstreams in estimating the efficiency of a firm, namely parametric
techniques and non-parametric techniques. Each technique has its own advantages and
weaknesses. This section describes the two mainstreams in estimating efficiency:
3.2.1.2.1. Non-parametric Techniques
There are some techniques applied using the non-parametric method, i.e. Data
Envelopment Analysis (DEA) and the Free Disposal Hull (FDH). The DEA approach is
the most widely used in this kind of research to date. An explanation about DEA in
more details is presented in the Chapter 4 in the methodology section.
FDH analysis was firstly introduced in a study of the relative efficiency of post office
operations in Deprins et al. (1984). Tulkens and Eeckaut (1995) then discussed more
advanced about issues, particularly about the technological change and shift in the
production possibility frontier. The central principle of the FDH is that, a producer is
relatively inefficient if another producer uses less or an equal amount of inputs to
generate more output (Gupta & Verhoeven, 2001). The Free Disposal Hull (FDH)
model is a special case of DEA, which has a different assumption than the CCR and
44
BCC models29
. Unlike those two models, the FDH does not operate with a convexity
assumption. In practice, the efficient reference point for an inefficient DMU is among
the existing DMUs and not chosen as a point of a continuous efficiency frontier. Thus,
this model has a discrete nature. Moreover, as stated by Tulkens (1993), the evolution of
the FDH has adopted the replications of the existing DMU in the extension of the
hypothetical DMUs.
3.2.1.2.2. Parametric Techniques
In general, a parametric approach accounts for the assumption of the error term in the
efficiency estimation. Various methods have been developed in this parametric
mainstream and the choice of methods is mainly dependent on the assumptions of the
distribution of the error components.
There have been some methods can be applied i.e. stochastic frontier approach (SFA),
distribution free approach (DFA) and thick frontier approach (TFA). Detail explanation
of the SFA is presented in the methodology section in Chapter 4.
The distribution free approach (DFA) which was first mentioned by Berger (1993) is a
method which does not make specific distributional assumptions on the error
components. DFA requires panel data and is based on a translog system of cost and
input cost share equations that generate estimates of cost inefficiency for each producers
in each time period (Kumbhakar and Lovell, 2000). The main advantage of this method
is that the estimation results do not depend on the distributional assumption on
(inefficiency term). The other advantage is as being based on a sequence of time
separate cross sectional regressions, it allows the structure of production technology to
vary flexibly through time (Kumbhakar and Lovell, 2000). While a disadvantage of
DFA is the prerequisite that cost efficiency is time invariant, and the assumption would
become less tenable as time increases.
29
Charnes, Cooper, and Rhodes (CCR in 1978) and Banker, Charnes and Cooper (BCC in 1984) models.
45
Some of the techniques that could be classified in this distributional free approach are
Corrected OLS (COLS)30
, and Corrected Mean Absolute Deviation (CMAD)31
. Both
techniques do not make distributional assumptions on the error components.
The thick frontier approach (TFA) was proposed by Berger and Humprey (1991). This
approach is mostly applied within the cost frontier model, although it may also be
employed for a production frontier model. This approach clusters samples into N
quartiles (e.g. 4 quartiles) based on the indicator of an observed efficiency i.e. the
average cost or the average output. In general, the advantage of the TFA approach is it
could obtain the inference of the cost inefficiency although it does not rely on particular
distributional assumptions. While some drawbacks of the approach are as follows: First,
it needs a large dataset; otherwise it would apply too small data because of the
stratifications. Second, it is not easy to stratify the data for multiple outputs.
3.2.2. The Empirical Literature on the Efficiency and Productivity Studies
There is abundant research on the efficiency and productivity of commercial banks.
However, the number of studies on emerging markets is fewer compared to studies on
developed ones (Fethi and Pasiouras, 2010).
Bonin et al. (2005) study the impact of ownership, particularly a strategic foreign owner,
on bank efficiency in eleven transition countries in Europe within the period 1996-2000.
They carry out the research by the reason that the asset majority of commercial banks in
those countries were dominated by foreign ownership during that period. They used data
of 225 banks from eleven countries (the Czech Republic, Hungary, Poland, Slovakia,
Bulgaria, Croatia, Romania, Slovenia, Estonia, Latvia, and Lithuania) and employ SFA
technique for finding the efficiency. Their results suggest that state-owned banks are not
less efficient than domestic private banks, meaning that privatization itself does not
guarantee that the commercial banks are more efficient. The other results indicate that
foreign-owned banks are more cost-efficient than their peer banks. Foreign banks also
30
COLS was proposed by Winsten (1957). The model is a frontier that is deterministic (non-stochastic),
since it excludes the statistical error . Hence, the deterministic frontier model is
31
CMAD is also known as the median regression. We can estimate efficiency with the same procedure as
the OLS, but the difference is that CMAD regression passes through the median, while the OLS
regression passes through the mean of the data.
46
provide better services in terms of their ability to gain more deposits and make more
loans than domestic private banks and government banks.
Mamatzakis et al. (2008) employ SFA to estimate commercial bank cost and profit
inefficiencies in the 10 new European members during the period 1998-2003. The 10
countries included in their research are Cyprus, the Czech Republic, Estonia, Hungary,
Latvia, Lithuania, Malta, Poland, the Slovak Republic, and Slovenia. Based on the
inefficiency scores for both cost and profit functions, the results indicate that in general
there has been la ow level of cost efficiency and profit efficiency and there are no
significant differences in inefficiency scores across the countries. The other findings
suggest that foreign banks are more profit efficient than their peers, while in terms of
cost efficiency, government-owned banks are the most efficient banks amongst the
others.
Hamiltona et al. (2010) undertook the research with the aim to examine cost and profit
efficiency in Jordanian banks over the period 1993-2006. They employed SFA to
estimate the banks efficiency using data from all 22 banks, which covers foreign and
domestic banks in Jordan. The observed banks are classified into three types of banks
namely Islamic, investment, and commercial banks. The findings of the study show that
the average levels of cost-efficiency in these types of banks are much higher than their
profit-efficiency. The levels of standard profit-efficiency of the commercial banks are
higher than those of the Islamic and investment banks. The last finding in terms of
alternative profit-efficiency is that the Islamic banks are more efficient than investment
banks and commercial banks.
The association between macro-economic factors and bank efficiency has been studied
by Chan and Karim (2010). Their study covers some developing countries in Asia, the
Middle East/North Africa, and Africa. They employ SFA to estimate the bank efficiency
of 43 developing countries during the period 2000-2005. The findings of the research
indicate that the impact of those macroeconomic factors on bank efficiency across the
regions is not uniform. Commercial banks in the Middle Eastern/North African region
are the most efficient in terms of cost efficiency with an average cost efficiency of 94.9
percent, whilst commercial banks in Asia are the most profit efficient with an average
efficiency score of 97.6 percent over the period. Lastly, commercial banks in the
47
African region are the least profit and cost efficient with an average level of 42.4
percent and 82.8 percent. In the Asia region, the GDP per capita, credit to private
sector32
, and market concentration33
have negative impacts, but trade openness34
has a
positive impact, on bank cost inefficiency respectively. In the Middle East/North Africa,
market concentration has a positive impact but trade openness has a negative impact,
both on the banks cost inefficiency. However, both market concentration and trade
openness have positive influences on the banks cost inefficiencies in the African region.
The role of ownership, market power, and institutional development in regards to
enhance bank efficiency has been examined by Fang et al. (2011). They examined the
profit and cost efficiency of commercial banks in South-Eastern Europe during the
period 1998-2008 using the SFA technique. They employed data of 171 banks in six
South-Eastern Europe countries (i.e. Albania, Bulgaria, Croatia, Macedonia, Romania,
and Serbia) and 209 banks in Central and Eastern Europe (i.e. the Czech Republic,
Estonia, Hungary, Latvia, Lithuania, Poland, Slovakia, and Slovenia) for the
comparison. The results suggest that the average bank cost efficiency in South-Eastern
Europe countries is 68.59 percent which is lower than those of Central and Eastern
Europe countries which accounts for 70.77 percent, but the average bank profit
efficiency of SEE countries denotes 53.87 percent which is higher than those of Central
and Eastern Europe countries as they mark only 52.95 percent. The other results show
that domestic private banks experience higher profit efficiency than state-owned banks.
Foreign-owned banks obtain lower cost-efficiency but higher profit-efficiency than their
peers. Nevertheless, the gap of efficiency scores over the foreign banks, domestic
private banks, and government-owned banks are not quite wide over the study period.
The regression results suggest that market power35
and institutional development36
have
positive impacts on bank efficiency. These findings reflect that banks with high market
power tend to gain more cost and profit efficient and the institutional developments
32
They employed the ratio of private credit to GDP to reflect the financial market depth in the countries. 33
They used the three firms‟ concentration ratio for representing a market structure where the commercial
banks operate. The three firms‟ concentration ratio is the proportion of the three largest banks‟ total assets
over the banks‟ total assets in the sample. 34
Trade openness is measured by the sum of export and import as a percentage of GDP (Fang et al.,
2011). 35
They use a learner index for a proxy for a market power. Learner index is calculated by the mark-up of
the price over marginal cost and divided by the price (Fang et al., 2011) 36
For representing the institutional development, they employ the development in banking regulatory
reforms, privatization, and enterprise corporate governance restructuring (Fang et al., 2011).
48
through banking reform, enterprise restructuring and privatization are important factors
to promote both banks profit and cost efficiency.
Besides SFA, many authors estimate bank efficiency using the DEA technique [See for
example, Sufian (2010), Fethi, et al. (2011), Gardener, et al. (2011), Andries (2011)].
Fethi, et al. (2011) undertook the research to assess the Egyptian performance banks
within a period characterised by changes in economic policies and investigated the
influence of the privatisation process at the end of 1995 on the efficiency and
productivity of the Egyptian banks. They also evaluated whether the liberalisation
affected the various kinds of bank ownership and size with consistent magnitude. The
DEA-based Malmquist technique was employed to measure the efficiency and
productivity of 25 commercial Egyptian banks over the period 1984-2002. The results
reveal that the productivity growth of Egyptian banking sector tends to be volatile with
a Malmquist Total Factor Productivity Index (MTFPI) of 0.9930 within the period.
Total Factor Productivity (TFP) growth of the Egyptian banking sector was -0.7 per cent
during the period with the high volatility in the pattern of Technological Change Index
(TCI).
Other findings suggest that the liberalisation policies directed to enhance bank
performance, evidenced by a growth of productivity achievement of 7.6 per cent
compared to the pre-liberalisation productivity regress. The Liberalisation had an effect
on dissimilar ownership and size with varied magnitudes. In the post liberalisation, the
most productive banks are small banks. In the post privatisation, all the privatised banks
went through productivity regress compared to the previous period, and the joint
venture banks which altered to foreign ownership inclined to gain better productivity
growth than any other ownership forms.
Andries (2011) undertook the research about bank efficiency in Central and Eastern
European countries with both approaches, SFA and DEA. The study examined factors
impacting on the banking system‟s efficiency and productivity in seven Central and
Eastern European countries over the period 2004-2008. He used a dataset which was
derived from the Bankers almanac database covering 112 banks in 7 countries:
Romania, Poland, Bulgaria, Hungary, Slovenia, the Czech Republic, and Slovakia. The
results of the research suggest that for the period 2004-2008 the average efficiency of
commercial banks in Central and Eastern European countries has increased. The highest
49
average technical efficiency based on the SFA technique is Romania with a score of
0.8644, whilst according to the DEA technique is the Czech Republic with an average
efficiency score of 0.9135. Meanwhile, the lowest average SFA efficiency is Slovakia
with a score of 0.6275 and for the DEA is Slovenia with a score of 0.6338.
The other findings reveal that based on the Malmquist index, the productivity of the
banks have increased over the study period by 24.27 per cent due to an increase in
technological modification. State-owned banks are less efficient than private banks, but
in terms of productivity, state-owned banks experience a larger increase than private
banks. Small banks (with assets of less than $1 billion) are the most efficient banks
compared to the medium banks (with assets greater than $1 billion but less than $10
billion) and large banks (with assets greater than $10 billion). But in terms of
productivity, the highest growing banks are the medium banks. The findings over the
OLS regression reveal that in order to enhance the efficiency, banks should improve
their asset quality by enhancing the quality their loan processing and implementing their
risk management to reduce their non-performing loans. The finding also suggests that
an increase in assets results in an increase in efficiency. From a macroeconomic
perspective, the finding suggests that government authorities should support the banking
liberalisation and reforms and exert some efforts to maintain the inflation rate in low
level.
3.2.2.1. Efficiency and Productivity Studies on Indonesian banks
There have been few studies about Indonesian bank efficiency (Harada and Ito (2005),
Margono et al. (2010), Hadad et al. (2011a,b), Suzuki and Sastrosuwito (2011), Besar
(2011)). These studies investigated the performance of Indonesian commercial banks by
employing the parametric and the non-parametric techniques.
Harada and Ito (2005) are among the first to estimate commercial banks efficiency in
Indonesia. They focused only on large commercial banks over the period 1999-2003
using the DEA method. Personnel expenses, general and administrative expenses, and
interest expenses are employed as inputs, while interest income and commission income
are chosen as outputs. The results show that the efficiency of those banks were on a
recovery trend as reflected by the average efficiency scores across the years. The
50
average efficiency score bottomed out at an average of 0.80 in 1999 and it then trended
to increase at the remaining years at an average of more than 0.90. It is also revealed
that the efficiency of private banks that had not accepted public funds after the 1997-
crisis performed well during the study period.
Margono et.al (2010) used the stochastic frontier approach (SFA) to estimate cost
efficiency, scale economies, technological progress, and the productivity growth of
Indonesian banks, over the period 1993 to 2000. They employ a flexible Fourier form to
attain the estimation of Indonesian bank cost efficiency with the intermediation
approach applied. They utilize two outputs namely total aggregate loans and total
aggregate securities respectively, together with the three prices of inputs, namely price
of labour, price of funds, and price of capital over the consistent data of 134 commercial
banks in Indonesia. In order to be able to identify the effect of the 1997 Asian crisis on
the efficiency, they also divide the two different periods, 1993-1997 (pre-crisis) and
1998-2000 (post-crisis).
The results show that the average cost efficiency of all Indonesian banks was 70 percent
during the research period, with 79.7 percent and 53.4 percent for the period of the pre-
Asian crisis and post-Asian crisis respectively. These findings indicate that the Asian
crisis of 1997-1998 extremely affected Indonesian bank performance as evidenced by
their reduced average cost efficiency in the period of after the crisis compared with the
period of before the crisis. In addition, there was an increase at an average annual rate of
banks‟ efficiencies for both periods, with an annual increase of 6.3 per cent for the
period 1993-1997 and 1.4 per cent for the period 1998-2000. Their findings also show
that joint venture/foreign banks were more efficient than private banks and state-owned
banks. State-owned banks were the lowest efficient banks for all the consecutive years.
Moreover, the relationship between cost efficiency and total assets suggests an optimum
bank asset size reached at between 500 to 750 billion rupiahs of annual asset size, since
the most efficient ones occurred at that interval size of assets.
Hadad et.al (2011a,b) used non-parametric slack-based DEA to examine the profit-
based technical efficiency and productivity of listed Indonesian banks and their market
performance. They found that listed banks‟ average efficiency varied widely over the
sample period between 34 percent to 97 percent and that banks‟ efficiency scores were
positively correlated with share prices and return on equity in all models. In addition,
51
they also analyze productivity changes for Indonesian banks over the period Q1 2003 to
Q2 2007. The research results show that the average productivity changes in Indonesian
banking industry tended to be driven by technological progress rather than by frontier
shift and that most of the balance sheet variables have had the expected impact on risk
management efficiency.
Suzuki and Sastrosuwito (2011) also used DEA to investigate the efficiency of
Indonesian commercial banks and examine productivity change during the period of
1994-2008. They employed 70 banks in Indonesia classified into four groups based on
their ownership, namely government-owned banks, local-government owned banks,
private owned banks, joint venture and foreign-owned banks. The results show that the
average efficiency of Indonesian banks over the study period was 0.866. The most
efficient banks were government-owned banks with an average efficiency of 0.953 and
then followed by joint venture and foreign owned banks with an average of 0.943. In
addition, the productivity growth of Indonesian banks increased by 0.5 percent annually
during the period. The achievement was supported by the technological change that had
grown by 1.7 percent whilst the efficiency change had declined by 1.1 percent annually
over the period.
Besar (2011) examined the cost efficiency of commercial banks in Indonesia employing
the SFA technique. He specifies his analysis on the influence of foreign ownership on
bank performance covering data of 119 commercial banks over the period September
2000 (2000Q3) to September 2009 (2009Q3). The intermediation approach was chosen
with two output quantities (total loans and total securities) and three input prices (the
price of labour, the price of physical capital and the price of deposits). The dependent
variable is total costs and the control variables are NPL and capital ratio. He classified
the observed banks into 4 groups: state owned banks, private domestic owned banks,
old foreign owned banks, and new foreign owned banks37
.
The results suggest that the total cost of Indonesian banking has decreased over time
with an average yearly decreasing rate of 2 percent. The highest cost efficient bank is
one of the old foreign banks that obtained an efficiency score of 92 percent, whilst the
37
Old foreign owned banks refer to banks that are majority owned by foreign owners that were
established before the crisis, whilst new foreign banks are the banks majority owned by foreign owners
that were established after crisis and as a result of foreign acquisitions.
52
lowest cost efficient was a domestic private owned bank with an efficiency score of
27.5 percent. In general, the highest average cost efficiency score was obtained by the
state owned banks, followed by the new foreign banks, domestic private owned banks
and lastly, the old foreign banks.
3.2.2.2. Efficiency and Productivity Studies in other Emerging Economies Banking
3.2.2.2.1. South East Asia (SEA)
The last decade witnessed a growing literature on bank performance in the South East
particularly discussing the effect of the financial crisis in 1997-1998 on bank
performance. However, there is still a paucity of bank efficiency and productivity
research compared to those of developed countries. The studies are fragmented mostly
in Malaysia, Thailand and Indonesia as many commercial banks are operating there [For
example, see Williams and Nguyen (2005), Sufian (2010), Chunhachinda and Li (2010),
Suhaemi et al. (2010), Gardener, et al. (2011), Abdul-Majid et al. (2011)].
Williams and Nguyen (2005) examined the comparison of bank performance and bank
governance in the South East Asia (SEA) nations during the period 1990-2003. They
measured bank governance in terms of its ownership and rank order alternative
efficiency, technical change and productivity as the measurement of bank performance.
The research highlights the important years in which the financial liberalisation was
done before the crisis and the financial restructuring was done afterwards. By using
dataset covering 231 commercial banks between 1990-2003 in South East Asia
(Indonesia, Korea, Malaysia, the Philippines, and Thailand), the results of the research
suggest that the privatisation has been successful in that region since the privatised
banks showed better performance for the post-privatisation period than before the
privatisation. In terms of ownership, private ownership banks outperformed the state-
owned banks. Furthermore, foreign acquisition banks were the best performing banks in
terms of profit efficiency. The closure and absorption banks had on average
underperformance and low profit efficiency.
53
Sufian (2010) examined the effect of 1997-financial crisis in Asian countries on bank
efficiency in Malaysia and Thailand. He chose these countries considering that the
countries are the two of four South East Asian (SEA) countries having severely
experienced crises besides Indonesia and Philippines. He adopted three different
approaches to find out the efficiency scores, namely the intermediation approach, the
value-added approach and the operating approach with the DEA technique.
In general, the results suggest that their technical efficiency scores are consistently
higher under the operating approach than those of the intermediation and value-added
approaches. The empirical findings also advise that in the post-crisis, Malaysian
banking experienced higher technical efficiency scores under the intermediation and
value-added approaches than that of under the operating approach, while the Thailand
banking sector exhibited a lower technical efficiency level under all approaches during
the post-crisis than those of the pre-crisis period. In the multivariate regression section,
it is revealed that some explanatory variables (i.e. bank size, expense behaviour and
liquidity) have negative relationships with banks efficiency in Malaysian banks. On the
other hand, bank size, loans intensity, and capitalization have positive relationships but
expense behaviour and liquidity have the positive signs with banks efficiency in
Thailand.
Chunhachinda and Li (2010) compared the profit and cost the efficiency of Thai
commercial banks during the 1997 crisis period (they sub-divided the period into the
pre-crisis, the crisis, and the post-crisis periods). The benefit of this research compared
to others is the usage of both parametric and non-parametric techniques in estimating
cost and profit efficiency scores through the SFA and DEA methods. The research
findings indicate that the average cost inefficiency for the pre-crisis, crisis, and post
crisis are 8 percent, 9 percent, and 15 percent respectively, and the average profit
inefficiency for the periods are 15 percent, 28 percent, and 17 percent respectively.
The results of the second stage empirical analysis suggest that annual real GDP has a
positive impact on expected profit efficiency while the age of bank has also a positive
impact on cost efficiency. Small banks experience lower cost efficiency levels compared
to medium and large banks. The average cost efficiencies of private banks are much
54
higher than those of state-owned banks. The ratio of loan provision38
has a negative
impact on profit efficiency, while loan deposit ratios39
have also a negative impact on
cost inefficiency. The other results suggest that equity to total assets40
has not any
impact on bank efficiency but the ratio of non-interest income to interest income41
has a
negative impact on cost efficiency.
Suhaemi et al. (2010) investigated the determinants of commercial banks profit
efficiency in Malaysia over the period 1995-2007. They employed SFA to find out the
efficiency scores for 23 commercial banks operating in Malaysia. They chose profit
efficiency instead of cost efficiency as their arguments stated that profit efficiency could
take into account the two effects, namely cost effect and revenue effect, although the
previous studies on bank efficiency have almost always employed cost efficiency. Cost
effects are related to the bank that do not apply the minimum input and revenue effects
are regarding the banks that do not reach the maximum output.
The findings reveal that the average profit efficiency for both local banks and foreign
banks tend to have a downtrend during the period with an overall average of 0.6916 in
1995 and 0.3557 in 2007. A downtrend of profit efficiency is identified due to the fact
that the interest income has faced the same downtrend within the period. The second
empirical evidence demonstrates that information and communication technology (ICT)
infrastructure has a positive impact on the profit efficiency although the effect is not
significant. The significant factors for bank profit efficiency are non ICT infrastructure,
market share, and ownership with each having a positive direction, whilst size has a
negative association with the profit efficiency.
Gardener, et al. (2011) study covers five South East Asian countries. They employed the
non-parametric DEA and TOBIT regression to answer the research questions. The study
aims to investigate bank efficiency in the South East Asian economies after the 1997
financial crisis and to assess the effect of post-crisis restructuring. The findings of this
38
Loan provision represents the credit risk of a bank.
39
Loans deposits ratio is a proxy for a bank liquidity risk.
40
Equity to total assets denotes a bank capital risk.
41
Non-interest income to interest income represents market risk for a bank.
55
research reveal that the structure of ownership has significantly affected bank efficiency
in South East Asia over the period 1998-2004. Foreign banks operate more efficiently
than domestic banks. Within each country, government-owned banks are recorded more
efficient than private banks. The other results suggest that the restructuring programme
implemented in that region (post the 1997-crisis) does not have a positive effect on bank
efficiency, reflected by the downtrend of the efficiency scores over the period 1998-
2004.
One of the most recent studies on bank efficiency in Malaysia was completed by Abdul-
Majid et al. (2011). They estimated the efficiency and the TFP (total factor productivity)
change of commercial banks in Malaysia during the period 1996-2002. The findings
indicate that the average net inefficiency of commercial banks in Malaysia was around
4-6% annually from 1996 to 2002. Foreign-owned banks are on average more efficient
than domestic ones. Mergers, Islamic full-pledge banks and conventional banks which
also operate the Islamic banking windows experienced less efficient. The results also
suggest that the banks‟ average productivity change of 2.37 percent was assigned to
technical change and it has declined during the period.
3.2.2.2. East Asia (EA)
Many studies about the efficiency of commercial banks have been completed in East
Asia region covering new industrial countries (i.e. China, Korea, and Taiwan) as well as
Japan which is recognized as one of the developed economies in the world [Among
others, see Altunbas et al. (2000), Drake and Hall (2003), Yao et al. (2008), Yeh (2011),
Shin and Kim (2011)].
Altunbas et al. (2000) examined risk and quality factors and their influence on bank
performance, especially bank cost efficiency and technical change using the SFA
method. They utilize data from commercial banks in Japan during the period 1993-1996
which covered 139 banks for 1993-1995 and 136 banks in 1996. By comparing the cost-
function specifications between addressing and not addressing the risk and quality
factors, the results show that when the function is not taking into account the risk and
quality factors, all Japanese banking except for the largest Japanese banks experience
56
scale economies. When the function is controlled by the risk and quality factors, the
majority of commercial banks experience diseconomies of scale excluding the very
smallest banks. The findings indicate that the scale of economies of bank has become
smaller for controlling the risk. This result also suggests that bank cost can be reduced
by technical change. However, it happens at a declining rate, during the period 1993-
1996.
Drake and Hall (2003) evaluated Japanese bank performance with a non-parametric
frontier approach, the DEA method. The aims of their research were to estimate the
technical and scale efficiency in Japanese banking employing a cross-section sample of
149 banks on March 1997. The results do not conform to the wave of mergers occurring
during the time since in terms of economies of scale the evidence indicates that
Japanese‟s large banks are operating well above the minimum efficient scale, thus the
mergers among them tended to intensify their diseconomies of scale. Moreover, the
findings also suggest that those large banks, especially the Long-Term Credit Banks
(LTCBs), incline to have the lowest points of pure technical inefficiency among all
Japanese banks. On the other hand, a contradictory result occurred for smaller banks
since the evidence of significant economies of scale for the smallest Japanese banks
during the period.
Yao et al. (2008) studied the technical efficiency of Chinese banks using DEA. They
concentrated on the efficiency issue regarding the new era of the pre and post-World
Trade Organization (WTO), by investigating the positive or negative reaction of
Chinese state commercial banks through the new challenges. The research employs data
from all commercial banks in China during the period 1998-2005. Besides using the
DEA technique in gaining efficiency scores, they also carried out a Malmquist index
analysis to find out bank productivity changes. The research findings show that the gap
of average efficiency scores from high and low levels of efficiency is at only 15 percent
indicating that Chinese national banks do not have significant differences in their
technical efficiency. Another finding is that the Chinese national banks react positively
and aggressively to the new reforms of ownership and challenges by foreign
competition since the total factor productivity (TFP) has increased by 5.6 percent
annually during the period 1998-2005.
57
The association between capital structure and cost efficiency in the banking industry
was accomplished by Yeh (2011). The study aimed to examine whether banks are able
to diminish their agency costs through the optimization of capital structure in enhancing
their efficiency. They adopted the stochastic frontier approach to estimate the bank cost
efficiency of 44 commercial banks in Taiwan from 1999 to 2004. They also employed
two-stages least squares to investigate the effect of capital structure on banks efficiency.
The findings of this research suggest that optimal capital structure can be used by
managers to reduce the agency problem and increase bank efficiency as a proxy of bank
performance. These results are evidenced by the positive association between debt ratio
and bank efficiency. Another finding is that insider share ownership has a negative
effect on bank efficiency. This means that by reducing managerial share-holding, banks
can decrease agency costs and finally enhance their performance. In addition, bank size
has a positive impact on bank efficiency reflecting that larger banks are more efficient.
Another study was done by Sufian (2011). His research aimed to examine the efficiency
of the Korean banking industry over the period 1992-2003. The advantage of this
research is the usage of three various approaches to obtain different efficiency scores as
the changes in input and output components. The three approaches are intermediation,
value added, and the operating approach. The study employed the DEA technique to
estimate bank efficiency on around data of 31 banks in Korea during the period 1992-
2003. The result shows that the technical efficiency estimates are constantly higher
under the operating approach than those of derived from the intermediation and value-
added approaches. The lowest level of technical efficiency scores are shown by the
intermediation approach.
Shin and Kim (2011) also carried out research about bank efficiency in Korea. The
objective of this research was to examine the relationship between the efficiency and
profitability of the Korean banking industry. The paper also investigated whether the
Korean banking efficiency increased when the bank restructuring program was started
in 1997. By using the DEA technique for estimating efficiency, the results of this
research suggest that the higher efficiency banks exhibit higher profitability and bank
efficiency influences positively on the market share and market concentration.
Nevertheless, in the period of the post-1997 crisis, bank profitability was not improved
by technological efficiency rather than by the increased scale efficiency.
58
3.2.3. A Brief Theory of Loan Management in Commercial Banks
Finance or lending is one of the most important activities in commercial banking.
Commercial banks play a unique role by channelling funds from depositors in the form
of loans to borrowers (Heffernan, 2009). The role reflects a commercial bank as a
financial intermediary institution between depositors and borrowers. Howells and Bain
(2008) formulate the main activities of a commercial bank in a balance sheet as follows:
Table 3.1. A commercial bank’s balance sheet
Assets Liabilities
Notes and coin Capital and shareholder's funds
Deposits at the central bank Customer deposits
Loans to the money markets
Investments
Loans to public sector
Loans to the general public
Source: Howells and Bain (2008)
A Commercial bank engages with the maturity transformation42
activities, thus it
manages the deposits collected on the liabilities side and provides some accounts on the
assets side through as bank reserves, as loans to money market or other
financial institutions, as investments in the form of securities and other investments.
denote loans to public sector and to non-bank private sector, respectively.
Loans to non-bank private sector are an important part in the asset side since it
commonly dominates bank‟ assets.
Banks make profit mainly by issuing loans or financing. Loans have mostly generated
more than half of bank revenues [Mishkin (2010), Saunders and Cornett (2012)].
However, loans are also recognized as the major sources of the credit43
and liquidity
risks44
in commercial banks. Considering the circumstances, it is considerably vital by
bank managers to manage loans correctly in order to create revenue with less risk
incurred.
42
It relates to arrangement of the short-terms deposits into longer-term loans and investments. 43
Credit risk or default risk is the event when borrowers failed to repay loans. 44
Liquidity risk, since loans cannot be turned into cash until the loans matures.
59
Loan portfolio management has been one of the most crucial aspects in bank
management in the sense that inappropriate management of the bank loan portfolios
could lead to a negative impact on bank performance, and generally on the banking
system of a country (Macerinskiene and Ivaskeviciute, 2008).
Banks can classify their financings into the many categories45
. Each loan category
entails a wide variety of characteristics that must be evaluated to determine the risk
involved and to decide whether they should grant loans to potential customers. For
instance, in the US, it is common practice to stipulate the total loans along with their
components, i.e. commercial and industrial loans, real estate loans, individual loans, and
other loans46
. Such classifications are very beneficial for bank managers to assess the
performance of each loan as well as the total loans of each category respectively.
Bank managers should carefully organise all loan processes starting from the search of
potential borrowers to the maintenance and monitoring of their loans. Banks will incur
intermediary costs in the process of searching, verifying, and monitoring potential
borrowers‟ creditworthiness (Heffernan, 2009). The intermediation process and its costs
also encompass administration cost and other costs of transactions related to savings and
financings products provided by banks. In this regard, the aim of a bank is to create
qualified loans or financings in order to cover all those expenses together along with the
expected return.
Bank managers apply different methods to assess the credit worthiness of potential
customers (loan applicants) based on the type of loans using the principle of 5 C‟s47
of
credit analysis. The emphasis of the assessment in the credit scoring can be relatively
different accordingly. For instance, in the process of assessing a mortgage loan,
commercial banks‟ decisions are mainly dominated by two considerations48
: the ability
and willingness of the applicant in making timely repayments of the principal and
interest, and the applicant‟s collateral value. Thus, collateral value is very important in
taking out mortgage loans, whereas for nonmortgage consumer loans, the focus is on the
45
According to the intended use, loans can be classified as investment loans, working capital loans and
consumer loans. Based on the size of the borrowers, the loans can be divided as small business loans,
medium-business loans, and large business loans. 46
See Saunders and Cornett (2012: 352) 47
5 C‟s of credit: character, capacity, capital, collateral, and conditions. 48
For more details, see Saunders and Cornett (2012).
60
individual‟s ability evidenced by the personal characteristics i.e. annual gross income
and the total debt service49
.
Regarding the evaluation of small business loans, Saunders and Cornett (2012)
recommend that credit scoring for the small business loans model should be more
sophisticated compared to those that are applied to mortgages and consumer loans. They
argue that the credit risk of small business loans is higher due to incomplete accounting
and financial information (information asymmetry). To mitigate the risk emerging from
information asymmetry a rather deeper analysis should be applied with regard to small
businesses. Moreover, the credit scoring should complement the computer-based
analysis of borrower‟s financial statements with the behavioural analysis of small
business owner.
Credit analysis for the larger segment of customers, namely mid-market corporate50
, is
quite different compared to small business loans credit analysis. The focus is on the
business itself besides the character of the company‟s management. Thus, it covers
several dimensions, including the five C‟s of credit, financial ratio analysis, analysis of
cash flow, and the comparison of financial statements.
Mishkin (2010) sheds light on the importance of banks in understanding the concept of
adverse selection51
and moral hazard52
problems that could lead loans more likely to
default. Adverse selection occurs when a bank mistakenly chooses potential borrowers.
Banks may choose those who are particularly at the very high risk because of
information asymmetry however they are likely to produce adverse outcome. Moral
hazard happens if borrowers have inducement to involve in undesirable activities from
the viewpoint of the lender. Hence, in this circumstance, bank may face the hazard of
default.
49
Total debt service ratio is calculated by total accommodation expenses plus all other debt service
payments divided by gross income (Saunders and Cornett, 2012). 50
In the US, mid-market corporates typically have sales revenues from $5 million to $100 million a year
(Saunders and Cornett, 2012). 51
Borrowers with very risky investment projects but they are chosen by a bank is an example of adverse
selection, since they are most likely to be unable to payback their loans (Mishkin, 2010). 52
For instance, when borrowers obtain loans, they tend to invest in high risky projects (that pay high
returns if successful). Nevertheless, the high risk project creates them less likely to be able to pay the loan
back (Mishkin, 2010).
61
To overcome the problem of adverse selection, Mishkin (2010) advises the lenders to
collect more reliable information about the potential borrowers and screen out the bad
credit risk from the good ones. By these actions, it is expected that bank would only
grant the loans or financings to the prospective borrowers. The other way to mitigate the
risk of adverse selection is becoming more specialized bank in lending53
. While in
reducing the risk of moral hazard, Mishkin suggested bank to apply restrictive
covenants54
into loan contract.
Regarding the association between bank size and lending activities, Saunders and
Cornett (2012) elaborate that small banks mostly concentrate on the retail side of the
business (i.e. making loans and issuing deposits to customers and small businesses),
whilst larger banks engage in both retail and wholesale banking55
. These circumstances
are underpinned by the relatively easy access of larger banks to purchased funds and
capital markets compared to small banks.
In reaching the goal of maximizing profits, banks would also deal with macroeconomic
risks that surrounding their circumstances. These environment factors could be directly
and indirectly influencing bank performance. For instance, inflation and recession in the
economy are among the risks that should be accounted for by bank managers
(Heffernan, 2009).
3.2.4. The Empirical Literature on the Bank Lending Propensities
The literature that discusses the lending propensities of commercial banks mostly
focuses on the tendency of banks to provide lending to small businesses. One of the
main objectives of this thesis is to discuss the trend of Indonesian commercial banks in
providing their lending facilities (total loans and small business loans) to their
customers (borrowers) as well as to examine the effect of the asset size of Indonesian
banks on their propensity to lend to their borrowers.
53
This means that a bank should concentrate its lending only to companies in specific industries until it
becomes more knowledgeable about the industries. 54
Covenants restrict borrowers from engaging in risky activities. 55
Wholesale banking is commercial oriented banking, providing commercial & industrial loans funded
with purchased funds (Saunders and Cornett, 2012).
62
Most of the literature in this area use surveys data for their studies [Jayaratne and
Wolken, (1999), Scott and Dunkelberg (2003), Berry and Grant (2004), Clarke et al.
(2005)]. Other studies discuss the determinants of small business lending using
secondary data from small businesses [Strahan and Wetson (1998), Peek and Rosengren
(1998), Berger et al. (2001), Shen, et al., (2009), McNulty, et al. (2013)]. This thesis
focuses on the lending propensity of commercial banks and their determinants.
Jayaratne and Wolken (1999) investigated the importance of small banks in providing
small business finance in the US. They employed data of the National Surveys on Small
Business Finance (NSSBF) in 1993, to test whether the wave of bank mergers which
occurred in 1990s in the US significantly affected on the probability of a small business
to have a credit line from a commercial bank. They also examined whether small
businesses experienced difficulties in paying their trade-credit. Their findings suggest
that the probability of a small firm to obtain credit facilities from banks did not decrease
in the long-run although in areas where there were only a few numbers of small banks
available. They also found that small businesses do not have difficulties in paying their
trade-credit reflected in their on-time payment records.
Scott and Dunkelberg (2003) scrutinized the impact of bank mergers on the
opportunities for small firms in obtaining financing in the US. The study employed data
resulted from small business survey which was undertaken by the National Federation
of Independent Business in 1995. The results of the study exhibit that there is no
influence of bank mergers on the ability of small businesses to obtain loans. Mergers
have also no significant effects on new loan interest rates except for some other non-
price terms such as collateral requirement, compensating balance, and other fees.
Berry and Grant (2004) conducted a survey on small business lending in the United
Kingdom. They used survey data to examine the extent of the involvement of European
Banks operating in the UK in providing financing for Small Medium Enterprises
(SMEs) in the UK. The study analysed the approach (i.e. the going concern56
approach
or the gone concern approach57
) employed by these banks as a lending practice.
56
Going concern approach is the assessment of the credit worthiness by looking at the future prospects of
potential customers which are reflected from their ability of future cash generation.
57
Gone concern approach, also called liquidation approach, is the evaluation method of creditworthiness,
with emphasis on what will happen in the worst case scenario, through the sale of the assets and other
security taken in the form of fixed and floating charges, guarantees, and restrictive covenants.
63
Additional information was gathered by interviewing the lending managers of all those
European banks that are presumably active in lending to SMEs in UK. Their results
reveal that only 10 European banks out of 97 banks were active in providing loans to
UK SMEs. All these banks are branch offices with headquarters in their home countries.
Most of them adopted a mixture both approaches in assessing loan feasibility.
Clarke et al. (2005) investigated the relation between bank origins and share and growth
of small business financing. The study employed commercial bank data of four Latin
American countries (Argentina, Chile, Colombia, and Peru) during the period of the late
1990s. The results show that on average, foreign banks in Chile and Peru experienced
less financing for small businesses compared to domestic banks. In Argentina and
Colombia, foreign and domestic banks lent to small businesses with a nearly equal
proportion. In addition, in the case of Chile and Argentina in general, the findings
suggest that large foreign banks exceed large domestic banks in terms of share and
growth rate for small business loans.
Strahan and Wetson (1998) investigated the relationship between small business
lending, bank size and bank consolidation in the US banking industry within the period
1993-1994. They aimed to examine whether the size and complexity of a bank 58
has an
impact on its ability to provide loans to small businesses. In addition, they also
investigated the influence of consolidation, which will result in a larger and more
complicated bank, on the availability to provide lending to small businesses. The results
of the study state that within the period, small business lending has increased in terms of
the percentage to total assets for the small banking companies until their assets reached
around $300 million, and it downturns after that point. Whereas for the larger banking
companies, small business lending has risen slowly as the size of banks increased with a
decline in portfolio percentage. The complexity of the bank has no significant
relationship with lending to small businesses. Consolidation through mergers and
acquisitions (M&A) among small banks have a positive impact on the increase in the
small business lending, but the M&A among large banking companies and small
banking ones have also a positive effect although the increase is not significant.
58
Strahan and Weston (1998) explained complexity of a bank as a more complex organizational structure
of a bank. They assume that the complexity of large banks may cause diseconomies in the organization
which provide relationship loans more expensive for them. The total numbers of subsidiaries held by a
banking company and the number of states in which the bank operates are the proxies for the complexity.
64
Peek and Rosengren (1998) investigated the influence of mergers on bank willingness59
to finance small businesses. The study corresponds to the concerns over public policy
stemming from the wave of mergers that caused the shrinkage of the number of banks
by nearly 30 per cent in the US over the period 1985-1995. They aim to explore whether
bank consolidation has an impact on the availability of loans to small businesses. In
order to answer the research questions, they employed a survey conducted annually
based on the June Bank Call Report over the period 1993-1996. The results demonstrate
that in around half of the mergers, the acquirers had more small business loans than
those of their target banks before mergers and they tended to enhance their small
business loans after the mergers. The results also reveal that small acquiring banks
tended to be more active in small business lending than the large acquirers.
Berger et al. (2001) examined the impact of bank size, foreign ownership60
and
distress61
on lending to informational opaque small businesses in Argentina. By
employing the data set of 61,295 non-financial firms and total loans from 115 different
commercial banks as of the end of 1998, the results suggest that large and foreign-
owned banks experienced some difficulties in extending relationship loans to opaque
small businesses. These conditions were caused by the higher expenses experienced by
those banks in providing relationship lending services to small businesses together with
other services to large companies. This was also magnified for small businesses with
delinquencies in the repayment of their debts. However, distressed banks experienced
no significant matters in providing loans for small firms during the period.
Shen, et al. (2009) examined the determinants of small business finance through
investigating the association between bank size and small business lending in some
commercial banks in China. They employed panel data of county level banks operating
in China assessing how bank size and other determinants namely discretion over credit,
incentive schemes, competition, and the institutional environment62
affected SMEs
59
Peek and Rosengren (1998) consider the „willingness of banks to finance the SBs‟ as reflected in the
percentage of small business loans to total assets. 60
Foreign ownership in this case means a branch or subsidiary in which its headquarters abroad or in
which at least 50% of the capital is foreign-owned (Berger et al., 2001). 61
Berger et al. (2001) define a distressed bank as an unhealthy bank or a bank experiencing financial
difficulties due to some reasons i.e. the credit crunch or the financial crisis. 62
Loans approval share is a proxy for the discretion over credit. Bank‟s loan market share is a proxy for
the competition. For the incentive scheme, the study uses the dummy of „linkage of wage with the loan
quality‟ with the hypothesis link wage and loan quality tend to increase SMEs loans. For the institutional
environments, they use dummies for government influence on the loans approval, degree of law
65
lending in China. The data covers 79 counties within 12 provinces detailing the
information the banks‟ structure of governance, loans and deposit policy, the system of
incentive, and banks‟ balance sheet over the period 2001-2004. The study provided the
treatment of instrumental variables for the endogeneity problem resulting from the
association between some explanatory variables and error term in the study. The results
show that bank size had no impact on the SMEs lending. Some variables (the linkage of
wage and loan quality, the competition, and institutional arrangement) significantly
affected the SMEs lending. In other words, if loan quality is connected with the
manager‟s wage, if there is greater competition, and if an institution has more self-loan
approval right, then financing to SMEs will be higher. In addition, law enforcement also
influenced the SMEs lending. The weak law enforcement leads to low SMEs lending.
The most recent study that elaborates the determinants of small business finance has
been undertaken by McNulty, et al. (2013). They investigated the relationship between
small business lending and bank size by using 2SLS cross sectional regressions with
fixed effects approach over 5,537 banks for 1993-2006. Their findings suggest that there
was a negative association between bank size and small business lending proportion,
meaning that the propensity to lend to small businesses declined when bank size
increased. Small business lending growth did not compensate for the growth in the size
of the bank. When a bank asset size increased from $ 1 billion to $ 100 billion, the
proportion of small business loans was reduced by 28 percentage points. The results
also suggest that small banks have recorded most of their loans to small businesses.
Small banks have made 28.4% of total banking loans for small businesses even though
they only accounted for 14.1% of total banking deposits and 9.7% of total banking
assets during the period 1993-2006. They also provide recommendations for continuing
research on lending propensity because it contains very valuable information as a proxy
of the supply of loans.
One of the chapters of this thesis examines the determinants of Indonesian bank lending
propensity (in terms of total loans and small business loans) on the commercial banks in
Indonesia over the period 2002-2010. It particularly explores the influence of size of the
banks on their total loans and small business loans. The study gives the contributions to
the literature since it would be the first study to exhibit the factors explaining total
enforcement when enterprises defaulted. All those data are particularly collected by the surveys on the
banks managers.
66
finance as well as small business finance in Indonesia whilst it has been done by authors
in several countries to date.
3.3. Conclusion
This chapter aim to provide brief background on theoretical and empirical literature on
banks efficiency, productivity and propensity to lending to small businesses. The
chapter covered studies from emerging economies, with particular focus on studies from
South East Asia and East Asia.
The first section of this chapter reviews the brief theory of production efficiency as well
as empirical evidence on bank efficiency in some emerging countries. The focus of the
empirical evidence presented is the efficiency and productivity studies on Indonesian
commercial banks as well efficiency and productivity studies on emerging economies
particularly in areas adjacent to Indonesia i.e. the South East Asia (SEA) and East Asia
(EA) countries. These studies collectively cover bank efficiency and productivity
studies in Indonesia and some emerging economies amounting to twenty six articles in
totals are reviewed.
The second section also provides a brief theory of loan management and the previous
literature on the determinants of bank lending propensity. It covers the importance of
loan portfolio management in managing commercial banks and the determinants of
bank lending propensity mainly in small business lending as one of the important banks
portfolio financing in some countries covering the US, the UK, Latin America Countries
and China.
Regarding the studies of bank efficiency in Indonesia and the adjacent regions,
concerning the seventeen studies reviewed, nine studies employed the non-parametric
technique DEA, seven studies employed the parametric technique SFA and only one
study used the SFA and DEA simultaneously.
The literature on Indonesian bank efficiency and productivity is limited. To the author
knowledge there are only five published studies on Indonesian bank efficiency, of which
three have employed the DEA ([Harada and Ito (2005), Hadad et al. (2011a,b), Suzuki
67
and Sastrosuwito (2011)], while the rest [Margono et al. (2010) and Besar (2011)] use
the SFA. Harada and Ito (2005) are the first authors examine the efficiency of 10
Indonesian banks over the period 1999-2003 with the DEA method. The results indicate
that the average efficiency score bottomed out in 1999 and it then trended to increase
during the remaining years. A distinctive feature of these studies is that they have
covered the performance of Indonesian banks over a limited number of years or a
limited number of banks.
It is worth noting that most of studies about small business lending are concerned with
the two issues: Firstly, investigating the involvement of commercial banks in providing
financing for small businesses. They mainly reveal survey or interview findings.
Secondly, examining the determinants of small business finance in commercial banks,
particularly exploring the association between bank size and willingness to provide
financing for small businesses.
Regarding the first issue, the studies‟ findings reveal that small banks experienced less
difficulties in providing small business finance and the probability of Small businesses
to have access to bank finance from small banks did not change (Jayaratne and Wolken,
1999), there was no influence of bank mergers on bank financing for small businesses
(Scott and Dunkelberg, 2003), there was a small amounts of European banks operating
in the UK which provided financing for small businesses (Berry and Grant, 2004).
Foreign banks in some Latin American Countries practiced less financing for Small
businesses (Clarke et al., 2005).
For the second issue, the results reveal that small banks tended to finance small
businesses and the consolidation between small banks increased small business finance
(Strahan and Wetson, 1998). Small acquiring banks were more likely to finance small
businesses compared to large acquiring banks (Peek and Rosengren, 1998). Large and
foreign-owned banks experienced some difficulties in extending relationship loans to
opaque small businesses (Berger et al., 2001). Bank size did not impact on SMEs
lending (Shen, et al., 2009). There was a negative association between bank size and
small business lending proportion (McNulty, et al., 2013). The last results mean that
that the propensity to lend to small businesses declined when bank size increased.
68
The present thesis contributes to the literature in six fold; first, it investigates the
performance of Indonesian banks by estimating both technical and cost efficiency.
Second, it utilises both the DEA and SFA techniques to provide more insight on the
performance of Indonesian banks using different methodological techniques. Third, it
covers a sample of Indonesian banks that represent 93% of the total banks operating in
Indonesia. Fourth, it is the first to investigate lending propensity to small businesses in
Indonesia. Fifth, the sample covers the years of change in banks‟ regulations as well as
the financial crisis in 2007, hence the study investigates the impact of the crisis and
regulatory change on the Indonesian banking sector. Sixth, it is the first to investigate
the effect of small business lending on bank technical and cost efficiency.
69
4. Chapter Four: Methodology
4.1. Introduction
This thesis employs the frontier approach in measuring bank performance. The frontier
approach is supposed to be robust compared to the traditional approach such as financial
ratio analysis (Berger and Humphrey, 1997). The advantage of frontier approach
compared to the traditional approach is that the frontier approach compares the
performance of banks to best practice frontier which is more beneficial in terms of
regulatory purposes. Therefore, it provides an analytical tool that is useful for policy
makers to investigate the impact of government policies (i.e. deregulation,
liberalization, privatization, mergers and acquisitions) on the performance of banking
sector. It is also a useful comparison tool in case of the occurrence of certain events e.g.
financial crisis or disruption [see Bauer et al. (1998), Shaban (2008)]. Frontier approach
uses either linear programming or econometric techniques to eliminate the effects of
exogenous market factors on inputs and inputs prices. Therefore, the method can
provide more reliable estimates of firms‟ performance (Shaban, 2008).
Technically, the frontier approach identifies an unobservable best practice
(benchmark(s)) bank(s) in terms of efficiency and measure the performance of other
banks in the sample relative to the best practice bank(s). This helps to provide
recommendation to the non-efficient banks (non-frontier banks) to improve their
performance to catch-up with the efficient one (frontier-banks).
There are two prominent methods in estimating the efficiency of commercial banks,
namely the non-parametric approach Data Envelopment Analysis (DEA) and the
parametric approach Stochastic Frontier Analysis (SFA). Each of these approaches has
its own advantages and disadvantage. So far there is no consensus in the literature on
which approach is the most preferred in determining the best practice frontier and
estimating the efficiency scores (Berger and Humphrey, 1997).
This thesis adopts both methodologies to obtain efficiency estimate for Indonesian
commercial banks during the period 2002-2010. The analysis of Indonesian banks‟
70
performance is extended by estimating non-parametric productivity using the DEA-
based Malmquist Total Factor Productivity Index/MTFPI (Färe et.al, 1994).
The thesis also employs TOBIT-regression to investigate the determinants of Indonesian
bank efficiency scores obtained from DEA. Preceding studies adopted TOBIT
regression on their efficiencies [For example, see (Chang et.al (1998), Nikiel and Opiela
(2002), Casu and Molyneux (2003), Havrylchyk (2005), Grigorian and Manole (2006),
Ariff and Can (2008), Sufian (2010), Gardener et.al (2011)].
The chapter is organised as follows. The first section discusses the comparison between
the non-parametric DEA approach and the parametric SFA approach, and the review on
the concept of DEA approach and DEA-based Malmquist productivity index
particularly for the production function on the technical efficiency. The second section
presents a review on the parametric SFA approach and some underlying estimation
models more specifically for the cost frontier panel data model. The third section
explains the estimation of the regression equations employed in this thesis covering the
TOBIT regression for the efficiency scores, multiple regressions with fixed-
effects/random-effects and/or the Two-stage least square (TSLS) regression on small
business finance63
. The fourth section explains the variables and the data employed in
the empirical chapters. The last section provides conclusions.
4.2. A Comparison of the DEA and SFA Approaches
A brief discussion about the evolution of the efficiency measurement revealing the
proposition of Farrel (1957)64
as the first to define technical efficiency, allocative
efficiency (price efficiency) and economic efficiency (overall efficiency) has been
presented in Chapter 3 in the literature review section.
63
The determinants of small business finance is presented on the third empirical chapter of this thesis (see
chapter 7) 64
Farrel (1957) was inspired by the work of Debreu (1951) and Koopmans (1951) to define those
efficiency terms covering technical efficiency, price efficiency, and overall efficiency with taking into
account the multiple inputs.
71
The Farrell‟s concept focuses on the best practices resulted from the data observation
suggesting the term „frontier‟ that all the observation would be compared to it. The
Farrell‟s work has brought to the two developed methodological mainstreams in
constructing the production frontiers, namely parametric (stochastic) approach and non-
parametric (non-stochastic) approach.
Shephard (1970) and Afriat (1972) are among the authors who were then interested in
the frontier concept and advised the usage of mathematical programming methods in
frontier estimation. The concept was first attractive to many authors since Charnes,
Cooper and Rhodes (1978) introduced the term data envelopment analysis (DEA) in
their paper. Since then the DEA approach has started to be widely used by authors
(Coelli et al., 2005).
Charnes, Cooper and Rhodes (1978) first introduced the DEA approach with input
oriented scheme and the constant returns to scale (CRS) for the technological
assumption. The model was widely known as the CCR model and it has been
extensively employed by authors until Banker, Charnes and Cooper (1984) proposed the
variable return to scale (VRS) in their assumption. The model was known as BCC
model and it has been more widely used to date. The idea behind the latter model is
underpinned by the weakness of the CCR model. The CCR model is only appropriate if
all firms are operating at an optimal scale. However, some conditions (i.e. government
regulations, imperfect competition, constraints on finance) could cause a firm not to
operate in the optimal scale. Hence, as the solution, the BCC model was introduced
(Coelli et al, 2005).
On the other hand, regarding the stochastic frontier analysis (SFA) evolution,
Kumbhakar and Lovell (2000) mention some significant contribution of authors [i.e.
Koopmans (1951), Debreu (1951), and Shepard (1953)]. These authors provided the
influential theoretical literature on productive efficiency. Kumbhakar and Lovell (2000)
remark that Koopmans‟ contribution was in the provision of the technical efficiency65
,
while Debreu and Shepard proposed the distance functions as a method of addressing
65
Technical efficiency was defined by Koopmans (1951) as “a producer is technically efficient if and
only if, it is impossible to produce more of any output without producing less of some other output or
using more of some input” (Kumbhakar and Lovel, 2000:6).
72
multiple-output technology, as well as measuring the radial distance of a producer from
a frontier.
The other authors having contributions to the development of SFA are Aigner and Chu
(1968), Seitz (1971), Timmer (1971), Afriat (1972) and Richmond (1974). These
authors estimated a production function with the deterministic method, either by
employing linear programming techniques or modifying the least squares techniques
(Kumbhakar and Lovell, 2000).
However, the significant articles acknowledged as the SFA origination are created by
Meeusen and Broeck (1977) and Aigner, Lovell, and Schmidt (1977), which are then
followed by Battesse and Corra (1977). These two aforementioned papers (MB and
ALS66
) were published almost simultaneously by two different teams of authors from
the two different continents. These three papers contain the composed error structure
and were developed in the context of production function. The inclusion of composed
error structure in the estimation of efficiency is then recognized as one of the important
features of the SFA compared to the other approach.
In general, both approaches (the DEA and SFA) have their own advantages and
disadvantages. DEA is one of the non-parametric techniques that uses a linear
programming concept in the operation and does not account for random errors. Thus, it
is known as deterministic (non-stochastic) method. Conversely, SFA is one of the
parametric techniques that requires functional form and takes into account random
errors. Hence, it is also recognized as a stochastic method.
In the operation, SFA necessitates the assumption for the error term distribution and
employs the functions i.e. production function, cost function, etc. For instance, in the
production function, the left-hand side of the equation (the output of a firm) is a
function of the right-hand side of the equation (it comprises a set of inputs, inefficiency
and random error).
Meanwhile DEA is very popular in the non-parametric mainstream in estimating
efficiency with some advantages as follows (Suzuki and Sastrosuwito, 2011): DEA is a
mathematical linear programming model and more flexible in terms of its ability to take
into accounts for multiple outputs and inputs in estimating the efficiency. Moreover,
66
MB stands for Meeusen and Broeck, and ALS is Aigner, Lovell, and Schmidt.
73
DEA works well with a relatively small sample, and unlike the SFA, it does not
necessitate predetermined assumptions about the distribution of inefficiency and does
not require a specific functional form on the data to determine the most efficient banks
(Fethi et al., 2011).
However, the DEA also has some disadvantages. The weakness of the DEA and non-
parametric methods in general is that the assumption of not allowing for random error
due to luck, data problems or other measurement errors. The problem might arise due to
some errors in one of the units on the efficient frontier can affect the efficiency scores of
other units, especially compared to efficient units (Berger and Humphrey, 1997). In
addition, Coelli et al (2005) emphasized that, because of ignoring statistical noise, the
efficiency estimation resulting from the DEA approach may be biased if the production
process is characterized largely by stochastic elements (Coelli et al., 2005).
On the other hand, SFA have some advantages compared to DEA are: first, it takes into
account statistical noise (random errors) and second, it allows carrying out some
statistical tests (Coelli, et al, 2005). Table 4.1 shows that SFA permits the carrying out
some statistical tests i.e. hypothesis test, variables inclusion test, distributional
assumptions test, and endogeneity bias test.
Nevertheless, SFA also has some drawbacks. Among those weaknesses are the
prerequisites of distributional form in determining inefficiency and the functional form
for estimating production function, cost function, etc. (Coelli, et al, 2005). These
weaknesses led to the SFA method is considered more complicated than the DEA.
Table 4.1 summarizes the comparison between SFA and DEA methods as follows:
Table 4.1.The Comparison between SFA and DEA methods
Conditions SFA DEA
Prior assumptions on production frontier needed Yes No Accommodates multiple inputs and outputs Yes Yes
Accounts for error terms Yes No
Prior assumptions for the error term Yes No Allows for environmental variables Yes Yes
Allows for hypothesis test Yes No
Allows for variables inclusion test Yes No Allows for distributional assumptions test Yes No
Multicollinearity problems Yes No Provides information on peers No Yes
Vulnerability to small number of observations Yes Moderate
Vulnerability to endogeneity bias Yes Yes Allows for endogeneity bias test Yes No
Source: Shaban (2008)
74
The noticeable similarities between DEA and SFA are: both methods accommodate
multiple inputs and outputs and allow for environmental variables in estimating the
efficiency. The same circumstance is also applied to both methods in terms of
vulnerability to endogeneity67
bias.
4.3. Non-parametric Methods: Data Envelopment Analysis (DEA)
Data Envelopment Analysis (DEA) idea was firstly inspired by seminal work of
Farrell‟s (1957) and was subsequently expanded by the works of Charnes et al. (1978)
and Banker et al. (1984). Through the DEA, it is estimated the best practice decision
making units (DMUs) in the frontier and relatively compared the rest of DMUs to the
frontier.
The Data Envelopment Analysis (DEA) approach is one technique that creates a non-
parametric piecewise surface (or frontier) over the data by using the methods of linear
programming. In estimating the efficiency of each observation, the surface becomes a
benchmark(s) where each of them is compared relative to it (Coelli et al., 2005).
Shaban (2008) defines DEA as a non-parametric linear programming technique used to
develop empirical production frontiers and to assess the performance of firms, or
Decision Making Units (DMUs).
Based on those definitions, it is worth noting that DEA uses the mathematical or
deterministic approach particularly the linear programming technique to estimate the
efficient frontier and accordingly the efficient performance of DMUs or firms can be
evaluated.
DEA is widely used to measure the efficiency of economic entities and has been
adopted by many studies. For instances, studies in banking Industry employed this
approach [Fethi et.al (2011), Shin and Kim (2011), Sufian (2010, 2011), Gardener et.al
(2011), Kenjegalieva et al. (2009), Yao et al. (2008), Kao and Liu (2004), Drake and
Hall (2003)].
67
Endogeneity means that there is a significant correlation between the parameter or variable and the
error term.
75
4.3.1. Technical Efficiency (DEA)
As the proposition of Farrell‟s (1957) states that the components of efficiency are
technical efficiency and allocative efficiency. Technical efficiency indicates the
firm‟s/DMU‟s ability to obtain maximal output from a given set of inputs, while
allocative efficiency reflects the firm‟s/DMU‟s ability to use inputs in optimal
proportions, given their respective prices and production technology (Coelli et al, 2005).
In estimating technical efficiency, there has two extended DEA models can be operated.
These are namely the CCR model and BCC model. Brief descriptions about the two
models are presented as follows:
4.3.2. The CCR and BCC Models
CCR Model stands for Charnes, Cooper and Rhodes (1978), the inventors of the model.
The CCR model assumes a constant return to scale (CRS) in the technology of
production and an input orientation. Then, some authors proposed some alternative
assumptions (i.e. Fare, Grosskopf and Logan (1983) and Banker, Charnes and Coopers
(1984) recommended the use of other assumption of the production technology, a
variable return to scale (VRS)).
Before explaining the CCR and BCC model, it is beneficial to discuss the basic concept
of the DEA in general. The DEA efficiency is closely related with the principal concept
of productivity with the notion of ratio of output to input for firm „i‟ as follows (Coelli
et al., 2005):
(4.1)
Where u is a vector of output weights and is the vector of input weights.
Then, the optimal weights are defined by solving the programming problem as follows
(Coelli et al., 2005):
Maximize u, v: (4.2)
Subject to
.
76
In order to maximize the efficiency of firm j, with subject to the restriction that the
efficiencies of all firms are less than or equal to one and all weights are non-negative,
then it requires the optimal weights. The objective function is homogeneous of degree
zero, so that any various weights will produce the same solution, then that condition
need to be treated by a normalization such as with a restriction such as = 1.
By transforming and simplifying that problem, it will produce (Coelli et al, 2005):
Maximize μ, v: (4.3)
Subject to
And for the minimizing form:
Minimize: (4.4)
Subject to
λ
Where: is a scalar representing the input oriented technical efficiency score for i-th
firm, while λ is a vector or constant. Thus, = exhibits the ability of
the firm to reduce inputs in gaining the same output – relatively compared to other firms
in the sample. In the sample, it could be found that some firms will have the efficiency
score of 1.0. This points out that the firms are considered efficient, otherwise,
the higher the score of the firm‟s (the score would be in between 0 and 1 or
), the more efficient the firm. In other words, if , it means that the firm-i is not
efficiently operated, and thus, it consequently requires the reduction of inputs level as
much as to grasp the frontier.
The above formulations are constructed based on the Constant Return to Scale (CRS)
assumption, which is founded by Charnes, Coopers and Rhodes (1978), thus it is well
known as CCR model.
77
For the BCC model, which is proposed by Banker, Charnes and Coopers (1984), it is
recognized as Variable Return to Scale (VRS) assumption. The assumption of VRS can
be applied to above equations by adding a restriction .
This VRS assumption has been applied by many authors in their estimations considering
that the VRS model considers the possibility of a firm to operate not in optimal
condition due to some reasons, i.e. it is caused by some factors: government
regulations, imperfect competition, constraints on finance, etc.
If we are able to find the efficiency scores derived from CRS and VRS assumptions,
then, the „scale efficiency‟ of each firm can be determined by the formula as follows:
(4.5)
Where is scale efficiency, denotes the technical efficiency measure obtained
under the CRS and represents the efficiency score gained by VRS counterpart.
The other alternative outlook of the optimization process is that the output oriented.
This type of the orientation assumes that the output could be increased by utilizing the
same inputs, corresponding with the standard of other firms in the sample. The output
orientation of DEA can be produced by this equation:
Maximize : (4.6)
Subject to
λ
Where denotes the output oriented technical efficiency score for i-th firm.
= exhibit the ability of the firm to maximize outputs obtained by employing the same
inputs, relatively compared to other firms in the sample. The equation was made under
the assumption of CRS.
78
The form of VRS can be applied by adding up the restriction . The measurement
of technical efficiency would be in between 0 and 1 or we can write as follow:
(4.7)
With the calculation, some firms would be identified to be technically efficient, while
the others are not technically efficient in their production.
4.3.3. Nonparametric Productivity Measurement: DEA-based Malmquist
Productivity Index
Through non-parametric DEA method, we could also measure productivity
measurement. Productivity in this case is referring to the concept of the total factor
productivity (TFP) which is accounted for measuring a productivity comprising all
factors of production (multiple inputs) in the process of production to produce multiple
outputs (Coelli et al., 2005).
TFP can be used as the comprehensive measurement of how well the company employ
all factors of production (materials, labour, and capital) to make its outputs (products
and services). TFP can be broken down into technical efficiency change and
technological change. Technical efficiency change entailed to the company related to
some factors such as the improvement of managerial practices, the betterment of
industrial relationships particularly with employees, or the dissemination of the new
technological knowledge across the departments in the company. Whereas,
technological change can be attributed to the introduction of new technology, in which
it guides to a better productivity level (higher outputs made given the same level of
inputs).
As a matter of fact, productivity can also be measured by using partial productivity
indices such as the productivity per labour employed in a factory, the productivity per
land available in farming industry, or the productivity per capital. Since this thesis
79
adopts the concept of transforming multiple inputs into multiple outputs, thus the TFP is
used.
There are three alternatives different indices can be used to evaluate the productivity
changes: the Fisher index, Tornqvist index, and Malmquist index. According to Suzuki
and Sastrosuwito (2011), the Malmquist index has three main advantages relative to the
Fisher and Tornqvist indices. First, it does not require profit maximization, or cost
minimization assumption. Second, it does not require information on the input and
output prices. Third, if researcher has panel data, it allows for the decomposition of
productivity changes into two components (i.e. technical efficiency change or catching
up and technical change or changes in the best practice). Beside its advantages, this
method has also its drawbacks. The major disadvantage is the necessity to compute the
distance functions, but fortunately, the DEA can be used to solve this problem.
4.3.3.1. The Malmquist Index of Productivity
Along with the DEA efficiency estimation, there are some different methods in
calculating productivity index. Those methods are Fisher index, Tornqvist index, and
Malmquist index. In this research, the Malmquist index is employed since the method
has some characteristics as follows (Sufian, 2009): First, it does not require the
assumption of profit maximization or cost minimization. Second, it does not require
information of the input and output prices.
Productivity index is an essential measure to find out the level of productivity and the
change of productivity. In multiple outputs and inputs, Multifactor or Total Factor
Productivity (TFP) is employed to determine the ratio of aggregate outputs produced
relative to aggregate inputs used.
The Malmquist index determines the total factor productivity change (TFP change) by
calculating the ratio of the distance of the observed output and input vectors between
two periods, relative to a reference technology (Coelli et al. 2005).
Färe et al. (1994) specifies the output-oriented Malmquist productivity change index as
expressed as follows:
80
(4.8)
Where:
Mj = Malmquist productivity index
Dj = Distance function
x and y = inputs and outputs, respectively, across time period t to t+1.
That equation exposes the components of the Malmquist productivity index. The ratio
outside the brackets is equal to the technical efficiency change
(EFFCH) between time t and t+1. This ratio represents the change in the relative
distance of the observed production from the maximum potential production. The
component inside the brackets of the Equation is the
geometric mean of the two productivity indices which represents the shift in production
technologies (technical change or TECHCH) between time t and t+1. We can state that
Total Factor Productivity Change (TFP Change) or the Malmquist Productivity change
is the product or the outcome of the two components (EFFCH and TECHCH).
In addition, Technical efficiency change (EFFCH) can be broken down into the two
other components, namely Pure Technical Efficiency Change (PEFFCH) and Scale
Efficiency Change (SECH) as in Fare et al (1994):
(4.9)
(4.10)
81
Where:
The ratio = the Pure Technical Efficiency Change (PEFFCH),
The ratio = Scale Efficiency Change (SECH)
Scale efficiency (SECH) is Technical Efficiency Change (EFFCH) divided by Pure
Technical Efficiency Change (PEFFCH).
Thus, from the above equations, we can derive the conclusion that:
MPI = Malmquist Productivity Index
MPI = EFFCH x TECHCH
MPI = PEFFCH x SECH x TECHCH.
4.4. Parametric Methods: Stochastic Frontier Analysis (SFA)
4.4.1. The Stochastic Production Frontier
Stochastic frontier analysis (SFA) has been employed widely in many industries since
Aigner et al (1977) and Meeusen and Van den Broeck (1977) developed it in 1977.
They originally proposed the stochastic frontier form for the production function model
as follows (in the case of cross sectional data):
(4.11)
Where represents the observed outcome of i-th firm (the observed output),
denotes a K x 1 vector consisting of logarithms of inputs, and is a vector of
unknown parameters. Denotes the deterministic part of the frontier while is a
symmetric random error to account for statistical noise ( is a stochastic part of the
frontier and ~ N [0, and is a non-negative random variables associated with
the technical inefficiency. = |U| and U~N[0, . Inefficiency means the quantity
(percentage) by which the frontier (the optimum) could not be attained.
82
The SFA has been extensively used in a large scope of studies encompassing various
models of goal attainment such as production frontier, cost frontier, revenue frontier,
and profit frontier efficiencies. The analysis of SFA technique usually has two stages:
First; it estimates the model parameter. It is mostly done by maximum likelihood
technique. Second; it estimates inefficiency or efficiency based on the model
construction. The result of efficiency estimation for each unit could be employed to
evaluate performance by comparing the efficiency of one unit to another.
4.4.2. Estimating the parameters
4.4.2.1. Ordinary Least Squares (OLS) Estimation
In estimating the parameters, there are some particular assumptions underpinned the
estimation. Some assumptions are as follows (Coelli et al, 2005): is distributed
independently of each and that those both errors are uncorrelated with the
explanatory variables (in ). The other assumptions are zero means ( ,
homoskedastic ( , uncorrelated ( ),
homoskedastic ( , uncorrelated ( ).
Above assumptions seem to be similar with those of the noise component in the
classical linear regression model, unless for since the inefficiency components have
no negative values ( Under those assumptions, the slope coefficients can be
obtained through ordinary least squares (OLS).
The OLS approach estimates the parameter (β) by minimising the sum of squared
deviation between the their means as reflected in the equation:
(4.12)
Through the first-order derivatives to zero, the solution to the first conditions with
respect to β is the OLS estimator:
(4.13)
83
However the OLS estimator of the intercept coefficient is biased downward, it is
suggested to use the other estimation tool namely corrected ordinary least square
(COLS) estimator (Coelli et al., 2005). In addition, many researchers employ maximum
likelihood (ML) estimators since ML estimators can provide the better solution for the
problems of distributional assumptions of the two error terms and have many desirable
large sample (i.e., asymptotic) properties.
4.4.2.2. Maximum Likelihood (ML) Estimation
The ML estimation concept is underpinned by the idea that a definite sample is more
likely to have been generated from some distributions than from others (Coelli et al.,
2005). The ML estimates an unknown parameter by maximizing the probability of
randomly drawing particular sample observations.
The usage of ML in estimating the parameters usually requires the assumption of
which states that errors are independently and identically distributed
normal random variables with 0 means and . Then, the joint density
function for the vector can be written as follows:
(4.14)
This equation articulates the likelihood of observing the sample observations as a
function of β and . The ML estimator of β can be derived by maximising the
logarithm of the likelihood function:
(4.15)
This is to maximise log-likelihood function with respect to β. In the last part of right-
hand side in the equation denotes the sum of squares function which is similar to the
OLS estimation, meaning that maximizing the log-likelihood with respect to β is
equivalent to minimizing the sum of squares (Coelli et al., 2005).
84
4.4.2.3. Estimating Efficiency
After estimating the model parameters, the other aim of SFA technique is to predict the
efficiency level or the degree of efficiency of each observation. The estimation of this
efficiency level is the main interest of the SFA study. There are many advantages to
have the estimate. Through the estimate, we can rank each individual (firm) or unit
within the groups and we can also identify the outperformed unit(s) compared to the
under-performing units.
The common output oriented measure of technical efficiency is the ratio of observed
output to the corresponding stochastic frontier output. Coelli et al. (2005) show it
through this equation:
(4.16)
Where is technical efficiency for i-th firm. The technical efficiency takes a value of
between 0 and 1. The equation measures the output of i-th firm relative to the output of
fully-efficient producing firm using the same input vector. If a firm obtains the value of
1, meaning that the firm is operated in a fully efficient way. Otherwise, if a firm get the
value of below 1, meaning that the firm operates inefficiently.
For the illustration, in the case of output oriented, the value of x 100% would
be the percentage of maximum output that is produced by the firm-i. Hence, If
x 100% = 87%, it means that the firm is producing only 87% of the
maximum possible (frontier) output. It also means that 100% - 87% = 13% is the
percentage of output which is lost due to technical inefficiency of the firm-i.
4.4.2.4. The Stochastic Cost Frontier
Beside technical efficiency in the production function, we could also estimate cost
efficiency by using a cost frontier efficiency function if price data are available. This
cost efficiency information is very important for the management of a company to take
85
a decision making regarding the possibility of the firm to cut down costs (expenses)
while keeping up the same level of output in the operation. The cost frontier efficiency
is also beneficial for the policy makers, such as government bodies and regulators, to
take any decisions and actions after considering the cost performances of all units or
firms under their supervision.
In the case of evaluating banks performance, the stochastic cost frontier efficiency can
be employed to find out, by comparing banks, which bank is outperformed in terms of
doing the operation with the lower cost and which bank is underperformed. We can also
identify the cost efficiency of each bank and the key drivers of the cost efficiency.
The general form of the cost frontier model (in the case of cross-sectional data) is as
follows (Coelli et al., 2005):
(4.17)
Where = the observed cost of the firm 1; = the N-th input price; = the M-th
output; while c(.) is a cost function. This cost function has the characteristics of linearly
homogeneous, non-decreasing, and concave in prices.
The first step to estimate efficiency following the equation (4.17) is determining the
functional form for the cost function. There are two most popular forms can be chosen,
the Cobb-Douglass form and trans-log form.
The Cost frontier model for Cobb-Douglass form is as follows (Coelli et al, 2005):
(4.18)
Where = a symmetric random variable (approximation errors and other statistical-
noise sources); = a non-negative variable (efficiency). If the are non-negative and
fulfil the constraint , the function would be linearly homogeneous, non-
decreasing, and concave in input.
86
By substituting the constraint into the formula (4.18), it results the homogeneity-
constrained Cobb-Douglass cost-frontier model as follows (Coelli et al., 2005):
(4.19)
The same way can be applied to obtain a trans-log model.
The next step is to find cost efficiency through the formula as follows (Coelli et al.,
2005):
(4.20)
Where = cost efficiency for i-firm. Cost efficiency reflects the ratio of minimum
cost to observed cost. The value of CE would always be in between 0 and 1. The value
closes to 1 means more efficient and the value closes to 0 is less-efficient.
Actually, there is also another model besides the technical efficiency (production
frontier) model and cost efficiency (cost frontier) models, and is namely the profit
efficiency (profit frontier) model. However, since this thesis only applies technical
efficiency for DEA approach and cost efficiency model for the SFA approach, thus the
profit efficiency model is not discussed.
4.4.2.5. Stochastic Frontier Models for Panel Data.
The stochastic frontier models can also be applied for panel data sets. Panel data sets
have usually more observations than cross-sectional data sets since they contain various
firms across several years.
Coelli et al. (2005: 275) remarks that panel data frontier models allow us to: relax some
of the strong distributional assumptions that were necessary to disentangle the separate
effects of inefficiency and noise; obtain consistent predictions of technical efficiencies;
87
and investigate changes in technical efficiencies (as well as the underlying production
technology) over time.
In addition, Kumbhakar and Lovell (2000) point out that there is a fundamental problem
with cross-section data compared to panel data in the econometric analysis. The
problem is that through the cross section data, we could only observe each firm once, so
that this could severely limit the confidence either for technical or cost efficiency
estimates.
The general form of the panel data model for stochastic frontier production function is
as follows (Aigner, et al, 1977):
(4.21)
The equation is almost the same with the equation (4.11) excepting that we append “t”
after i to denote time. is the observed outcome (goal attainment) of the i-th firm at
time-t; is the deterministic part of the frontier; is the stochastic part and v ~
N[0, ]; while is inefficiency of the i-th firm at time-t. For a stochastic cost frontier
model, would be changed by .
In the estimation purposes, we should determine which structure would be chosen either
time-invariant or time-varying inefficiency models to account for the assumption of
independent distribution of
In the assumptions of time invariant cost-efficiency in the single equation case, the cost
efficiency model for the panel data would be (Kumbhakar and Lovell, 2000):
(4.22)
Where denotes the expenditure incurred by producer (firm)-i at time-t, is the
output of the i-th firm at time-t; is a vector of input prices
faced by firm-i at time-t; is random statistical noise; is time invariant cost-
88
inefficiency. In addition, =1 is applied to ensure the homogeneity of the cost
frontier in input prices.
If we have a panel data set employed, Kumbhakar and Lovell (2000:170) emphasize
that the longer the panel, the less tenable is the assumption that cost efficiency is time
invariant. Thus, all the estimation procedure (Least Square Dummy Variables,
Generalised Least Square, Maximum Likelihood Estimator) can be employed and
modified to accommodate time-varying cost efficiency as well as they were in the
estimation of technical efficiency.
4.4.2.5.1. Time-invariant Inefficiency Models
Time-invariant inefficiency model simply bases the assumption of (Coelli et al, 2005):
(4.23)
Where is treated as either a fixed parameter (in the fixed effects model) or a random
variable (in the random effects model). A standard regression framework can be applied
to estimate the fixed effects model using dummy variables, whilst least-squares or
maximum likelihood techniques can be employed to estimate the random effects model.
Regarding the assumptions of , Pitt and Lee (1981) assumed a half normal
distribution, while Battese and Coelli (1988) regarded the more general truncated
distribution (Coelli et al, 2005).
Pitt and Lee (1981)68
employ the maximum likelihood estimation (MLE) technique to
estimate random effects model over the panel data with time invariant technical
efficiency. The random effects for their panel data are specified as follows:
(4.24)
68
Pitt and Lee (1981) employ panel data on fifty Indonesian weaving firms over the years 1972, 1973 and
1975 in the estimation of a stochastic frontier Cobb-Douglas production function.
89
Where S = +1 for a technical (production) efficiency model, and -1 for a cost efficiency
model. Pit and Lee (1981) assume that the component of inefficiency is to be invariant
with the normal-half normal model:
(4.25)
Battese and Coelli (1988)69
suggested the more general truncated normal distribution for
inefficiency term to be applied on panel data. With this specification, the general
production function for the panel data is similar with the formula (4.26) but the
assumption of inefficiency has become:
(4.26)
This normal-truncated normal model relaxes an implicit restriction in the normal-half
normal model that the mean of the underlying inefficiency variable is zero. Thus, by this
model, it allows , the mean of to be non-zero. The production function of this panel
data model can also be estimated using the MLE technique.
4.4.2.5.2. Time-varying Inefficiency Models
Since time invariant inefficiency models are considered rather restrictive, there could be
the other models that are more relaxed in order managers to be able to take lessons from
experience and the changes of the technical efficiencies over time (Coelli et al, 2005).
The time varying inefficiency models take the basic form:
(4.27)
Where f(.) is a function that decides the time-varying inefficiency. There are two
alternative models (Coelli et al, 2005) f(t) as follows:
69
Battese and Coelli (1988) examine the prediction of the firm-level technical efficiencies with a
Generalised frontier production function over the panel data of the New South Wales and Victorian dairy
industries in Australia over the three financial years, 1978-79, 79-80, and 80-81.
90
Kumbhakar (1990): (4.28)
Battese and Coelli (1992) (4.29)
The difference between the Kumbakar (1990) model and Battese and Coelli (1992)
model are as follows (Coelli et al, 2005:278):
Kumbhakar (1990) Battese and Coelli (1992)
The function lies in the unit interval and can be
non-increasing, non-decreasing, concave or convex
depending on the signs and magnitude of α and β.
The function involves only one unknown parameter
and partly as consequence, is less flexible.
It has the properties f(t) ≥0 and f(T)=1 and is either
non-increasing or non-decreasing depending on the
sign . It is also convex for all values of .
In addition, both models can be estimated through fixed effects as well as random
effects framework using maximum likelihood technique under the assumption that
has a truncated normal distribution (Coelli et al., 2005):
(4.30)
4.4.2.5.3. Environmental effects model
In the estimation of efficiency, we can put the only inputs in the calculation process and
it then obtains the results as outputs. However, we can also take into account some
environmental variables considering that those variables could also influence the
outputs. Environmental variables are exogenous variables that could be divided into
stochastic variables and non-stochastic variables.
Coelli et al. (2005) define that non-stochastic variables are observable variables that
exist when the decisions on the key production process are made (e.g. degree of
government regulation, type of firm ownership, age of the labour force, etc.). On the
other hand, the unforeseen stochastic variables are defined as sources of production risk,
such as weather, pest infestations, etc.
91
In regard to account for the non-stochastic environmental variables, there have been
different opinions whether to incorporate them directly into the equation of the
production function or exclude them in the function and put those variables in the
second stage analysis.
Coelli et al. (2005) explain that the easiest way to incorporate non-stochastic
environmental variables in the model of efficiency is directly integrating those variables
into the non-stochastic part of the production frontier. For illustration, the form of a
model would be (i.e. in the case of cross sectional data):
(4.31)
Where denotes a vector of environmental variables, while γ is a vector of unknown
parameters. Then we could estimate the technical efficiency of i-th firm by the equation
(Coelli et al., 2005):
(4.32)
Where is a function of and all together. Hence, the outcome (the firm technical
efficiency) would vary in accordance with the common inputs and the environmental
variables.
Some authors allow the inclusion of those environmental variables directly affect the
stochastic component of production frontier. Kumbhakar et al. (1991) and Battese and
Coelli (1993, 1995) are those who supported the model either for cross sectional model
and panel data model, respectively. Through this concept, Kumbhakar et al. (1991)
assume the equations as follows:
(4.33) and
γ, (4.34)
Through these equations, it can be seen that the inefficiency effects have distributions
that vary together with .
92
Conversely, a number of authors put environmental variables as the predictors of
technical efficiency using the two-stage approach. For instance, Pitt and Lee (1981)
accounts for those variables in the second stage by examining the relationship between
environmental variables and predicted technical efficiencies. Through this two-stage
approach, they first estimate a frontier model without inserting environmental variables.
Then for the second stage they include the regression of the predicted efficiencies on the
environmental variables.
4.4.2.5.4. Battese and Coelli 1995 Model
There are three popular Battese and Coelli (1988, 1992, and 1995) models in estimating
the SFA efficiency (BC88, BC92, and BC95). The model of BC92 has been briefly
explored in the section of time varying inefficiency model (section 4.4.2.5.2), while
BC88 model has also been presented in the section time-invariant inefficiency models
(section of 4.4.2.5.1).
Battese and Coelli (1995) expand the time varying efficiency model by generalizing in
the panel data models in the BC95 model. Unlike BC92 model, the model accounts for
the observable environmental variables directly influence the stochastic component in
the production function, thus this model employ the only one-stage simultaneously
analysis.
In this model, the non-negative technical inefficiency effects are assumed to be a
function of firm-specific variables and time and to be independently distributed as
truncations of normal distributions. The distribution assumptions of the inefficiency are
also assumed with constant variance, but with means which are a linear function of
observable variables (Battese and Coelli, 1995).
Thus, the technical inefficiency effects in the model of this stochastic frontier are
specified in the equation as follows (Battese and Coelli, 1995):
(4.35)
93
Where: represents the observable environmental variables that influence on the
inefficiency . , defined by the truncation of the normal
distribution with zero mean and variance .
Chapter six of this thesis employs SFA method by BC92 model thus the model uses
two-stage analysis comprising efficiency estimation and examining the determinants of
the efficiency.
4.5. Multiple Regression Estimation
Corresponding to some of the research questions set out in this thesis which the two of
them are to find out the determinants of efficiency and the determinants of small
business finance, this thesis uses multiple regressions. A multiple regression model is a
regression model with more than one explanatory variable (Gujarati and Porter, 2010).
The use of this kind of regression is important due to the fact that very few economic
phenomena can be explained by only a single explanatory variable, thus we need some
other explanatory variables included in the model. There are various types of multiple
regressions in practice70
and the type of regression used must be adapted to the
characteristics of the data and the regression techniques71
chosen should also be based
on the prevailing econometric theory.
The section of this thesis explains the basic concept of the multiple regression model as
well as some statistical aspects of the regressions and their applications in the panel data
set in accordance with the research purposes in this thesis.
4.5.1. Multiple Regressions for Panel Data
Panel data set comprises the observations of the same units over a number of periods. In
other words, these data covers the combination between cross-sectional and time-series
aspects of the unit observations.
70
There are multiple regressions for continuous data and regressions for the the limited dependent
variable, such as LOGIT, PROBIT, and TOBIT models, etc. 71
Regression techniques: Ordinary Least Squares, Generalized Least Squares, with fixed-effects/random-
effects, Two-Stage Least Squares, Three-Stage Least Squares, etc.
94
Panel data are commonly the collection of the micro-economic level data consisting of
some individuals, households, or firms in one country but it has become increasingly
familiar to pool individual time series of a number of countries or industries and analyse
them simultaneously (Verbeek, 2012:372).
The advantages of panel data over the cross sectional data or time-series are as follows:
It allows the economists to specify and estimate more complicated and more
realistic models than a single cross section and a single time series would do
(Verbeek, 2012).
It allows the researcher great flexibility in modelling differences in behaviour
across individuals. It provides such a rich environment for the development of
estimation techniques and theoretical results (Greene, 2008).
It gives a richer source of variation which allows for more efficient estimation of
the parameters. It is more informative that one can get more reliable estimates
and test more sophisticated behavioural models with less restrictive assumptions
(Baltagi, 2002).
The standard linear model for the panel data can be written as follows (Verbeek, 2012;
Koop, 2008):
(4.36)
Where denotes the observation of the dependent variable for firm-i at time-t; is
the intercept; is a K-dimensional vector of explanatory variables; is the error
term; this error term varies over individuals and time, and captures all unobservable
factors that affect .
To estimate the parameter model using ordinary least square, there have some basic
assumptions that should be fulfilled to achieve efficiency, consistency and unbiased
model. Those assumptions are almost the same with the cross section or time series
data. If it is assumed that complies with the classical assumption, then all results we
obtain under the assumptions will be valid and OLS would be BLUE (best linear
unbiased estimator). But, if we have a case that satisfies the classical assumptions
95
except that heteroskedasticity exists, then the Generalised Least Square methods are
applicable (Koop, 2008).
In addition, the problem exists when some basic classical assumptions could not be
fulfilled. Verbeek (2012) gives some illustrations i.e. If and ,
the OLS estimator is consistent for and β under weak regularity conditions. Due to
the fact that we observe the same units repeatedly, it seems unrealistic to assume that
error terms from different period are uncorrelated (Verbeek, 2012). Concerning the
problem, it is worthwhile to account for the models that are commonly used in panel
data analysis, namely REM (random effects model) and FEM (fixed effects model).
4.5.2. Random Effect Model
In estimating the panel data set through multiple regression techniques, we should be
aware to determine the treatment for the possibilities of the individual effect in the
model. Individual effect means that each individual has a different effect. Theoretically,
there are two main individual effects models in the panel data analysis: the fixed effect
model and the random effect model (Koop, 2008).
In panel data model it is assumed that:
(4.37)
Where it is assumed that is time invariant and homoskedastic across individuals;
is also homoscedastic and not correlated over time. The model and assumptions belong
to the formula (4.36 and 4.37) represent the Random Effects Model.
The Random Effect Model (REM), also known as a (one-way) error component model,
is a regression model assuming that all factors affecting the dependent variable, but are
still not included in the model as explanatory variables, can be properly summarized by
a random error term (Verbeek, 2012).
Koop (2008) puts the light on that REM takes the assumption that the individual effect
is a random variable, so that it does not use dummy variables.
96
REM can be written as follows:
(4.38)
Where it is assumed that, ; ; which means
that error term comprises two components: an individual specific component which
does not vary over time, and a remainder component is assumed to be uncorrelated over
time (Verbeek, 2012).
We could observe in the model that in the REM, it only has an intercept and k
explanatory variables and it does not require N-dummy explanatory variables like the
fixed-effects model (Koop, 2008).
Regarding the estimation technique in the REM, Koop (2008) asserts that OLS
(ordinary least square) can be employed if the model fulfils the classical assumption.
However, if the classical assumptions are violated, then OLS is no longer BLUE. As the
consequence, we could use GLS (generalised least square) estimation technique or OLS
technique with panel-corrected standard errors.
4.5.3. Fixed Effect Model
Fixed Effect Model (FEM) is a model in the panel data analysis which takes into
account the existing of each individual effect of the observations in the model.
The individual effects exist when we assume that each individual can have different
intercept in the model. As it can be seen in this equation (Koop, 2008):
(4.39)
It is almost similar with the common pooled model. The different lies in , which
varies across individuals. Each intercept for individuals, denotes as a fixed
(individual) effect. As the consequence, each individual can have a different regression
line.
97
The fixed effect model employs variables of dummy to account for the individual effect
in the model (Koop, 2008). A dummy variable has a value of either 0 or 1.
Stock and Watson (2012: 396) defines the fixed effect regression as a method for
controlling for omitted variables across entities (firms, states, etc.) but do not change
over time. The fixed effect regression model has different intercepts as many as the total
number of firms (entities).
The fixed effects regression model is as follows (Stock and Watson, 2012):
(4.40)
Where is the value of the first independent variable for the firm-i at time-t (1,
2,…,k = a number of independent variables); = ,…, are firm (entity) specific
intercepts.
Equally, in terms of a common intercept, the fixed effects regression model can also be
presented as follows (Stock and Watson, 2012):
(4.41)
Where =1 if i=2 and =0 otherwise, and so forth.
A problem of the FEM specification is that it will sometimes head to a regression model
with a large numbers of explanatory variables. Hence, it could lead another problem of
uneasy to compute due to the fact that numerous coefficients are being estimated (Koop,
2008).
In conjunction with the application of those REM and FEM in the panel data regression,
the Hausman-test could be employed to choose between them. The idea behind the
importance of the Hausman-test is as follows (Koop, 2008): If (the individual effect
is uncorrelated with any of explanatory variables) is true, then both REM and FEM
estimators are consistent and provide approximately similar result. But, if is false,
98
then REM will be inconsistent, whilst FE will be consistent, and those results could be
quite different.
4.5.4. TOBIT Regression
In some cases, we can encounter the model of regression in which the dependent
variable is not continuous but it is quite restrictive in various ways. One of the cases we
may face is for the case of efficiency estimate. Since the efficiency scores derived from
DEA or SFA techniques are in between 0 and 1, it is obvious that the scores are of the
censored dependent variables.
In this section of the thesis, it discusses the fittest model for the censored dependent
variable, especially for the efficiency estimate score which is named the TOBIT
regression.
TOBIT regression model is one of the models that take into account the limited
dependent variables. There are three alternative models considering the limited
dependent variable, they are namely: LOGIT72
, PROBIT73
, and TOBIT74
models.
However, TOBIT model is appropriate to examine the regression on the efficiency
because efficiency scores have properties that correspond to the TOBIT model which
are censored at zero with positive probability but is roughly continuously distributed
over strictly positive values (Wooldridge, 2013)
Koop (2008) specifies that the TOBIT model is a regression model that has the
dependent variable which is censored at zero. The regression equation for the TOBIT
model can be written as follows (Verbeek, 2012; Koop, 2008):
(4.42)
72
LOGIT model is a model for binary responses where the response probability is the LOGIT function
evaluated at a linear function of the explanatory variables (Wooldridge, 2013: 846). 73
PROBIT model is a model for binary responses where the response probability is the standard normal
cumulative distribution function (cdf) evaluated at a linear function of the explanatory variables
(Wooldridge, 2013:849). 74
TOBIT model is a model for a dependent variable that takes on the value zero with positive probability
but is roughly continuously distributed over strictly positive values (Wooldridge, 2013: 853).
99
With the assumption: if and if . is
unobserved or latent variable and is observed dependent variable for observation
unit-i at time-t.
In the TOBIT regression model, we also make some traditional assumptions regarding
parameters and error term. Those assumptions are (Verbeek, 2012): and are i.i.d.
normally distributed, independent of , with the means of zero and variances
of respectively.
Regarding the estimator, Koop (2008) stipulates that OLS method is not appropriate for
the TOBIT regression since the OLS estimator is biased for the censored observations.
The more points we have in the censored observations, the worse the bias will be.
Hence, the parameters of the TOBIT regression model could be properly estimated by
the estimator taking into account the censored nature of the dependent variable. The
estimator is the Maximum likelihood method.
The likelihood function for TOBIT regression model can be written as follows
(Verbeek, 2012):
(4.43)
Where and is given by:
If ,
If (4.44)
For the case of examining the determinants of efficiency, the TOBIT regression is the
fittest one since the assumption used if and if is
appropriate and reflect the circumstances of the efficiency values.
100
4.5.5. Two-Stage Least Squares (TSLS) Regression
Two-stage least squares (TSLS) is applied in the regression estimation when there is a
problem of endogeneity. Endogeneity means that there is a correlation between the
right-hand side regressors and the error terms (disturbances). Endogeneity occurs due to
some possible reasons i.e. measurement error, the omission of relevant variables,
sample selectivity, or other reasons (Baltagi, 2005).
Baltagi (2005) and Koop (2008) point out that Endogeneity causes inconsistency of the
usual OLS estimates or the OLS would be biased, and then for curing the problem it
requires instrumental variable (IV) methods i.e. TSLS to obtain consistent parameter
estimates. Koop explains instrumental variable (or instrument) is a random variable that
has a correlation with the explanatory variable but is not correlated with the regression
error.
Gujarati (2006) explains that the TSLS method involves two successive applications of
OLS in the process. The principal idea underpinning the TSLS is to substitute the
explanatory variable which has a correlation with the error term of the equation by a
variable that is not so correlated. Such a variable is named a proxy or instrumental
variable. Through this process, the parameter estimators resulted from the TSLS are
consistent estimators.
In practice, to create instrumental variables that they are then employed to replace
endogenous variables, it firstly requires running a regression on the reduced form of the
right-side endogenous variables. Then, or fitted values form those reduced
regressions can be used as instrumental variables (Studenmund, 2011).
4.6. Data and Variables
This thesis uses data of all commercial banks in Indonesia, which is provided at the
Central Bank of Indonesia (BI) and data provided by PT. Ekofin Konsulindo with the
criteria as follows :
101
The sample is all commercial banks, which consist of conventional banks and
sharia banks in Indonesia.
Only banks which have complete data (about the bank's financial statements
over the period 2002-2010) to be included in the observation data, except for
sharia banks. As the number of sharia banks is relatively very small compared to
conventional banks, then the data of Islamic banks established in the period
2002-2012 are also included.
The data are presented in annual basis for the period of 2002-2010.
After the data are collected and sorted out, there have been 116 banks fullfiling the
criteria comprising 109 conventional banks and 7 islamic banks. All these 109
conventional banks consist of 4 State-owned banks (SOB), 55 Private banks (PB), 26
provincial/Local-government owned banks (LGOB), 15 joint-venture banks (JVB) and 9
foreign-owned banks (FB). The total of islamic banks is 7 banks and are all privately
owned banks.
4.6.1. The Variables Specification and Definition
In estimating the technical efficiency of a bank through the DEA method, there has two
prominent approaches can be used to determine the inputs and outputs. They are namely
the production approach and the intermediation approach. The production approach
treats banking industry similar to any kinds of industries that produce goods and
services. Therefore it observes material, capital and human resources as production
factors to provide outputs. The inputs for the production approach are usually general
and administrative expenses, labor cost, price of capital, etc and the output are deposits,
loans and others.
On the other hand, the intermediation approach considers bank as financial intermediary
between depositors and lenders, so that deposits are placed as input beside other
resources used and loans and any other income are put as the outputs.
While in the SFA estimation, the objective function is cost efficiency (CE) which stems
from the Total Costs of banks (operating costs and non operating costs). Total operating
102
costs consist of interest expenses and non-interest expenses (for conventional banks) or
profit-sharing expenses and non-profit sharing expenses (for islamic banks). Non-
operating costs comprise all expenses except for the operating costs (e.g. labour
expenses, rent expenses, depreciation expenses, etc.).
4.6.1.1. The Inputs and Outputs Specification
For the first empirical chapter where the DEA method is used as the main estimation
technique, this thesis employs 3 models in estimating bank technical efficiency with the
specifications are as follows:
Model 1, using total finance, securities and investments, other income as outputs and
general and administrative expenses, fixed assets, total deposits as inputs.
Model 2, using small business finance, other finance, securities and investments, other
income as outputs and general and administrative expenses, fixed assets, and total
deposits as inputs.
Model 3, using other finance, securities and investments, other income as outputs and
general and administrative expenses, fixed assets, and total deposits as inputs.
Model 1 assumes that all commercial banks ignore the compositions of their finance,
whether it derives from small business or other sectors of finance, so that the ultimate
goals of the banks are to service all customers without viewing customers‟ background.
Model 2 decomposes the finance into small business finance and other finance in order
to recognize of the loan potfolios. This model takes into account small business finance
as one of the important goals in providing services to customers. Contrasted with the
preceding models, the last model (model 3) does not include small business finance in
the components of outputs so that it only reflects other finance and securities and
investments in the outputs. For all models, securities and investments have also become
the output, since those activities are also important in gaining profit.
In addition, for the second empirical chapter (SFA estimate), it employs a couple of
outputs (total finance, securities and investments for the Model 1) and (small business
finance, other finance, securities and investments for the Model 2). Alongside the
103
outputs, it is also employed 3 prices of inputs (price of funds, price of labour and price
of capital)
4.6.1.2. Inputs and Outputs Definition
The definitions of the components of outputs and inputs in the models used are as
follows:
Output components (for DEA and SFA):
Total loans (total finance): Total loans (finance) on all sectors of each commercial bank
at the end of year position.
Securities and investments: Total investment in securities and all placements of each
commercial bank at the end of year position.
Small business loans (small business finance): Total small business loans (finance) of
each commercial bank at the end of year position.
Other loans (other finance): Total other loans (finance) outside small business finance
of each commercial bank at the end of year position.
Other income: Other income generated by each commercial bank at the end of year
position.
Cost component (for SFA):
Total costs (total expenses) : operating costs and non-operating costs.
Input components (for DEA):
General and administrative expenses: Total general and administrative expenses
occurred and experienced by each of commercial bank for the certain period i.e. over
one year (in the income statement).
Fixed assets: Total net fixed assets that each commercial bank has as of the particular
position i.e. at the end of year (in the balance sheet).
Total deposits: total deposits in the particular position (i.e. at the end of year) that each
commercial bank obtains from depositors (in the balance sheet).
Price of Inputs components (for SFA):
Price of funds (pf): total deposits expenses divided by total deposits.
104
Price of labour (pl): total labour expenses divided by total assets.
Price of capital (pc): total cost for the banks equity divided by bank‟s equity.
4.6.2. Environmental and Bank specific variables
There are several variables included in the analysis comprising of some bank specific
variables and some environmental variables which reflect the macroeconomic condition.
4.6.2.1. Bank specific variables
LNTA: Natural logarithm of total assets (LNTA) as a proxy for bank size.
ROA: Return on assets (ROA) is assigned to represent bank profit.
CAR: Capital adequacy ratio is assigned to embody bank capital. Capital adequacy
ratio (CAR) is measured by calculating bank capital divided by risky-assets.
LDR: Loans to deposits ratio (LDR) is allocated to demonstrate bank liquidity. Bank
liquidity means the ability of a bank to cover its short-term liabilities.
NPL : Non-performing loans (NPL) is assigned to represent bank risk.
4.6.2.2. Environmental variables
INFL: Annual inflation rate. This measures the overall percentage increase in the
consumer price index for all goods and services n Indonesia
GDPGR: Annual real GDP (gross domestic product) growth rate. The growth of
Indonesia‟s total goods and services adjusted for inflation.
UNEMP: Annual unemployment rate in Indonesia.
MKTINDEX: Indonesian Composite Index in Indonesia Stock Exchange.
USDRATE: United States Dollar (USD) exchange-rate against Indonesian Rupiahs
(IDR)
4.7. Software:
DEAP 2.1 and LIMDEP 10 software were employed to generate efficiency scores for
non-parametric method (DEA). The parametric efficiency estimation was also achieved
by LIMDEP 10 and FRONTIER4.1. The descriptive statistics as well as some other
graphs were obtained and displayed through STATA 12 and Microsoft Excel 2007.
105
5. Chapter Five: Small Business Finance and Indonesian Commercial Banks’
Technical Efficiency: DEA Approach
5.1. Introduction
This Chapter aims to assess the performance of commercial banks in Indonesia using
the popular linear programming non-parametric technique Data Envelopment Analysis
(DEA) to estimate banks‟ technical efficiency and total factor productivity index. The
analysis is extended by examining the determinants of Indonesian banks‟ efficiencies
during the period 2002-2010. The determinants observed in this study are the specific
factors of Indonesian commercial banks and macroeconomic indicators during the
period.
The analysis covers 116 commercial banks in Indonesia composed of 109 conventional
and 7 Islamic banks. Those conventional banks comprise 4 state-owned banks (SOB),
55 private banks (PB), 26 provincial/Local-government owned banks (LGOB), 15 joint-
venture banks (JVB) and 9 foreign-owned banks (FB), whilst the 7 Islamic banks are all
also of private banks (PB).
The intermediation approach is adopted in estimating the DEA technical efficiency. The
study uses input-oriented approach since that the factors of production are more
controllable for managers in a highly regulated industry. Three models are estimated
using different classifications in the outputs components: model 1 uses total finance
(TF), securities and investment (SI), and other income (OI); model 2 uses small
business finance (SBF), other finance (OF), securities and investment (SI), and other
income (OI); model 3 only uses other finance (OF), securities and investment (SI), and
other income (OI); The inputs employed in the study are total deposits (TD), fixed
assets (FA) and general and administrative expenses (GA).
The organisation of this chapter are as follows: The first section presents the descriptive
statistics of the data sample with their analysis to portray the characteristics of the
Indonesian banking industry as well as the macroeconomic indicators over the study
period. The second section shows the non-parametric efficiency and productivity of the
commercial banks along with the comparison of their group of banks‟ performance
106
during the period. The third section discusses the determinants of commercial banks‟
efficiency over the period. Finally, the last-section gives the conclusion.
5.2. Descriptive Statistics
This section discusses the descriptive statistics of the data covering two purposes:
estimating efficiency and predicting the determinants of efficiency. The data required for
obtaining the efficiency estimate are: the inputs and outputs components (employed in
estimating DEA efficiency). While the data needed for revealing the determinants of the
efficiency are: bank specific variables and macroeconomic variables. The characteristics
of the data are explored in this section.
5.2.1. Data of Inputs and Outputs
Table 5.1 shows the data of inputs and outputs employed in estimating the technical
efficiency of Indonesian commercial banks over the period 2002-2010. The study
estimates technical efficiency by adopting the intermediation approach, hence including
total deposits as one of the inputs.
Table 5.1. Inputs and Outputs
IDR Millions
Total
Finance
Small
Buss.
Finance
Other
Finance
Securities
&
Investments
Other
Income
General
& Admin
Fixed
Assets
Total
Deposits
Mean 8,170,709 1,005,995 7,164,714 5,498,728 256,670 463,405 225,000 12,294,227
Std Dev 22,376,636 4,376,527 19,388,371 16,975,907 761,355 1,260,603 656,037 35,659,342
Maximum 246,968,128 75,374,673 207,555,181 137,260,529 8,529,607 15,645,936 5,290,384 332,727,856
Minimum 1,229 0.001 53 5,981 36 2,060 112 156
Sources: Data observed
Table 5.1 summarises key discriptive statistics for the inputs and outputs used in the
analysis. There is disparity of banking operation in Indonesia from the small-scale
banks and the large-banks. The average finance of each commercial bank for a year was
around IDR 8.2 trillion where IDR 1.0 trillion was allocated to finance the small
business sector in Indonesia during the period of 2002-2010.
107
One important information that could be revealed from the data is that there are still
some banks that had zero75 outstanding balance in small business finance over the period
of 2002-2010. In other words, when the Central Bank of Indonesia (BI) relaxed the
regulation of minimum small business finance (SBF) portion at the particular
percentage and adjusted the SBF proportion to each bank‟ condition and willingness76
,
some commercial banks tended to disregard the finance for small businesses. The third
one, the average of SBF was still far below the average of securities and investments
(IDR 1.0 trillion vs. IDR 5.5 trillion). This implies that out of the deposits collected
from customers, commercial banks have more preferred to invest in securities and
investments rather than in SBF.
The high standard deviation for general and administrative expenses (GA) reflects the
high disparity in operations across the banks. The average total deposits collected from
depositors were about IDR 12.3 trillion, but the total finance was only about IDR 8.2
trillion. It suggests that the financial intermediary function was not optimally run. This
low portion of total finance over the deposits indicates that Indonesian banks‟ managers
applied a very prudent lending policy over the period.
Table 5.2. The Proportion of Small Business Finance
Type 2002 2003 2004 2005 2006 2007 2008 2009 2010 Avg.
FB 0.00% 0.00% 0.00% 0.00% 0.00% 0.20% 0.27% 0.05% 0.28% 0.09%
JVB 3.01% 3.46% 1.79% 1.62% 1.11% 1.39% 1.23% 1.04% 1.56% 1.80%
LGOB 43.73% 39.94% 34.39% 31.29% 23.19% 19.60% 17.84% 16.93% 16.85% 27.09%
PB 19.64% 18.66% 18.23% 16.67% 15.81% 14.42% 12.47% 13.04% 11.16% 15.57%
SHARIA 19.21% 16.79% 13.76% 11.84% 30.80% 32.78% 27.02% 30.41% 27.58% 23.35%
SOB 29.00% 26.78% 25.85% 24.43% 25.47% 25.76% 24.62% 24.96% 26.47% 25.93%
Avg. 19.10% 17.61% 15.67% 14.31% 16.06% 15.69% 13.91% 14.41% 13.98% 15.64%
FB is foreign-owned banks; JVB is joint venture banks; LGOB is local government-owned banks; PB is private banks; SHARIA is
Islamic banks; SOB is state-owned banks.
75
For the small business finance item, some banks in the sample do not provide finances to small
businesses a fixed value of 0.001 is added to the variable all over the sample to be able to estimate the
models. This is a common procedure in the literature in order to avoid the loss of data. 76
The regulation about the minimum SBF requirements for Indonesian commercial banks (The Package
of January 1990) was revoked by BI Regulation Number 3/2/PBI/2001 about Small Business Finance on
January 4, 2001. The regulation states that the provision of small business finance is readjusted based on
the ability of each bank. In other words, commercial banks are not compelled to provide the particular
percentage of their finance portfolios devoted to small businesses. See chapter 2 in details.
108
Table 5.2 demonstrates some particular information regarding small business finance of
Indonesian banks during 2002-2010. The average proportion of SBF out of the total
finance of all Indonesian commercial banks was about 15.64 percent over the period. It
indicates that on average, the proportion of SBF was still below 22.5 percent during the
period. It means that the change of bank regulation (by the GOI) to relax the minimum
threshold of small business finance proportion (adjusting to each bank condition)77
has
influenced the decision of Indonesian banks to finance small businesses over the period.
It also remarks that the focus of the Indonesian bank financing during the period was
still on the other finance, which consisted of corporate finance and consumer finance. It
is unfortunate that, data about the portfolios for both corporate finance and consumer
finance could not be found in banks‟ financial statements.
The lower portion of SBF created by Indonesian commercial banks during the period
signifies the impression that on average, management of the commercial banks in
Indonesia are still reluctant to finance the small business sector. This could strengthen
the argument that they mostly perceive small business sector as high-risk investment
because of fundamental weaknesses adhering to small business sector, particularly in
terms of lack of collateral and information asymmetry.
The local government owned banks (LGOB), Islamic banks (SHARIA) and state-owned
banks (SOB) could still provide more than 22.5 per cent of their lending portfolio to
small businesses given the fact that the regulatory of minimum SBF threshold was
scrapped in 2001. The LGOB on average are the highest provider of SBF as proportion
of their loan portfolio. They are primarily located in the province with their branches
spread in every district (regency). This gave them the advantage to exploit their location
efficiently in providing financing for the small business sector. However, in terms of
lending propensity, the LGOB‟s portfolio financing for small businesses decreased over
the period.
In addition, the SOB are recorded as the second highest group of banks that provided
financing small businesses for over 22.5 per cent during the period. One of the main
reasons for such a record is that the high performance and reputation of PT. Bank
77
BI Regulation Number 3/2/PBI/2001 about Small Business Finance on January 4, 2001.
109
Rakyat Indonesia (BRI). BRI has been recognized as one of the largest and most
successful microfinance institutions in the world and recognized more resilient across
the crises [Patten, et al., (2001), Seibel et al. (2010)]. BRI has more than four thousands
branches of which around 324 large branches with 3,694 small branches that provide
microbanking system covering the Indonesian whole country (Paten et al., 2001). It is
also worth noting that SOB was the only one group of banks which could constantly
maintain the proportion of SBF of over 22.5 per cent in every year.
Foreign-owned banks (FB) and joint-venture banks (JVB), each had a very low average
SBF portfolio during the period 2002-2010, 0.09 per cent and 1.80 per cent, respectively.
It seems that they disregarded SBF in their lending portfolio and concentrated their
financing mostly on corporate and consumer sectors over the period.
5.2.2. Data of Bank-Specific Variables and Macroeconomic Variables
The data of bank specific factors and macroeconomic indicators are very important to
be explored in regards to identify which variables explained the variation of the
Indonesian banks average efficiency during the period 2002-2010. The data descriptive
of those variables are presented in Table 5.3 as follows:
Table 5.3. Bank-Specific and Macroeconomic Variables
Variable Obs. Mean Std. Dev. Min Max
lnta 1011 14.9074 1.8291 9.7722 19.83
roa 1011 0.0228 0.0640 -0.829 0.32
car 1011 0.2420 0.1875 -0.223 3.494
ldr 1011 0.7698 0.3734 0.02 3.35
npl 1011 0.0224 0.0297 0.00 0.529
infl 1011 8.0658 3.9359 2.80 17.1
gdpgr 1011 5.3920 0.6446 4.50 6.3
unemp 1011 9.1770 1.1954 7.10 11.2
usdrate 1011 9340 499 8576 10398
LNTA is the natural logarithm of total assets; ROA is return on assets; CAR is capital adequacy ratio; NPL is
non-performing loans; INFL is annual inflation rate; GDPGR is Gross Domestic Product at real prices growth
rate; UNEMP is annual unemployment rate; USDRATE is United States Dollars (USD) exchange-rate in
terms of Indonesian Rupiahs (IDR).
The natural logarithm of total assets (LNTA) represents the size of bank. The expected
result in the hypothesis testing is the bigger the size of a bank, the more efficient the
bank will be, and vice versa, the smaller the size of the bank, the less efficient the bank.
110
This hypothesis is related to the structure conduct performance theory which states that
as the larger the size of the company, the more efficient the utilisation of the company‟s
resources, and it then improves the company‟s performance.
ROA (return on assets) is an indicator for bank profitability which is expected that the
more profitable the bank, the more efficient the bank will be. During the study period,
the average ROA of 116 commercial banks in Indonesia was 2.28 percent meaning that
the commercial banks in Indonesia enjoyed the profit during the period 2002-2010 after
suffering from the worst crisis: 1997-1998. The minimum ROA score was attained by
one of the closed banks during the period.
Capital adequacy ratio (CAR) is a proxy for the bank capital and it is expected that the
higher the capital of a bank, the more efficient the bank operated. The average CAR of
Indonesian commercial banks for the period 2002-2010 was 24.20 percent. It is
considerably sound performance since it was higher than the minimum of 8 percent
CAR requirement. The higher CAR also reflects the strong condition of the bank in
facing financial and business risks.
Loans to deposits ratio (LDR) is a representative for bank liquidity. It is also expected to
have a positive sign on bank efficiency. It means that as the higher the LDR, the bigger
the opportunity of a bank to generate profit and attain better efficiency, and vice versa.
The average LDR of 76.98 percent of Indonesian banks during the period 2002-2010
indicates that the function of financial intermediary has increased over the period
although it had not reached the optimal level yet.
Non-performing loans (NPL) as an indicator for bank risk. It is presumed to have a
negative sign on bank efficiency. The higher the NPL of a bank, the higher the equity of
the bank would be reduced to cover the loss experienced by the bank. The NPL of
Indonesian banks during the period 2002-2010 was 2.24 percent. It was extremely good,
since it was still below 5 percent. The more prudent lending policy adopted by banks
seems to have a good impact on the lower non performing loans during the period.
111
Three macroeconomic factors are employed in this study: inflation (INF), Gross
Domestic Product growth (GDPGR) and Unemployment rate (UNEMP), Stock market
index/Indonesia composite index (MKTINDEX), USD exchange rate against
Indonesian Rupiah/IDR (USDRATE).
The study employed some macroeconomic indicators adding to bank specific factors as
explanatory variables in the multiple regression models examined. This aims to
investigate the influence of environmental or external variables on Indonesian bank
efficiency during the period. The macroeconomic factors included in this study are:
inflation (INF), Gross Domestic Product growth (GDPGR) and Unemployment rate
(UNEMP), Stock market index/Indonesia composite index (MKTINDEX), and USD
exchange rate against Indonesian Rupiah/IDR (USDRATE).
Inflation rate (INF) is one of the important macroeconomic variables reflecting price
stability in a country. The rate of inflation is presumed to have a negative association
with bank efficiency. The higher the inflation rate, the lower the efficiency of a bank,
and vice versa. GDP growth (GDPGR) is a measurement of the level of an economic
growth in a country. It is also an indicator of the economic activities in the country.
GDP based economic growth is presumed to have a positive effect on bank efficiency.
The unemployment rate (UNEMP) is a measure of the level of unemployment in a
country. This indicator is also important to be observed to identify the employment
absorption rate of the labour force in the country. The association between UNEMP and
bank efficiency is assumed to have a negative sign.
USD exchange rate against IDR (USDRATE) is the price of one US Dollar currency
against Indonesian Rupiahs (IDR). If IDR weakens, more rupiah amounts should be
prepared to be exchanged with the USD. For instance, if this number increases (USD 1
= IDR 9,000 to USD 1 = IDR 10,000), or IDR weakens, it is expected that bank
efficiency will decrease, and vice versa, if IDR increases, bank efficiency will also
increase.
112
5.3. Efficiency and Productivity of Indonesian Commercial Banks
This section reveals the technical efficiency of Indonesian commercial banks that are
derived by non-parametric technique (the DEA approach) during the period 2002-2010.
This section also explains the productivity level of Indonesian commercial banks as well
as the productivity change. The efficiency and productivity level of the Indonesian
banks reflect the real performance of Indonesian banking Industry during the period.
5.3.1. Technical efficiency of Indonesian commercial banks
5.3.1.1. Overal performance of Indonesian banks
Figure 5.1 ilustrates the trend of technical efficiency of Indonesian banks during the
period 2002-2010. The three lines represent the estimated technical efficiency (TE)
from the three models (Model 1, 2 and 3 as explained in section 5.1.).
Figure 5.1. Average technical efficiency of the Indonesian banks (Model 1,2 and 3)
Source: Data Observed
There is a declining trend in Indonesian bank technical efficiency over the period.
However, due to the different specifications of the three models there is a disparity
between the average efficiency obtained from these models. The average efficiency
2002 2003 2004 2005 2006 2007 2008 2009 2010
TE-Model 1 0.6922 0.6812 0.6962 0.6884 0.6708 0.6639 0.5786 0.5194 0.5330
TE-Model 2 0.8075 0.7964 0.7755 0.7773 0.7275 0.7290 0.6707 0.6392 0.6403
TE-Model 3 0.6563 0.6481 0.6472 0.6575 0.6361 0.6304 0.5496 0.4845 0.4983
-
0.1000
0.2000
0.3000
0.4000
0.5000
0.6000
0.7000
0.8000
0.9000
Ave
rage
Eff
icie
ncy
Sco
re
Average Technical Efficiency (Model 1, 2, 3)
113
obtained from model 2 is the highest amongst the models78
. This may imply the effect
of economies of scope; hence in model 2 the study splits the total finance into two
outputs, small businesses finance and other finance. There could be two explanations of
the different levels of efficiency obtained from the models. One, that the main
contributor to the difference on average efficiency is the effect of economies of scope.
The second is the unobserved technical issues related to the DEA method, that is the
more outputs or inputs might be included in the model, the higher the efficiency scores
generated. Hence, it is not clear which of these effects rather prevail over the other, we
opt to consider the differences as economies of scope effects. Nonetheless, these three
models confirming the declining trends in Indonesian banks efficiency were also
supported by the finding of Suzuki and Sastrosuwito, (2011) and Hadad et.al (2011b).
Thus, the results imply that Indonesian banks need to minimise inputs given the level of
outputs they produce in order to improve their technical efficiency.
The results provide insight on the effect of diversifying loan portfolio on bank
performance. Technical efficiency gained due to diversification may result in using the
banks‟ resources more efficiently in engaging in small businesses finance79
. This
approach might have a positive impact on the banks performance in terms of
improvement in technical efficiency. However, it might also augment the risk associated
with banks‟ operation by allocating high proportion of the loan portfolio to small
businesses due to information asymmetry problems. It is important to further investigate
the effect of loans and risk on banks‟ efficiency using a second stage analysis. The last
section of this chapter provides such an analysis in order to understand the determinants
of banks‟ efficiency.
Financial Crisis and efficiency:
It is evident from Figure 5.1 that the decline in banks technical efficiency became more
steeper post the financial crisis on 2007. Nonetheless, the decline in performance was
not sharp as was the case for other countries, i.e. European and Australian banks
(Alzubaidi and Bougheas, (2012), Forughi and De Zoysa (2012). The smoother decline
78
Average efficiency of those models over the period 2002-2010 are 0.6352 for model 1, 0.7285 for
model 2, and 0.6001 for model 3.
79
Diversification of products could reduce the unit cost, since only the component of variable cost
changes while the total fixed cost does not.
114
in efficiency can be explained as a reflection of the resilience and experience of
Indonesian banks in dealing with the crisis. The Indonesian banks suffered dramatically
during the crisis in 1997 compared to the slight decline in performance post 2007 crisis.
Similar to many emerging markets, Indonesian banks were more capitalised compared
to European banks during the crisis in 2007 (Fethi et al, 2012). The high level of
capitalisation has played an important role in absorbing the effect of the crisis. Post the
crisis of 1997, the BI has implemented various reform programmes that aim to increase
the resilience of Indonesian banks, by increasing capital, improving competition and
improving banks‟ efficiency (see more details in section 2.2.3 and 2.2.4).
Another factor that could have helped banks absorb the shock of the financial crisis in
2007 is the diversification of loan portfolio. One of the main criticisms to the
Indonesian banks post the 1997 crisis was the higher exposure to corporate loans. These
corporate were highly vulnerable to the financial crisis contributed to the default risk
because of the high level of bankruptcy risk associated. Despite the lifting of BI
regulation that imposed at least 20% of banks‟ loan portfolios to be allocated to small
businesses; many banks‟ groups maintained a relatively high proportion of their loan
portfolios to small businesses.
5.3.1.2. Ownership and Banks’ efficiency
It is commonly perceived in both the theoretical and empirical literature that state-
owned banks tend to underperform private and foreign banks in terms of efficiency. The
perception stems from the high agency problem and lack of incentive to improve
performance associated with state-owned enterprise managers. In the case of Indonesian
banks, the results reveal that state-owned banks seem to be the most efficient banks in
the banking sector (Table 5.4). The state-owned banks ranked first followed by foreign
and joint venture banks. Local government banks and private banks ranked as the least
efficient according to the results obtained from all models.
These results provide interesting evidence that despite those foreign banks has almost 0%
of small business loans they still attain a higher level of efficiency compared to local
115
government banks, the highest provider of small business loans. This implies that the
efficiency of foreign banks comes from other sources rather than loan portfolio
diversification. For instance, they tend to have lower investment in branches and
number of employees (i.e. factors of production) compared to SOBs, LGOBs and PBs.
The private banks represent a controversial performance in the Indonesian banking
sector. One would expect higher performance for private banks unlike what the result
implies. Perhaps the technology of FBs and the sheer size of SOBs are factors that
played an important role in clustering the performance of banks as seen in Table 5.4.
The results coincide with the evidence provided by Fethi et al. (2011) that find state-
owned banks are the most efficient banks in Egypt.
The result contradicts the findings of Yildirim (2002) and Chen et al. (2005) for the case
of Turkish and Chinese banks respectively. They found that state-owned banks are less
efficient in Turkey and China respectively.
Another classification of banks in the sample is obtained by clustering the banks into
conventional and Sharia banks (Islamic banks). As seen from the Table 5.4 that Sharia
banks seem to underperform conventional banks according to the results from model 1
and 3. According to the efficiency scores obtained from model 2 by splitting the loans
into two types of outputs namely small business finance and other finance, Sharia
banks tend to outperform conventional banks. This result implies the positive effect of
loan portfolio diversification in the Sharia banks on their technical efficiency. This
deduction can be further verified from the high growth in small business finance
performed by Sharia banks (Shaban et al., 2014).
The ranking of the Indonesian banks efficiency based on the three criteria (ownership,
operation, and FE exposure) for the three consecutive models are provided in the table
5.4 as follows:
116
Table 5.4. Ranking of Indonesian Banks Efficiency (2002-2010)
Ownership: Efficiency
Model 1
Ranking Efficiency
Model 2
Ranking Efficiency
Model 3
Ranking
SOB 0.966 1 0.989 1 0.861 2
FB 0.912 2 0.914 2 0.914 1
JVB 0.773 3 0.785 3 0.770 3
PB 0.574 4 0.678 5 0.535 4
LGOB 0.545 5 0.706 4 0.500 5
Operation: Efficiency
Model 1
Ranking Efficiency
Model 2
Ranking Efficiency
Model 3
Ranking
CONV 0.636 1 0.725 2 0.604 1
SHARIA 0.595 2 0.834 1 0.465 2
FE Operation: Efficiency
Model 1
Ranking Efficiency
Model 2
Ranking Efficiency
Model 3
Ranking
FEB 0.663 1 0.742 1 0.632 1
NFEB 0.534 2 0.678 2 0.482 2
For the ownership, there are state-owned banks (SOB), foreign-owned banks (FB), joint-venture banks (JVB), private
banks (PB), and local government-owned banks (LGOB). For the operation, there are conventional banks (CONV)
and islamic banks (SHARIA). For the foreign exchange (FEB) operation, there are also two groups of banks, namely
foreign exchange banks (FEB) and non foreign exchange banks (NFEB).
Based on ownership criteria, for model 1 and model 2, SOBs have been the most
efficient banks with the average efficiency scores of 0.966 and 0.989. While in the
model 3, SOBs were only posed in the second position below the FBs, with an average
score of 0.861. These results are very sensible considering that in the model 1 and 2, the
calculation of efficiency took into account the total finance (model 1) and its
decomposition (model 2) in the output components. This is important since SOBs were
undergoing various financing schemes for both the corporate level as well as for small
and medium businesses, so that their exposures on various financing schemes could
enhance bank efficiency. Whereas when the model only considered other finance (OF)
and abandoned SBF in the output components, the most efficient one was foreign banks
(FBs). This result is also sensible because FB did not provide adequate exposure for
SBF and they seemed to concentrate on other finance in their lending exposures. Those
results have confirmed the results of previous studies (Suzuki and Sastrosuwito, 2011;
Haddad, 2011a) particularly about the most efficiency‟s bank group in Indonesia.
In addition, both in model 1 and 2, FBs have constantly occupied the second position
except for model 3 where they successfully occupied the top position. JVBs were also
117
consistent in the third place in all models provided. The least efficient banks in terms of
ownership during the period were private banks (PB) and local government-owned
banks (LGOB) which had an average efficiency of below 0.750. Private Banks were still
better than the LGOBs since they ranked the fourth in the two models.
In terms of their operations, conventional banks (CONV) were more efficient in two
models, while Islamic banks (SHARIA) could outperform in model 2. For the Sharia
banks, we could find that when they considered small business finance into their
accounts in the estimation of efficiency, they obtained higher efficiency scores and
outperformed the conventional banks. In model 2, Sharia banks reached 0.834 on their
average efficiency but in model 3, when small business finance was excluded in the
estimation, their average efficiency dropped to 0.465. It is quite interesting to explore
why Sharia banks had a better efficiency score when they paid more attention to
providing finance for small businesses in Indonesia during the period 2002-2010 (in
model 2). This following figure seems to answer the question:
Figure 5.2. Numbers of sharia banks and Efficiencies Model 2
Source: Data Observed
Figure 5.2 demonstrates that the number of sharia banks in Indonesia has increased
within the period 2002-2010 and that the increasing number of banks tends to have an
association with the attainment of their efficiency score. Their efficiency scores were
below 0.800 for the period 2002-2005 when the number of sharia banks was only two
banks. Since the number of banks increased in 2005, average efficiency has also
increased and they tend to get their best performance within 2006-2008. The
phenomenon is quite interesting before the decline in the year 2009.
0.632
0.801
0
1
2
3
4
5
6
7
8
0.000
0.200
0.400
0.600
0.800
1.000
1.200
2002 2003 2004 2005 2006 2007 2008 2009 2010
No. of SHARIA Banks
CONV
SHARIA
118
Another figure (figure 5.3) could also explain the phenomenon of raising the
performance of sharia banks when they involved more in small business finance.
Figure 5.3. SBF percentage and Efficiencies Model 2
Source: Data Observed
The figure above displays the average percentage of small business finance within their
portfolios for sharia and conventional banks. Small business finance percentage
(SBFPEC) for conventional banks tend to decrease from 21.82 percent in 2002 to 10.86
percent in 2010 with the average SBFPEC at 15.52 percent, while SBFPEC for sharia
banks tend to increase even though a decline occurred in 2003-2005. The average
SBFPEC for Sharia banks in 2002 was 19.21 percent and it has considerably increased
in 2006 and since then the percentage has always been above 25 percent, with the
average of 23.35 percent within the period. That figure also shows that when Sharia
banks added the proportion of small business finance in 2006, efficiency scores also
increased until the drop occurred in 2009 and 2010. Based on the figure, we could
suppose that the efficiency score of sharia banks tends to increase when they increase
the proportion of their small business finance. Another finding from the figure is that the
decrease in small business finance proportion of the conventional banks has also caused
a decline in their efficiency within the period.
On the last criteria, in terms of foreign exchange services provided, foreign exchange
banks (FEB) have consistently gained higher efficiency scores than those of non foreign
exchange banks (NFEB). We can conclude that banks which provided foreign exchange
transactions in their operation could get benefit from the transactions, not only to
0.00%
5.00%
10.00%
15.00%
20.00%
25.00%
30.00%
35.00%
0
0.2
0.4
0.6
0.8
1
1.2
2002 2003 2004 2005 2006 2007 2008 2009 2010
CONV_SBFPEC SHARIA_ SBFPEC CONV_EFF SHARIA_EFF
119
increase their profit but also to enhance their efficiency.
The graphs of the evolution of the Indonesian banks‟ technical efficiency among the
different group of ownership during 2002-2010 can be observed in figure 5.4, 5.5., and
5.6. (See the Appendices 3, 4, and 5).
5.3.1.2. Size and Banks’ efficiency
In terms of the size of bank, the average technical efficiency levels of the Indonesian
banks are presented in Figure 5.4, 5.5, and 5.6 for the three consecutive models. This
thesis classifies the banks into four bank peer groups based on asset size. Bank peer
group 1 are those which each has total assets of less than IDR 10 trillion. Bank peer
group 2 consists of banks which have more than IDR 10 trillion until IDR 50 trillion.
Bank peer group 3 are those with each has the asset size of greater than IDR 50 trillion
and less than and equal to IDR 100 trillion. The last group (Bank peer group 4) are those
banks which each has more than IDR 100 trillion in asset size.
Table 5.5. Distribution of Technical Efficiency By Bank Peer Group
Peer Group 1 2 3 4 TOTAL
Average Technical Efficiency of each group (MODEL 1) 0.5654 0.7842 0.8891 0.9835 0.6352
Average Technical Efficiency of each group (MODEL 2) 0.6808 0.8249 0.9142 0.9883 0.7285
Average Technical Efficiency of each group (MODEL 3) 0.5277 0.7563 0.8498 0.9614 0.6001
Number of banks in Peer group (as of 2010) 70 33 6 7 116
Peer Group 1 (Total Assets <= IDR 10 Trillion);
Peer Group 2 (Total Assets > IDR 10 Trillion and <= IDR 50Trillion);
Peer Group 3 (Total Assets > IDR 50 Trillion and <= IDR 100 Trillion); Peer Group 4 (Total Assets > IDR 100 Trillion);
120
Table 5.5 reveals the results that the largest banks (peer group 4) are the most technical
efficient banks in Indonesia over the period 2002-2010 under the model 1, 2 and 3 with
an average technical efficiency of 0.9835, 0.9883, and 0.9614, respectively. These
results suggest the positive impact of the economies of scale on bank efficiency given
the fact that the largest banks tend to be more efficient than other peer groups. These
results are supported in practice by the very large number of customers served by the
large banks operated in Indonesia i.e. PT. Bank Mandiri, PT. Bank BCA, and PT Bank
BNI with sophisticated bank technology including cash machines and internet banking
facilities. These findings coincide with the studies of Shaban (2008), Demir et al (2005)
and Isik and Hassan (2003b) in the case of Egypt and Turkey. They found that large
banks had a better performance compared to the other size of banks.
The second highest efficient banks are peer group 3 and the least efficient banks are the
smallest banks (peer group 1). The smallest banks experienced the least efficient
operated in Indonesia during the period 2002-2010. These results are also sensible given
the fact that the smallest banks are majority of the private banks (PBs) which find some
difficulty in obtaining customers‟ confidence post-the crisis 1997-98. As known, the
crisis of 1997-98 testified the considerable funds‟ movement from PBs to state-owned
banks.
Figure 5.7. Average Technical Efficiency By Bank Peer Group (Model 1)
Source: Data Observed
Figure 5.7 exhibits the evolution of their technical efficiency levels attained by four
different groups for model 1. Almost all bank peer groups except peer group 4
experienced the decline pattern of the average technical efficiency over the period 2002
0.0000
0.2000
0.4000
0.6000
0.8000
1.0000
1.2000
2002 2003 2004 2005 2006 2007 2008 2009 2010
Ave
rage
Eff
icie
ncy
Sco
re
Average Technical Efficiency Model 1 by Bank Peer Group
Peer Group 1 (Total Assets <=IDR 10 Trillion)
Pee Group 2 (Total Assets > IDR10 Trillion and <= IDR 50Trillion)
Peer Group 3 (Total Assets > IDR50 Trillion and <= IDR 100Trillion)
Peer Group 4 (Total Assets > IDR100 Trillion)
121
to 2010. The largest banks (bank peer group 4) emerge to be the most stable banks
attaining a technical efficiency level of around 1.000. This constant performance has
delivered the banks to become the most technical efficient banks in Indonesia during the
period. It is also worthwhile to state that the crisis of 2007-08 had a lower impact on the
largest banks (bank peer group 4) than those of their peers of group 1, 2 and 3. This is
also evidence that the largest banks‟ group in Indonesia was more resilient to the current
world economic crisis. Almost similar findings obtained by the other models can be
found in the figure 5.8 and 5.9 (See Appendix 6 and 7).
5.3.2. The Productivity of Indonesian commercial banks
The productivity of Indonesian commercial banks during the period 2002-2010 can be
observed through Table 5.6 as follows:
Table 5.6. Malmquist Index Summary over the period 2002-2010
Year EFF.CH TECH.CH PE.CH SE.CH TFP.CH
Model 1 0.954 1.043 0.963 0.990 0.995
Model 2 0.964 1.026 0.968 0.996 0.989
Model 3 0.951 1.043 0.961 0.990 0.991
EFF.CH is the efficiency change; TECH.CH is the technological change; PE.CH is pure efficiency
change; SE.CH is scale efficiency change; TPF.CH is total factor productivity change.
In general, Indonesian commercial banks underwent the decreasing productivity growth
(productivity regress) over the period 2002-2010. All models witness the productivity
regress over the period. Model 1, 2, 3 observe the regress of 0.05%, 1.1%, and 0.9% on
average annually. This productivity regress was caused by the non-existence of
efficiency change although there had a progress in the technology change. The
technology change experienced a progress of 4.3%, 2.6%, and 4.3% yearly on average
for the three consecutive models respectively over the study period. The positive
technological change implies an improvement in technology adopted by banks over the
period 2002-2010. These results provide strong evidence of the success of Indonesian
banks in terms of facilitating the newest technology including internet banking and text
message banking in their services.
122
While the evidence that there was no efficiency change in Indonesian banks during the
observed period, has supported the findings of the downward trend of technical
efficiency experienced by Indonesian banks over the period. This led Indonesian banks
to attain the productivity regress over the period. Hence, the recommendation for
Indonesian banks to improve their technical efficiency by minimising the inputs given
the level of outputs they produce is also very relevant to enhance their productivity level
(see section 5.3.1.1).
Table 5.7. presents in details on annual productivity progress/regress over the period for
the model 1.
Table 5.7. Malmquist Index Summary of Annual Means (Model 1)
Year EFF.CH TECH.CH PE.CH SE.CH TFP.CH
2002/2003 0.959 0.969 0.989 0.970 0.930
2003/2004 1.055 0.949 1.030 1.025 1.002
2004/2005 1.021 0.988 0.992 1.030 1.009
2005/2006 0.948 1.083 0.961 0.987 1.027
2006/2007 0.846 1.207 0.989 0.855 1.021
2007/2008 0.846 1.160 0.832 1.017 0.981
2008/2009 0.916 1.066 0.903 1.014 0.976
2009/2010 1.065 0.953 1.027 1.037 1.015
Mean 0.954 1.043 0.963 0.990 0.995
EFF.CH is the efficiency change; TECH.CH is the technological change; PE.CH is pure efficiency
change; SE.CH is scale efficiency change; TPF.CH is total factor productivity change.
In model 1, during the covered period, on average the productivity growth decreased by
0.5 % annually. We can clearly identify that productivity growth has declined for a
couple of years: the year 2002/2003, 2007/2008 and 2008/2009. The decline for the
period of 2002/2003 was due to the fact that commercial banks had been still in a
recovery period after the big crisis of 1997-98, whilst the consecutive decline in the
period of 2007/2008 and 2008/2009 seem to be affected by the global financial crisis in
2007/2008. This decline of productivity growth of Indonesian banks that occurred
during the crises of 2007-2008 is similar with the findings of the previous study (Suzuki
and Sastrosuwito, 2011). In addition, the productivity growth of Indonesian banks
during the period 2002-2010 was basically supported by the technological change that
grew 4.3% annually and there was no efficiency change within the period.
123
Meanwhile, the annual productivity progress/regress over the same period for model 2
and model 3 can be seen at the Table 5.8and 5.9 (at the Appendix 8 and 9).
5.4. Determinants of Efficiency
This section reveals the determinants of technical efficiency of Indonesian banks during
the period 2002-2010. There are three models examined. The dependent variable for
Model 1 (TE-1) is the technical efficiency obtained by the DEA approach considering
total finance, and securities and investment as outputs. The dependent variable for
Model 2 (TE-2) is the technical efficiency obtained by the DEA approach considering
small business finance, other finance, securities and investment as outputs. The last one,
the dependent variable for Model 3 (TE-3), is the technical efficiency obtained by the
DEA approach which takes into account only other finance, and securities and
investment as outputs. In addition, all models use the same components of inputs,
namely general and administrative expenses, fixed assets, total deposits as inputs. The
explanatory variables included in the regression models are some specific factors of
Indonesian banks as well as the country macroeconomic Indicators.
The TOBIT-regression is employed to estimate the determinants of Indonesian bank
efficiency since the efficiency scores are constrained between zero and one. A lot of
previous studies use the same kind of regression in revealing the determinants of
efficiency [(Chang et.al (1998); Nikiel and Opiela (2002); Casu and Molyneux (2003);
Havrylchyk (2005); Grigorian and Manole (2006); Ariff and Can (2008); Sufian (2010);
Gardener et.al (2011)].
The standard TOBIT-regression model is defined as follows:
Where is a latent variable, is technical efficiency score obtained from DEA
approach. is the set of parameters to be estimated. is the error term.
The equation of the regression is specified as follows:
124
= b0 + b1 + b2 + b3 + b4 + b5 + b6 +
b7 + b8 + b9 + b10 .
Where:
The technical efficiency of i-th bank in period t obtained by the DEA
approach.
Natural logarithm of total assets (LNTA) as a proxy for bank size.
Return on assets (ROA) is assigned to represent bank profit.
Capital adequacy ratio has been assigned to embody bank capital.
Capital adequacy ratio (CAR) is calculated by the formula of bank
capital divided by risk-weighted assets.
Bank liquidity means the ability of a bank to cover its short-term
liabilities. Loans to deposits ratio (LDR) is supposed to demonstrate
bank liquidity.
Non-performing loans (NPL) is assigned to represent the bank risk.
Annual inflation rate. This measure reflects the overall percentage
increase in the consumer price index for all goods and services in
Indonesia.
Annual real GDP (gross domestic product) growth rate. The growth of
Indonesia‟s total goods and services produced in Indonesia in annual
term.
Annual unemployment rate in Indonesia.
Indonesian Stock Composite Index in the Indonesia Stock Exchange.
US Dollar exchange rate, in terms of IDR (Indonesian Rupiah).
Error term
All variables employed in the model are to reveal the determinants of efficiency in the
Indonesian banking industry. The total banks included in this study are 116 banks which
covers around 96 per cent of the population, within the 9-year study period from 2002-
2012, resulting in 1011 total observations.
The correlation matrix amongst independent variables is presented as follows:
125
Table 5.10. Correlation Matrix of Independent Variables-1 lnta roa car ldr npl infl gdpgr unemp mktindex
lnta 1
roa 0.0742 1
car -0.2228 0.0426 1
ldr -0.0116 0.0331 0.1126 1
npl -0.1030 -0.4565 0.0877 0.0831 1
infl -0.0523 0.0095 -0.0525 -0.0061 0.0346 1
gdpgr 0.1477 -0.048 0.0275 0.1198 -0.1518 0.3182 1
unemp -0.1646 0.0756 -0.0521 -0.1345 0.0784 0.4866 -0.1168 1
mktindex 0.2200 -0.069 0.0615 0.1461 -0.168 -0.3461 0.5662 -0.6426 1
usdrate 0.1147 -0.027 0.0051 0.1062 -0.0829 0.1306 -0.1054 -0.2222 0.1913
LNTA is the natural logarithm of total assets; ROA is return on assets; CAR is capital adequacy ratio; NPL is non-performing loans;
INFL is annual inflation rate; GDPGR is Gross Domestic Product at real prices growth rate; UNEMP is annual unemployment rate;
MKTINDEX is the market index, a composite index in Indonesia stock exchange; USDRATE is United States Dollars (USD)
exchange-rate in terms of Indonesian Rupiahs (IDR).
Since the variable MKTINDEX has a high correlation either with GDPGR (0.5662) and
UNEMP (-0.6426), it is decided to drop the MKTINDEX in the model. The significant
test for the correlations can be observed on appendix 26. Thus, the correlation matrix
between the variables incorporated in the model is presented on the following table:
Table 5.11. Correlation Matrix of Independent variables-2
lnta roa car ldr npl infl gdpgr unemp usdrate
lnta 1.0000
roa 0.0361 1.0000
car -0.2060 -0.1484 1.0000
ldr -0.0115 0.0504 0.1104 1.0000
npl -0.1031 -0.3794 0.3160 0.0831 1.0000
infl -0.0524 0.0156 -0.0481 -0.0061 0.0346 1.0000
gdpgr 0.1478 -0.0175 0.0388 0.1198 -0.1518 0.3182 1.0000
unemp -0.1646 0.0354 -0.0686 -0.1345 0.0784 0.4866 -0.1168 1.0000
usdrate 0.1147 -0.0008 -0.0029 0.1062 -0.0829 0.1306 -0.1054 -0.2222 1.0000
LNTA is the natural logarithm of total assets; ROA is return on assets; CAR is capital adequacy ratio; NPL is non-performing loans;
INFL is annual inflation rate; GDPGR is Gross Domestic Product at real prices growth rate; UNEMP is annual unemployment rate;
USDRATE is United States Dollars (USD) exchange-rate in terms of Indonesian Rupiahs (IDR).
Table 5.11 shows that there is no significant association between each independent
variable (all the correlation coefficients are below 0.5). Therefore, we conclude that
there is no multicollinearity in the model. The correlation coefficient between each
variable is very important to be observed since a strong association between each of the
explanatory variables would demolish the model and it is considered not a BLUE (best,
linear, unbiased estimator) model.
126
The determinants of Indonesian banks‟ technical efficiency over the period 2002 to
2010 are presented in table 5.10 as follows:
Tabel 5.12. Determinants of Indonesian banks’ technical efficiency (2002-2010)
VARIABLES TE-1 TE-2 TE-3
lnta 0.0953*** 0.0766*** 0.0965***
(0.00453) (0.00500) (0.00482)
roa 0.114 0.206 0.0906
(0.130) (0.140) (0.140)
car 0.492*** 0.334*** 0.582***
(0.0497) (0.0531) (0.0535)
ldr 0.0988*** 0.131*** 0.0731***
(0.0219) (0.0248) (0.0233)
npl 0.703** 0.226 0.801**
(0.326) (0.351) (0.349)
infl 0.00779*** 0.0114*** 0.00791***
(0.00262) (0.00288) (0.00281)
gdpgr -0.0801*** -0.112*** -0.0749***
(0.0143) (0.0158) (0.0153)
unemp 0.0545*** 0.0367*** 0.0539***
(0.00829) (0.00908) (0.00889)
usdrate -0.000118*** -0.000127*** -0.000113***
(1.71e-05) (1.87e-05) (1.83e-05)
Constant -0.00188 0.798*** -0.129
(0.233) (0.256) (0.250)
No. of Observations 1,011 1,011 1,011
Pseudo R-square 0.6223 0.3802 0.5235
Standard errors in parentheses
*** p<0.01, ** p<0.05, *p<0.1
The dependent variable for Model 1 (TE-1) is the DEA technical efficiency scores obtained considering total finance, securities and
investment as outputs. The dependent variable for Model 2 (TE-2) is the DEA technical efficiency scores obtained considering
small business finance, other finance, securities and investment as outputs. The dependent variable for Model 3 (TE-3) is the DEA
technical efficiency scores obtained with takes into account only other finance, and securities and investment as outputs. All models
use the same components of inputs, namely general and administrative expenses, fixed assets, total deposits.
The explanatory variables are: LNTA is the natural logarithm of total assets; ROA is return on assets; CAR is capital adequacy
ratio; NPL is non-performing loans; INFL is annual inflation rate; GDPGR is Gross Domestic Product at real prices growth rate;
UNEMP is annual unemployment rate; USDRATE is United States Dollars (USD) exchange-rate in terms of Indonesian Rupiahs
(IDR).
Based on table 5.10, it is observed that the regression on model 1 has been the fittest
model reflected by the Pseudo_R-square of 0.6223. This means the 62.23% variation of
the efficiency scores can be explained by the explanatory variables in the model.
Meanwhile, the lowest R-square of about 38.02 percent is presented by the regression
on model 2.
127
LNTA as the proxy for bank size has been a significant factor for Indonesian banks‟
efficiency with a positive sign. LNTA is significant for all models with a confidence
level of 99 percent. These results suggest that the bigger the size of a bank, the bank
operates more efficiently. These findings also confirmed the Structure Conduct
Performance hypothesis that if the size of a company increases, the company will
operate more efficient (due to the benefit of economies of scale), and it would finally
enhance the company‟s profit. In terms of the case of Indonesian banks, these results
also witnessed the right policy adopted by the Indonesian banks‟ authority (BI) during
the period by merging some commercial banks. The purposes of the mergers
programme during the period were to have bigger, strong and sturdy banks and become
more resilient to crises that may occur. In addition, these findings are consistent with
other results [see e.g. Andries (2011), Vu and Turnel, (2011), Yildirim and Philippatos
(2007)].
CAR has also affected Indonesian bank efficiency with the positive direction. CAR is
significant across the three models with the confidence level of 99 percent. These
findings suggest that it is improved for banks to add their capital considering that as the
capital adequacy increases, the efficiency of bank also increases. The CAR of
Indonesian commercial banks over the study period was 24.20 per cent. Their high level
of capital adequacy reflects high compliance of Indonesian banks to the regulation
regarding the anticipation to encounter all bank risks including credit risk and solvency
risk. This findings coincide with other literatures [Yildirim and Philippatos (2007),
Andries (2011); Thangavelu and Findlay, C. (2013)].
LDR has also a significant impact on the Indonesian bank efficiency with the positive
association. This LDR is also significant at a confidence level of 99 percent. This result
suggests that as lending or finance exposure increases, the efficiency of bank also
increases. In other words, the additional proportion of credit or finance over deposit can
enhance bank efficiency. The average LDR across all Indonesian commercial banks
during the period was 76.98 per cent. This means that there are more chances for
Indonesian commercial banks to enhance their lending exposure (since the average LDR
was less than 80 per cent) to generate more profit and obtain a higher efficiency level.
Obviously, the management of banks should also practise more prudential principle in
implementing the policy of expanding their lending portfolios.
128
The NPL variable has no conclusive results for the three models. NPL affects banks‟
efficiency in model-1 and model-3 with the confidence level of 95 percent, but it failed
to significantly affect bank efficiency in model-2. The insignificant effects of the NPL
on banks‟ efficiency is similar with the findings of previous literatures [Andries (2011),
Thangavelu and Findlay (2013)].
Surprisingly, the ROA variable as the proxy for bank profitability has no significant
effect on Indonesian bank efficiency in the three models although the relation between
the variable with efficiency is positive. It means that the more profitable the bank
operates, the more efficient the bank will be. However, the positive relationship is not
significant. These results coincide with the findings of Andries (2011).
All macroeconomic variables, namely INFL, GDPGR, UNEMP and USDRATE have
also significantly affected Indonesian banking efficiency during the period 2002-2010.
All macroeconomic indicators affected Indonesian bank efficiency with an alpha of one
percent.
In general, we can conclude that from all the models examined, the variable of LNTA,
CAR, and LDR are the bank-specific variables that consistently affected Indonesian
bank efficiency over the period 2002-2010. These results could also be used as an
important recommendation for all management of the banks that the size of bank,
capital adequacy, and bank liquidity are the three important keys for enhancing
efficiency. These findings are also important for policy makers, particularly for the
Indonesian commercial bank authority to create the regulations which are empirically
beneficial for commercial banks.
5.5. Conclusions
The results of the study show that the average technical efficiency of commercial banks
in Indonesia has tended to decline over the period 2002-2010. State-owned banks were
the most efficient banks in Indonesia when considering small business finance (SBF) as
an output in the calculation of efficiency, but when SBF was excluded, Foreign-owned
banks were the most efficient ones. Sharia banks outperformed conventional banks in
129
one model when the total finance was splitted to small business finance and other
finance in the output components.
The productivity growth of Indonesian banks has decreased during the study period. No
change in the average banks‟ efficiency levels, has led to their decreased productivity
growth attained, although there was an improvement in technological change. These
findings suggest the importance of Indonesian banks to minimise their inputs given the
outputs to enhance their efficiency level as well as their productivity.
Across all models, it is verified that LNTA, CAR and LDR are bank specific variables
that consistently affect Indonesian banks‟ technical efficiency. Meanwhile all
macroeconomic variables in this study [INFL (inflation rate), GDPGR (Gross Domestic
Product Growth), UNEMP (unemployment rate) and USD RATE (USD rate)] have also
significantly affected Indonesian banks‟ technical efficiency during the period 2002-
2010.
The implications of these results are considered as suggestions for the policy makers in
Indonesia as follows: First, it is not a guarantee that privatisation is beneficial. The case
of Indonesian banks has indicated that state-owned banks (SOBs) are much more
efficient than their private counterparts. Second, small business finance (SBF) is also an
important factor in enhancing bank efficiency. Third, bank size, bank capital, and bank
liquidity as well as bank lending/finance portfolio proportions are confirmed to be of the
banks‟ specific variables that significantly affected the Indonesian banks‟ efficiency. All
these findings suggest that the merger policy, the provision to maintain the capital
adequacy, and the provision to manage adequate liquidity as well as lending portfolio
have become very crucial aspects in managing commercial banks. Fourth, although
there was a productivity regress of Indonesian banks due to the non-existence of their
efficiency change, the technological change implemented in Indonesian commercial
banks has increased the productivity level of Indonesian commercial banks during the
study period. This suggest that extensive new technologies (including the
implementation of cash machines, phone-banking, internet banking, etc) have
beneficially improved the level of productivity of Indonesian commercial banks during
the period.
130
6. Chapter Six: Small Business Finance and Indonesian Commercial Banks’ Cost
Efficiency: SFA Approach
6.1. Introduction
This chapter examines the performance of Indonesian commercial banks using the most
popular technique in the parametric method, namely the stochastic frontier analysis
(SFA) to estimate banks‟ cost efficiency over the period 2002-2010. This chapter
focuses only on the cost efficiency estimates given the fact the cost efficiency model is
more widely used than the SFA method in the banking industry. This chapter also
provides an extensive analysis by examining the determinants of Indonesian bank cost
efficiency over the study period observing bank-specific factors and macroeconomic
indicators as explanatory variables.
The study includes 111 commercial banks in Indonesia consisting of 109 conventional
banks and 2 Islamic banks during the study period. In terms of their ownership, these
banks are composed of 4 State-owned banks (SOB), 55 private banks (PB), 26
provincial/Local-government owned banks (LGOB), 15 joint-venture banks (JVB), 9
foreign-owned banks (FB). All banks are conventional banks except the two banks
which are classified as Islamic banks (SHARIA). These Islamic banks are also private
banks (PB).
In estimating cost efficiency, this study uses a translog function for the total cost as an
objective of the function along with some outputs and prices of inputs variables. The
study presents three different models classified by different components in outputs.
model 1 uses total finance (TF), securities and investment (SI), and other income (OI);
model 2 use small business finance (SBF), other finance (OF), securities and
investment (SI), and other income (OI); model 3 uses other finance (OF), securities and
investment (SI), and other income (OI). The input prices employed in this study are
price of fund (PF), price of labour (PL), and price of capital (PC). All variables are
presented in a logarithm form.
The organisation of this chapter is as follows: The first section demonstrates the
descriptive statistics of the data sample completed with a brief analysis of each variable.
131
The second section presents the parametric SFA cost efficiency estimates of Indonesian
banks as well as the comparison of their groups of banks‟ performance over the study
period. The third section discusses the determinants of Indonesian banks‟ cost efficiency
over the period. Finally, the last section provides the conclusion.
6.2. Descriptive Statistics
This section exhibits the descriptive statistics of the data sample employed in this thesis
to estimate Indonesian banks’ SFA cost efficiency as well as to predict the determinants
of bank cost efficiency. To estimate banks‟ cost efficiency we exploit the data of total
costs, outputs and the prices of inputs, while in examining the determinants of banks‟
cost efficiency, we use bank specific variables and macroeconomic indicators as
explanatory variables.
6.2.1. Data for Estimating Cost Efficiency
Table 6.1 exposes the data of outputs employed in estimating the cost efficiency of
Indonesian commercial banks over the period 2002-2010.
Table 6.1. Output Components
IDR Millions
Total Finance
Small Buss.
Finance
Other
Finance
Securities &
Investments Other Income
Mean 8,239,970 1,005,694 7,234,276 5,554,995 259,357
Std. Dev 22,500,000 4,401,679 19,500,000 17,100,000 765,516
Minimum 1,229 0.001 53 5,981 36
Maximum 247,000,000 75,400,000 208,000,000 137,000,000 8,529,607
Source: Data processed
The above table summarises the key descriptive statistics for the outputs used in the
analysis. There is a disparity of banking operation in Indonesia ranging from the small-
scale banks to the large-banks. The average annual finance of each commercial bank
during the period 2002-2010 was around IDR 8.2 trillion where IDR 1.0 trillion of the
number was allocated to financing the small business sector in Indonesia. In percentage
term, the proportion of small business finance was below 22.5 percent.
132
As explained in chapter five, a number of banks had zero80
outstanding balance in their
small business finance (SBF) over the study period. This circumstance occurred over
time during the period 2002-2010 after the Central Bank of Indonesia (BI) did not
impose all commercial banks to maintain the minimum threshold of SBF proportion at
the particular percentage rather than readjusting the proportion to each bank‟s condition
and willingness in 200181
.
Meanwhile, the characteristics of total cost and the price of inputs are presented at the
table as follows:
Table 6.2. Total Cost and Price of Inputs
IDR Millions
Total Costs
(IDR Million)
Price of
funds
Price of
Labour
Price of
Capital
Mean 1,417,452.00 0.0672 0.0208 3.2566
Std Dev 3,682,207.00 0.0300 0.0174 4.5489
Minimum 1,438.00 0.0070 0.0020 0.0752
Maximum 35,000,000.00 0.2950 0.3360 29.8478
Source: Data processed
Table 6.2 summarizes the data of total costs and prices of inputs. Total costs cover all
expenses incurred by all commercial banks i.e. operating expenses and non-operating
expenses. Operating expenses cover total interest expenses for conventional banks or
operating expenses from profit sharing for Islamic banks, provision for losses on
earnings assets, provision for losses on commitments and contingencies, general &
administrative expenses, salaries and employee benefits, losses from decline in fair
value of trading account securities, losses on foreign exchange, etc. The average annual
total costs for a bank over the study period was IDR 1.42 trillion with a minimum of
IDR 1.44 billion and a maximum of IDR 35 trillion. The disparity of the total costs of
Indonesian banks represents the difference of the scale of the operations over large
banks and small banks in Indonesia.
80
For the small business finance items, some banks in the sample do not provide finance to small
businesses, and then a fixed value of 0.001 is added to the variable all over the sample to be able to
estimate the models. This is a common procedure in the literature in order to avoid loss of data.
81
The regulation about the minimum SBF requirements for Indonesian commercial banks (The Package of
January 1990) was revoked by BI Regulation Number 3/2/PBI/2001 about Small Business Finance on
January 4, 2001. See chapter 2 for details.
133
Price of funds (PF) is calculated by total interest expenses divided by total deposits. It
can be observed that the average price of funds is around 6.72% out of total deposits
collected. Price of labour (PL) is calculated by personnel expenses divided by total
assets and price of capital (PC) is calculated by non-interest expenses divided by fixed
assets. On the average, price of labour of Indonesian banks was around 2.08% of the
total assets. For calculating the price of capital, the components of non-interest expenses
are calculated by total expenses minus interest expenses and personnel expenses. The
maximum data for PF, PL and PC are derived from the data of one commercial bank
which was closed down at the end of the study period.
In terms of banks‟ ownership and operations, the total cost of Indonesian commercial
banks over the study period is presented in the following table:
Table 6.3. Total Costs to Total Assets (2002-2010)
Types 2002 2003 2004 2005 2006 2007 2008 2009 2010 Avg.
Ownership:
FB 0.0993 0.0651 0.0553 0.0683 0.0847 0.0804 0.0761 0.0614 0.1868 0.0864
JVB 0.1020 0.0780 0.0678 0.0571 0.0747 0.0646 0.0699 0.0647 0.1086 0.0764
LGOB 0.1138 0.1080 0.0919 0.0863 0.0865 0.0884 0.0948 0.0948 0.1033 0.0964
PB 0.1361 0.1168 0.0927 0.1014 0.1167 0.0960 0.1252 0.1059 0.1179 0.1121
SOB 0.1301 0.1144 0.0879 0.0922 0.1029 0.0851 0.0835 0.0846 0.0799 0.0956
Operational
System:
CONV 0.1163 0.0964 0.0791 0.0811 0.0931 0.0829 0.0899 0.0823 0.1193 0.0934
SHARIA 0.0867 0.0947 0.0866 0.0977 0.1102 0.1016 0.1004 0.0998 0.0832 0.0957
State-owned banks (SOB), foreign-owned banks (FB), joint-venture banks (JVB), private banks (PB), local government-
owned banks (LGOB), conventional banks (CONV) and islamic banks (SHARIA).
Table 6.3 illustrates the profile of expenditures experienced by commercial banks in
Indonesia during the period 2002-2010. Based on types of ownership, joint venture
banks (JVB) and foreign banks (FB) were the first and second lowest expenditure banks‟
group in Indonesia in terms of average total costs to total assets (TCTA). They attained
an average TCTA of 7.64% and 8.64% respectively. These results support the belief that
foreign banks or joint venture banks operate more efficiently than their local peers
(domestic banks). These facts are very sensible considering that they employ fewer
employees for similar roles compared to domestic banks. In addition, they could also
minimize the expenses of the usage of fixed assets by employing less branches and
134
offices in supporting their operations. All these conditions reduced the total cost of their
operations. Meanwhile, private banks (PB) witnessed the least efficient among all
groups in terms of TCTA with a proportion of total expenditures over total assets
reaching more than 10 percent. These results are also sensible given the fact that they
spend more resources to win consumer confidence compared to other groups. Since the
crisis of 1997-98, PBs obtained less trustworthiness from customers compared to their
peers. Since then they have applied a number of strategies to attract customers i.e.
giving more gifts to depositors and employing more sales force to promote their deposit
and lending products.
In terms of their operational system, conventional banks (CONV) consumed lower costs
on average than their peers from Islamic banks (SHARIA). They experienced 9.34%
and 9.57% of TCTA on average respectively over the period. These results are also
reasonable for the period since conventional banks benefited from their long
establishment in the Indonesian banking Industry compared to their peers of Islamic
banks. This profile coincides with Shaban et al. (2014) stating that Islamic banks
experienced overpricing behaviour during the period.
In addition, Table 6.4 exhibits the decomposition of the total cost experienced by
Indonesian banks by ownership and operational system during the period 2002-2010.
Table 6.4. Expenses of Indonesian banks (2002-2010)
OWNERSHIP Types: OPERATIONAL
Systems:
SOB PB JVB LGOB FB Conv. Sharia
Obs 36 513 135 234 81 981 18
Avg Total Costs
(IDR Million) 15,300,000 1,125,210 385,052 492,589 1,479,997 1,423,883 1,066,982
Avg Interest Exp82
(IDR Million) 7,923,090 597,617 173,957 235,385 488,748 714,684 492,020
Avg Non-Interest
Exp (IDR Million)
4,580,650
344,037
149,970
127,932
769,566
455,378
399,177
Avg Personnel Exp
(IDR Million) 2,820,548 185,049 61,459 129,272 223,231 254,775 175,786
Avg Total
Costs/Assets 0.0956 0.1121 0.0764 0.0964 0.0864 0.0934 0.0957
State-owned banks (SOB), foreign-owned banks (FB), joint-venture banks (JVB), private banks (PB), local
government-owned banks (LGOB), conventional banks (CONV) and islamic banks (SHARIA).
82
Interest expense is applied to conventional banks; meanwhile for sharia (Islamic banks), operating
expenses on profit or revenue sharing contract is applied.
135
State Owned Banks (SOBs) experienced the highest average expenditure for all three
kinds of expenses (interest expenses, non-interest expenses, and personnel expenses)
incurred in their banks. These results are very sensible given the facts that they are
mainly from the largest banks in terms of total assets. They also employ a considerable
amount of employees across their branches all over Indonesia. JVBs experienced the
lowest expenditures for the components of interest expenses and personnel expenses
whilst the lowest expenditures on non-interest expenses were experienced by the
LGOBs. Overall, conventional banks underwent higher average expenditures than those
of their peers from Islamic banks. This result is different with the findings of the TCTA
terms since this is measured in IDR currency.
The following three tables reveal the costs for each of the collected funds, employed
labours and consumed capital assets. The price of funds indicates the average cost of
funds as the main cost for banks in their operation.
Table 6.5. Price of Funds (2002-2010)
2002 2003 2004 2005 2006 2007 2008 2009 2010 Avg.
Ownership Type:
FB 0.0833 0.0382 0.0316 0.0502 0.0800 0.0641 0.0661 0.0626 0.0462 0.0580
JVB 0.0645 0.0424 0.0313 0.0501 0.0763 0.0590 0.0621 0.0517 0.0657 0.0559
LGOB 0.0750 0.0680 0.0439 0.0435 0.0513 0.0523 0.0528 0.0562 0.0583 0.0557
PB 0.1184 0.0880 0.0562 0.0668 0.0879 0.0680 0.0731 0.0719 0.0659 0.0774
SOB 0.1255 0.0993 0.0593 0.0615 0.0778 0.0563 0.0518 0.0565 0.0455 0.0704
Operation system:
CONV 0.0991 0.0739 0.0481 0.0575 0.0769 0.0626 0.0658 0.0643 0.0622 0.0678
SHARIA 0.0565 0.0515 0.0505 0.0595 0.0670 0.0470 0.0500 0.0555 0.0420 0.0533
State-owned banks (SOB), foreign-owned banks (FB), joint-venture banks (JVB), private banks (PB), local government-owned
banks (LGOB), conventional banks (CONV) and islamic banks (SHARIA).
LGOBs, JVBs and FBs experienced the lowest price of funds in the cost structure of
Indonesian banking over the period of 2002-2010. These results depict the success of
the LGOBs to gain the cheapest funds benefiting from their locations. LGOBs are
mainly situated in provinces and districts all over Indonesia. For the JVBs and FBs,
these results are not quite surprising considering that they are known as superior banks
in interest-rate management. Conversely, the PBs spent more to pay interest to their
depositors. These results are very sensible given the fact that they found difficulties in
obtaining more funds from depositors due to less trustworthiness perceived by
customers compared to their peers from foreign banks and state-owned banks. In
136
addition, Islamic banks offered lower prices on their funds collected compared to their
peers from the conventional banks.
Table 6.6. Price of Labour (2002-2010)
2002 2003 2004 2005 2006 2007 2008 2009 2010 Avg.
Ownership Type:
FB 0.0162 0.0137 0.0120 0.0113 0.0124 0.0136 0.0111 0.0112 0.0118 0.0126
JVB 0.0157 0.0209 0.0181 0.0121 0.0125 0.0105 0.0104 0.0123 0.0209 0.0148
LGOB 0.0261 0.0263 0.0285 0.0244 0.0223 0.0241 0.0295 0.0292 0.0294 0.0266
PB 0.0180 0.0196 0.0212 0.0219 0.0215 0.0194 0.0222 0.0239 0.0332 0.0223
SOB 0.0140 0.0168 0.0185 0.0195 0.0183 0.0180 0.0170 0.0148 0.0145 0.0168
Operation system:
CONV 0.0193 0.0209 0.0218 0.0203 0.0197 0.0188 0.0214 0.0223 0.0285 0.0214
SHARIA 0.0190 0.0150 0.0145 0.0160 0.0155 0.0155 0.0145 0.0155 0.0155 0.0157
State-owned banks (SOB), foreign-owned banks (FB), joint-venture banks (JVB), private banks (PB), local government-
owned banks (LGOB), conventional banks (CONV) and islamic banks (SHARIA).
In terms of price of labour, FBs and JVBs experienced the lowest costs on their
personnel employment. These results are very sensible considering that they can operate
more efficiently in using their employees compared to other peer groups. On the other
hand, LGOBs spend more money on paying salaries and other related expenses to their
employees. These might happen due to higher salaries applied to their employees and or
more employees were hired for a similar position compared to other peers.
Unfortunately, the data concerning the number of employees from all banks is not
available. However, the possibility of hiring more employees might be the main factor
of their high cost of labour. In addition, based on their operational system, Islamic banks
experienced lower expenses on their human resources compared to their peers from
conventional banks.
Table 6.7. Price of Capital (2002-2010) 2002 2003 2004 2005 2006 2007 2008 2009 2010 Avg.
Ownership Type:
FB 6.40 7.34 9.43 13.78 14.68 14.19 15.80 14.94 16.62 12.58
JVB 5.97 7.00 6.99 5.97 5.37 5.45 4.24 3.21 7.65 5.76
LGOB 1.99 1.83 1.75 1.98 2.23 2.27 2.26 2.03 2.68 2.11
PB 1.40 1.51 1.56 1.52 1.60 1.86 1.90 1.89 2.29 1.72
SOB 1.73 1.18 1.43 1.55 1.97 1.77 2.48 3.22 3.97 2.15
Operation system:
CONV 2.61 2.79 3.00 3.23 3.34 3.43 3.44 3.20 4.35 3.27
SHARIA 0.70 2.48 1.31 2.96 3.00 4.07 4.00 3.55 3.02 2.79
State-owned banks (SOB), foreign-owned banks (FB), joint-venture banks (JVB), private banks (PB), local
government-owned banks (LGOB), conventional banks (CONV) and islamic banks (SHARIA).
137
The last component of the right-hand side of the equation in the estimation models for
Indonesian banks‟ cost efficiency is the price of capital. This ratio is achieved by total
non-interest expenses divided by fixed assets. The data indicates that PBs are the most
efficient banks while FBs are the least efficient ones. These results are contrasted with
the preceding results on price of funds and price of labour. These results might be
caused by a lower level of fixed assets employed by the FBs to the PBs. On the other
hand, Sharia banks turned out to experience lower average price of capital than those of
conventional ones.
6.2.2. Small Business Finance of Indonesian commercial banks
The three different models examined in this thesis reside in the output components
where in model 2, total finance is broken down in to small business finance (SBF) and
other finance (OF). Table 6.8 presents the profile of small business finance undertaken
by Indonesian commercial banks classified based on ownership and operation:
Table 6.8. Banks’ Small Business Finance by Different Ownership & Operation
(2002-2010)
OWNERSHIP Types: OPERATIONAL Systems:
SBF-SOB SBF-PB SBF-JVB SBF-LGOB SBF-FB SBF-Conv SBF-Sharia
Obs. 36 513 135 234 81 981 18
Mean
(IDR Million) 16,800,000 485,770 27,689 615,106 22,691 997,652 1,443,998
Std Dev
(IDR Million) 15,900,000 1,344,016 73,873 961,757 111,866 4,437,158 1,493,210
Minimum
(IDR Million) 2,883,722 235 0 7,526 0 0 174,114
Maximum
(IDR Million) 75,400,000 13,100,000 558,277 7,280,370 676,037 75,400,000 5,506,198
Avg SBFPCT 0.2593 0.1564 0.0180 0.2709 0.0009 0.1552 0.1766
SBF-SOB: small business finance of State-owned banks; SBF-PB: small business finance of private banks; SBF-JVB: small
business finance of joint-venture banks; SBF-LGOB: small business finance of local government-owned banks; SBF-FB:
small business finance of foreign-owned banks; SBF-Conv: small business finance of conventional banks; SBF-Sharia: small
business finance of Islamic banks.
LGOBs and SOBs successfully maintained their lending (finance) portfolio devoted to
small businesses of around 27.09% and 25.93% respectively over the period 2002 to
138
2010 although the GOI had revoked the regulation of SBF threshold in 200183
. Their
achievements imply their success in utilizing their vast network coverage over the
provinces and districts in Indonesia. Moreover, this also reflects the high involvement of
state-owned banks in providing financing to small businesses in Indonesia. As known
that PT. Bank Rakyat Indonesia (Bank BRI) is one of the state-owned banks that has
given significant contributions to the development of small businesses in Indonesia by
providing financing to micro and small businesses.
In terms of the operational system, Islamic banks (SHARIA) allocated their financing
portfolios to small businesses well above the achievement of their peers from
conventional banks. Islamic banks benefited from the murabaha84
contract in their
financing. The advantage of this scheme compared to the working capital or investment
loans of conventional banks is that in the murabaha contract, a small business client
does not need to provide any collateral in advance whilst conventional banks prevail
collateral as the main pre-condition for loans (Shaban et al., 2014).
Figure 6.1 presents the propensity of Indonesian banks to finance small businesses
across different groups of banks.
Figure 6.1. Banks’ Small Business Finance Proportions (2002-2010)
State-owned banks (SOB), foreign-owned banks (FB), joint-venture banks (JVB), private banks (PB),
local government-owned banks (LGOB), and Islamic banks (SHARIA).
83
The prevailing regulations during the period (BI Regulation Number 3/2/PBI/2001 about Small
Business Finance on January 4, 2001) does not oblige all commercial banks in Indonesia to finance small
businesses a minimum of 22.5 percent anymore, rather it is readjusted based on the ability of each bank.
84
Murabaha (baia or sale) contract is a transaction where a bank buys an asset on behalf of the client and
selling it to the client at a mark-up price.
0.00%
10.00%
20.00%
30.00%
40.00%
50.00%
2002 2003 2004 2005 2006 2007 2008 2009 2010
SBF
SBF Percentage (Indonesian Commercial Banks Period 2002-2010)
FB
JVB
LGOB
PB
SOB
SHARIA
139
Although LGOBs are recorded as the largest providers of small businesses over the
period, their SBs financing tended to decline over time during the period 2002 to 2010.
SOBs emerged to be the only one group that constantly provides SBF of above 22.5%
over time during the period. This attainment reflects their high commitment to support
small businesses development in Indonesia over the period. On the other hand, JVBs
and FBs seem to be still reluctant to finance small businesses. They achieved their SBF
of 1.80% and 0.09% on average over time respectively. This might be caused by some
reasons. First, they have less experience in providing financing to small businesses.
Second, small business finance require specific lending technology i.e. relationship
lending which is for some banks considered more costly. Third, as Owualah (1990) and
Petersen and Rajan (1994) explained that the characteristics embedded in small
businesses (asymmetry of information, lack of collateral, and inexperienced
management) have been the major impediments to obtain external financings including
from commercial banks.
Sharia banks are recorded as a group of banks which had an increasing trend of
financing small businesses in Indonesia particularly in the last five years of the study
period. They recorded their financing to small businesses above 20% in 2010 with an
annual average financing of 17.66% over the period 2002 to 2010. The condition
coincides with the study undertaken by Shaban (2014) which states that sharia banks
(Islamic banks) were inclined to finance small businesses in Indonesia during the same
period.
6.2.3. Data of Bank-Specific Variables and Macroeconomic Variables
Bank specific factors are derived from the observed data of 109 commercial banks
consisting of 107 conventional banks and 2 sharia banks in Indonesia for the period
2002-2010, whilst the macroeconomic Indicators were collected from the Bank
Indonesia, Statistics of Indonesia, Department of National Planning, Government of
Indonesia as well as the Asian Development Bank (ADB). Table 6.9 presents the
characteristics of the Indonesian bank specific variables as well as the macroeconomic
Indicators included in the analysis in this chapter.
140
Size of bank is represented by natural logarithm of total assets (LNTA). This variable is
adopted to determine whether a bank could take benefit from the economic of scale of
its operation. It is hypothesised that the bigger a bank the more efficient the bank will be.
The result of the hypothesis testing would also testify whether the wave of mergers
among commercial banks is a right decision to enhance bank efficiency. In addition,
return on assets (ROA) represents banks‟ profitability. On the observed dataset, average
ROA of 109 commercial banks in Indonesia was 2.24 percent over the study period.
This is regarded a successful achievement after they suffered during the crisis of 1997-
98. During the crisis, the majority of banks experienced losses in their operations.
Table 6.9. Bank-Specific and Macroeconomic Variables
Variable Obs. Mean Std. Dev. Min Max
lnta 999 14.9090 1.8381 9.7722 19.8264
roa 999 0.0229 0.0643 -1.53 0.32
car 999 0.2401 0.1584 -0.2230 1.0900
ldr 999 0.7669 0.3741 0.0001 3.3500
npl 999 0.0225 0.0298 0.0001 0.5290
infl 999 8.08 3.94 2.80 17.10
gdpgr 999 5.39 0.64 4.50 6.30
unemp 999 9.19 1.19 7.10 11.20
mktindex 999 1,714 1,020 425 3,704
usdr 999 9,338 498 8,576 10,398
LNTA is the natural logarithm of total assets; ROA is return on assets; CAR is capital
adequacy ratio; NPL is non-performing loans; INFL is annual inflation rate; GDPGR is
Gross Domestic Product at real prices growth rate; UNEMP is annual unemployment rate;
MKTINDEX is the market index, a composite index in Indonesia stock exchange;
USDRATE is United States Dollars (USD) exchange-rate in terms of Indonesian Rupiahs
(IDR).
Capital adequacy ratio (CAR) is a representative for bank capital and this reflects the
ability of the bank to cover all risks including financial and solvency risks. The average
CAR of those 109 banks was 24.01 per cent meaning a strong condition of Indonesian
banks‟ equity over the period 2002 to 2010. In addition, loan to deposits ratio (LDR) is a
proxy for bank liquidity as well as bank lending exposure. The average LDR of
Indonesian banks over the period was 76.69 percent, indicating that the condition of the
banks was quite liquid and the financial intermediary function of the banks increasingly
improved during the period 2002 to 2010 after suffering from the Asian financial crisis
of 1997-98.
141
Non-performing loans (NPL) as an indicator of the bank risk. The average NPL of
Indonesian banks during the period was 2.25 percent which is considered a good
performance85
. This achievement reflects sound and better loan management undertaken
by Indonesian bank management post the crisis of 1997-98.
As these are also applied in the first empirical chapter (chapter five), this chapter also
adopts some macroeconomic indicators namely inflation (INF), Gross Domestic Product
growth (GDPGR), unemployment rate (UNEMP), Indonesia stock market index
(MKTINDEX), United State Dollars against Indonesian Rupiahs (USDRATE). These
factors are considered as the environmental factors. The inclusion of these factors in the
regression models is to investigate their association with Indonesian banks‟ cost
efficiency during the period 2002 to 2010.
6.3. Cost Efficiency of Indonesian Commercial Banks
6.3.1. Average Cost Efficiency of Indonesian Commercial Banks
This section presents an analysis of the cost efficiency of Indonesian commercial banks
over the period 2002-2010. The analysis provides two estimation methods namely
standard pooled estimation (pooled) and Battese and Coelli 1992 (BC92) methods. The
pooled estimation method is a standard-model using pooled data in estimating the
efficiency scores in the stochastic frontier approach (SFA) with the assumption of time
variant. Meanwhile, BC92 is an estimation method to obtain efficiency estimates using
time variant with different assumption of inefficiency (μ is assumed to be truncated
normal distribution).
85
Indonesian bank Authority (BI) determines 5% NPL as cut-off point for a commercial bank to be
classified as a well-performing bank in managing its loans.
142
6.3.1.1. Overall Performance of Indonesian Banks’ Cost Efficiency
Figure 6.2 exhibits the tendency of cost efficiency of Indonesian banks during the
period 2002-2010. The estimated average cost efficiency of the banks across the three
models are represented by the three lines inside the figure.
Figure 6.2. Average Cost Efficiency of Indonesian Banks (Pooled-Model 1,2,3)
CEPOOL1: Average SFA cost efficiency scores obtained by standard pooled method for model 1;
CEPOOL2: Average SFA cost efficiency scores obtained by standard pooled method for model 2;
CEPOOL3: Average SFA cost efficiency scores obtained by standard pooled method for model 3.
The three lines depict a slightly increasing trend of Indonesian banks‟ cost efficiency
over the period 2002 to 2010. There is a disparity in the average cost efficiency obtained
across the three models. These results are caused by different specifications at the
output components of the three models. The average cost efficiency obtained from
model 1 (cepool1) is the highest amongst the three models86
. Model 1 estimates cost
efficiency with the specifications in the output components where total finance was used.
In addition, model 3 records the lowest average cost efficiency. This result is very
sensible considering that the model only accounts for other finance and disregards small
business finance.
It is worth noting that despite that there was an incline tendency of the average cost
efficiency attained by Indonesian banks, they recorded declines over the year 2003,
2004, 2008 and 2010 compared to the preceding year respectively. The decline that
occurred in the period 2003-2004 might have been because of the economic conditions
during the time where the average banks‟ lending rate were trending to decrease from
86
Average cost efficiency of the three models over the period 2002-2010 are 0.8464, 0.8392, and 0.8115
for the model 1, 2 and 3 respectively.
2002 2003 2004 2005 2006 2007 2008 2009 2010
cepool1 0.8394 0.8331 0.8324 0.8415 0.8557 0.8558 0.8556 0.8611 0.8429
cepool2 0.8287 0.8243 0.8253 0.8325 0.8502 0.8491 0.8451 0.8533 0.8445
cepool3 0.8009 0.7956 0.7953 0.8061 0.8232 0.8244 0.8203 0.8273 0.8100
0.7600 0.7800 0.8000 0.8200 0.8400 0.8600 0.8800
Effi
cie
ncy
Sco
re
Cost Efficiency - Pooled - Model 1,2,3
143
18.76% in 2002 to 16.48% and 14.68% in 2003 and 2004 respectively. The decrease of
the lending rate seems to affect bank income as well as bank cost efficiency to decline
during the time. Meanwhile, the decline occurred in 2008 was mainly caused by the
global economic crisis where Indonesian economic growth shrunk from 6.30% in 2007
to 6.00% in 2008. In addition, the decline which occurred in 2010 was caused by the
overheated economy in 2010 reflected by the increasing in the inflation rate from 2.80%
in 2009 to 7.00% in 2010.
Figure 6.3. Average Cost Efficiency of Indonesian Banks (BC92-Model 1,2,3)
BC92_1: Average SFA cost efficiency scores obtained by Battese-Coelli 1992 (BC92) method for model 1;
BC92_2: Average SFA cost efficiency scores obtained by Battese-Coelli 1992 (BC92) method for model 2;
BC92_3: Average SFA cost efficiency scores obtained by Battese-Coelli 1992 (BC92) method for model 3.
Interesting results are obtained from the Battese and Coelli 1992 (BC92) estimation
method (see figure 6.3). This estimation also provides an increasing trend over a period
similar to the standard pooled estimation method. The increasing pattern of SFA cost
efficiency estimate could be observed by the sign of Eta ( )87
which has a positive sign
over all BC2 method (See appendices 19, 21 and 23).
The differences between the BC92 and standard pooled estimation are as follows: first,
the average cost efficiency estimation under BC92 constantly increased over time
during the period. Second, model 2 which accounts for small business finance and other
87 Eta ( ) is the only one parameter estimated in the BC92 approach. Eta reflects the improvement or
decline of the efficiency attainment by observed companies (banks) over time during the study. A positive
Eta constitutes that there has an improvement of the efficiency attainment by the companies (banks) (in
this case in terms of cost efficiency attainment) on average.
2002 2003 2004 2005 2006 2007 2008 2009 2010
bc92_1 0.6616 0.6758 0.6896 0.7029 0.7158 0.7284 0.7405 0.7522 0.7634
bc92_2 0.7437 0.7607 0.7768 0.7920 0.8064 0.8199 0.8326 0.8446 0.8558
bc92_3 0.5844 0.5998 0.6150 0.6298 0.6443 0.6584 0.6722 0.6856 0.6987
-
0.1000
0.2000
0.3000
0.4000
0.5000
0.6000
0.7000
0.8000
0.9000
Effi
cie
ncy
Sco
re
Cost Efficiency - BC92 - Model 1,2,3
144
finance in their outputs becomes the highest average cost efficient amongst the models88
under BC92. This latter result coincides with the findings of the first empirical chapter,
stating that model 2 achieved the highest average technical efficiency amongst the
models (see chapter 5 for details).
The achievement of model 2 to outperform other two models has empirically confirmed
the effect of diversifying the loan portfolio on bank performance in terms of cost
efficiency. The cost efficiency obtained due to diversification might have resulted from
more efficient usage of bank resources related to small business finance. However, it is
important to investigate whether loans and risks affected banks‟ cost efficiency. For the
empirical evidence, the last section of this chapter provides an analysis of the
determinants of banks‟ cost efficiency where loans and risks are adopted as some
explanatory variables in the regression models.
Overall, all estimations obtained using both the standard pooled and BC92 estimation
methods provide similar findings that Indonesian banking cost efficiency enjoyed an
increasing trend over the period 2002 to 2010. These findings supported the argument
that Indonesian banks have taken lessons from past Asian Financial Crisis of 1997-98
and become more resilient over the study period.
Financial Crisis and Cost Efficiency:
It is clearly observed from figure 6.2 and 6.3 that Indonesian banks enjoyed an
increasing trend of cost efficiency levels over the period 2002 to 2010. Those figures
also illustrate Indonesian bank cost efficiency levels attained were relatively stable
during the years 2007-2009. The very tiny decline occurred in 2008. These results
verified the better performance of Indonesian banks over the global financial crisis of
2008 compared to other countries i.e. European and Australian banks (Alzubaidi and
Bougheas, (2012), Forughi and De Zoysa (2012).
The smoother decline in the cost efficiency levels could also be a reflection of the
resilience of Indonesian banks in dealing with the crisis of 2007. These findings also
88
Average cost efficiency of the three models derived from BC92 estimation over the period 2002-2010
are 0.7145 for model 1, 0.8036 for model 2, and 0.6431 for model 3.
145
reflect that Indonesian banks took lessons learnt from the dramatic suffering of the
previous crisis of 1997-98. Emerging markets including the Indonesian banks were
more capitalised compared to European banks during the crisis in 2007 (Fethi et al,
2012). The high level of capitalisation has played an important role in absorbing the
effect of the crisis. Due to the success of the GOI in implementing the numbers of
various programmes [by increasing bank capital, improving bank competition and
efficiency (see the section 2.2.3 and 2.2.4)], these have resulted in the resilience of
Indonesian banks to successfully get through the financial crisis of 2007.
6.3.1.2. Ownership and Banks’ Cost Efficiency
Figure 6.4 displays the different performance of Indonesian banks cost efficiency across
different ownership based on the standard pooled (pool) and Battese & Coelli 1992
(BC92) estimation methods for model 1, 2, and 3 respectively. It seems that the
disparity of the performance is more observable on the BC92 method rather than those
of the pooled method. For instance, for model 1, the range of the efficiency score for
BC92 is 0.2285 (from 0.8903-0.6618) while for the pooled method is 0.0237 (from
0.8606-0.8370). Similar results are also applied for models 2 and 3. These different
results are possible due to different methods applied. This study adopts both methods in
the analysis of the findings and conclusions.
Figure 6.4. Cost Efficiency of the Indonesian Banks: By Different Ownership
CEPOOL1: Average SFA cost efficiency scores under standard pooled method for model 1; CEPOOL2: Average
SFA cost efficiency scores under standard pooled method for model 2. CEPOOL3: Average SFA cost efficiency
scores under standard pooled method for model 3. BC92_1: Average SFA cost efficiency scores under Battese-Coelli
1992 (BC92) method for model 1; BC92_2: Average SFA cost efficiency scores under Battese-Coelli 1992 (BC92)
method for model 2; BC92_3: Average SFA cost efficiency scores under Battese-Coelli 1992 (BC92) method for
cepool1 cepool2 cepool3 bc92_1 bc92_2 bc92_3
FB 0.8606 0.8445 0.8365 0.8903 0.8626 0.8907
JVB 0.8561 0.8228 0.8389 0.8237 0.8108 0.8119
LGOB 0.8546 0.8521 0.8083 0.7017 0.8205 0.5696
PB 0.8370 0.8354 0.8025 0.6618 0.7813 0.5970
SOB 0.8590 0.8593 0.8004 0.7436 0.8517 0.5890
- 0.2000 0.4000 0.6000 0.8000 1.0000
Effi
cie
ncy
Sco
re
Average Cost Efficiency - Model 1,2,3 (By Ownership)
146
model 3. Banks’ Ownership: State-owned banks (SOB), foreign-owned banks (FB), joint-venture banks (JVB),
private banks (PB), local government-owned banks (LGOB).
The ranking of Indonesian banks‟ cost efficiency based on their ownership and
operations during the study period is presented in the following table:
Table 6.10. Ranking of Indonesian Banks Cost Efficiency (2002-2010)
Ownership: POOL1 Rank POOL2 Rank POOL3 Rank BC92_1 Rank BC92_2 Rank BC92_3 Rank
FB 0.8606 1 0.8445 3 0.8365 2 0.8903 1 0.8626 1 0.8907 1
JVB 0.8561 3 0.8228 5 0.8389 1 0.8237 2 0.8108 4 0.8119 2
LGOB 0.8546 4 0.8521 2 0.8083 3 0.7017 4 0.8205 3 0.5696 5
PB 0.8370 5 0.8354 4 0.8025 4 0.6618 5 0.7813 5 0.5970 3
SOB 0.8590 2 0.8593 1 0.8004 5 0.7436 3 0.8517 2 0.5890 4
System
Operation: POOL1 Rank POOL2 Rank POOL3 Rank BC92_1 Rank BC92_2 Rank BC92_3 Rank
CONV 0.8469 1 0.8394 1 0.8124 1 0.7155 1 0.8033 2 0.6449 1
SHARIA 0.8190 2 0.8289 2 0.7629 2 0.6576 2 0.8182 1 0.5491 2
FE Operation: POOL1 Rank POOL2 Rank POOL3 Rank BC92_1 Rank BC92_2 Rank BC92_3 Rank
FEB 0.8473 1 0.8392 2 0.8152 1 0.7218 1 0.8079 1 0.654 1
Non FEB 0.8432 2 0.8394 1 0.7978 2 0.6879 2 0.7882 2 0.6039 2
For the ownership, they compose of state-owned banks (SOB), foreign-owned banks (FB), joint-venture banks
(JVB), private banks (PB), and local government-owned banks (LGOB). For the system operation, they consist of
conventional banks (CONV) and islamic banks (SHARIA). For the foreign exchange (FE) operation, they include
two groups of banks, namely foreign exchange banks (FEB) and non foreign exchange banks (Non FEB).
Based on ownership criteria, the results expose that foreign banks (FBs) emerge as the
most cost efficient banks in the Indonesian banking sector (Table 6.10). The FBs ranked
first at four criteria [the standard pooled method for model 1 (pool1), and BC92 method
for model 1, 2, and 3 respectively (bc92_1, bc92_2, and bc92_3)]. It is then followed by
SOBs and JVBs which ranked first at one criterion (pool2 and pool3) respectively.
Lastly, private banks (PBs) ranked as the least cost efficient according to the results
obtained from all criteria evaluated. They recorded the lowest rank at three criteria
(pool1, bc92_1 and bc92_2) and the fourth rank at two criteria (pool2 and pool3).
The success of FBs to be the most cost-efficient banks coincides with some empirical
evidence that FBs tend to outperform other group of banks in terms of efficiency [See
Bonin et al (2005) for the case of eleven transition countries in Europe and Gardener et
147
al (2011) for the case of Southeast Asian banking]. It is worthwhile to explain that the
sound performance of FBs does not come from their activities in loan portfolio
diversification given the fact that they only posed at the third rank at the standard
pooled method for model 2 (pool2)89
. These results also coincide with the findings in
chapter 5 of this thesis. Their highest ranked performance seems to be derived from
their lower investment in branches and number of employees (i.e. factors of production)
as well as their competitive cost of funds compared to SOBs, LGOBs and PBs.
In addition, SOBs posed the first rank at one criterion (the standard pooled method for
model 2). This result suggests that their deeply involvement in small business finance
has resulted to be the most cost efficient banks in that criterion. These findings coincide
with the results of the DEA technical efficiency (see chapter five) as well as the
evidence provided by Fethi et al. (2011) which found that state-owned banks were the
most efficient banks in Egypt. Meanwhile, private banks experienced a controversial
performance by achieving the lowest SFA cost efficiency on average. These findings
dissatisfy the belief that private banks operate more professionally than their peers from
state-owned banks. These findings are different from the study results of Margono, et al
(2010).
In terms of their system operations, conventional banks (CONV) were more efficient in
all criteria (models and methods) evaluated, except for the BC92 method of model 2
(bc92_2) where Islamic banks (SHARIA) achieved a slightly higher cost efficient than
their peers of CONV. These findings are very sensible considering that conventional
banks have been more established than Islamic banks in terms of their experience in the
industry. These results coincide with the findings of Hamiltona et al (2010) that Islamic
banks were less cost efficient in the Jordanian banking sector.
Finally, in terms of foreign exchange services, foreign exchange banks (FEB) constantly
obtained higher efficiency scores than their peers from non-foreign exchange banks
(NFEB), except for the pooled method for the model 2 (pool2). Hence, we can derive
the conclusion that foreign exchange banks benefited from the economies of scope by
89
Model 2 splits total finance into small business finance and other finance in outputs for the calculation
of cost efficiency estimates.
148
providing foreign exchange transaction in addition to domestic currency one in their
operations.
The Evolution of Indonesian Banks’ Cost Efficiency over the Period 2002-2010:
The evolution of average cost efficiency attained by different groups of ownerships and
different operational systems across Indonesian commercial banks during the period
2002 to 2010, can be observed in the figure as follows:
Figure 6.5. The Evolution of Indonesian Banks’ Cost Efficiency – Pooled - Model 1
State-owned banks (SOB), foreign-owned banks (FB), joint-venture banks (JVB), private banks
(PB), local government-owned banks (LGOB), conventional banks (CONV) and islamic banks
(SHARIA).
Figure 6.5 demonstrates the fluctuation of their cost efficiency over the period 2002 to
2010. All banks underwent fluctuations with an average cost efficiency was over 0.800.
FBs and SOBs experienced the highest average cost efficiency during the period with an
average of 0.861 and 0.859 respectively. FBs‟ cost efficiency reached their peak
performance of 0.907 in the year 2009 but turned out to be steeply down in 2010. SOBs
did very well in the first two years, but systematically deteriorated on average in the
following years. Although PBs recorded attaining the lowest cost efficiency level on
average, they fortunately had an upward trend in their cost efficiency attainment over
time during the period. It is worthwhile to note that all banks experienced a slight
decrease in 2008 except the LGOBs and SOBs. These reflect that all government-owned
banks (local and central government banks) in Indonesia performed better and more
2002 2003 2004 2005 2006 2007 2008 2009 2010
FB 0.8808 0.8447 0.8764 0.8336 0.8736 0.8776 0.8678 0.9077 0.7835
JVB 0.8537 0.8501 0.8222 0.8712 0.8779 0.8697 0.8668 0.8610 0.8327
LGOB 0.8310 0.8365 0.8419 0.8419 0.8582 0.8631 0.8708 0.8764 0.8720
PB 0.8293 0.8223 0.8217 0.8336 0.8470 0.8462 0.8444 0.8469 0.8411
SOB 0.8920 0.8756 0.8628 0.8587 0.8400 0.8456 0.8469 0.8599 0.8493
CONV 0.8388 0.8340 0.8338 0.8424 0.8558 0.8564 0.8562 0.8617 0.8430
SHARIA 0.8732 0.7835 0.7585 0.7964 0.8486 0.8241 0.8221 0.8305 0.8344
0.6500
0.7000
0.7500
0.8000
0.8500
0.9000
0.9500
Co
st E
ffic
ien
cy
Cost efficiency of Model 1: Pooled
149
resilient during the crisis of 2007-08.
In addition, conventional banks experienced a slight increase in the trend of their cost
efficiency and they performed better in terms of cost efficiency performance compared
to Sharia banks over the period. This means that their management successfully
managed in employing their experiences in the industry to increase their cost efficiency
level.
Figure 6.6. The Evolution of Indonesian Banks’ Cost Efficiency – BC92 - Model 1
State-owned banks (SOB), foreign-owned banks (FB), joint-venture banks (JVB), private banks
(PB), local government-owned banks (LGOB), conventional banks (CONV) and islamic banks
(SHARIA).
In contrast to the data pooled estimation, the BC92 generates the results that all groups
experienced an increasing trend in their average cost efficiency levels over time during
the period. FBs are the most cost efficient banks and were followed by JVBs and SOBs
in the second and third places. Similar to the previous results, PBs demonstrated the
lowest cost efficiency level although the trend was positive. In addition, Conventional
banks experienced better average cost efficiency compared to Sharia banks over the
period.
The evolutions of Indonesian banks‟ cost efficiency under model 2 and 3 respectively
are presented on figures 6.7, 6.8, 6.9, and 6.10 (see the appendices 10-13).
2002 2003 2004 2005 2006 2007 2008 2009 2010
FB 0.8664 0.8730 0.8794 0.8854 0.8912 0.8967 0.9019 0.9069 0.9116
JVB 0.7908 0.7997 0.8082 0.8166 0.8246 0.8324 0.8399 0.8471 0.8541
LGOB 0.6453 0.6605 0.6752 0.6894 0.7032 0.7165 0.7294 0.7418 0.7537
PB 0.6005 0.6169 0.6328 0.6482 0.6633 0.6778 0.6920 0.7056 0.7188
SOB 0.6933 0.7069 0.7201 0.7328 0.7451 0.7569 0.7682 0.7791 0.7896
CONV 0.6629 0.6770 0.6907 0.7040 0.7169 0.7294 0.7414 0.7531 0.7643
SHARIA 0.5942 0.6111 0.6277 0.6437 0.6592 0.6743 0.6888 0.7029 0.7164
-
0.1000
0.2000
0.3000
0.4000
0.5000
0.6000
0.7000
0.8000
0.9000
1.0000
Co
st E
ffic
ien
cy
Cost efficiency of Model 1: BC92
150
6.3.1.3. Size and Banks’ Cost Efficiency
This section presents the average cost efficiency of Indonesian banks under the standard
pooled and BC92 methods for the three models over the period 2002 to 2010. Banks are
classified into four bank peer groups based on their asset size. Bank peer group 1 covers
banks‟ asset size of less than IDR 10 trillion. Bank peer group 2 includes banks with
their asset size of more than IDR 10 trillion and less than and equal to IDR 50 trillion.
Bank peer group 3 composes of banks with asset size of greater than IDR 50 trillion and
less than and equal to IDR 100 trillion. Bank peer group 4 consists of banks with asset
size of more than IDR 100 trillion. The distribution of the group is presented in the table
6.11.
Table 6.11. Distribution of Average Cost Efficiency By Bank Peer Group
Peer Group 1 2 3 4 TOTAL
Average Cost Efficiency (Pooled) MODEL 1 for each group 0.8459 0.8468 0.8443 0.8539 0.8464
Average Cost Efficiency ( bc92_1) MODEL 1 for each group 0.7049 0.7439 0.7104 0.7358 0.7145
Average Cost Efficiency (Pooled) MODEL 2 for each group 0.8388 0.8403 0.8339 0.8446 0.8392
Average Cost Efficiency ( bc92_2) MODEL 2 for each group 0.7947 0.8261 0.8357 0.8280 0.8036
Average Cost Efficiency (Pooled) MODEL 3 for each group 0.8114 0.8103 0.7991 0.8258 0.8115
Average Cost Efficiency ( bc92_3) MODEL 3 for each group 0.6342 0.6732 0.6292 0.6562 0.6431
Number of banks in Peer group (as of 2010) 65 33 6 7 111
Peer Group 1 (Total Assets <= IDR 10 Trillion); Peer Group 2 (Total Assets > IDR 10 Trillion and <= IDR 50Trillion);
Peer Group 3 (Total Assets > IDR 50 Trillion and <= IDR 100 Trillion); Peer Group 4 (Total Assets > IDR 100 Trillion).
Bank peer group 4 (the large banks) experienced the highest average cost efficiency
under the pooled method for model 1, 2, and 3 with an average cost efficiency of 0.8539,
0.8446, and 0.8258 respectively, while bank peer group 3 suffered the lowest average
cost efficiency during the period 2002 to 2010. These results reflect the success story of
the largest banks in Indonesia in managing their operation benefiting from the
economies of scale. PT. Bank Mandiri, PT. Bank BNI, PT. Bank BRI, and PT. Bank
BCA are those of the group members that are consistently marked sound performance
over the period. These results also coincide with the findings of Shaban (2008), Demir
et al (2005) and Isik and Hassan (2003b) in the case of Egypt and Turkey. They found
that large banks were more efficient compared to other sizes of bank.
151
On the other hand, there is a disparity of the results obtained from the BC92 method
revealing that bank peer group 2 recorded the most cost efficient banks for model 1 and
3, whilst for model 2, bank peer group 3 ranked first during the period.
Figure 6.11. Cost Efficiency – Pooled - Model 1 (by Bank Peer Group)
Source: Data processed
Figure 6.11 displays the evolution of their cost efficiency levels attained by their groups
for model 1. Bank peer group 3 and 4 experienced a declining trend of the average cost
efficiency while bank peer group 2 and 1 underwent an increasing trend over the period
2002-2010. However, bank peer group 4 came to be more stable with the attainment of
their average cost efficiency level of above 0.8400. This made the group become the
most cost-efficient banks during the period.
Figure 6.12. Cost Efficiency – BC92 - Model 1 (by Bank Peer Group)
Source: Data processed
Meanwhile, under the BC92 estimation method, peer group 1, 2 and 4 enjoyed the
0.7800
0.8000
0.8200
0.8400
0.8600
0.8800
0.9000
0.9200
2002 2003 2004 2005 2006 2007 2008 2009 2010
Ave
rage
Co
st E
ffic
ien
cy
Average Cost Efficiency Pooled-Model 1 by Bank Peer Group
Peer Group 1 (Total Assets <= IDR 10Trillion)
Pee Group 2 (Total Assets > IDR 10Trillion and <= IDR 50Trillion)
Peer Group 3 (Total Assets > IDR 50Trillion and <= IDR 100 Trillion)
Peer Group 4 (Total Assets > IDR 100Trillion)
0.0000
0.1000
0.2000
0.3000
0.4000
0.5000
0.6000
0.7000
0.8000
0.9000
2002 2003 2004 2005 2006 2007 2008 2009 2010
Ave
rage
Co
st E
ffic
ien
cy
Average Cost Efficiency BC92-Model 1 by Bank Peer Group
Peer Group 1 (Total Assets <= IDR10 Trillion)
Pee Group 2 (Total Assets > IDR 10Trillion and <= IDR 50Trillion)
Peer Group 3 (Total Assets > IDR 50Trillion and <= IDR 100 Trillion)
Peer Group 4 (Total Assets > IDR100 Trillion)
152
increasing trend of their average cost efficiency over the period whilst peer group 3
experienced a decline trend. Bank peer group 2 reached the highest average cost
efficiency level of 0.7439 while the lowest average cost efficiency level was undergone
by bank peer group 1 with an average level of 0.7049.
The other results obtained for models 2 and 3 are presented in figure 6.13, 6.14, 6.15,
and 6.16 (see the appendices 14-17).
6.4. Determinants of Cost Efficiency
This section reveals the determinants of cost efficiency of Indonesian banks during the
period 2002-2010. There are three models (model 1, model 2 and model 3) examined
with the two different methods (the pooled data method and the BC92 method). The
cost efficiency estimates obtained from the two methods are dependent variables which
are then regressed with banks‟ specific factors as well as macroeconomic Indicators as
their explanatory variables.
This thesis uses TOBIT-regression to reveal the determinants of Indonesian banks‟ cost
efficiency given the fact efficiency scores are constrained between zero and one. The
preceding studies also use this method to examine the determinants of bank efficiency.
[See i.e. Chang et.al (1998); Nikiel and Opiela (2002); Casu and Molyneux (2003);
Havrylchyk (2005); Grigorian and Manole (2006); Ariff and Can (2008); Sufian (2010);
Gardener et.al (2011)].
The standard Tobit-regression model is defined as follows:
Where is a latent variable, is cost efficiency estimate obtained by SFA parametric
approach. is the set of parameters to be estimated. is the error term.
The equation of the regression is specified as follows:
153
= b0 + b1 + b2 + b3 + b4 + b5 + b6 +
b7 + b8 + b9 + b10 .
Where:
The dependent variable is which denotes the SFA cost efficiency estimate of i-th
bank in period t. is the natural logarithm of total assets, a proxy for bank size.
is return on assets, a proxy for bank profitability. is capital adequacy
ratio, represents bank capital. This ratio also reflects bank provisions to cover bank
solvency risk. is loan deposit ratio, a proxy of bank intermediation as well as
bank liquidity. is non-performing loans, represents bank risk. is an
annual inflation rate. This reflects the overall percentage increase in the consumer price
index for all goods and services in Indonesia. is gross domestic product at real
price growth rate. This is a sign of the growth of Indonesia‟s total goods and services
produced in Indonesia in annual term. is an annual unemployment rate in
Indonesia. is the stock composite index in the Indonesia Stock Exchange
(ISE). is United State Dollar exchange rate, in terms of Indonesian Rupiah
(IDR). is error term.
The sample dataset covers 109 banks consisting of around 93 per cent of the population
of Indonesian banking Industry, with the 9-year study period from 2002-2012, resulting
999 total observations. The criteria for bank inclusion in this study are those banks
which have been in existence over the period 2002 to 2010 and their annual financial
statements data are complete over the period.
Table 6.12. Correlation Matrix of Independent Variables-1
lnta roa car ldr npl infl gdpgr unemp mktindex
lnta 1
roa 0.0742 1
car -0.2228 0.0426 1
ldr -0.0116 0.0331 0.1126 1
npl -0.1030 -0.4565 0.0877 0.0831 1
infl -0.0523 0.0095 -0.0525 -0.0061 0.0346 1
gdpgr 0.1477 -0.048 0.0275 0.1198 -0.1518 0.3182 1
unemp -0.1646 0.0756 -0.0521 -0.1345 0.0784 0.4866 -0.1168 1
mktindex 0.2200 -0.069 0.0615 0.1461 -0.168 -0.3461 0.5662 -0.6426 1
usdrate 0.1147 -0.027 0.0051 0.1062 -0.0829 0.1306 -0.1054 -0.2222 0.1913
LNTA is the natural logarithm of total assets; ROA is return on assets; CAR is capital adequacy ratio; NPL is non-performing loans;
INFL is annual inflation rate; GDPGR is Gross Domestic Product at real prices growth rate; UNEMP is annual unemployment rate;
154
MKTINDEX is the market index, a composite index in Indonesia stock exchange; USDRATE is United States Dollars (USD)
exchange-rate in terms of Indonesian Rupiahs (IDR).
Table 6.12 presents the correlation matrix amongst independent variables. This aims to
justify that all explanatory variables have low correlation between each other. In other
words, this is an effort to avoid a multicollinearity problem. It is observed from the table
that the variable MKTINDEX has a high correlation both with GDPGR (0.5662) and
with UNEMP (-0.6426). The significant test for the correlations can be observed on
appendix 26. To solve the problem of this multicollinearity, the MKTINDEX variable
is then dropped from the model.
Table 6.13. Correlation Matrix of Independent variables-2 lnta roa car ldr npl infl gdpgr unemp usdrate
lnta 1.0000
roa 0.0361 1.0000
car -0.2060 -0.1484 1.0000
ldr -0.0115 0.0504 0.1104 1.0000
npl -0.1031 -0.3794 0.3160 0.0831 1.0000
infl -0.0524 0.0156 -0.0481 -0.0061 0.0346 1.0000
gdpgr 0.1478 -0.0175 0.0388 0.1198 -0.1518 0.3182 1.0000
unemp -0.1646 0.0354 -0.0686 -0.1345 0.0784 0.4866 -0.1168 1.0000
usdrate 0.1147 -0.0008 -0.0029 0.1062 -0.0829 0.1306 -0.1054 -0.2222 1.0000
LNTA is the natural logarithm of total assets; ROA is return on assets; CAR is capital adequacy ratio; NPL is non-
performing loans; INFL is annual inflation rate; GDPGR is Gross Domestic Product at real prices growth rate; UNEMP is
annual unemployment rate; USDRATE is United States Dollars (USD) exchange-rate in terms of Indonesian Rupiahs
(IDR).
The table 6.13 constitute the results that there is no significant association between these
independent variables, hence, it can be concluded thta there is no multicollinearity in the
model. All the correlation coefficients are below 0.5. The determinants of Indonesian
banks‟ cost efficiency of the model 1 is presented in table. 6.14.
The regression models present CEPOOL1 (Cost efficiency estimates obtained from the
standard-pooled data method for Model 1) and BC92-1 (cost efficiency estimates under
Battese-Coelli 1992 method for Model 1) as dependent variables. In general, the results
of the first regression (CEPOOL1) suggest that ROA, CAR, LDR, and NPL are of the
bank specific factors which significantly determine the cost efficiency of Indonesian
banks over the period 2002-2010. Meanwhile, INFL, GDPGR, and USDR are of the
macroeconomic indicators which recorded as significant factors for the Indonesian
banks‟ cost efficiency over the period.
155
Tabel 6.14. Determinants of banks cost efficiency – Model 1 (1) (2)
VARIABLES cepool1 bc92_1
lnta 0.000636 0.00890***
(0.00161) (0.00155)
roa 0.275*** 0.00105
(0.0260) (0.00981)
car 0.0414*** -0.000495
(0.0123) (0.00517)
ldr 0.0289*** 0.00952***
(0.00554) (0.00241)
npl -0.126** 0.0472**
(0.0607) (0.0236)
infl -0.00134*** -0.00291***
(0.000502) (0.000195)
gdpgr 0.00627** 0.0320***
(0.00280) (0.00127)
unemp 0.00244 -0.00576***
(0.00159) (0.000606)
usdr 1.37e-05*** 3.08e-05***
(3.33e-06) (1.40e-06)
Constant 0.628*** 0.189***
(0.0463) (0.0247)
sigma_u 0.0274*** 0.124***
(0.00242) (0.00838)
sigma_e 0.0445*** 0.0163***
(0.00106) (0.000387)
Observations 999 999
Number of id 111 111
Standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
The dependent variables are: CEPOOL1: SFA cost efficiency scores
obtained by the standard pooled data method for Model 1; and BC92-1:
SFA cost efficiency scores obtained by Battese-Coelli 1992 method for
Model 1.
The explanatory variables are: LNTA is the natural logarithm of total
assets; ROA is return on assets; CAR is capital adequacy ratio; NPL is
non-performing loans; INFL is annual inflation rate; GDPGR is Gross
Domestic Product at real prices growth rate; UNEMP is annual
unemployment rate; USDRATE is United States Dollars (USD) exchange-
rate in terms of Indonesian Rupiahs (IDR).
On the other hand, the findings derived from the BC92 regression reveal that LNTA,
LDR, and NPL significantly affect the cost efficiency of Indonesian banks over the
period. In addition, all macroeconomic variables (INFL, GDPGR, UNEMP, and USDR)
also impact on the cost efficiency of Indonesian banks over the period.
Size of bank which is represented by LNTA has a positive impact on Indonesian banks‟
cost efficiency in the second regression (BC92) with a confidence level of 99%. This
156
result suggests that the bigger the size of a bank, the more efficient the bank will be.
This finding complies with the SCP hypothesis, stating that as the size of a company‟s
operation increases, the company will operate more efficiently due to the economies of
scale and benefit more profit. This result also verifies the strategic policies exerted by
the Indonesian bank authority to merge commercial banks in Indonesia in regards to
establish many sturdy banks which are more resilient to crises. LNTA has also a positive
association with cost efficiency under the first regression (CEPOOL1) although the
association is not significant. The positive association between bank size and bank
efficiency coincides with the previous results (Yildirim and Philippatos (2007), Andries
(2011), Vu and Turnel, (2011).
ROA as the proxy for bank profitability has a positive impact on Indonesian banks‟ cost
efficiency in the first regression with a confidence level of 99%. This implies that the
profitability gained by Indonesian banks has a significant contribution on the
enhancement of the cost efficiency level of the banks during the period. Meanwhile in
the second regression, ROA has also a positive impact on Indonesian bank cost
efficiency although the the relationship is not significant. This finding is in opposite to
the findings of Yildirim and Philippatos (2007), Andries (2011), and Thangavelu and
Findlay, C. (2013). They did not find a significant association between ROA and bank
efficiency.
Bank capital which is represented by CAR has also affected Indonesian banks‟ cost
efficiency with a positive direction in the first regression. This variable is significant
with an alpha of 1%. This implies that capital adequacy is very important for banks to
support their operation to be more cost efficient. It is worthwhile to explain that the high
compliance of Indonesian banks in terms of capital adequacy fulfilment has a positive
impact on the attainment of their cost efficiency level90
. This finding coincides with
previous studies [Yildirim and Philippatos (2007), Andries (2011); Thangavelu and
Findlay, C. (2013)].
90
The average CAR of Indonesian commercial banks over the study period was 24.01 per cent (it was
higher than the minimum CAR imposed by the Indonesian bank authority of 8 per cent).
157
Meanwhile bank liquidity as represented by LDR has also a significant impact on
Indonesian bank cost efficiency with a positive association. It occurred impressively in
both approaches. These findings suggest the importance on bank liquidity as well as the
degree of lending exposure maintained by banks in enhancing cost efficiency. These
findings witnessed the success of Indonesian banks to manage their expenses when their
loan portfolio increased over the study period. Average LDR across all Indonesian
commercial banks during the period was 76.69 per cent. There is still a chance for
Indonesian commercial banks to enhance the finance proportion in their investment
portfolios since the average LDR during the period was less than 80 percent. It is also
important to state that an increase in lending exposure should always be accompanied
by the implementation of prudential principles in processing and monitoring the loans in
order to guarantee the performing loan quality. Loan quality management is very critical
for bank management.
NPL as a proxy for bank credit risk has also an impact on Indonesian bank cost
efficiency in both regressions, although the directions of associations are different. NPL
has a negative impact on bank cost efficiency in the first regression with a confidence
level of 95%, but it significantly affected bank cost efficiency in the second regression
with a positive sign. Based on the value of the coefficient of the regression of that
variable, it is obvious that the results in the first regression are stronger than those of the
second. These findings are different to the findings of Andries (2011), Thangavelu and
Findlay (2013) in that they did not find a significant association between non-
performing loans and bank efficiency.
It is also worth noting that from the macroeconomic indicators, INFL, GDPGR, and
USDR (USD rate in terms of IDR/Indonesian Rupiahs) are significant factors across the
two methods. INFL has a negative association with cost efficiency, while GDPGR and
USDR have positive associations with the cost efficiency of Indonesian banks. These
findings are very sensible, since if the inflation rises, bank would experience an increase
in costs or expenses, and it finally makes the bank less efficient, and vice versa.
GDPGR as the indicator of economic growth has a positive association with cost
efficiency. This means that the higher the economic growth in Indonesia, the
commercial banks in the country experience more cost efficient, and vice versa.
158
In general, it can be concluded that from the regressions of the model 1 across the two
methods examined, the variables LNTA, ROA, CAR, LDR, and NPL are bank specific
factors that significantly affected Indonesian bank efficiency over the period 2002-2010
at least in one regression. These results could also be used as an important
recommendation for all management of banks that the size of a bank, the capital
adequacy, the liquidity or the finance expansion indicator, and the credit risk
management are the important keys for enhancing efficiency. These findings are also
important for policy makers, particularly for the Indonesian commercial bank authority
in setting regulations regarding those significant factors91
.
Table 6.15 demonstrates the determinants of Indonesian banks‟ cost efficiency over the
model 2. The table witnessed ROA and LDR as the significant factors determining the
cost efficiency of Indonesian banks over the period 2002-2010. Meanwhile, INFL,
GDPGR, and USDR (of the macroeconomic indicators) also determine Indonesian bank
cost efficiency over the period. These results are derived from the pooled data method.
On the other hand, for the BC92 method, LNTA and LDR significantly determine the
cost efficiency of Indonesian banks over the study period. Meanwhile (of the
macroeconomic variables), all factors (INFL, GDPGR, UNEMP, and USDR) are
significant in predicting the cost efficiency of the Indonesian banks over the period.
Similar to model 1, LNTA as the proxy for the size of bank has been a significant factor
for Indonesian bank efficiency with a positive sign in the second regression (BC92
method). LNTA is significant for the second regression with a confidence level of 99%.
This result implies that the bigger the size of a bank, the bank will be more cost efficient.
This finding complies with the SCP hypothesis that if the size of a company‟s operation
increases, the company will benefit from economic of scale and exhibit more efficient.
This result also fits with the current policy of the Indonesian bank authority to merge
commercial banks in regards to establishing strong and sturdy banks and also be
resilient to crises. LNTA has also a positive association with cost efficiency in the first
regression (the Pooled data method) albeit it is not significant.
91
The policies that relate to those factors i.e. bank mergers and acquisitions, capital adequacy requirement,
and bank liquidity and risk management.
159
Tabel 6.15. Determinants of banks cost efficiency – Model 2 (1) (2)
VARIABLES cepool2 bc92_2
lnta 0.000920 0.0122***
(0.00209) (0.00186)
roa 0.308*** -0.00695
(0.0314) (0.0121)
car 0.0219 -0.00454
(0.0150) (0.00636)
ldr 0.0233*** 0.0102***
(0.00678) (0.00297)
npl -0.0806 -0.0182
(0.0732) (0.0290)
infl -0.00166*** -0.00321***
(0.000605) (0.000239)
gdpgr 0.00977*** 0.0346***
(0.00339) (0.00155)
unemp 0.00129 -0.00509***
(0.00191) (0.000746)
usdr 1.27e-05*** 3.33e-05***
(4.02e-06) (1.72e-06)
Constant 0.627*** 0.192***
(0.0561) (0.0282)
sigma_u 0.0368*** 0.102***
(0.00307) (0.00691)
sigma_e 0.0534*** 0.0201***
(0.00127) (0.000477)
Observations 999 999
Number of id 111 111
Standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
The dependent variables are: CEPOOL2: SFA cost efficiency scores
obtained by the standard pooled data method for Model 2; and BC92-2:
SFA cost efficiency scores obtained by Battese-Coelli 1992 method for
Model 2.
The explanatory variables are: LNTA is the natural logarithm of total
assets; ROA is return on assets; CAR is capital adequacy ratio; NPL is
non-performing loans; INFL is annual inflation rate; GDPGR is Gross
Domestic Product at real prices growth rate; UNEMP is annual
unemployment rate; USDRATE is United States Dollars (USD) exchange-
rate in terms of Indonesian Rupiahs (IDR).
Also, similar to the results of model 1, ROA affected bank cost efficiency in the first
regression (the pooled data method) with a positive sign. It implies the importance of
banks to obtain profit from the operation since the higher profitability of a bank would
lead that bank to be more cost efficient.
Different with the findings of model 1, CAR is not recorded as a significant factor
affecting Indonesian banks‟ cost efficiency during the period. These results are
160
interesting given the fact that in model 2 the output components were broken down to
be small business finance and other finance. This might be caused by the advantage of
the diversifications which lead banks‟ management not necessary to provide more
capital since the risk embedded on the small business finance could be considered lower.
LDR has impressively affected the Indonesian bank cost efficiency in both regressions
with a positive association. These findings imply the importance of lending exposure on
the enhancement of bank cost efficiency. In other words, the additional proportion of
credit or finance over the deposit could enhance bank cost efficiency. Average LDR
across all Indonesian commercial banks during the period was 76.69 percent. There is
still a chance for Indonesian commercial banks to enhance the finance proportion in
their investment portfolios since the average LDR during the period was less than 80 per
cent. It is also worth noting that banks should also maintain the quality of their finance
portfolios in order to achieve low non-performing loans (NPL less than 5%). The
average NPL for this period was 2.25% which is considered good. Unfortunately, NPL
was not recorded as significant factors affecting banks‟ cost efficiency in this model
although the direction of the association is in accordance with the hypothesis. The
association between NPL with cost efficiency is negative for both regressions.
Similar to the first model, from the macroeconomic variables, INFL, GDPGR, and
USDR (USD rate in terms of IDR/Indonesian Rupiahs) are significant factors across the
two approaches. INFL has a negative association with the cost efficiency, while GDPGR
and USDR have positive associations with the cost efficiency of Indonesian banks.
These findings are very sensible, since if inflation rises, banks would exert increasing
costs or expenses, and it finally makes the bank less efficient, and vice versa. GDPGR
as the indicator of economic growth has a positive association with cost efficiency. It
means that with a higher level of the economic growth, banks would also undergo the
higher efficiency due to the growing economic activities in the real sectors. Meanwhile
UNEMP has significantly affected the cost efficiency of the second approach (BC92)
with the negative sign.
It is concluded from the regressions of model 2, that the variable of LNTA, ROA, and
LDR are of bank specific factors that significantly affected the Indonesian bank cost
efficiency over the period 2002-2010 at least in one regression. These results lead to the
161
important recommendation either for banks‟ management as well as the policy makers
that the size of a bank, the profitability of a bank, and the liquidity or the finance
expansion indicator are the important keys for enhancing cost efficiency particularly for
the model considering the diversifications of the financing or loans.
Table 6.16 depicts the determinants of Indonesian bank cost efficiency over the model 3
when the model only accounts for other finance as one of the output and disregards
small business finance.
For the standard pooled data method, ROA, CAR and LDR are recorded as the
significant factors determining the cost efficiency of Indonesian banks over the period
2002-2010 from the bank specific factors. Meanwhile, of the macroeconomic indicators,
INFL, GDPGR, and USDR each has also an impact on Indonesian bank cost efficiency
over the period. On the other hand, for the BC92 method, it is verified that LNTA, LDR,
and NPL each has a significant impact on the cost efficiency of Indonesian banks over
the study period. While of all the macroeconomic indicators, all factors (INFL, GDPGR,
UNEMP, and USDR) each has an impact on the cost efficiency of the Indonesian banks
over the period.
LNTA has a positive impact on Indonesian bank cost efficiency with a positive sign in
the second regression (BC92 method). LNTA is significant for the second regression
with a confidence level of 99%. This result also supports the SCP hypothesis that the
company would experience more efficiency in operation when the size of the company
increases. LNTA has also a positive association with cost efficiency in the first
regression (the Pooled data method) although the association is not significant.
ROA has also an impact on the bank cost efficiency in the first regression (the pooled
data method) with a positive sign. This finding suggests the importance of the
enhancement of the ROA attainment since this will lead banks to be more cost efficient.
In the second regression (the BC92 method), ROA has also a positive association with
cost efficiency, but the relationship is not significant.
162
Tabel 6.16. Determinants of banks cost efficiency – Model 3 (1) (2)
VARIABLES cepool3 bc92_3
lnta 0.000466 0.00993***
(0.00229) (0.00169)
roa 0.254*** 0.00472
(0.0313) (0.0106)
car 0.0711*** 0.00193
(0.0153) (0.00559)
ldr 0.0137** 0.0115***
(0.00696) (0.00261)
npl 0.0109 0.0707***
(0.0733) (0.0255)
infl -0.00188*** -0.00323***
(0.000602) (0.000210)
gdpgr 0.0121*** 0.0357***
(0.00341) (0.00137)
unemp 0.00287 -0.00677***
(0.00190) (0.000655)
usdr 1.80e-05*** 3.44e-05***
(4.02e-06) (1.52e-06)
Constant 0.526*** 0.0588**
(0.0565) (0.0277)
sigma_u 0.0427*** 0.153***
(0.00339) (0.0103)
sigma_e 0.0530*** 0.0176***
(0.00126) (0.000418)
Observations 999 999
Number of id 111 111
Standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
The dependent variables are: CEPOOL3: SFA cost efficiency scores
obtained by the standard pooled data method for Model 3; and BC92-3:
SFA cost efficiency scores obtained by Battese-Coelli 1992 method for
Model 3.
The explanatory variables are: LNTA is the natural logarithm of total
assets; ROA is return on assets; CAR is capital adequacy ratio; NPL is
non-performing loans; INFL is annual inflation rate; GDPGR is Gross
Domestic Product at real prices growth rate; UNEMP is annual
unemployment rate; USDRATE is United States Dollars (USD) exchange-
rate in terms of Indonesian Rupiahs (IDR).
CAR has also affected Indonesian bank efficiency with a positive direction in the first
regression. This result implies the importance of capital in the cost efficiency attainment
when banks ignored small business finance in their portfolio financing. CAR has been
significant for the first regression with a confidence level of 99%. CAR also has a
positive association with cost efficiency in the second regression albeit not significant.
163
LDR has significantly affected Indonesian bank cost efficiency with a positive
association. This occurred impressively in both methods with a confidence level of 95%
on the first regression and 99% on the second. These findings mean that as much as the
finance exposure increases, the cost efficiency of a bank would also increase, and vice
versa. Average LDR across all Indonesian commercial banks during the period was
76.69 per cent. The average LDR of 76.69 percent is not satisfactory during the period
(less than 80%) and bank managers have the opportunity to increase the proportion of
LDR in regards to stimulate the financial intermediary function of commercial banks.
In addition, NPL has significantly affected bank cost efficiency in the second regression
(BC92 method), with a positive signed association. This result coincides with the
second regression of model 1.
From the macroeconomic indicators, INFL, GDPGR, and USDR are significant factors
across the two regressions. INFL has a negative association with cost efficiency, while
GDPGR and USDR have positive associations with the cost efficiency of Indonesian
banks. These findings are very sensible, since if the inflation rises, banks would exert
increasing costs or expenses that lead to be less cost efficient. GDPGR as the indicator
of economic growth has a positive impact on the banks cost efficiency. This implies that
higher level of the economic growth would lead Indonesian banks to be more cost
efficient. Meanwhile, UNEMP has affected bank cost efficiency in the second
regression with a negative association. This implies an increasing number of
unemployed in the country would lead commercial banks to be less cost efficient.
It is concluded from the regression results across the two methods in the model 2 that
the variables of LNTA, ROA, CAR, LDR, and NPL are of bank specific factors that
significantly affected Indonesian bank efficiency over the period 2002-2010 at least in
one regression. These results could also be used as an important recommendation for all
managers of banks that the size of a bank, the bank profitability, the capital adequacy,
the liquidity or finance expansion, and the credit risk management are the important
keys for enhancing efficiency. These findings are also important for policy makers,
particularly for the Indonesian commercial bank‟s authority to set the proper regulations
that are considered relevant to the factors (i.e. the policies of mergers and acquisitions,
capital adequacy requirement, etc).
164
In general, we can conclude that from all models examined, LDR has emerged as the
only one variable from bank specific factors that constantly affected Indonesian bank
cost efficiency over the study period. This implies that bank loan proportions is the most
important factor affecting cost efficiency of Indonesian banks over the study period.
Meanwhile, in terms of macroeconomic indicators, some variables which consistently
affected Indonesian bank cost efficiency across the three different models are inflation
rate (INFL), GDP Growth (GDPGR), and USD rate (USDR).
6.5. Conclusions
The average cost efficiency of Indonesian commercial banks are likely to increase over
the period 2002-2010. These findings are derived by three models (model 1, model 2,
and model 3) and two different efficiency estimation methods (the Pooled-data standard
method and BC92 method).
In general, FBs (foreign-owned banks) were constantly on the top position across the
three different models in terms of cost efficiency. They were the most efficient banks in
the two methods in model 1 (the Pooled data method and BC92), one method in model
2 (BC92) and one method in model 3 (BC92). SOBs (state-owned banks) experience the
most efficient in one method in model 2 (the Pooled data method), and JVBs (joint
venture banks) were the most efficient also in one method in model 3 (the Pooled data
method).
In model 2, when considering SBF (small business finance) as an output in the
calculation of efficiency, SOBs were the most efficient banks in Indonesia in one
method ( the Pooled data method). This finding strengthens the argument that SOBs
which have PT Bank BRI as the market leader of SBF were successful in employing
their SBF in enhancing their cost efficiency. However, when SBF was excluded (in
model 3), JVBs were the most efficient ones in the same method. CONV (conventional
banks) outperformed SHARIA (Islamic banks) except for the only in one method (BC92)
in model 2 when total business finance was broken down into small business finance
and other finance in the output components. These findings indicate the success of
165
conventional banks using their experience in the industry and confirm the importance of
small business finance experienced by sharia banks. FEB (foreign exchange banks)
were more efficient than Non FEB (Non foreign exchange banks) in all methods across
the three models except for one method (the pooled data method) in model 3.
Large banks (peer group 4) experienced the highest average cost efficiency under the
pooled method for models 1, 2, and 3 during the period 2002 to 2010. These results
reflect the success of the largest banks in Indonesia in managing their operation
benefited from the economies of scale. PT. Bank Mandiri, PT. Bank BNI, PT. Bank BRI
and PT. Bank BCA are those of the group members that are consistently marked sound
performance over the period. These results coincide with the findings of Shaban (2008),
Demir et al (2005) and Isik and Hassan (2003b) for the case of Egypt and Turkey. They
found that large banks were more efficient compared to other sizes of banks.
The findings of examining the determinants of Indonesian commercial banks‟ cost
efficiency over the period 2002-2010 reveal that in general, some variables which
consistently affected Indonesian bank cost efficiency across the three different models
during the period 2002-2010 are Loans to deposits ratio (LDR) as a proxy from the bank
specific factors, and inflation rate (INFL), GDP Growth (GDPGR), and USD rate
(USDR) as the proxies for the macroeconomic variables.
The implications of those results as considered suggestions for policy makers in
Indonesia are as follows: First; it is not a guarantee that privatisation is beneficial. The
case of Indonesian banks has indicated that SOBs (state-owned banks) are much more
efficient than the private ones. Second; Small business finance (SBF) is still very
important to be maintained by commercial banks, considering that it does not only have
the social effects, but also it can enhance the banks‟ cost efficiency. SOBs and Sharia
banks are of the cases that enjoy the success of enhancing efficiency by means of
providing small business finance in their portfolio financing. Third; Liquidity ratio
(LDR) has proven to be the most consistent factors that determine the cost efficiency of
Indonesian banks. These imply that the policies of managing adequate liquidity as well
as loan expansions are very beneficial in the improvement of bank performance in terms
of cost efficiency.
166
7. Chapter seven: The Lending Propensities of Indonesian Commercial Banks over
the Period 2002-2010
7.1. Introduction
This chapter specially discusses the lending propensity of Indonesian banks in terms of
the trend of the total loans (TL) and small business loans (SBL) experienced by
Indonesian commercial banks over the period 2002-2010. The outputs of the study
reveal the propensity of lending facilities provided by Indonesian banks in terms of their
ownership and their size. This study also provides an extensive analysis by exposing the
significant factors affecting bank lending propensity.
The study about the lending propensity of Indonesian banks is motivated by three
reasons. First, the topic of bank lending propensity has been currently one of the
important issues in banking and financial research [See McNulty (2013), Berger and
Black (2011), Rraci (2010) Laderman (2008), Frame et al. (2004), Akhavein et al.
(2004), Berger et al (1998, 2007)]. Second, lending or financing is one of the most
essential activities in Indonesian commercial banks considering that the Indonesian
economy is still very dependent on bank financing. Third, studies about the lending
propensity of commercial banks in emerging economies particularly in Indonesia are
still scarce to date. Hence, this is the first study to examine bank lending propensity in
Indonesia.
The analysis covers 109 commercial banks in Indonesia where these banks are
categorized as conventional banks over the period 2002-2010. Islamic banks (Sharia
banks) are excluded in this study considering that Islamic banks have different nature of
the financing scheme compared to the conventional ones92
.
This chapter is organized as follows: The first section discusses previous studies on
lending propensity in commercial banks and the importance of lending propensity in
commercial banks‟ activities. The second section presents the methodology adopted.
92
Islamic banks do not provide lending facilities with an interest instrument charged to their customers,
rather than using financing facilities with the scheme of profit or revenue sharing.
167
The third section reveals the empirical findings of lending propensity of Indonesian
banks. Finally, the last section concludes the chapter.
7.2. Previous studies
Empirical studies on examining bank lending propensity are growing fast in the
academic literatures. For details, see among others [Strahan and Wetson (1998), Peek
and Rosengren (1998), Goldberg and White (1998), Berger et al. (2001, 2007),
Akhavein et al. (2004), Laderman (2008), Shen et al. (2009), Berger and Black (2011),
McNulty (2013)].
The study about the lending propensity of commercial banks mostly discusses the
tendency of banks to provide small business loans as the main point of view. This is
applied due to the fact that the type of the loan is quite unique93
, hence not all
commercial banks are able to serve small businesses in the their provision of loans.
Strahan and Wetson (1998) examined the impact of bank size and level of complexity of
the organizational structure of the bank and the banking consolidation on bank ability to
provide small business loans. They employed sample dataset on commercial banks in
the U.S. during the period 1993 to 1994. The findings of their study reveal that the
amounts of small business loans has increased for both small banks and large banks
over the study period but the growth of the loans has tended to decrease for large banks.
The complexity of the organizational structure94
has nothing to do with a bank's small
business loans. Mergers and acquisitions (M & A) as a form of consolidation among
smaller banks have a positive influence on the increase in small business lending, while
the opposite result occurs when the M & A experienced by the large banks and small
banks.
93
It is rather different with commercial and corporate loans in terms of lending technologies applied.
Commercial and corporate loans usually apply financial statement analysis and credit scoring systems in
the loan evaluation process, whilst the majority of small business loans apply the relationship lending
technology in the process.
94
The complexity of organizational structure of the banks is reflected by the form of their business entity
whether they are a sole bank company or a holding bank company.
168
Peek and Rosengren (1998) evaluated the impact of bank mergers on the lending
propensity of small businesses in the US banking industry over the period 1993-1996.
Their research was motivated by a huge wave of bank mergers which occurred in the
US over the period 1985-1995. The findings suggest that for around half of merger
occasions, the small business lending of acquiring banks before mergers was higher than
those of the acquired banks and they tended to increase their small business lending
portfolio after mergers. Their findings also advocate that small acquiring banks have a
propensity to more actively lend to small businesses than those of the large acquiring
banks.
Regarding the wave of mergers which took place in the US banks within the 1990s
decade, Goldberg and White (1998) also raised concerns whether widespread
consolidation through mergers had an impact on the lower bank availability to lend to
small businesses. Their concerns were implemented by their study to make a
comparison between de novo banks‟ small business lending with those of similarly sized
incumbent banks95
. They employed the dataset from Federal Deposit Insurance
Corporation (FDIC) call reports over the period 1987-1994. The results show that the
small business lending portfolios of de novo banks‟ were significantly higher than those
of similar portfolios of incumbents. These results suggest the possibility of novo banks
to become one of the solutions to the problem of less availability of small business loans
from the consolidated banks.
Berger et al. (2001) investigated the effect of bank size and foreign ownership on
lending propensity to small businesses in Argentina. They employed 115 commercial
banks during the period 1998. They found that major and foreign-owned banks did
experience some difficulties in lending relationships with small businesses opaque.
These situations are caused by the higher costs incurred by banks. Banks spent more
expenses in lending relationships to small businesses‟ customers. In addition, the
circumstance was magnified by a delinquency of small businesses in the payment of
their debts. However, another finding suggests that the distressed banks did not
95
A de novo bank is a bank that has received a charter and started operations within am aximum of three
years from the year of observation. A bank that has been in operation for longer than three years from the
year of observation is treated as an incumbent (Goldberg and White, 1998). They focused on banks with
an asset size of $5 million to $100 million.
169
experience significant matters in lending to small businesses. This is supported by the
fact that small businesses are also likely to borrow from other banks to reduce their risks
of having a relationship with the distressed banks.
Another study exploring banks‟ small business lending was done by Akhavein et al.
(2004). They examined the lending propensity of small banks to small firms in the
agricultural industry in the US over the period 1987-1994. The study focused only on
small banks with an asset size of greater than $ 5 million and less than $100 million.
The findings of the study suggest that the relationships, represented by the length of
tenure of farm operators, had a positive impact on bank lending. De novo banks96
were
inclined positively to lend to small businesses. However, de novo banks experienced
some difficulties in obtaining new customers that had longer and stronger relationships
with other banks. Another finding suggests that for the category of small banks, small
business lending to small farms tended to decrease when bank size increased.
Rao et al. (2006) analysed the propensity of banks‟ lending to small scale industry in
India over the period 1992-2003. They also examined the impact of the size and the
performance of banks on loans to small scale industry in India. They employed the
dataset of 97 commercial banks which are classified into four groups: State Bank of
India and its associates, nationalised banks, foreign banks and other commercial banks.
Those banks are also widely categorized into three size classes (small, medium, and
large banks) based on the total assets as on March 31, 2003. The results reveal that loans
for small scale industry have a lower annual growth compared to loan growths of all
industries and all sectors over the study period (11.4 percent compared to 14.4 percent
and 16.3 percent, respectively). The findings also suggest that across different groups,
the growth rates of loans for small-scale industries are lower than those of non-small
scale industry. The findings also suggest that the size of a bank has a negative impact on
small scale industry loans for public-sector banks, while the opposite result occurred to
their peers from the private banks. In addition, the performance of banks represented by
the return on assets has a positive effect on small scale industry loans for public and
private banks albeit significant only for private banks.
96
They employ the definition from Goldberg and White (1998) that de novo banks are banks that operate
no more than three years of age.
170
Meanwhile, Laderman (2008) investigated whether the presence of a bank‟s branch
affect its propensity on small business lending in the United States. They employed data
on small business lending flow from the Community Reinvestment Act reports97
over
the year 2004 in the US. The findings suggest that the majority of small business
lending was coming from the banks which have a local branch. This reflects that the
proximity is advantageous (conducive) to lending. She also found that only around 10
percent of small business lending came from banks with no branch in the local market
and around half of these seemed to be from a bank with branches in the same country.
These results also support the role of proximity in small business lending. The last
findings reveal that the average size of the loans were almost similar between out-of
market and in-market98
and those were likely to be backed up by commercial real estate
as collateral.
Shen, et al. (2009) examined the impact of bank size on the lending propensity of
commercial banks to small medium enterprises (SMEs) in China. They employed a
dataset covering commercial banks of 79 counties in 12 provinces in China over the
period 2001-2004. The study found that the size of the banks do not have an impact on
SMEs lending. Some of the variables i.e. the relationship of wages and loan quality,
competition, and institutional arrangements significantly affect SME lending. In other
words, if loan quality is connected with a manager‟s wage, if there is a greater
competition, and if an institution has more self-loan approval right, then financing to
SMEs will be higher. In addition, law enforcement has also a positive impact on SMEs
lending.
Chernykh and Theodossiou (2011) examined the determinants of banks‟ propensity to
provide long-term business loans in Russia. They focused on long-term business loans99
since they assume that the total loans and in particular long-term loans are the main
97
The reports include data containing 5,000 banks delivered in accordance with the Community
Reinvestment Act (CRA). He used the CRA reports to only focus on small business loans (loans under $
1 million for businesses with revenue under $ 1 million).
98
She defines out-of-market as for banks that make loans in the market (metropolitan statistical area or
MSA) but do not have branches there. Conversely, in-market loans mean that the banks provide loans
with a physical presence (branches) in the MSA. 99
They define long-term business loans as loans with a term at least three years to maturity.
171
factors affecting economic development and economic growth in emerging markets,
including Russia. The study employed a cross-sectional dataset of 881 commercial
banks operating in Russia during the year 2007. The findings suggest that the Russian
banks‟ propensity to lend long-term to companies were influenced by their size, capital,
and the availability of their long-term liabilities (deposits) as their sources of funding.
Bank size, bank capital, and bank long-term liabilities had a positive impact on the
propensity of banks to make long-term loans in Russia throughout the year.
A recent study by Berger and Black (2011) shed light on the investigation whether the
current paradigm in the study of small business lending is applied for their observed
data. The paradigm suggests that small banks are benefiting by financing smaller, less
transparent businesses (employing „soft information100
‟), whilst large banks are likely to
finance relatively large companies because of their transparent information (using the
„hard information101
‟). They made a comparison between large and small banks‟
advantages in using their lending technologies. The results suggest that the predictions
of the current paradigm do not apply because large banks do not have the same benefits
of using hard technology102
in their loans than small banks. In addition, the advantages
of large banks do not increase monotonically following an increase in the size of the
bank103
. Other findings show that small banks have a comparative advantage in
relationship lending technology; however these kinds of advantages can also be
implemented to finance large companies.
The most recent study outlining the determinants of small business lending was
conducted by McNulty, et al. (2013). They examined the effect of bank size on banks‟
100
Soft information means qualitative information which is actually difficult to quantify and process by
loan officers i.e. the circumstances of the firm, the owners of the company, and its management. 101
Hard information means quantitative information processed by loan officers, i.e. financial ratios over
audited financial statements, collateral values, and credit scores. 102
The hard lending technology applied in this case is using financial statement as the main tools for
evaluating loan application (financial statement lending technology).
103
The assumption suggests that “large banks‟ comparative advantage in using hard-information lending
technologies should be monotonically increasing in the size of the firm. As firms increase in size, they
tend to have higher-quality financial statements, yielding an implied increasing advantage in hard
technologies.” (Berger & Black, 2011:725). Therefore, the results are contrary to the assumption.
172
propensity on small business loans in 5537 banks in the United States during the period
1993-2006. Their findings suggest that there was a negative relationship between bank
size and the proportion of small business loans. These imply that the tendency to lend to
small businesses has declined as the size of the bank increases. Small business loan
growth did not keep pace with the growth of the size of the bank. The results also show
that small banks were recorded to be more inclined to lend to small businesses.
7.3. The Importance of Banks’ Lending Propensities
The study about bank lending propensity is very essential to be investigated. The
findings of the study could be used as indicators of banks‟ willingness to finance all
business sectors (reflected by total loans) and small businesses (represented by small
business loans).
The discussion about bank lending propensity became much more interesting when
Berger et al. (2007) questioned the use of the small business loans to total assets
(SBLTA) as the proxy for representing bank‟s lending propensity on small businesses.
They doubt the majority findings of the previous research regarding the large banks
experienced lower ratio of SBLTA than those of their peers from small banks over the
study period. They argue that there was a possibility that large banks had a lower ratio
of SBLTA just because the denominator was expanded and not because of the
contraction in the numerator. In addition, they also explain that the expansion of the
denominator originated from some activities (i.e. making large business loans or other
investments) that magnify their assets, while at the same time, small banks were not
able to expand their loan portfolios on large credits due to the problems of the legal
lending limit or highly large corporate lending risks.
Corresponding to Berger et al.‟s opinion, McNulty et al (2013) provided the alternative
solutions to represent bank lending propensities issued by offering another ratio namely
small business loans to total loans and leases (SBLTLL) in their article. They suggest
that the lending propensities of a bank to small businesses is still very important issue to
date to explore considering that commercial banks‟ lending is still an important source
of supporting small businesses.
173
This thesis adopts six variables to represent the bank lending propensity of Indonesian
commercial banks. The first three variables reflect banks‟ total lending (total loans)
propensity [these are namely a natural logarithm of total loans (LNTL), total loans to
total assets (TLTA) and total loans to total deposits (TLTD)] and three other variables
represent bank lending propensity only on small businesses [namely natural logarithm
of Small business loans (LNSBL), small business loans to total assets (SBLTA) and
small business loans to total loans (SBLTF)104
].
7.4. Methodology
7.4.1. Demand and Supply Model of Small Business Finance
The main research question of this study is to examine the impact of bank size on the
supply of total loans (TL) and small business loans (SBL) provided by Indonesian
commercial banks. However, in economic theory, the supply and demand models are
mainly related to one variable, namely price. In other words, both the quantity
demanded and the quantity of the supply of loans, each of which is a function of a single
explanatory variable, called, price (McNulty et al., 2013). Since the propensity of banks
to provide those types of loans reflects the supply quantity of the total loans and total
small business loans on each, the models should take into account price variable.
McNulty et al. (2013) as well as Peek et.al (2003) argue that the demands for loans from
companies are also associated with the demands for the companies‟ products. Hence,
taking into account the aggregate demand factors, it is also important to control the loan
demand function. Therefore, in the demand model specification, loan demand by
businesses is a function of the price of the loans ( and economic variables that
affect the productivity of business entities or enterprises ( . The structural form
of the demand function can be written as follows:
(7.1)
104
The use of SBLTA and SBLTF is to address the previous studies undertaken by many authors
including Berger et al. (2007) and McNulty et al (2013).
174
Where: is a constant, are coefficients of the explanatory variables and is
the error term. For controlling the aggregate demand function, they proposed the
variables representing the economic factors ( : the number of establishments in
the bank‟s county domicile, the unemployment rate in the county, and the year-end
value for the industrial production index. They employed the data of county economic
factors adjusted with the data covered.
This study uses a similar model as above with the adjustment in the macroeconomic
factors employed. Since the dataset employed in this study covers all commercial banks
in the country of Indonesia, this study uses three variables that represent the Indonesian
macroeconomic conditions, namely (1) GDP Growth (GDPGR) as proxy for Indonesian
economic activity, (2) Inflation rate (INFL) as a proxy of price stability in Indonesia,
and (3) Unemployment rate (UNEMP)105
as a proxy for Indonesian prevalence for
unemployment. These variables are employed as the control variables for the function of
loan demanded by debtors as well as the supply of loans by banks. Peek et.al (2003)
state that using such variables i.e. GDP growth, inflation and unemployment rates, as a
part of forecast errors could be more effective ways to control for loan demand.
Referring to some previous studies106
examining the loans supply, it is specified that the
loan supply structural model is as follows (McNulty et al., 2013):
(7.2)
Where denotes the quantity of supplied loans (in this case, total loans, and small
business loans), is loan rate or price at time-t, is the size of each bank at time-
t, is the bank capital of each bank at time-t, and is the bank
deposit of each bank at time-t, and is the error term.
This study also employs a similar model with the dataset of the Indonesian banks as
follows is total loans (TL) and small business loans (SBL) provided by
commercial banks. Banks‟ propensity on total loans is represented by natural logarithm
105
Unemployment rate is calculated by dividing the number of unemployed individuals by all individuals
currently in the labor force. The data obtained from the Statistics of Indonesia. http://www.bps.go.id.
106
See for example: Berger et al. (1998), Kishan and Opiela (2000), Akhavein et al. (2004), Berger and
Black (2011), among others.
175
of total loans (LNTL), total loans total loans to total assets (TLTA) and total loans to
total deposits (TLTD) while banks‟ propensity to lend to small businesses is represented
the three variables, namely natural logarithm of small business loans (LNSBL), small
business loans to total assets (SBLTA) and small business loans to total loans
(SBLTF). Meanwhile, is the price of loans (lending rate data from the Central Bank
of Indonesia). The data of the price of loans are the average lending rates for different
groups of banks operated in Indonesia over the period 2002-2010. Since there is no
difference between the loan rates for the two types of loans, the data of the price of
loans are used for both models. Bank size ( ) is represented by the natural
logarithm of bank’ total assets; bank capital ( is represented by the bank‟
capital to total assets ratio; and bank deposit ( ) is represented by the natural
logarithm of bank’ total deposits.
7.4.2. Two-Stage Least Square (TSLS) Technique
Most of the previous studies examined use a two-stage regression technique of least
squares (TSLS) to estimate the supply of loans. This technique is taken as a response to
and treatment in accordance with the natural conditions around the model, namely the
simultaneity problem. The problem is characterized by the simultaneity condition due to
feedback or a simultaneous relationship between the supply of loans and price variables.
In other words, these two variables affect each other. The problem is also considered by
the econometric terms as endogeneity, because there is an endogenous variable (in this
case, price) between the explanatory variables in determining the supply of the loans
that have a high correlation with the error term. As a consequence, the resulting
estimates of OLS will be biased and inconsistent.
TSLS is applied to handle the problem thus there is no more endogeneity and the
estimation will be unbiased and consistent. Studenmund (2011) states that TSLS is the
most widely used technique to mitigate this problem. He added that the TSLS is
operated by running a regression on the reduced form on the right hand side endogenous
variable that needs to be replaced, and then the (the fitted values resulted from the
176
reduced-form regression) is used as an instrumental variable to replace the endogenous
variable in the main equation.
Murray (2006) stated the advantages of the TSLS method. He explained that the TSLS
estimator does not suffer measurement error bias due to two special natures of the
TSLS. Both of these properties are: 1) Fitted values resulted from OLS regressions are
always uncorrelated with the related residuals; 2) Each fitted value in TSLS is a linear
combination of explanatory variables from the first stage regression. Hence, each fitted
value in the second stage is not correlated with all miss measurement conditions caused
by employing fitted values replacing the actual values in the second stage of the
TSLS107
.
The example of the equations that reflect the TSLS can be observed as follows [Gujarati
(2006), Studenmund (2011), Wooldridge (2013)]:
(7.3)
(7.4)
We have one main equation in (7.3) that is a function of and . Since is
also affected by , thus there is a problem of simultaneity and become endogenous
variable as well as . To mitigate the problem, we regress the equation (7.4) until we
obtain the prediction value for , that is hese fitted values are then used as the
instrumental variable to replace in the main equation (7.3).
For the identification purposes, especially for the case of the loan supply equation (7.2),
we can identify that the variable of price ( seems to be endogenous variable in the
place of right-hand side of the equation. It occurs since is not only explaining
the , but also is explaining the . The mutual influences reflect that both are
endogenous variable. Then, as the consequence, we should find the instrumental
107
See Murray (2006:591) for more details.
177
variable to replace in the (7.2). The instrumental variable is originated by finding the
fitted values through the regression equation as follows:
(7.5)
From this regression of this equation, we can obtain the values of . Then, the (as
the instrumental variable) can be used in the equation (7.2) to replace the .
(7.6)
The regression estimation resulted from this equation would be unbiased and consistent
to mitigate the problem of endogeneity and simultaneity in this case [See McNulty et al.
(2013) and Gujarati (2006)].
7.5. Data
This study employs data of all commercial banks in Indonesia, which are provided by
the Central Bank of Indonesia and the data provided by PT. Ekofin Konsulindo with the
following criteria: (1) the sample data are all commercial banks, particularly
conventional banks operating in Indonesia during the period 2002-2010. (2) Each of the
banks‟ financial statements should be available completely within the period from 2002
to 2010. There have 109 banks to meet the criteria. It comprises of 4 state-owned banks
(SOB), 55 private banks (PB), 26 provincial / local government-owned banks (LGOB),
15 joint venture banks (JVB) and 9 foreign-owned banks (FB).
7.6. Empirical results
7.6.1. Descriptive Statistics
This section exhibits the descriptive statistics of the data sample employed to examine
the lending propensity of Indonesian commercial banks during the period 2002-2010. It
comprises two important parts. First, it describes the characteristics of some key
variables used to estimate the lending propensity of Indonesian commercial banks.
Second, it provides the classification of the data by peer group based on asset size. The
178
classification is very beneficial to provide more insights on the attitude of banks in
terms of their lending (finance) propensity based on the size of each bank during the
sample period.
7.6.1.1. Descriptive Statistics of the key variables.
The descriptive statistics of the key variables is presented in table 7.1. There is a wide
variation of banking operations in terms of asset size of commercial banks in Indonesia.
There is a bank that operates with only IDR 17.5 billion of assets whilst another bank
operates its asset up to IDR 408 trillion. An average asset size of Indonesian banks
included in this study is around IDR 16.9 trillion over the period.
Table 7.1. Descriptive statistics of the key variables
Variable N Mean Std. Dev. Minimum Maximum
Total Assets (TA) (in IDR Million) 981 16,100,000 44,900,000 17,540 408,000,000
Total Loans (TL) (in IDR Million) 981 8,231,294 22,700,000 1,229 247,000,000
Total Loans to Total Assets (TLTA) 981 0.5375 0.1852 0.0055 1.0810
Total Loans to Total Deposits (TLTD) 981 0.7866 0.4777 0.0090 3.4329
Small Business Loans (SBL) (in IDR Million) 981 997,652 4,437,158 0 75,400,000
Small Business Loans to Total Assets
(SBLTA) 981 0.0789 0.1015 0 0.6005
Small Business Loans to Total Loans
(SBLTL) 981 0.1552 0.1950 0 0.9994
Price of Loans (P) 981 0.1459 0.0211 0.1039 0.1891
Total Deposits (DEP) (in IDR Million) 981 12,500,000 36,200,000 1,503 333,000,000
Capital Ratio (CAPRATIO) 981 0.1419 0.1180 - 0.2749 1.0274
Source: Data observed
In terms of total loans achieved by the observed commercial banks in Indonesia over the
study period, they could achieve IDR 8.23 trillion on average, with ratios of total loans
to total assets (TLTA) and total loans to total deposits (TLTD) are 53.75 per cent and
78.66 percent respectively. The ratio of TLTD is also known as loans to deposits ratio
(LDR). The percentage achievement of TLTD or LDR reflects that Indonesian
commercial banks have managed well in operating the intermediary function over the
study period108
. This achievement could have also contributed to the average GDP
108
Bank Indonesia Regulation No. 15/7/PBI/2013 on the reserve requirement states that the LDR target of
each bank should reach 78-100 percent in position before December 1, 2013 and 78-92 percent on
December 2, 2013 and thereafter.
179
growth over the period which reached 5.39 percent.
It is important to explain that in terms of small business loans, some banks do not have
lending exposure to small businesses, whilst in reverse some banks put their lending
portfolios majority on small businesses. An average small business loan to total assets
(SBLTA) is 7.89 percent while the average small business loan to total loans (SBLTL) is
15.52 percent. It is worthwhile to state that the change of the banking regulation by
eliminating the minimum particular percentage threshold of SBLTL for each bank109
has
affected the ratio of SBLTL over the period 2002-2010.
Price of loans (P) is the average lending rate for different groups of banks imposed on
commercial banks in Indonesia during the period. The lending rate prevailed at the
commercial banks in Indonesia over the period was between 10 percent and 19 percent
with an average of about 14.59 percent. In addition, total deposits (DEP) collected by
commercial banks in Indonesia vary with an average of IDR 12.5 trillion. Capital ratio
(CAPRATIO) also ranges with an average ratio of 14.19 percent.
7.6.1.2. Distribution by Peer group
In accordance with the main purpose of this study which is to examine the impact of
bank size on the lending propensity of Indonesian commercial banks, the information
contained in the table 7.2 assists us to understand the condition of loan propensity of
Indonesian banks associated with their asset size.
Table 7.2 presents the distribution of some of the key variables across 4 (four) banks‟
groups classified by asset size. The classification of the banks is as follows: Peer group
1 comprises banks with the asset size of less than or equal to IDR 10 trillion (TA <=
109
The regulations about the minimum threshold of small business loans for each commercial bank in
Indonesia (20 per cent on the package of January 1990 and 22.5 per cent on The Directors of BI Decree
No. 30/4/KEP/DIR on April 1997) are then revoked by BI Regulation Number 3/2/PBI/2001 about Small
Business Finance on January 4, 2001. The regulation states that the provision of small business loans is
then readjusted based on the ability of each bank. In other words, since then, commercial banks are not
compelled to provide a minimum particular percentage of their finance portfolios devoted to small
businesses.
180
IDR10 trillion). Peer group 2 consists of banks with asset size of more than IDR 10
trillion and less than or equal to IDR 50 trillion (IDR 10 trillion < TA <= IDR 50
trillion). Peer group 3 includes the banks with asset size of more than IDR 50 trillion
and less than or equal to IDR 100 trillion (IDR 50 trillion < TA <= IDR 100 trillion).
The last, Peer group 4 covers banks with asset size of more than IDR 100 trillion. The
number of banks varies across the years since some banks changed their classifications
into different groups in subsequent years. However, as shown in the table below, as of
the year 2010, there are 65, 31, 6, and 7 banks, classified in the peer groups 1, 2, 3, 4,
respectively.
Table 7.2. Distribution by Peer group
Peer Group 1 2 3 4 TOTAL
Average Total Deposits (IDR Million) 1,621,552 15,887,459 49,176,734 166,000,000 232,685,745
Peer Group average deposits as a proportion of all
peer groups' average deposits 0.70% 6.83% 21.13% 71.34% 100.00%
Average Total Assets (IDR Million) 2,230,373 21,520,513 66,363,032 205,900,000 296,013,918
Peer Group average assets as a proportion of all
peer groups' average assets 0.75% 7.27% 22.42% 69.56% 100.00%
Average Total Loans (IDR Million) 1,221,311 11,386,730 38,976,264 98,463,083 150,047,388
Peer Group average TL as a proportion of all peer
groups' average TL 0.81% 7.59% 25.98% 65.62% 100.00%
Average Total Loans/Total Assets (TLTA) 0.5430 0.5235 0.5871 0.4786 0.5331
Average Total Loans/Total Deposits (TLTD) 0.8110 0.7352 0.7991 0.6003 0.7364
Average Small Business Loans (IDR Million) 162,812 1,042,671 5,001,519 13,191,781 19,398,783
Peer Group average SBL as a proportion of all peer
groups' average SBL 0.84% 5.37% 25.78% 68.00% 100.00%
Average Small Business Loans/Total Assets
(SBLTA) 0.0884 0.0488 0.0690 0.0668 0.0682
Average Small Business Loans/Total Loans
(SBLTL) 0.1743 0.0980 0.1182 0.1241 0.1287
Number of banks in Peer group (as of 2010) 65 31 6 7 109
Peer Group 1 (Total Assets <= IDR 10 Trillion); Peer Group 2 (Total Assets > IDR 10 Trillion and <= IDR 50Trillion);
Peer Group 3 (Total Assets > IDR 50 Trillion and <= IDR 100 Trillion); Peer Group 4 (Total Assets > IDR 100 Trillion).
Average total deposits collected by the largest banks (peer group 4) were accounted for
71.34% while the smallest banks (peer group 1) only accounted for 0.70%. Peer group 2
and 3 enjoyed average deposits of about 6.83% and 21.13% respectively. These figures
reflect that the concentration of deposit accounts of commercial banks in Indonesia is
located on the large banks.
Similar conditions are also reflected in terms of asset size. The proportion of average
181
assets was also dominated by the largest banks which accounted for 69.56% out of an
average asset of all peer groups. The smallest banks simply enjoyed average assets of
0.75%. These results indicate that large banks in Indonesia have greater opportunities to
use their resources in terms of assets to lend to their customers. However, the magnitude
of the tendency of banks to provide loans is depended on the willingness of their
management and management's perception of the risk of their lending portfolios.
In terms of total loans, the largest banks accounted for 65.62% of all average loans
provided by Indonesian commercial banks over the study period. It valued an IDR 98.46
trillion on average. Nonetheless, in terms of average TLTA and TLTD, the smallest
banks had a better portion of the total loans than those of the largest ones. The small
group accounted for an average of 54.30% and 81.10% for TLTA and TLTD, compared
to the largest peer which only accounted for 47.86% and 60.03%. It is worth noting that
in the perspective of the financial intermediary, the smallest banks performed superior
than those of other peers with an average TLTD achievement of 81.10%.
Meanwhile, in terms of small business loans, the average SBL of the smallest banks
accounted for only 0.84%t which is much lower than the largest banks which accounted
for 68%. These seem to be because of the different magnitude of their assets. However,
these results indicate that large banks in Indonesia also give contributions to the
development of small businesses by allocating their portfolio financing devoted to small
businesses. As known, a number of large banks employ their extensive network of
branches to increase their absolute value of small business loans. A state-owned bank
namely PT. Bank BRI, is one of the large banks that uses a large number of branches
throughout Indonesia to provide small business loans to its customers. In addition, PT.
Bank Danamon Indonesia, a private bank in the large category as well, has also
expanded their financing for small businesses by establishing a large number of small
lines of business namely Danamon Savings and Loans (DSP)110
. The combination
between the two contributors along with other banks has delivered the group to reach an
average SBLTL and SBLTA of about 12.41% and 6.68% respectively over the study
110
PT. Bank Danamon Indonesia through one of its lines of business, namely Danamon Savings and
Loans/ Danamon Simpan Pinjam (DSP), specifically to serve and help the development of micro and
small enterprises, especially those in the surrounding communities and the market, such as market traders,
retailers, and others. DSP was first established in 2004.
182
period. Of course, their achievement is still below the attainment of the smallest banks
in terms of SBLTL and SBLTA which recorded 17.43% and 8.84%. These results
coincide with the findings of Shaban et al (2014) that Indonesian large banks were
generally less inclined to provide small business loans over the study period.
In addition, the achievement of the smallest banks to have higher SBLTL and SBLTA
than those of their large peers indicate that on average, the smallest banks‟ propensity to
finance small businesses are higher than those of the large banks. These findings
coincide with the study results of McNulty, et al (2013) which suggests that small banks
have a higher propensity to lend to small businesses compared to those of the large
banks.
The evolution of total lending exposures (total loans) among different groups of banks
during the period 2002-2010 can be observed in some of the following figures (figure
7.1, 7.2, and 7.3):
Figure 7.1. Average TLTA across different groups
Source: Data processed
Figure 7.1 demonstrates the total lending propensity of commercial banks in Indonesia
in terms of TLTA over the period 2002-2010. It shows an upward trend of total bank
loans from under 50 per cent in 2002 to more than 60 per cent in 2010. The smallest
banks (peer group 1) posed at the top spot over the first five years, whilst peer group 3
overtook the position for the rest of the following years.
The large banks recorded the lowest performance on average, but the slope of their
credit curve looks steeper than their counterparts. The lowest attainment of their TLTA
-
0.1000
0.2000
0.3000
0.4000
0.5000
0.6000
0.7000
2002 2003 2004 2005 2006 2007 2008 2009 2010
Year
Total Loans to Total Assets (TLTA) by Bank Peer Group
Peer Group 1 (Total Assets<= IDR 10 Trillion)
Pee Group 2 (Total Assets >IDR 10 Trillion and <= IDR50Trillion)
Peer Group 3 (Total Assets >IDR 50 Trillion and <= IDR100 Trillion)
Peer Group 4 (Total Assets >IDR 100 Trillion)
183
in 2002 is due to fact that the year was of the beginning revival period after the Asian
financial crisis of 1997-1998. Meanwhile, the global crisis of 2008 did not appear to
significantly affect the achievement of loans in 2009-2010.
Figure 7.2. Average TLTD across different groups
Source: Data processed
Figure 7.2 shows the lending propensity of Indonesian banks in terms of total loans to
total deposits (TLTD) over the period 2002 to 2010. This figure is very important to be
evaluated since TLTD is as a representative for the function of banks‟ financial
intermediary111
. The figure shows the upward trend of the ratio meaning that the
intermediary function of the banks increased over the period. The smallest banks
recorded high willingness to provide loans to small businesses over the period compared
to their peers. Meanwhile, the largest banks experienced the lowest average ratio over
the period albeit the slope of the curve is steeper than those of other peers.
111
TLTD is similar with the loans deposits ratio (LDR), the most commonly used ratio term applied in
banking to reflect the financial intermediary functions of commercial banks. The banks‟ intermediary
functions are associated with activities of collecting funds (deposits) from depositors and granting credit
(loans) to borrowers.
-
0.1000
0.2000
0.3000
0.4000
0.5000
0.6000
0.7000
0.8000
0.9000
1.0000
2002 2003 2004 2005 2006 2007 2008 2009 2010
Year
Total Loans to Total Deposits (TLTD) by Bank Peer Group
Peer Group 1 (Total Assets <=IDR 10 Trillion)
Pee Group 2 (Total Assets >IDR 10 Trillion and <= IDR50Trillion)
Peer Group 3 (Total Assets >IDR 50 Trillion and <= IDR 100Trillion)
Peer Group 4 (Total Assets >IDR 100 Trillion)
184
Figure 7.3. Average TL across different groups
Source: Data processed
From the perspective of the average amount of the total loans across the groups, figure
7.3 illustrates that the largest banks showed a high increase of average loans over time
across the years, whilst other peers tended to keep maintaining the loans constant in
magnitude. The increase in the nominal average loans of the large banks (peer group 4)
significantly contributed to the increase of their attainment on average TLTA and TLTD.
In addition, the evolution of small business loans achieved by a different group of bank
during the period 2002-2010 can be observed by the following figures (figure 7.4, 7.5,
and 7.6):
Figure 7.4. Average SBLTL across different groups
Source: Data processed
-
20,000,000
40,000,000
60,000,000
80,000,000
100,000,000
120,000,000
140,000,000
160,000,000
20
02
20
03
20
04
20
05
20
06
20
07
20
08
20
09
20
10
IDR
Mill
ion
Year
Average Total Loans (TL) by Bank Peer Group
Peer Group 1 (TotalAssets <= IDR 10Trillion)
Pee Group 2 (TotalAssets > IDR 10 Trillionand <= IDR 50Trillion)
Peer Group 3 (TotalAssets > IDR 50 Trillionand <= IDR 100 Trillion)
Peer Group 4 (TotalAssets > IDR 100Trillion)
-
0.0500
0.1000
0.1500
0.2000
0.2500
0.3000
0.3500
0.4000
2002 2003 2004 2005 2006 2007 2008 2009 2010
Year
Average Small Business Loans to Total Loans (SBLTL) by Bank Peer Group
Peer Group 1 (Total Assets <=IDR 10 Trillion)
Pee Group 2 (Total Assets > IDR10 Trillion and <= IDR 50Trillion)
Peer Group 3 (Total Assets > IDR50 Trillion and <= IDR 100Trillion)
Peer Group 4 (Total Assets > IDR100 Trillion)
185
Figure 7.4 illustrates that, in general, commercial banks in Indonesia tend to reduce their
financing portfolios to small businesses. A regulatory change in 2001112
appears to have
a significant impact on the propensity of Indonesian banks‟ lending to small businesses
with a downward direction during the period 2002-2010. In particular, the sharp decline
in small business lending occurs in peer groups 1 and 3 where they dropped their
proportion of small business financing to below 20 per cent after years of 2003-2004.
Fortunately, peer group 4 (largest banks) shows the opposite performance. They tend to
increase the proportion of small business financing to nearly 15 percent on average in
2010. It seems that they have considered the potential advantages of financing the small
business sector. In practice, one of the largest banks, PT. Danamon Indonesia
successfully formed a special business unit working on micro and small business loans
conducted in 2004. The efforts of the bank along with PT. Bank BRI and PT. BNI113
appear to have a considerable contribution to the attainment of the largest banks in
financing small businesses.
Figure 7.5. Average SBLTA across different groups
Source: Data processed
Similar patterns are shown by the above figure. In terms of SBLTA, the downward trend
experienced by peer group 1 and 3, while the opposite results achieved by peer groups 4
and 2.
112
BI Regulation Number 3/2/PBI/2001 about Small Business Finance on January 4, 2001.
113
PT. Bank Rakyat Indonesia (BRI) and PT. Bank Negara Indonesia (BNI) are two state-owned banks
that constantly provide small business loans in a sizeable proportion among the other large banks during
the period 2002-2010.
-
0.0200
0.0400
0.0600
0.0800
0.1000
0.1200
0.1400
0.1600
0.1800
0.2000
2002 2003 2004 2005 2006 2007 2008 2009 2010
Year
Average Small Business Loans to Total Assets (SBLTA) by Bank Peer Group
Peer Group 1 (Total Assets <=IDR 10 Trillion)
Pee Group 2 (Total Assets >IDR 10 Trillion and <= IDR50Trillion)
Peer Group 3 (Total Assets >IDR 50 Trillion and <= IDR 100Trillion)
Peer Group 4 (Total Assets >IDR 100 Trillion)
186
Figure 7.6. Average SBL across different groups
Source: Data processed
Meanwhile, in terms of the absolute amount of the average small business loans
achieved by commercial banks in Indonesia over the period 2002-2010, it appears that
there was a sharp rise shown by Peer Group 4. They reached the average SBL above
IDR 20 trillion in 2010. The SBL of the Peer Group 3 has fluctuated up to IDR 5 trillion
in 2010. SBF of banks in Peer group 1 and 2 are still in the range of about IDR 1 trillion
and less. These results make sense given the fact that small banks have a restricted
capacity of financing due to asset size. However, when assessed by the proportion of
SBF, the banks included in Peer Group 1 and 2 achieved higher SBF ratios than those of
the large banks.
7.6.2. Regression Results
This section presents the regression results obtained from the two-stage least squares
regression on the data of the Indonesian commercial banks to estimate the supply of
total loans (TL) and total small business loans (SBL) over the period 2002-2010. Fixed
effects panel data regression is employed in this study. The result of the Hausman-test
indicates that fixed effects are more appropriate to apply. Since the focus of the
discussion of this section is the determinants of the lending propensity of Indonesian
commercial banks, especially to investigate the association between asset size and total
loans as well as small business loans, the first regression is not described. However the
first regression results reveal that the F-statistics and t-statistics of the reduced form
equations are statistically different from zero, meaning that the shift variables are
-
5,000,000
10,000,000
15,000,000
20,000,000
25,000,000
2002 2003 2004 2005 2006 2007 2008 2009 2010
IDR
Mill
ion
Year
Average Small Business Loans (SBL) by Bank Peer Group
Peer Group 1 (Total Assets <= IDR10 Trillion)
Pee Group 2 (Total Assets > IDR10 Trillion and <= IDR 50Trillion)
Peer Group 3 (Total Assets > IDR50 Trillion and <= IDR 100 Trillion)
Peer Group 4 (Total Assets > IDR100 Trillion)
187
statistically significant.
The results of the regressions encompass the three models. Model 1, 2, and 3 present the
results when each uses gross domestic product growth (GDPGR), inflation rate (INFL),
and unemployment rate (UNEMP) as the shift variable for the demand function in the
reduced equations respectively.
Table 7.3 shows the regression results where the lending propensity of Indonesian
commercial banks are expressed by the natural logarithm of total loans (LNTL), the
ratio of total loans to total assets (TLTA), and the ratio of total loans to total deposits
(TLTD).
Table 7.3. TSLS Regression Results 1 MODELS: 1 2 3 1 2 3 1 2 3
VARIABLES: LNTL LNTL LNTL TLTA TLTA TLTA TLTD TLTD TLTD
p -13.53*** 1.221 -10.99** -3.657** 0.611 -4.752*** -9.729** 0.714 -13.26***
-5.12 -1.581 -4.576 -1.576 -0.516 -1.55 -4.247 -1.345 -4.251
lnta 0.558*** 0.765*** 0.593*** -0.0731** -0.0132 -0.0885*** 0.963*** 1.110*** 0.914***
-0.104 -0.0725 -0.0967 -0.0321 -0.0237 -0.0328 -0.0864 -0.0617 -0.0899
capratio -0.896** -0.0398 -0.749** -0.239** 0.00894 -0.302*** -1.647*** -1.040*** -1.851***
-0.374 -0.228 -0.344 -0.115 -0.0743 -0.116 -0.31 -0.194 -0.319
lndep 0.370*** 0.416*** 0.378*** 0.0663*** 0.0795*** 0.0629*** -1.132*** -1.100*** -1.143***
-0.0745 -0.0669 -0.0713 -0.0229 -0.0218 -0.0242 -0.0618 -0.0569 -0.0662
Constant 2.588 -3.434*** 1.554 1.230* -0.513** 1.677*** 4.562*** 0.299 6.002***
-2.118 -0.716 -1.896 -0.652 -0.234 -0.642 -1.756 -0.609 -1.762
Observations 981 981 981 981 981 981 981 981 981
Number of id 109 109 109 109 109 109 109 109 109
Standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Model 1 applied GDPGR (Gross Domestic Products Growth) as a shift variable for the demand function
Model 2 applied INFL (inflation rate) as a shift variable for the demand function
Model 3 applied UNEMP (unemployment rate) as a shift variable for the demand function
The Dependent variables are: the natural logarithm of total loans (LNTL), the ratio of total loans to total assets (TLTA), and ratio
of total loans to total deposits (TLTD) for the three consecutive models. All reflects the lending propensities of Indonesian commercial banks over the study period. The Explanatory variables are: Price of loans (P) is the average lending rate for different
groups of banks imposed at commercial banks in Indonesia during the period. LNTA is the natural logarithm of bank‟s total assets.
LNDEP is the natural logarithm of bank‟s total deposits. Capital ratio (CAPRATIO) is the ratio of bank‟s capital over bank‟s total assets.
188
The relationship between bank size and total loans is positive and significant in terms of
LNTL and TLTD as dependent variables. They are all significant at the 1% level. The
positive coefficient means that as bank size increases, the propensity of Indonesian
banking loans also increases. In terms of LNTL models, the positive coefficients on the
LNTA across the three models are below 1.00 (i.e. 0.56, 0.76, and 0.59 for the three
models respectively). This means that an increase in their total loans does not keep pace
with the growth size of the bank. These results reflect the growth in the bank size is
higher than the lending growth. These results confirm the findings of McNulty et al
(2013). Meanwhile, the association between asset size and the TLTA is negative and
significant in the two models (model 1 and 3). These results imply that the lower growth
of bank loans compared to the growth of asset size has led a relationship between these
variables to be negative.
The variable of bank deposits has also a significant impact on commercial banks‟ loans
tendency in Indonesia over the study period. In terms of LNTL and TLTA as the
dependent variables, the variable natural logarithm of deposits (LNDEP) has a positive
and significant relationship with LNTL and TLTA respectively. These results indicate
that the increase in total bank deposits has caused the total volume of loans to increase
as well. Similar results are also responding to the ratio of total loans to total assets.
However, a LNDEP coefficient of less than 1.0 in all models for the LNTL and TLTA
regression results means that the total increase from both variables is less than
proportional to the growth of the banks‟ deposits. The inverse findings occurred for the
TLTD models. The difference between the lending and deposit growths seems to be the
reason for the negative association between those variables.
In addition, the variable of capital ratio (CAPRATIO) is statistically significant with
negative signs in 7 out of 9 models presented in terms of total loans (LNTL, TLTA, and
TLTD variables). These findings mean as bank capital increases, the size of bank
lending decreases and vice versa. McNulty et al (2013) pointed out that if there is a
negative association found between the capital ratio and bank lending, it seems that
those banks are willing to pursue risky loans by increasing lending although the bank
capital decrease. These results confirmed the study findings of McNulty et al (2013) in
189
the LNSBL114
models.
Table 7.4 presents the regression results where the lending propensity of Indonesian
banks to small businesses are represented by the natural logarithm of total small
business loans (LNSBL), small business loans to total assets (SBLTA) and small
business loans to total loans (SBLTL).
Table 7.4. TSLS Regression Results 2 MODELS: 1 2 3 1 2 3 1 2 3
VARIABLES: LNSBL LNSBL LNSBL SBLTA SBLTA SBLTA SBLTL SBLTL SBLTL
p -52.94 2.737 -44.68 1.305* 0.0809 0.391 3.448** 0.0744 1.2
-36.25 -11.46 -33.22 -0.679 -0.217 -0.598 -1.391 -0.445 -1.214
lnta -1.496** -0.714 -1.380** -0.0239* -0.0411*** -0.0367*** -0.0116 -0.0590*** -0.0432*
-0.737 -0.526 -0.702 -0.0138 -0.00994 -0.0127 -0.0283 -0.0204 -0.0257
capratio 0.0843 3.316** 0.564 0.0374 -0.0336 -0.0156 0.0688 -0.127** -0.0617
-2.648 -1.649 -2.495 -0.0496 -0.0312 -0.0449 -0.102 -0.0641 -0.0911
lndep 2.317*** 2.490*** 2.343*** 0.0282*** 0.0244*** 0.0253*** 0.0107 0.000274 0.00376
-0.527 -0.485 -0.518 -0.00987 -0.00917 -0.00932 -0.0202 -0.0188 -0.0189
Constant 5.416 -17.32*** 2.042 -0.171 0.329*** 0.202 -0.34 1.037*** 0.577
-14.99 -5.189 -13.77 -0.281 -0.0981 -0.248 -0.575 -0.202 -0.503
Observations 981 981 981 981 981 981 981 981 981
Number of id 109 109 109 109 109 109 109 109 109
Standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Model 1 applied GDPGR (Gross Domestic Products Growth) as a shift variable for the demand function
Model 2 applied INFL (inflation rate) as a shift variable for the demand function
Model 3 applied UNEMP (unemployment rate) as a shift variable for the demand function
The Dependent variables are: the natural logarithm of small business loans (LNSBL), the ratio of small business loans to total assets
(SBLTA), and ratio of small business loans to total loans (SBLTL) for the three consecutive models. All reflects the lending propensities of Indonesian commercial banks over the study period. The Explanatory variables are: Price of loans (P) is the average
lending rate for different groups of banks imposed at commercial banks in Indonesia during the period. LNTA is the natural
logarithm of bank‟s total assets. LNDEP is the natural logarithm of bank‟s total deposits. Capital ratio (CAPRATIO) is the ratio of bank‟s capital over bank‟s total assets.
In terms of LNSBL, SBLTA, and SBLTL as dependent variables, the majority findings
of the three models (Model 1, 2 and 3) specify that the variable of asset size has a
significant impact on small business loans with negative sign. The coefficient of
negative means that as asset size increases, the size of small business loans decreases.
114
Natural logarithm of small business loans (LNSBL).
190
The results supported the figures above concerning that the propensity of Indonesian
commercial banks to lend to small businesses has decreased over the period. It seems
that the change of the regulation to eliminate minimum threshold of small business
loans of each bank in 2001115
has reduced the propensity of Indonesian banks to lend to
small businesses.
Negative coefficient of greater than 1.0 for both models in the LNSBL regressions
(model 1 and model 3) indicate that as banks increase in asset size, their total volume of
small business finance decreases with a magnitude higher than their increasing level of
the asset size. However, in terms of SBLTA and SBLTL, the negative coefficients are
not large enough although they are significant in 5 out of 6 models presented. The
negative association between bank size and small business loans is similar with the
study findings of McNulty et al (2013) especially for a model of SBLTLL116
.
All models show a positive and highly significant117
relationship between LNDEP and
small business loans in terms LNSBL and SBLTA as dependent variables. This suggests
that the increase in total deposits is associated with an increase in small business loans.
A high relationship is reflected in the models of LNSBL with coefficients above 1.0.
These results also confirm the findings of McNulty et al (2013) in the model of LNSBL.
The relationship between LNDEP with SBLTL is also positive in all three models
although not significant.
In addition, the association between capital ratios (CAPRATIO) with the small business
loans is not significant in the majority models except for model 2 of the LNSBL and
SBTL. These findings also confirmed the study results of McNulty et al (2013).
7.7. Conclusions
This chapter investigates the effect of bank size and other variables on the lending
propensity of Indonesian commercial banks during the period 2002-2010. Proxies for
115
BI Regulation Number 3/2/PBI/2001 about Small Business Finance on January 4, 2001.
116
SBLTLL stands for small business loans to total loans and leases.
117
It is significant at 1% level.
191
banks‟ lending propensity are the total loans (LNTL, TLTA, and TLTD) and small
business loans (LNSBL, SBLTA, and SBLTL). Two-stage least square regression with
fixed-effects is adopted to overcome the simultaneity problem caused by the
relationship between supply and demand equations that are both associated with or each
is a function of the price variable.
In the section of descriptive statistics, it is explained that in terms of the average amount
of total loans (TL), the largest banks (peer group 4) dominate the share of loans which
reached 65.62% of average loans granted by commercial banks in Indonesia over the
study period. However, in terms of average TLTA and TLTD, the smallest group (peer
group 1) has a better portion of the total loans than those of the largest one. The group
accounted for an average of 54.30% and 81.10% for TLTA and TLTD, compared to the
largest banks which only accounted for 47.86% and 60.03% respectively. It is worth
noting that in the perspective of the financial intermediary, the smallest bank group is
better than other peers by achieving an average of 81.10% TLTD.
In terms of the average amount of small business loans (SBL), the largest bank group
(Peer group 4) also dominates the share of small business loans across all peer groups
which accounted for 68% compared to the smallest peers which only accounted for
0.84% on average. This large amount of SBL achieved by the largest group is the result
of the efforts of some large banks that use an extensive network of branch offices
throughout Indonesia i.e. PT. Bank BRI, PT. Bank BNI and PT. Bank Danamon
Indonesia). However, in terms of the ratio of small business loans (SBLTA and
SBFTL), the smallest bank group (peer group 1) exceeds the largest bank group with the
attainment of 8.84% and 17.43% for SBLTA and SBLTL (compared to the largest
banks which only reach 6.68% and 12.41% respectively). These results indicate that on
average, the propensity of small banks to lend to small businesses are higher than those
of the large banks. These findings are consistent with McNulty, et al (2013).
In general, by observing the evolution of total lending exposures (total loans) among
different group of banks over the period 2002-2010, it is concluded that all commercial
banks tended to increase the average loan exposures to their customers. The upward
propensity of total commercial bank loans in Indonesia indicate the revival period has
192
happened after hitting by the Asia financial crisis in 1997-98. This also means that the
financial intermediation function of commercial banks in Indonesia operated well
during the period.
On the other hand, from the perspective of the development of small business loans, it is
concluded that commercial banks in Indonesia tend to reduce their portfolios to small
business loans during the period. A regulatory change in 2001118
appears to have a
significant impact on the trend of Indonesian bank loans to small businesses with a
downward direction during the period 2002-2010.
Meanwhile, on the regression results, which employs six variables: LNTL, TLTA, and
TLTD (each as a proxy for total banks‟ lending propensities) and LNSBL, SBLTA, and
SBLTD (each as a proxy for banks‟ lending propensities to small businesses), it is
concluded that the relationship between bank size and total loans is positive and
significant in terms of LNTL and TLTD as dependent variables. They are all significant
at the 1% level. A positive coefficient means that as bank size increases, the propensity
of Indonesian banking loans also increases. However, the increase in their total loans
does not keep pace with the growth of bank size because the coefficient is under 1.0.
These results confirm the findings of McNulty et al (2013). Meanwhile, the relationship
between asset size and the TLTA is negative and significant in two models (model 1 and
3). These results seem to occur resulting from the lower growth of bank lending
compared to the growth of asset size has led to the relationship between these variables
is negative.
The variable of bank deposits also has a significant impact on the lending propensities
of commercial banks in Indonesia. In terms of LNTL and TLTA as the dependent
variables, the variable natural logarithm of deposits (LNDEP) has a positive and
significant relationship with LNTL and TLTA respectively. These results indicate that an
increase in the amount of bank deposits, the total volume of loans increases as well.
Similar results are also applied to the ratio of total loans to total assets. However, the
LNDEP coefficient of less than 1.0 in all models for the LNTL and TLTA regression
118
BI Regulation Number 3/2/PBI/2001 about Small Business Finance on January 4, 2001.
193
results means that the total increase from both variables are less than proportional to the
growth of bank deposits. Meanwhile, inverse findings occurred for the TLTD models.
The difference between loans and deposits growth seems to be the reason for the
negative relationship between the variables.
In addition, the variable of capital ratio (CAPRATIO) is statistically significant with a
negative sign on 7 of 9 models presented in terms of total loans (LNTL, TLTA, and
TLTD) as the dependent variables. These findings mean as bank capital increases, the
size of bank loans decreases and vice versa. McNulty et al (2013) pointed out that a
negative association between capital ratio and bank loans is caused by the banks that are
willing to pursue risky loans by increasing lending despite a decline in bank capital.
These results confirmed the findings of McNulty et al (2013) in the LNSBL119
models.
On the other hand, in the case of LNSBL, SBLTA, and SBLTL as the dependent
variables, the findings of the majority of the three models (Model 1, 2 and 3) determine
that the variable of asset size has a significant impact on small business loans with a
negative sign. The negative coefficient means that as the size of the asset increases, the
size of the small business loans declines. The results justified some figures concerning
small business loans (figure 7.4 and 7.5) that the propensity of Indonesian commercial
banks to lend to small businesses has declined over the period. A regulatory change to
eliminate the 20 per cent minimum threshold of small business loans for each bank in
2001120
seems to have reduced the propensity of Indonesian banks to lend to SBs.
The negative coefficient of greater than 1.0 for both models in the LNSBL regressions
(model 1 and model 3) indicate that when bank assets grow, the total volume of small
business loans decrease with a higher magnitude than the increase in asset size.
However, in the case of SBLTA and SBLTL, there is a significant negative coefficient
on 5 of the 6 models presented. The negative relationship between bank size and small
business loans is similar to the findings of McNulty et al (2013) especially for the
SBLTLL model121
.
119
Natural logarithm of small business loans (LNSBL).
120
BI Regulation Number 3/2/PBI/2001 about Small Business Finance on January 4, 2001.
121
SBLTLL stands for small business loans to total loans and leases.
194
All models showed a significant122
positive relationship between LNDEP and small
business loans in terms LNSBL and SBLTA as the dependent variable. This suggests
that the increase in total deposits is associated with an increase in small business loans.
A high relationship is also reflected in the model LNSBL with coefficients above 1.0.
These results also confirm the findings of McNulty et al (2013) in the model LNSBL.
The relationship between LNDEP with SBLTL is also positive in all three models
although not significant. While the relationship between bank capital ratios
(CAPRATIO) with small business loans is not significant in the majority of the models
except for model 2 of the LNSBL and SBTL. This finding also confirms the study
results of McNulty et al (2013).
The implication of the findings of the study as a suggestion for the policy makers in
Indonesia is as follows: First, Indonesia's banking supervisory agency123
should
continue to monitor the achievement of commercial banks in providing loans to their
customers so that economic activities in Indonesia operate as expected. However, such
monitoring should also include the development of credit quality given that low quality
of credit will not only reduce bank assets but also the bank's capital. Second, the
Indonesian government's decision to re-enact the minimum threshold of lending to small
businesses in 2012124
[although the new regulations is dedicated to small and medium
enterprises (SME)] is right, given that when the government relaxed small business loan
portfolios during the period from 2002 to 2010, the commercial banks tended to lower
the portion of the small business loans in their portfolios. By implementing the new
regulations, it is expected that the role of Indonesian commercial banks in providing
loans or financing for small businesses or SMEs, will continue to rise in support of the
economic development of Indonesia.
122
Significant at 1% level,
123
As stipulated by Indonesia‟s Government Law No. 21 of 2011 on the Financial Services Authority, as
of December 31, 2013, the task of banking regulation and supervision of the Bank Indonesia has been
transferred to the Financial Services Authority (FSA or OJK).
124
BI Regulation No.14/22/PBI/2012 on 21 December, 2012 regarding the credit or financing for micro,
small, and medium enterprises by commercial banks and technical assistance for the development of these
businesses. This regulation imposed on all commercial banks in Indonesia to finance SMEs gradually up
to reaching 20 percent in the end of 2018.
195
8. Chapter eight: Conclusion
This chapter summarizes the empirical findings obtained in the pursuit to answer the
research questions identified in Chapter 1 of this thesis. This thesis applied two
prominent methodologies, namely the Stochastic Frontier Analysis (SFA) and Data
Envelopment Analysis (DEA) to estimate the efficiency of Indonesian banks during the
period 2002-2010. The thesis also investigated the impact of bank size and other
variables on Indonesian banks‟ propensity to provide lending to their borrowers.
To the best of my knowledge, this is the first study that applies the two methodologies:
non-parametric DEA and parametric SFA for estimating Indonesian commercial banks‟
efficiency, especially using bank production function through DEA and cost function
through SFA over the study period. The thesis also examines bank specific factors and
macroeconomic factors that determine the efficiency of commercial banks in Indonesia.
The thesis also contributes to the literature by examining small business finance as one
of the output components in estimating bank efficiency in one of the models presented.
The last, this thesis is also the first in the literature that evaluates the lending propensity
of Indonesian banks over the period 2002-2010.
This chapter is organized as follows: the first section provides a summary of the
findings obtained from the three empirical chapters. The next section provides a number
of policy implications that might be of considerations for policy makers. Finally, the last
section provides conclusions with the limitations and recommendation for future
research.
8.1. Empirical Findings
In general, the efficiency of Indonesian commercial banks in terms of DEA technical
efficiency tended to decline over the period 2002-2010. The decline tendency is justified
by the declining records during the years of 2006, 2008 and 2009. The global financial
crisis of 2008 seems to have a little effect on the performance of the Indonesian banking
sector especially in terms of DEA technical efficiency as well as SFA cost efficiency.
These results coincide with the findings of previous studies carried out by Suzuki and
196
Sastrosuwito (2011) and Hadad et.al (2011).
Across all groups of banks by ownership, in general, state-owned banks (SOB)
experienced the most efficient banks in the majority of the models presented in the DEA
efficiency section. These results also confirmed the findings of Suzuki and Sastrosuwito
(2011) and Hadad et al (2011). These results suggest that SOBs were financially more
resilient in facing the global crisis of 2008. The results also imply the traditional
reliance from Indonesian households to SOBs as the most trustworthy place for
depositing their money across the period, similar to the condition which occurred during
the Asian financial crisis of 1997-98125
. The exemption was only happened on one
model when the outputs neglect small business finance, where the foreign owned banks
(FOB) overtook the top position ahead of the SOBs.
Meanwhile, in terms of the different operational system offered to customers,
conventional banks (CONV) are more efficient than Islamic banks (SHARIA) in the
majority of the models examined. These results indicate the success of the banks in
employing their long experience to operate more efficiently during the period. However,
Islamic banks could outperform conventional banks in the model two where the
efficiency estimation takes into accounts small business finance as one of the important
outputs in operation. These results also strengthen the findings that Islamic banks have a
higher propensity to finance small businesses compared to their peers from conventional
banks.
Across all models examined, it is found that that the productivity growth of Indonesian
commercial banks tended to decrease over the study period. These results were
empowered by the decrease over the following periods: 2002/2003, 2007/2008, and
2008/2009. The decline for the period of 2002/2003 occured due to the fact that
commercial banks were still in the beginning years of the recovery period after suffering
from the Asian crisis of 1997-98, whilst the consecutive declines over the period
2007/2008 and 2008/2009 seem to be caused by the effects of the global financial crisis
in 2007/2008. The decline of Indonesian banks‟ productivitivty growth over the periods
125
The crisis of 1997-1998 testifies the huge movement of funds from private commercial banks to state-
owned banks in Indonesia during the period.
197
coincides with the findings of the previous study (Suzuki and Sastrosuwito, 2011). In
addition, productivity growth of Indonesian banks over time during the period 2002-
2010 was mainly caused by technological change and there was no efficiency change
over the period. The findings were evidenced by the consistent findings across the three
models for the Malmquist Index Summary of annual means.
It is obtained from DEA efficiency regressions, that across all models examined, the
variables of LNTA, CAR, and LDR were of the bank-specific variables that consistently
affected the Indonesian bank efficiency over the period 2002-2010. All three variables
have positive signs associated with banks‟ efficiency. These imply that the increase in
assets, capital, and the volume of loans would definitely enhance the efficiency of the
Indonesian banks. These findings could also be used as an important recommendation
for all the banks‟ management that they should highlight those factors (the size of bank,
the capital adequacy, and the bank liquidity/the bank lending exposure) in enhancing
efficiency. The policies of merging these banks, managing their adequate capital, and
managing bank liquidity/lending exposure seem to be important factors to enhance
Indonesian banks‟ efficiency. In addition, all macroeconomic indicators included in this
study [INFL (inflation rate), GDPGR (Gross Domestic Product Growth), UNEMP
(unemployment rate) and USD RATE (USD rate)] significantly affected Indonesian
bank efficiency during the study period. These suggest the importance of the
government of Indonesia to maintain the stability of the macroeconomic indicators to
guarantee the efficiency of Indonesian commercial banks to be in the best level.
In the case of the SFA cost efficiency estimation, the results of the study show that in
general, the average cost efficiency of Indonesian commercial banks across the three
majority models tended to increase over the period 2002-2010. Foreign-owned banks
(FB) appeared to be the most efficient banks across the three different models in terms
of cost efficiency. They were most efficient in the two methods in model 1 [Pooled Data
and Battese and Coelli 1992 (BC92)], one method in model 2 (BC92) and one method
in model 3 (BC92). Meanwhile, state-owned banks (SOB) and joint venture banks (JVB)
have been the most efficient ones in only one method respectively (model 2 of the
Pooled Data method for SOB and model 3 of the Pooled Data method for JVB). In
addition, when considering small business finance (SBF) as one of the outputs in the
198
estimation of efficiency, SOBs were the most efficient banks in Indonesia in the Pooled
Data method. Meanwhile, when SBF was excluded (in model 3), JVBs were the most
efficient ones in the Pooled Data method as well. Conventional banks are more efficient
in almost all models compared to Islamic banks (SHARIA) except in model 2 of the
BC92 method where the model considered small business finance.
From the regression results of the SFA cost efficiency models, it is found that in general,
LDR was the only one from the bank specific factors which consistently affected
Indonesian banks‟ cost efficiency over the period 2002-2010. These findings could
become some important recommendation for all the banks‟ management in Indonesia
that bank lending exposure or bank liquidity is the most important factor in enhancing
bank cost efficiency. Again, the policies of managing bank lending portfolio and or bank
liquidity seem to be the crucial aspects to enhance Indonesian banks‟ cost efficiency.
Meanwhile, of the macroeconomic indicators, inflation rate (INFL), GDP Growth
(GDPGR), and USD rate (USDR) were also factors that significantly impacted on
Indonesian banks‟ cost efficiency over the period.
The findings of the study on examining the impact of bank size and other variables on
Indonesian banks‟ lending propensity over the period 2002-2010, show that Indonesian
commercial banks‟ lending tended to increase over time during the study period. The
upward propensity of Indonesian banks‟ total loans implies that the recovery period post
the Asia financial crisis of 1997-98 has occurred during the study period. This also
means that the function of bank as financial intermediary has begun to operate well
during the period (particularly in terms of the lending provision). In reverse, the
Indonesian banks‟ lending propensity in terms of small business loans tended to
decrease over time during the study period. A regulatory change in 2001126
appears to
have a significant impact on the declining trend of Indonesian banks‟ loans to small
businesses over the period. It is worth noting that government regulations on banking
would significantly affect banks‟ management in Indonesia. However, this controversy
policy adopted over the 2001-2012 was rectified by the new regulation issued in 2012
where all commercial banks in Indonesia should be returned back to finance small
126
BI Regulation Number 3/2/PBI/2001 about Small Business Finance on January 4, 2001.
199
businesses as they are considered one significant element of micro, small and medium
enterprises127
.
In addition, the findings of the lending propensity regression models, which employ six
variables: LNTL, TLTA, and TLTD128
(each as a proxy for total banks‟ lending
propensity) and LNSBL, SBLTA, and SBLTD129
(each as a proxy for banks‟ lending
propensity to small businesses), it is found that the relationship between bank size and
total loans is positive and significant in terms of LNTL and TLTD as dependent
variables. They are all significant a 99% confidence level. Positive coefficient of the
variable means that as bank size increases, the propensity of Indonesian banking loans
also increases. However, the increase in their total loans was not similar in magnitude
with the growth of bank size because the coefficient is under 1.0. These results coincide
with the findings of McNulty et al (2013) in the case of the US banks. Meanwhile, the
relationship between asset size and the TLTA is negative and significant in two models
(model 1 and 3).
The variable of bank deposits has also a significant impact on the lending propensity of
commercial banks in Indonesia. In terms of LNTL and TLTA as the dependent variables,
the variable natural logarithm of deposits (LNDEP) has a positive and significant
relationship with LNTL and TLTA respectively. These results imply that an increase in
the amount of the banks‟ deposits caused the total volume of their loans to increase as
well. Similar results are also occurred to the ratio of the total loans to total assets.
However, the LNDEP coefficient of less than 1.0 in all models for the LNTL and TLTA
regression results means that the total increase in both variables (LNTL and TLTA) are
less than proportional to the growth of the bank deposits. Meanwhile, the opposite
results are found for the TLTD models. The difference between the growth of loans and
deposits seems to be the reason for the negative relationship between the variables.
127
BI Regulation No.14/22/PBI/2012 on 21 December, 2012 regarding the commercial banks‟ loans or
financings for micro, small and medium enterprises (MSMEs), has compelled the Indonesian commercial
banks to provide minimum 20% of their financing allocated to MSMEs. See more details in the chapter
two.
128
LNTL (natural logarithm of total loans), TLTA (total loans to total assets), and TLTD (total loans to
total deposits).
129
LNSBL (natural logarithm of small business loans, SBLTA (small business loans to total assets), and
SBLTD (small business loans to total deposits).
200
In addition, the variable of capital ratio (CAPRATIO) is statistically significant with a
negative sign on 7 of 9 models presented in terms of total loans (LNTL, TLTA, and
TLTD) as the dependent variables. These findings mean as bank capital increases, the
size of bank loans decrease and vice versa. McNulty et al (2013) pointed out that a
negative association between capital ratio and bank loans is caused by the banks that are
willing to pursue risky loans by increasing lending despite a decline in bank capital.
These results confirmed the findings of McNulty et al (2013) in the LNSBL130
models.
On the other hand, in the case of LNSBL, SBLTA, and SBLTL as the dependent
variables, the majority findings of the three models (Model 1, 2 and 3) demonstrate that
the variable of asset size has a significant impact on small business loans with a
negative sign. The negative coefficient means that as the size of the asset increases, the
size of the small business loans decline. The results justified some figures concerning
small business loans (figure 7.4 and 7.5) that the propensity of Indonesian commercial
banks to lend to small businesses has declined over the period. A regulatory change to
eliminate the minimum threshold of small business loans for each bank in 2001131
seems to have a significant impact on the declining propensity of Indonesian banks to
lend to small businesses.
The negative coefficient of greater than 1.0 for both models in LNSBL regressions
(model 1 and model 3) indicate that when bank assets grew, the total volume of small
business loans decreased with a higher magnitude than an increase in asset size.
However, in the case of SBLTA and SBLTL, there is a significant negative coefficient
on 5 of the 6 models presented. The negative relationship between bank size and small
business loans is similar to the findings of McNulty et al (2013) especially for the
SBLTLL model in the case of the US banks132
.
130
Natural logarithm of small business loans (LNSBL).
131
BI Regulation Number 3/2/PBI/2001 about Small Business Finance on January 4, 2001.
132
SBLTLL stands for small business loans to total loans and leases.
201
All models show a significant133
positive relationship between LNDEP and small
business loans in terms of LNSBL and SBLTA as the dependent variables. These
findings suggest that the increase in total deposits is associated with an increase in small
business loans. A high relationship between those variables is also reflected in the model
LNSBL with the coefficients above 1.0. These results also coincide with the findings of
McNulty et al (2013) in the model LNSBL. The relationship between LNDEP with
SBLTL is also positive in all three models albeit not significant.
Meanwhile, the association between capital ratios (CAPRATIO) with small business
loans is not significant in the majority of models except for model 2 of the LNSBL and
SBTL. These findings also coincide with the study results of McNulty et al (2013) for
the case of the US banks. Meanwhile, the relationship between bank capital ratios
(CAPRATIO) with small business loans is not significant in the majority models except
for model 2 of the LNSBL and SBTL. The findings coincide with the study results of
McNulty et al (2013).
8.2. Policy implications
The findings of this thesis imply that the policies undertaken by the Government of
Indonesia (GOI) has succeeded in increasing the cost efficiency of Indonesian banks but
failed to enhance their technical efficiency over the period. These achievements are
stemmed from the fact that the period was the revival phase after suffering from the
Asian financial crisis of 1997-98. It is also worth noting that the majority of commercial
banks in Indonesia seem to have implemented the cost efficiency programme in their
operations as the internal management policies to avoid failure in their operations post
the crisis period.
Meanwhile, in terms of the input-output process in the operations, it seems that the
majority of commercial banks were successful in collecting funds but failed to enhance
earning assets. The prudential policies implemented in their lending activities (reflected
by the low attainment of loans deposits ratio), although it has increased over time during
133
Significant at 1% level,
202
the period, it could not offset the attainment of funding activities. The results of the
productivity study of the DEA Malmquist index supported the evidence that the
productivity of Indonesian commercial banks tended to decrease over time during the
period.
The policy of GOI released in 2001 which relaxed the compulsory threshold of a
minimum particular percentage of banks financing to small business finance has also
affected the performance of Indonesian banks over the period. This policy has affected
the downward trend (declining path) of small business loans in commercial banks in
Indonesia over the period 2002 to 2010. Although the total lending propensity (in terms
of the total loans) has increased, it could not have enhanced the technical efficiency of
the banks over the period.
The GOI policy to empower state-owned banks (SOB) by merging some particular
banks after the crisis period is also successful in maintaining customer trust. The policy
has also resulted that SOBs is the most efficient banks in terms of technical efficiency,
although in terms of cost efficiency, foreign-owned banks (FOB) performed better.
Some particular reasons that could be traced to SOBs in terms of their cost efficiency
performance are that the overstaffing, the highly fixed assets employment and the
market segmentations. These factors as well as any other factors (endogenous and
exogenous) should be importantly considered by GOI to enhance the performance of the
SOBs. However, the high performance of the SOBs in terms of their technical efficiency
should be maintained over time in the future.
In addition, the importance of small business lending (finance) should be of
considerations to the policy makers given the fact that in some cases explained in this
thesis the involvement of small businesses has increased the attainment of bank
efficiency. The performance of SOBs to outperform all other bank groups in all majority
models in DEA and in one model in SFA (when considering small business finance)
witnessed the importance of the small business finance in the banks‟ financing
portfolios. The success of Sharia banks to outperform conventional banks in the model
of taking into account small business finance in the output components is also very
interesting to be considered by the policy makers.
203
On the other hand, although foreign banks (FB) did better performances in terms of cost
efficiency, they failed to be the best performer in the technical efficiency. The reluctance
to expand their business in providing small business lending seems to be one of the
major reasons to support the circumstances. Lack of experience in small business
lending technologies as well as lack of expertise in dealing with customers in the
segment of small businesses seem to be the main factors they have not deeply involved
in the financing of small businesses. The collaborations with some particular domestic
banks that have successful experiences in managing small business loans can be one of
the solutions to begin providing lending facilities to this segment in their portfolios.
Similar circumstances are also applied to joint venture banks (JVB).
One of the influential factors for enhancing bank technical efficiency performance is
bank size. The findings of the study in this thesis reveal the positive relationship
between bank size and efficiency. It seems that the policy of Bank Indonesia in the new
architecture of Indonesian banking to establish a number of anchor banks that could
compete in the global market is relevant with the findings of the study. Establishing the
large size of banks through mergers and acquisitions seem to be the alternative solution
to reach the objectives. This is most appealing given the fact that Indonesia would face
the ASEAN free trade area in 2015. Two other important factors in determining bank
technical efficiency are loans to deposits ratio (LDR) and capital adequacy ratio (CAR).
Meanwhile, in terms of cost efficiency, the most important determinant is loans to
deposits ratio (LDR). These imply the importance of policy makers (banking
supervisory and authority board) to set policies regarding capital adequacy management
as well as lending exposure management to guarantee the sound performance of all
commercial banks in Indonesia.
In addition, in terms of enhancing bank productivity, the findings of the study reveal
that technological change implemented in commercial banks has supported the
productivity level of Indonesian commercial banks during the study period. This implies
the extensive new technologies (including the implementation of cash machines, phone-
banking, internet banking, etc.) application has a significant contribution to the attained
level of productivity of Indonesian commercial banks. Considering the fact, it is very
advantageous for bank managers to expand their operations by implementing new
technologies that could affect to the enhancement of their productivity.
204
From the last empirical chapter examined, there are two recommendations to policy
makers regarding the lending facilities provided by Indonesian commercial banks. First,
Indonesia's banking supervisory board134
should continue to monitor the achievement of
commercial bank lending to support the economy of Indonesia. The monitoring process
should also consider the quality of each lending facility given that the low quality of
credit will not only reduce a bank's assets but also a bank's capital. Second, the
Indonesian government's decision to re-enact the minimum threshold of lending to small
businesses in 2012135
[although the new regulations is dedicated to micro, small and
medium enterprises (MSMEs)] is absolutely important, to support the enhancement of
commercial banks‟ involvement in providing loans or financing to small businesses as
well as micro and medium enterprises in Indonesia.
8.3. Limitations
The limitations of the results of this thesis can be summarised as follows. More detailed
data on the observed banks would help expand the analysis with the possible various
models to examine. Some data which are not available are the number of employees, the
number of branches, and the individual interest rate for each of type of lending facilities
provided. The availability of the numbers of employees‟ data would be beneficial to be
used as one of the input components (as the labour input). The number of branches data
could also be used to understand the effect of excessive branches on bank performance
included as control variables. The individual interest rate for each type of loans could
also be utilized to expand the various analyses in the last empirical chapter.
This thesis uses the Battese and Coelli 1992 (BC92) in estimating SFA cost efficiency
(in the second empirical chapter) since the study aims to adopt two-stage effects model
134
As stipulated by Indonesia‟s Government Law No. 21 of 2011 on the Financial Services Authority, as
of December 31, 2013, the task of banking regulation and supervision of Bank Indonesia has been
transferred to the Financial Services Authority (FSA or OJK).
135
BI Regulation No.14/22/PBI/2012 on 21 December, 2012 regarding the credit or financing to micro,
small, and medium enterprises by commercial banks and technical assistance for the development of these
businesses. This regulation imposed on all commercial banks in Indonesia to finance SMEs gradually
started in 2012. It is then expected that by the end of 2018 all commercial banks should be maintaining
the SMEs finance minimum of 20 percent.
205
and only account for an average estimation over the period. However, this method is
less flexible compared to the BC1995 method in terms of capturing the yearly rank
order over time.
8.4. Future Research
This thesis results have brought to some recommendations that could be undertaken as
the potential research in the future as follows:
It is very beneficial to compare the performance of Indonesian banks in terms of
operational systems (conventional banks and sharia banks) considering that the number
of sharia banks has been increasing and the share of their assets are also growing to date.
It is also very challenging to compare bank performance across the Association of South
East Asian Nations (ASEAN) countries considering that the region could be facing a
free trade area agreement in 2015 for its members (Indonesia, Malaysia, Thailand, the
Philippines, Singapore, Brunei Darussalam, Laos, Cambodia, Vietnam, and Timor
Leste). The research about the comparison performance both in bank efficiency and
bank lending propensity of the ASEAN‟s commercial banks would help map a
constellation of the competition among the banks operating across the region. In
addition, the study results could be of considerations to explore the possible cooperation
among commercial banks across the region.
In terms of the estimation of SFA cost efficiency, it is recommended for further research
to use the Battese and Coelli 1995 (BC95) considering that the method is more flexible
and has an advantage in terms of its ability to observe the yearly rank order over the
observations. It is also beneficial for further research to conduct profit efficiency of
commercial banks given the fact that that profit efficiency considers not only expenses
but also revenues, thus the results could be more comprehensive in explaining the entire
efficiency of the bank compared to cost efficiency.
206
Bibliography
Abdul-Majid, M., Saal, D.S, and Battisti, G. (2011). Efficiency and Total Factor
Productivity Change of Malaysian Commercial banks. The Service Industries journal
Vol. 31, No. 13: 2117-2143.
Afriat, S. N. (1972). Efficiency Estimation of Production Functions. International
Economic Review Vol.13, No. 3: 568-98.
Aigner D.J, Lovell, C.A.K. and Schmidt, P. (1977). Formulation and estimation of
stochastic frontier production function models. Journal of econometrics, 6: 21-37.
Aigner, D.J. and Chu, S.F. (1968). On Estimating the Industry Production Function.
American Economic Review, 58, 4: 826-39.
Akhavein, J., Goldberg, L.G, and White, L.J. (2004). Small Banks, Small Business, and
Relationships: An Empirical Study of Lending to Small Farms. Journal of Financial
Services Research, Vol. 26 Issue: 3: 245-261.
Akhavein, J., Frame, W.S, and White, L.J. (2005). The Diffusion of Financial
Innovations: an Examination of the Adoption of Small Business Credit Scoring by Large
Banking Organizations. Journal of Business, 78:577–596.
Altunbas, Y., Liu, M.H., Molyneux, P., and Seth, R. (2000). Efficiency and Risk in
Japanese Banking. Journal of Banking and Finance 24: 1605-1628.
Alzubaidi, H. and Bougheas, S. (2012). The Impact of the Global Financial Crisis on
European Banking Efficiency. Working Paper 12/05. Centre for Finance and Credit
Markets, School of Economics, University of Nottingham.
Andries, A. M. (2011). The Determinants of Bank Efficiency and Productivity Growth
in the Central and Eastern European Banking Systems. Eastern European Economics,
Vol. 49, No. 6: 38–59.
Ariff, M and Khalid, A. M. (2005). Liberalization and Growth in Asia: 21st century
challenges. Edward Elgar Publishing Limited, Inc.
Ariff, M. and Can, L. (2008). Cost and profit efficiency of Chinese banks: A non-
parametric analysis. China Economic Review, 19(2): 260-273.
Assaf and Josiassen (2012). Time-varying production efficiency in the health care
foodservice industry: ABayesian method. Journal of Business Research, 65:617-625.
Baltagi, B.H. (2002). Econometrics. Third Edition. Springer-Verlag Berlin Heidelberg
New York.
Baltagi, B. H. (2005). Econometric Analysis of Panel Data, Third Edition. John Wiley
& Sons Ltd.
207
Banker, R.D., Charnes, A. and Cooper, W.W. (1984). Some Models for Estimating
Technical and Scale Inefficiencies in Data Envelopment Analysis. Management Science,
30:1078-1092.
Basri, M.C. (2011). “The Impact of Global Financial Crisis on the Indonesian
Economy” in “Managing Economic Crisis”, edited by Saw Swee-Hock. Institute of
Southeast Asian Studies (ISEAS) Publishing.
Basri, M.C. and Siregar, R.Y. (2009). Navigating Policy Responses at the National
Level in the Midst of the Global Financial Crisis: The Experience of Indonesia. Asian
Economic Paper 8, no. 3: 1-35.
Battese, G.E. and Coelli, T. J. (1988). Prediction of Firm-Level Technical Efficiencies
with a Generalised Frontier Production Function and Panel Data. Journal of
Econometrics, 38: 387-399.
Battese, G.E. and Coelli, T. J. (1992). Frontier Production Functions, Technical
Efficiency and Panel Data: With Application to Paddy Farmers in India. Journal of
Productivity Analysis, 3: 153-169.
Battese, G.E. and Coelli, T. J. (1995). A Model for Technical Inefficiency Effects in a
Stochastic Frontier Production Function for Panel Data. Empirical Economics, 20:
325-332.
Battesse, G.E. and Corra, G.S. (1977). Estimation of Production Frontier Model: With
Application to the Pastoral Zone of Eastern Australia. Australian Journal of
Agricultural Economics, 21: 169-179.
Baum, C. F. (2006). An Introduction to Modern Econometrics Using Stata. A Stata Press
Publication, StataCorp LP, College Station, Texas.
Beck, T. and Kunt, A.D. (2006). Small and medium-size enterprises: Access to finance
as a growth constraint. Journal of Banking & Finance 30: 2931–2943.
Berger, A.N., and Humprey, D.B. (1991). The Dominance of Inefficiencies over Scale
and Product Mix Economies in Banking. Journal of Monetary Economics 28: 117-148.
Berger, A.N. (1993). Distribution Free Estimates of Efficiency in the US Banking
Industry and Tests of the Standard Distributional Assumptions. Journal of Productivity
Analysis, Volume 4, Issue 3: 261-292.
Berger, A.N. and Humphrey, D.B. (1997). Efficiency of financial institutions:
International survey and direction for future research. European Journal of Operational
Research, 1977:175-212.
Berger A.N., Saunders, A., Scalise, J.M., Udell, G.F. (1998). The Effects of Bank
Mergers and Acquisitions on Small Business Lending. Journal of Financial Economics
50:187–229.
208
Berger, A.N. and Udell, G. (1998). The Economics of Small Business Finance: the
Roles of Private Equity and Debt Markets in the Financial Growth Cycle. Journal of
Banking and Finance 22: 613–673.
Berger, A.N., Klapper, L.F., and Udell, G.F. (2001). The Ability of Banks to Lend to
Informationally Opaque Small Businesses. Journal of Banking & Finance 25: 2127-
2167.
Berger A.N., Rosen, R.J., Udell, G.F. (2007). Does Market Size Structure Affect
Competition: the Case of Small Business Lending. Journal of Banking and Finance
31:11–33.
Berger, A.N. and Black, L.K. (2011). Bank Size, Lending Technologies, and Small
Business Finance. Journal of Banking & Finance 35: 724–735.
Berry, A. and Grant, P. (2004). European Bank Lending to the UK SME Sector.
International Small Business Journal Vol. 22 (2): 115-130.
Besar, D.S. (2011). Essays on Indonesian Banking: Competition, Efficiency, and its
Role in Monetary Policy Transmission. Unpublished Doctoral thesis, City University
London.
Bonin, J.P., Hasan, I., Wachtel, P. (2005). Bank Performance, Efficiency and Ownership
in Transition Countries. Journal of Banking & Finance 29: 31-53.
Casu, B. and Molyneux, P. (2003). A Comparative Study of Efficiency in European
Banking. Applied Economics, 35: 1865-1876.
Chan, S.G. and Karim, M.Z.A. (2010). Bank Efficiency and Macro-economic Factors:
The Case of Developing Countries. Global Economic Review: Perspectives on East
Asian Economies and Industries, 39:3, 269-289.
Chang, C.E., Hasan, I., & Hunter, W. (1998). Efficiency of Multinational Banks: An
Empirical Investigation. Applied Financial Economics, 8: 689-696.
Charnes, A., Cooper, W., and Rhodes, E. (1978). Measuring the Efficiency of Decision
Making Units. European Journal of Operation Research, 2: 429-44.
Chen, X., Skully, M., and Brown, K. (2005). Banking efficiency in China: Application
of DEA to Pre- and Post-Deregulation Eras: 1993–2000. China Economic Review, 16:
229–245.
Chernykh, L. and Theodossiou, A.K. (2011). Determinants of Bank Long-term Lending
Behavior: Evidence from Russia. Multinational Finance Journal, vol. 15, no. 3/4: 193–
216.
Chunhachinda, P. and Li, Li. (2010). Efficiency of Thai Commercial Banks: Pre-
vs.Post-1997 Financial Crisis. Review of Pacific Basin Financial Markets and Policies,
Vol. 13, No. 3: 417-447.
209
Clarke, G., Cull, R., Peria, M.S.M., and Sanchez, S.M. (2005). Bank Lending to Small
Businesses in Latin America: Does Bank Origin Matter?. Journal of Money, Credit and
Banking, Vol. 37, No. 1: 83-118.
Coelli, T., Perelman, S. and Romano, E. (1999). Accounting for Environmental
Influences in Stochastic Frontier Models: With Application to International Airlines.
Journal of Productivity Analysis, 11:251-273.
Coelli, T.J., Rao, D.S.P., O‟Donnell, C.J., and Battese, G.E. (2005). An Introduction to
Efficiency and Productivity Analysis. Second Edition. Springer Science and Business
Media, LLC.
Debreu, G. (1951). The Coefficient of Resource Utilisation. Econometrica, Journal of
the Econometric Society, Vol. 19 No. 3: 273-292.
Demir, N., F. Mahmud, S, and Babuscu, S. (2005). The Technical Inefficiency Effects of
Turkish Banks after Financial Liberalisation. The Developing Economics, XLIII, 3:
396-411.
Deprins, D., Simar, L., & Tulkens, H. (1984). “Measuring Labor-Efficiency in Post
Offices”. In: M. Marchand, P. Pestieau, & H. Tulkens (Eds.), “The Performance of
Public Enterprises: Concepts and Measurement”. Amsterdam: North-Holland.
Devaney, M., and Weber, W. L. (2002). Small Business Lending and Profit Efficiency in
Commercial Banking. Journal of Financial Services Research 22:3, 225-246.
Drake, L. and Hall, Maximilian J. B (2003). Efficiency in Japanese banking: An
empirical analysis. Journal of Banking & Finance 27: 891-917.
Edmiston, K. (2007). The Role of Small and Large Businesses in Economic
Development. Economic Review, Federal Reserve Bank of Kansas City, (Second
Quarter), 73–97.
Enoch, C., Baldwin, B., Frecaut, O., and Kovanen, A. (2001). Indonesia: Anatomy of a
Banking Crisis Two Years of Living Dangerously 1997-99. IMF Working Paper No. 52,
Washington D.C, International Monetary Fund.
Fane, G., and McLeod, R. H. (2002). Banking Collapse and Restructuring in Indonesia,
1997-2001. Cato Journal Vol. 22. Cato Institute.
Fare, R., Grosskopf, S. and Logan, J. (1983). The Relative Efficiency of Illinois Electric
Utilities. Resources and Energy, 5: 349-367.
Färe, R., Grosskopf, S., Norris, M. and Zhang, Z. (1994). Productivity Growth,
Technical Progress, and Efficiency Change in Industrialized Countries. American
Economic Review, 84: 66-8.
210
Forughi, S. & De Zoysa, A. (2012). Australian Banks Performance during the Global
Financial Crisis: an Analysis on the Efficiency and Productivity. Saarbrucken,
Germany: Lambert Academic Publishing.
Fang, Y., Hasan, I. and Marton, K. (2011). Bank Efficiency in South-Eastern Europe:
The Role of Ownership, Market Power, and Institutional Development. Economics of
Transition Volume 19 (3): 495-520.
Farrel, M.J. (1957). The Measurement of Productive Efficiency. Journal of the Royal
Statistical Society, Series A (General), Vol. 120, No. 3: 253-290.
Fethi, M.D. and Pasiouras, F. (2010). Assessing Bank Efficiency and Performance with
Operational Research and Artificial Intelligence Techniques: A Survey. European
Journal of Operational Research 204: 189–198.
Fethi, M.D., Shaban, M., Jones, T.W. (2011). Liberalisation, Privatisation and the
Productivity of Egyptian Banks: A Non-Parametric Approach. The Service Industries
Journal Vol. 31, No. 7, May 2011: 1143–1163.
Fethi, M.D., Shaban, M., Jones, T.W. (2012). Turkish Banking Recapitalization and the
Financial Crisis: An Efficiency and Productivity Analysis. Emerging Markets Finance
& Trade, Vol. 48, Supplement 5: 76–90.
Frame W.S., Padhi, M., Woosley, L. (2004). Credit Scoring and the Availability of
Small Business Credit in Low and Moderate-Income Areas. Financial Review 39:35–54.
Gardener, E., Molyneux, P. and Nguyen-Linh, H. (2011). Determinants of Efficiency in
South East Asian Banking. The Service Industries Journal. 31:16, 2693-2719.
Goldberg, L.G. and White, L. J. (1998). De Novo Banks and Lending to Small
Businesses: An Empirical Analysis. Journal of Banking and Finance Vol. 22, Issues 6-8:
851-867.
Green, W.H. (2008). Econometric Analysis. Sixth Edition. Pearson International
Edition.
Greene, W.H. (2008). “The Econometric Approach to Efficiency Analysis” in “The
Measurement of Productive Efficiency and Productivity Growth. Edited by H.O. Fried,
C.A.K. Lovell and S.S. Schmidt. Oxford University Press.
Greene, W.H. (2013). Econometric Estimation of Frontier Functions and Economic
Efficiency. Unpublished Training Hand Out. Cemmap & ESRC, UCL London, 21-22
January, 2013.
Grigorian, D.A. and Manole, V. (2006). Determinants of Commercial Banks
Performance in Transition: An Application of Data Envelopment Analysis. Comparative
Economic Studies, 48: 497-522.
211
Gujarati, D. N. (2006). Essentials of Econometrics. Third Edition. McGraw Hill
International Edition.
Gujarati, D.N. and Porter, D.C. (2010). Essentials of Econometrics. Fourth Edition.
McGraw Hill.
Gup, B.E. (2010). “Global financial crisis” in “The Financial and Economic Crises: An
International Perspective”. Edited by Benton E. Gup. Edward Elgar Publishing Limited.
Gupta, S., and Verhoeven, M. (2001). The Efficiency of Government Expenditure
Experiences from Africa. Journal of Policy Modelling, 23: 433– 467.
Hadad, M.D, M.J.B. Hall, K.A. Kenjegalieva, W. Santoso, R. Satria, and R. Simper,
(2011). Banking Efficiency and Stock Market Performance: An Analysis of Listed
Indonesian Banks. Rev Quant Finan Acc 37:1-20.
Hadad, M.D, M.J.B. Hall, K.A. Kenjegalieva, W. Santoso, and R. Simper (2011).
Productivity Changes and Risk Management in Indonesian banking: Malmquist
Analysis. Applied Financial Economics, 21: 847-861.
Hamiltona, R., Qasrawib, W., and Al-Jarrah, I. M. (2010). Cost and Profit Efficiency in
the Jordanian Banking Sector 1993-2006: A Parametric Approach. International
Research Journal of Finance and Economics, Issue 56.
Han, L. and Benson, A. (2010). The Use and Usefulness of Financial Assistance to UK
SMEs. Environment and Planning C: Government and Policy 2010, vol.28: 552–566.
Harada, K. & Ito, T. (2005). Rebuilding the Indonesian Banking Sector: Economic
Analysis of Bank Consolidation and Efficiency. Japan Bank for International
Cooperation Institute Review (JBICI), 12: 32–59.
Havrylchyk, O. (2005). Efficiency of the Polish Banking Industry: Foreign versus
Domestic Banks. Journal of Banking and Finance, 30: 1975-1996.
Heffernan, S. (2009). Modern Banking. John Wiley & Sons, Ltd.
Hock, S.S. (2011). “The Global Financial Crisis: Impact and Response in Southeast
Asia, in the book of Managing Economic Crisis. Edited by Saw Swee Hock. Institute of
Southeast Asian Studies, Singapore.
Howells, P., and Bain, K. (2008). The Economics of Money, Banking and Finance: A
European Text. FT Prentice Hall, Pearson Education.
Isik, I. and Hassan, M. K. (2003). Financial Deregulation and Total Factor Productivity
Change: An Empirical Study of Turkish commercial banks. Journal of Banking and
Finance, 27: 1455-1485.
212
Isik, I. and Hassan, M. K. (2003). Efficiency, Ownership and Market Structure, Control
and Governance in Turkish Banking Industry. Journal of Business Finance and
Accounting 30 (9) & (10): 1363- 1421.
Isik, I. and Hassan, M. K. (2002). Technical, Scale and Allocative Efficiencies of
Turkish Banking Industry. Journal of Banking and Finance, 26: 719-766.
Jayaratne, J. and Wolken, J. (1999). How Important are Small Banks to Small Business
Lending? New Evidence from a Survey of Small Firms. Journal of Banking and
Finance, 23: 427-458.
Kao, C. and Liu, S.T. (2004). Predicting Bank Performance with Financial Forecasts:
A case of Taiwan Commercial Banks. Journal of Banking & Finance 28: 2353-2368.
Kenjegalieva, K. A., Simper, R., and Weyman-Jones, T.G. (2009). Efficiency of
Transition Banks: Inter-country Banking Industry Trends. Applied Financial
Economics, 19: 1531-1546.
Kishan, R. P. and Opiela, T. P. (2000). Bank Size, Bank Capital, and the Bank Lending
Channel. Journal of Money, Credit and Banking, Vol. 32, No. 1:121-141.
Koop, G. (2008). Introduction to Econometrics. John Wiley & Sons, Ltd.
Koopmans, T.C. (1951). “An analysis of Production as an Efficient Combination of
Activities” in “Activity Analysis of Production and Allocation”. Cowles Commission for
Research in Economics. Edited by T.C. Koopmans, (Ed.). Monograph No. 13, John
Wiley and Sons, New York.
Kumbhakar, S.C. (1990). Production Frontiers, Panel Data and Time-Varying
Technical Inefficiency. Journal of Econometrics, 46:201-211.
Kumbhakar, S.C., and Lovell, C.A.K (2000). Stochastic Frontier Analysis. Cambridge
University Press.
Kumbhakar, S.C., Ghosh, S. and J.T. McGuckin, J.T. (1991). A Generalized Production
Frontier Approach for Estimating Determinants of Inefficiency in US Dairy Farms.
Journal of Business and Economic Statistics, 9: 279-286.
Laderman, E. S. (2008). The Quantity and Character of Out-of-Market Small Business
Lending. Economic Review-Federal Reserve Bank of San Francisco, 31–39.
Macerinskiene, I., Ivaskeviciute, L. (2008). The Evaluation Model of a Commercial
Bank Portfolio. Journal of Business Economics and Management, 4:269-277.
Mamatzakis, E., Staikouras, C., and Filippaki, A.K. (2008). Bank Efficiency in the New
European Union Member States: Is There Convergence?. International Review of
Financial Analysis, 1156-1172.
213
Margono, H., and Sharma, S.C., and Melvin II, P.D. (2010). Cost Efficiency, Economies
of Scale, Technological Progress and Productivity in Indonesian Banks. Journal of
Asian Economics 21: 53-65.
McNulty, J. E., Murdock, M., and Richie, N. (2013). Are Commercial Bank Lending
Propensities Useful in Understanding Small Firm Finance?. Journal of Economics and
Finance, 37: 511–527.
Meeusen, W., and Van den Broeck, J.(1977). Efficiency Estimation from Cobb-
Douglass Production Functions with Composed Error. International economic review,
18: 435-444.
Mishkin, F.S. (2010). The Economics of Money, Banking and Financial Markets. Ninth
Edition. Pearson Education, Inc.
Murray, M. P. (2006). Econometrics: A Modern Introduction. Pearson, International
Edition.
Nikiel, E.M. and Opiela, T.P. (2002). Customer Type and Bank Efficiency in Poland:
Implications for Emerging Banking Market. Contemporary Economic Policy, 20: 255-
271.
Orea, L. (2002). Parametric Decomposition of a Generalized Malmquist Productivity
Index. Journal of Productivity Analysis: 17, 5-22.
Owualah, S. (1990). Providing the Necessary Economic Infrastructure for Small
Businesses: Whose Responsibility. International Small Business Journal, 6: 10-30).
Peek, J. and Rosengren, E.S. (1998). Bank Consolidation and Small Business Lending:
It‟s not just Bank Size that Matters. Journal of Banking & Finance 22: 799-819.
Peek, J., Rosengren, E.S., and Tootell, G.M.B (2003). Identifying the Macroeconomic
Effect of Loan Supply Shocks. Journal of Money Credit Bank, 35:931–946.
Petersen, M.A., and Rajan, R. G. (1994). The Benefits of Lending Relationships:
Evidence from Small Business Data. The Journal of Finance XLIX: 3-37.
Pitt, M.M. and Lee, L.F. (1981). Measurement and Sources of Technical Inefficiency in
the Indonesian Weaving Industry. Journal of Development Economics, 9: 43-64.
Rao, K.S.R., Das, A., and Singh, A.K. (2006). Commercial Bank Lending to Small-
Scale Industry. Economic and Political Weekly, Money, Banking and Finance, Vol. 41,
No. 11: 1025-1033.
Richmond, W. (1974). Estimating the Efficiency of Production. International Economic
Review 15:515–521.
Rraci, O. (2010). The Effect of Foreign Banks in Financing Firms, Especially Small
Firms, in Transition Economies. Eastern European Economics vol. 48, No. 4: 5–35.
214
Sato, Y. (2005). Bank Restructuring and Financial Institution Reform in Indonesia. The
Developing Economies, XLIII-1: 91-120.
Saunders, A., and Cornett, M.M. (2012). Financial Markets and Institutions. Fifth
Edition. McGraw Hill International Edition.
Scott, J.A. and Dunkelberg, W.C. (2003). Bank Mergers and Small Firm Financing.
Journal of Money, Credit and Banking, Vol. 35, No. 6, Part 1: 999-1017.
Seibel, H. D., and Rachmadi, A., and Kusumayakti, D. (2010). Reform, Growth and
Resilience of Savings-Led Commercial Microfinance Institutions: the Case of the Micro
banking Units of Bank Rakyat Indonesia. Savings and Development, Vol. 34, No. 3:
277-303.
Seitz, W.D. (1971). Productive Efficiency in the Steam-Electric Generating Industry.
Journal of Political Economy, 79, 4: 878-86.
Shaban, M. (2008). Financial Liberalisation, Privatisation and Productivity in Banking:
The Experience of Two Emerging Economies. PhD Thesis, University of Leicester.
Shaban, M., Duygun, M., Anwar, M., Akbar, B. (2014). Diversification and Banks’
Willingness to Lend to Small Businesses: Evidence from Islamic and Conventional
Banks in Indonesia. Journal of Economic Behavior & Organization 103: S39-S55.
Shen, Y., Shen, M., Xu, Z., and Bai, Y. (2009). Bank Size and Small- and Medium-sized
Enterprise (SME) Lending: Evidence from China. World Development Vol. 37, No. 4:
800–811.
Shephard, R.W. (1953). Cost and Production Functions. Princeton University Press,
Princeton.
Shephard, R.W. (1970). The Theory of Cost and Production Functions. Princeton
University Press, Princeton.
Shin, D.J. and Kim, B.H.S. (2011). Efficiency of the Banking Industry Structure in
Korea. Asian Economic Journal 2011, Vol. 25 No. 4, 355-373.
Stock, J.H., and Watson, M.M. (2012). Introduction to Econometrics. Third Edition.
Pearson Education Limited.
Strahan, P. E., and Weston, J. P. (1998). Small Business Lending and the Changing
Structure of the Banking Industry. Journal of Banking and Finance 22: 821-845.
Studenmund, A. H. (2011). Using Econometrics: A Practical Guide. Sixth Edition.
Pearson, International Edition.
Sufian, F. (2009). The impact of Off-balance sheet Items on Banks' Total Factor
Productivity: Empirical Evidence from the Chinese Banking Sector. American Journal
of Finance and Accounting, Vol. 1, 3: 213-238.
215
Sufian, F. (2010). The Impact of the Asian Financial Crisis on Bank Efficiency: the
1997 Experience of Malaysia and Thailand. Journal of International Development, 22:
866-889.
Sufian, F. (2011). Benchmarking the Efficiency of the Korean Banking Sector: a DEA
Approach". Benchmarking: An International Journal, Vol. 18 Iss.: 1:107 – 127.
Suhaemi, R., Abdullah, F., and Saban, G. (2010). Factors affecting profit efficiency of
commercial banks in Malaysia. International Conference on Science and Social
Research (CSSR 2010), December 5-7, 2010, Kuala Lumpur, Malaysia
Suzuki, Y. & Sastrosuwito, S. (2011). Efficiency and Productivity Change of the
Indonesian commercial banks. International Conference on Economics, Trade and
Development. IPEDR Vol. 7, 2011, IACSIT Press, Singapore.
Thangavelu, S. M., and Findlay, C. (2013). Bank Efficiency, Regulation and Response
to Crisis of Financial Institutions in Selected Asian Countries. International Conference
Recent Developments in Asian Trade Policy and Integration, 20-21 February 2013.
Centre for Research on Globalisation and Economic Policy (GEP), Kuala Lumpur
Teaching Centre for the University of Nottingham Malaysia Campus, Malaysia.
Timberg, T.A. (1999). Small and Micro-Enterprise Finance in Indonesia: What Do We
Know?. USAID-funded Partnership for Economic Growth (PEG) Projects. ECG,
USAID Jakarta.
Timmer, C.P. (1971). Using a Probabilistic Frontier Function to Measure Technical
Efficiency. Journal of Political Economy, 79: 579-597.
Tulkens, H. (1993). On FDH efficiency analysis: Some Methodological Issues and
Applications to Retail Banking, Courts, and Urban Transit. Journal of Productivity
Analysis, Vol.4, Issue 1-2: 183-210.
Tulkens, H., & Eeckaut, P.V. (1995). Non-Parametric Efficiency, Progress and Regress
Measures for Panel Data: Methodological Aspects. European Journal of Operational
Research, 80: 474– 499.
Verbeek, M. (2012). A Guide to Modern Econometrics. Fourth Edition. A John Wiley &
Sons, Ltd., Publication.
Vu, H., and Turnell, S. (2011). Cost and Profit Efficiencies of Australian Banks and the
Impact of the Global Financial Crisis. The Economic Record, Vol. 87, No. 279: 525-
536.
Williams, J. and Nguyen, N. (2005). Financial Liberalisation, Crisis, and
Restructuring: a comparative Study of Bank Performance and Bank Governance in
South East Asia. Journal of Banking & Finance 29: 2119–2154.
216
Winsten, C. B. (1957). Discussion on Mr. Farrell’s Paper. Journal of the Royal
Statistical Society, Series a-Statistics in Society 120(3): 282-284.
Wooldridge, J. M. (2013). Introductory Econometrics: A Modern Approach, Fifth
Edition. South Western, CENGAGE Learning, International Edition.
Yang, C.C. (2011). An Enhanced DEA Model for Decomposition of Technical
Efficiency in Banking. Annals of Operation Research, 82, 233–249. Springer
Science+Business Media.
Yao, S., Han, Z. and Feng, G. (2008). Ownership reform, foreign competition and
efficiency of Chinese Commercial Banks: A non-parametric approach. The World
Economy (2008): 1310-1326.
Yeh, T.L. (2011). Capital Structure and Cost efficiency in the Taiwanese banking
Industry. The Service Industries Journal, Vol. 31, No. 2, 237-249.
Yildirim, C. (2002). Evolution of Banking Efficiency within an Unstable
Macroeconomic Environment: The Case of Turkish Commercial Banking. Applied
Economics, 34, 18: 980-88.
Yildirim, H. S., and Philippatos, G.C. (2007). Efficiency of Banks: Recent Evidence
from the Transition Economies in Europe, 1993-2000. The European Journal of
Finance, 13: 123-143.
Bank Indonesia (BI) (Central Bank of Indonesia). http://www.bi.go.id
Bappenas RI (Ministry of National Planning). http://www.bappenas.go.id.
The Directors of BI Decree No. 30/4/KEP/DIR on April 1997 on Small Business Loans.
BI Regulation No. 3/2/PBI/2001 on Small Business Loans (Finance).
BI Regulation No.14/22/PBI/2012 on Commercial Banks‟ Loans (Finance) for MSMEs
and its technical assistance.
BI Regulation Number 15/7/PBI/2013 on Reserve Requirement on Commercial Banks
in Indonesia.
Emergency Law Republic of Indonesia No. 18/1957 about Bank Farmers and
Fishermen.
Indonesian Bank Statistics, Various volumes, Monthly-Edition for the year 2002-1010.
Bank Indonesia, Jakarta.
Indonesia's Economic Report, 2001-2010 (Various Editions). Bank Indonesia
Indonesian Government Law no. 14/1967 on Banking.
217
Indonesian Government Law no. 13/1968 on Central Bank.
Indonesian Government Law No. 23/1999 on Bank Indonesia.
Indonesian Government Law No. 20/2008 on Micro and SMEs.
Indonesian Government Law No. 21 of 2011 on the Financial Services Authority.
International Finance Corporation. www.ifc.org/ . Accessed on November 7th
, 2013.
Key Indicator for Asia and the Pacific 2011. www.adb.org/statistics.
Statistics Indonesia. http://www.bps.go.id.
Technical Guidelines People's Business Credit (PBS / KUR) for the Agricultural Sector
(2012). Published by Directorate of Agricultural Financing Infrastructure Directorate
General of Agriculture, Ministry of Agriculture, Republic of Indonesia.
The History of Bank Indonesia: Indonesian Banking period 1983-1997.
http://www.bi.go.id/SejarahPerbankanPeriode19831997.pdf. Accessed on 9th
, June
2012.
The History of Bank Indonesia: Indonesian Banking period 1953-1959.
http://www.bi.go.id/ SejarahPerbankanPeriode19531959.pdf. Accessed on 9th
, June
2012.
The History of Bank Indonesia: Indonesian Banking period 1959-1966.
http://www.bi.go.id/ /SejarahPerbankanPeriode19591966.pdf. Accessed on 9th
, June
2012.
World Economic Outlook (International Monetary Fund). April 2012 Edition.
218
APPENDICES:
Appendix 1. Figure 1.1. Thesis Objectives
Appendix 2. Table 2.6. Indonesian Government Policy Responses in 1998 and
2008
The 1998 Crisis The 2008 Crisis
1.
Monetary policy: extremely strict. Bank of Indonesia
increased interest rate levels to very high levels.
Deposit account interest rates reached 60 per cent in the peak crisis period. As regards liquidity, the government
implemented a liquidity squeeze.
1. Monetary policy: Bank Indonesia interest rate was
reduced by 300 basis points from 9.5 per cent to 6.5
per cent. Liquidity was relaxed.
2. Fiscal policy: to being with these was a budget surplus then this was revised by permitting a large budget
deficit.
2. Fiscal policy: the stimulus was implemented, the budget deficit enlarged, taxes reduced.
3. Banking health: Prudential banking regulations were extremely weak. NPLs reached 27 per cent. LDR
became more than 100 per cent.
3. Banking health: Prudential banking regulations were relatively tight. NPL less than 4 per cent,
LDR 77 per cent, CAR around 17 per cent.
4. Response toward banking: closure of sixteen banks, which then led to rushes.
4. Response towards banking: deposit insurance increased from Rp.100 million to Rp.2 billion per
account.
5. Policies focused towards structural reform by carrying out economic liberalizations, getting rid of monopolies
and licensing.
5. Safeguarded relatively open trade regime.
6. Exchange rate regime: managed floating. Economic players were not used to exchange rate risk changes and
did not carry out Hedging.
6. Exchange rate regime: flexible. Economic players start to become used to exchange rate risk changes.
Source: Basri (2011)
219
Appendix 3. Figure 5.4. Average technical efficiency of Indonesian banks (model 1)
Source: Data Observed
Appendix 4. Figure 5.5. Average technical efficiency of Indonesian banks (model
2)
Source: Data Observed
Appendix 5. Figure 5.6. Average technical efficiency of Indonesian banks (model
3)
Source: Data Observed
0.000
0.200
0.400
0.600
0.800
1.000
1.200
2002 2003 2004 2005 2006 2007 2008 2009 2010
eff
icie
ncy
year
Average technical efficiency by ownership (model 1)
SOB
PB (FEB)
PB (NFEB)
LGOB
JVB
FOB
0.000
0.200
0.400
0.600
0.800
1.000
1.200
2002 2003 2004 2005 2006 2007 2008 2009 2010
eff
icie
ncy
year
Average technical efficiency by Ownership (model 2)
SOB
PB (FEB)
PB (NFEB)
LGOB
JVB
FOB
0.000
0.200
0.400
0.600
0.800
1.000
1.200
2002 2003 2004 2005 2006 2007 2008 2009 2010
eff
icie
ncy
year
Average techical efficiency by ownership (model 3)
SOB
PB (FEB)
PB (NFEB)
LGOB
JVB
FOB
220
Appendix 6. Figure 5.8. Average Technical Efficiency By Bank Peer Group (Model 2)
Source: Data processed
Appendix 7. Figure 5.9. Average Technical Efficiency By Bank Peer Group (Model 3)
Source: Data processed
Appendix 8. Table 5.8. Malmquist Index Summary of Annual Means (Model 2)
Year EFF.CH TECH.CH PE.CH SE.CH TFP.CH
2002/2003 0.972 0.968 0.987 0.984 0.941
2003/2004 0.972 1.024 0.973 0.999 0.996
2004/2005 1.020 0.976 1.004 1.016 0.995
2005/2006 0.919 1.085 0.943 0.974 0.997
2006/2007 0.872 1.154 0.991 0.880 1.006
2007/2008 0.896 1.095 0.875 1.023 0.981
2008/2009 1.005 0.990 0.946 1.062 0.995
2009/2010 1.072 0.935 1.028 1.043 1.002
Mean 0.964 1.026 0.968 0.996 0.989
Source: Data Observed
-
0.2000
0.4000
0.6000
0.8000
1.0000
1.2000
200220032004200520062007200820092010
Ave
rage
Eff
icie
ncy
Sco
re
Average Technical Efficiency Model 2 by Bank Peer Group
Peer Group 1 (Total Assets<= IDR 10 Trillion)
Pee Group 2 (Total Assets >IDR 10 Trillion and <= IDR50Trillion)
Peer Group 3 (Total Assets> IDR 50 Trillion and <= IDR100 Trillion)
-
0.2000
0.4000
0.6000
0.8000
1.0000
1.2000
Ave
rage
Eff
icie
ncy
Sco
re
Average Technical Efficiency Model 3 by Bank Peer Group
Peer Group 1 (TotalAssets <= IDR 10 Trillion)
Pee Group 2 (Total Assets> IDR 10 Trillion and <=IDR 50Trillion)
Peer Group 3 (TotalAssets > IDR 50 Trillionand <= IDR 100 Trillion)
221
Appendix 9. Table 5.9. Malmquist Index Summary of Annual Means (Model 3)
Year EFF.CH TECH.CH PE.CH SE.CH TFP.CH
2002/2003 0.921 0.989 0.980 0.939 0.910
2003/2004 1.059 0.938 1.012 1.046 0.993
2004/2005 1.036 0.978 1.018 1.018 1.013
2005/2006 0.969 1.076 0.966 1.003 1.043
2006/2007 0.837 1.224 0.986 0.849 1.025
2007/2008 0.840 1.145 0.825 1.018 0.962
2008/2009 0.909 1.079 0.892 1.020 0.981
2009/2010 1.066 0.946 1.027 1.038 1.008
Mean 0.951 1.043 0.961 0.990 0.991
Source: Data Observed
Appendix 10. Figure 6.7. Evolution of Indonesian Banks’ Cost Efficiency – Pooled
- Model 2
Source: Data processed
Appendix 11. Figure 6.8. The Evolution of Indonesian Banks’ Cost Efficiency –
BC92 - Model 2
Source: Data processed
0.6500
0.7000
0.7500
0.8000
0.8500
0.9000
0.9500
2002 2003 2004 2005 2006 2007 2008 2009 2010
Co
st E
ffic
ien
cy
Cost efficiency of Model 2 : Pooled
FB
JVB
LGOB
PB
SOB
CONV
SHARIA
-
0.2000
0.4000
0.6000
0.8000
1.0000
2002 2003 2004 2005 2006 2007 2008 2009 2010
Co
st E
ffic
ien
cy
Cost efficiency of Model 2 : BC92
FB
JVB
LGOB
PB
SOB
CONV
222
Appendix 12. Figure 6.9. Evolution of Indonesian Banks’ Cost Efficiency – Pooled
- Model 3
Source: Data processed
Appendix 13. Figure 6.10. The Evolution of Indonesian Banks’ Cost Efficiency –
BC92 - Model 3
Source: Data processed
-
0.1000
0.2000
0.3000
0.4000
0.5000
0.6000
0.7000
0.8000
0.9000
1.0000
2002 2003 2004 2005 2006 2007 2008 2009 2010
Co
st E
ffic
ien
cy
Cost efficiency of Model 3 : Pooled
FB
JVB
LGOB
PB
SOB
CONV
SHARIA
-
0.1000
0.2000
0.3000
0.4000
0.5000
0.6000
0.7000
0.8000
0.9000
1.0000
2002 2003 2004 2005 2006 2007 2008 2009 2010
Co
st E
ffic
ien
cy
Cost efficiency of Model 3: BC92
FB
JVB
LGOB
PB
SOB
CONV
SHARIA
223
Appendix 14. Figure 6.13. Cost Efficiency – Pooled - Model 2 (by Bank Peer
Group)
Appendix 15. Figure 6.14. Cost Efficiency – BC92 - Model 2 (by Bank Peer Group)
0.7000
0.7500
0.8000
0.8500
0.9000
0.9500
2002 2003 2004 2005 2006 2007 2008 2009 2010
Ave
rage
Co
st E
ffic
ien
cy
Average Cost Efficiency Pooled-Model 2 by Bank Peer Group
Peer Group 1 (Total Assets <= IDR10 Trillion)
Pee Group 2 (Total Assets > IDR 10Trillion and <= IDR 50Trillion)
Peer Group 3 (Total Assets > IDR 50Trillion and <= IDR 100 Trillion)
Peer Group 4 (Total Assets > IDR100 Trillion)
0.0000
0.1000
0.2000
0.3000
0.4000
0.5000
0.6000
0.7000
0.8000
0.9000
1.0000
2002 2003 2004 2005 2006 2007 2008 2009
Ave
rage
Co
st E
ffic
ien
cy
Average Cost Efficiency BC92-Model 2 by Bank Peer Group
Peer Group 1 (Total Assets <= IDR10 Trillion)
Pee Group 2 (Total Assets > IDR 10Trillion and <= IDR 50Trillion)
Peer Group 3 (Total Assets > IDR 50Trillion and <= IDR 100 Trillion)
Peer Group 4 (Total Assets > IDR100 Trillion)
224
Appendix 16. Figure 6.15. Cost Efficiency – Pooled - Model 3 (by Bank Peer
Group)
Appendix 17. Figure 6.16. Cost Efficiency – BC92 - Model 3 (by Bank Peer Group)
0.7200
0.7400
0.7600
0.7800
0.8000
0.8200
0.8400
0.8600
0.8800
2002 2003 2004 2005 2006 2007 2008 2009 2010
Ave
rage
Co
st E
ffic
ien
cy
Average Cost Efficiency Pooled-Model 3 by Bank Peer Group
Peer Group 1 (Total Assets <=IDR 10 Trillion)
Pee Group 2 (Total Assets >IDR 10 Trillion and <= IDR50Trillion)
Peer Group 3 (Total Assets >IDR 50 Trillion and <= IDR 100Trillion)
Peer Group 4 (Total Assets >IDR 100 Trillion)
0.0000
0.1000
0.2000
0.3000
0.4000
0.5000
0.6000
0.7000
0.8000
2002 2003 2004 2005 2006 2007 2008 2009 2010
Ave
rage
Co
st E
ffic
ien
cy
Average Cost Efficiency BC92-Model 3 by Bank Peer Group
Peer Group 1 (Total Assets <= IDR10 Trillion)
Pee Group 2 (Total Assets > IDR 10Trillion and <= IDR 50Trillion)
Peer Group 3 (Total Assets > IDR 50Trillion and <= IDR 100 Trillion)
Peer Group 4 (Total Assets > IDR100 Trillion)
225
Appendix 18. Table 6.17. Coefficient of Cost Efficiency Estimation Variables –
Pooled Model 1 -----------------------------------------------------------------------------
Limited Dependent Variable Model - FRONTIER
Dependent variable LTC
Log likelihood function 190.44549
Estimation based on N = 999, K = 23
Inf.Cr.AIC = -334.9 AIC/N = -.335
Model estimated: Sep 10, 2013, 07:30:47
Stochastic frontier based on panel data
Estimation based on 111 individuals
Variances: Sigma-squared(v)= .02913
Sigma-squared(u)= .18854
Sigma(v) = .17067
Sigma(u) = .43421
Sigma = Sqr[(s^2(u)+s^2(v)]= .46655
Gamma = sigma(u)^2/sigma^2 = .86618
Var[u]/{Var[u]+Var[v]} = .70167
Stochastic Cost Frontier Model, e = v+u
LR test for inefficiency vs. OLS v only
Deg. freedom for sigma-squared(u): 1
Deg. freedom for heteroscedasticity: 0
Deg. freedom for truncation mean: 0
Deg. freedom for inefficiency model: 1
LogL when sigma(u)=0 39.52751
Chi-sq=2*[LogL(SF)-LogL(LS)] = 301.836
Kodde-Palm C*: 95%: 2.706, 99%: 5.412
--------+--------------------------------------------------------------------
| Standard Prob. 95% Confidence
LTC| Coefficient Error z |z|>Z* Interval
--------+--------------------------------------------------------------------
|Deterministic Component of Stochastic Frontier Model
Constant| 3.43482*** .62279 5.52 .0000 2.21416 4.65547
LTF| .57733*** .09170 6.30 .0000 .39760 .75705
LSI| -.18000* .09591 -1.88 .0605 -.36798 .00797
LOI| .38607*** .08289 4.66 .0000 .22361 .54853
LTFSQ| .14138*** .01228 11.51 .0000 .11731 .16546
LSISQ| .22873*** .01937 11.81 .0000 .19076 .26669
LOISQ| .09386*** .00618 15.18 .0000 .08174 .10598
LPFPC| .23626* .12782 1.85 .0645 -.01426 .48679
LPLPC| .66082*** .10880 6.07 .0000 .44758 .87405
LPFPCSQ| .16672*** .02343 7.11 .0000 .12079 .21265
LPLPCSQ| .07845*** .01641 4.78 .0000 .04629 .11061
LTFSI| -.12279*** .01257 -9.76 .0000 -.14743 -.09814
LTFOI| -.02044*** .00618 -3.31 .0009 -.03255 -.00834
LSIOI| -.08518*** .01027 -8.30 .0000 -.10530 -.06506
LPFCPLC| -.11586*** .01876 -6.18 .0000 -.15263 -.07908
LTFPFPC| .03339** .01363 2.45 .0143 .00666 .06011
LTFPLPC| -.00482 .01014 -.48 .6345 -.02469 .01505
LSIPFPC| .00824 .01539 .54 .5923 -.02193 .03841
LSIPLPC| -.00055 .01380 -.04 .9683 -.02759 .02649
LOIPFPC| -.02459** .00997 -2.47 .0136 -.04412 -.00506
LOIPLPC| -.02206** .01044 -2.11 .0346 -.04252 -.00160
|Variance parameters for compound error
Lambda| 2.54414*** .71292 3.57 .0004 1.14685 3.94143
Sigma(u)| .43421*** .07484 5.80 .0000 .28754 .58089
--------+--------------------------------------------------------------------
Note: ***, **, * ==> Significance at 1%, 5%, 10% level.
-----------------------------------------------------------------------------
ltc = Log (tc/pc); ltf = Log(tf); lsi = Log(si);loi = Log(oi); ltfsq = .5*log(tf)^2; lsisq = .5*log(si)^2; loisq =
.5*log(oi)^2; lpfpc = Log(pf/pc); lplpc = Log(pl/pc); lpfpcsq = .5*log(pf/pc)^2; lplpcsq =.5*log(pl/pc)^2; ltfsi =
log(tf)*log(si); ltfoi = log(tf)*log(oi); lsioi = Log(si)*log(oi); lpfcplc = log(pf/pc)*log(pl/pc); ltfpfpc =
log(tf)*log(pf/pc); ltfplpc = log(tf)*log(pl/pc); lsipfpc = log(si)*log(pf/pc); lsiplpc = log(si)*log(pl/pc); loipfpc
= log(oi)*log(pf/pc); loiplpc = log(oi)*log(pl/pc)
226
Appendix 19. Table 6.18. Coefficient of Cost Efficiency Estimation Variables –
BC92 Model 1 -----------------------------------------------------------------------------
Limited Dependent Variable Model - FRONTIER
Dependent variable LTC
Log likelihood function 219.13477
Estimation based on N = 999, K = 24
Inf.Cr.AIC = -390.3 AIC/N = -.391
Model estimated: Sep 10, 2013, 07:31:39
Stochastic frontier based on panel data
Estimation based on 111 individuals
Variances: Sigma-squared(v)= .02779
Sigma-squared(u)= .09950
Sigma(v) = .16672
Sigma(u) = .31544
Sigma = Sqr[(s^2(u)+s^2(v)]= .35679
Gamma = sigma(u)^2/sigma^2 = .78165
Var[u]/{Var[u]+Var[v]} = .56538
Stochastic Cost Frontier Model, e = v+u
Battese-Coelli Models: Time Varying uit
Time dependent uit=exp[-eta(t-T)]*|U(i)|
LR test for inefficiency vs. OLS v only
Deg. freedom for sigma-squared(u): 1
Deg. freedom for heteroscedasticity: 0
Deg. freedom for truncation mean: 0
Deg. freedom for inefficiency model: 1
LogL when sigma(u)=0 39.52751
Chi-sq=2*[LogL(SF)-LogL(LS)] = 359.215
Kodde-Palm C*: 95%: 2.706, 99%: 5.412
--------+--------------------------------------------------------------------
| Standard Prob. 95% Confidence
LTC| Coefficient Error z |z|>Z* Interval
--------+--------------------------------------------------------------------
|Deterministic Component of Stochastic Frontier Model
Constant| 3.18476*** .72487 4.39 .0000 1.76405 4.60547
LTF| .47467*** .10062 4.72 .0000 .27745 .67189
LSI| -.08646 .11049 -.78 .4339 -.30302 .13010
LOI| .42440*** .09468 4.48 .0000 .23884 .60996
LTFSQ| .15213*** .01164 13.07 .0000 .12932 .17495
LSISQ| .22072*** .02218 9.95 .0000 .17725 .26419
LOISQ| .10024*** .00671 14.93 .0000 .08709 .11340
LPFPC| .11411 .13291 .86 .3906 -.14639 .37462
LPLPC| .79846*** .10715 7.45 .0000 .58845 1.00846
LPFPCSQ| .13804*** .02531 5.45 .0000 .08844 .18763
LPLPCSQ| .06890*** .01819 3.79 .0002 .03325 .10455
LTFSI| -.12015*** .01390 -8.65 .0000 -.14739 -.09291
LTFOI| -.02413*** .00618 -3.91 .0001 -.03624 -.01203
LSIOI| -.08852*** .01207 -7.33 .0000 -.11218 -.06486
LPFCPLC| -.09788*** .02025 -4.83 .0000 -.13756 -.05820
LTFPFPC| .03605*** .01329 2.71 .0067 .01001 .06209
LTFPLPC| -.00625 .01098 -.57 .5691 -.02777 .01527
LSIPFPC| .01034 .01500 .69 .4903 -.01905 .03973
LSIPLPC| -.00577 .01260 -.46 .6472 -.03047 .01894
LOIPFPC| -.02349** .01150 -2.04 .0411 -.04603 -.00095
LOIPLPC| -.02083* .01097 -1.90 .0576 -.04233 .00068
|Variance parameters for compound error
Lambda| 1.89206*** .06778 27.92 .0000 1.75922 2.02490
Sigma(u)| .31544*** .00639 49.37 .0000 .30292 .32796
|Eta parameter for time varying inefficiency
Eta| .05574*** .00799 6.98 .0000 .04008 .07139
--------+--------------------------------------------------------------------
Note: ***, **, * ==> Significance at 1%, 5%, 10% level.
----------------------------------------------------------------------------- ltc = Log (tc/pc); ltf = Log(tf); lsi = Log(si);loi = Log(oi); ltfsq = .5*log(tf)^2; lsisq = .5*log(si)^2; loisq =
.5*log(oi)^2; lpfpc = Log(pf/pc); lplpc = Log(pl/pc); lpfpcsq = .5*log(pf/pc)^2; lplpcsq =.5*log(pl/pc)^2; ltfsi =
log(tf)*log(si); ltfoi = log(tf)*log(oi); lsioi = Log(si)*log(oi); lpfcplc = log(pf/pc)*log(pl/pc); ltfpfpc =
log(tf)*log(pf/pc); ltfplpc = log(tf)*log(pl/pc); lsipfpc = log(si)*log(pf/pc); lsiplpc = log(si)*log(pl/pc); loipfpc
= log(oi)*log(pf/pc); loiplpc = log(oi)*log(pl/pc)
227
Appendix 20. Table 6.19. Coefficient of Cost Efficiency Estimation Variables –
Pooled Model 2
-----------------------------------------------------------------------------
Limited Dependent Variable Model - FRONTIER
Dependent variable LTC
Log likelihood function 205.64236
Estimation based on N = 999, K = 30
Inf.Cr.AIC = -351.3 AIC/N = -.352
Model estimated: Sep 10, 2013, 08:06:09
Stochastic frontier based on panel data
Estimation based on 111 individuals
Variances: Sigma-squared(v)= .03157
Sigma-squared(u)= .06352
Sigma(v) = .17768
Sigma(u) = .25202
Sigma = Sqr[(s^2(u)+s^2(v)]= .30836
Gamma = sigma(u)^2/sigma^2 = .66797
Var[u]/{Var[u]+Var[v]} = .42232
Stochastic Cost Frontier Model, e = v+u
LR test for inefficiency vs. OLS v only
Deg. freedom for sigma-squared(u): 1
Deg. freedom for heteroscedasticity: 0
Deg. freedom for truncation mean: 0
Deg. freedom for inefficiency model: 1
LogL when sigma(u)=0 101.43897
Chi-sq=2*[LogL(SF)-LogL(LS)] = 208.407
Kodde-Palm C*: 95%: 2.706, 99%: 5.412
--------+--------------------------------------------------------------------
| Standard Prob. 95% Confidence
LTC| Coefficient Error z |z|>Z* Interval
--------+--------------------------------------------------------------------
|Deterministic Component of Stochastic Frontier Model
Constant| 3.69130*** .65914 5.60 .0000 2.39941 4.98319
LSBF| .14775*** .01798 8.22 .0000 .11252 .18299
LOF| .41616*** .08166 5.10 .0000 .25612 .57621
LSI| -.30155*** .10875 -2.77 .0056 -.51469 -.08841
LOI| .59995*** .08794 6.82 .0000 .42760 .77231
LSBFSQ| .00992*** .00094 10.53 .0000 .00807 .01176
LOFSQ| .13127*** .00554 23.71 .0000 .12041 .14212
LSISQ| .21010*** .01901 11.05 .0000 .17285 .24736
LOISQ| .11595*** .00812 14.27 .0000 .10003 .13187
LPFPC| .20906 .16767 1.25 .2124 -.11956 .53769
LPLPC| .54858*** .13824 3.97 .0001 .27763 .81952
LPFPCSQ| .14580*** .02715 5.37 .0000 .09259 .19901
LPLPCSQ| .11596*** .02567 4.52 .0000 .06564 .16628
LSBFOF| -.00858*** .00113 -7.58 .0000 -.01080 -.00636
LSBFSI| -.00412** .00191 -2.16 .0310 -.00787 -.00038
LSBFOI| .00386*** .00124 3.11 .0018 .00143 .00629
LOFSI| -.08483*** .01005 -8.44 .0000 -.10453 -.06513
LOFOI| -.04336*** .00452 -9.60 .0000 -.05221 -.03451
LSIOI| -.09520*** .01236 -7.70 .0000 -.11943 -.07097
LPFCPLC| -.13185*** .02179 -6.05 .0000 -.17457 -.08913
LSBFPFPC| .00913*** .00260 3.52 .0004 .00404 .01421
LSBFPLPC| -.00505* .00284 -1.78 .0757 -.01062 .00052
LOFPFPC| .03913** .01655 2.36 .0181 .00668 .07157
LOFPLPC| -.00388 .01573 -.25 .8051 -.03471 .02695
LSIPFPC| -.01020 .01595 -.64 .5224 -.04147 .02107
LSIPLPC| .02010 .01523 1.32 .1870 -.00976 .04996
LOIPFPC| -.02436* .01260 -1.93 .0533 -.04906 .00034
LOIPLPC| -.02782** .01270 -2.19 .0285 -.05272 -.00292
|Variance parameters for compound error
Lambda| 1.41838*** .21623 6.56 .0000 .99458 1.84218
Sigma(u)| .25202*** .03158 7.98 .0000 .19014 .31391
--------+--------------------------------------------------------------------
Note: ***, **, * ==> Significance at 1%, 5%, 10% level.
-----------------------------------------------------------------------------
ltc = Log (tc/pc); lsbf = Log(sbf); lof = Log(of); lsi = Log(si); loi = Log(oi); lsbfsq = .5*log(sbf)^2; lofsq =
.5*log(of)^2; lsisq = .5*log(si)^2; loisq = .5*log(oi)^2; lpfpc = Log(pf/pc); lpfpc = Log(pf/pc); lplpc = Log(pl/pc);
lpfpcsq = .5*log(pf/pc)^2; lplpcsq =.5*log(pl/pc)^2; lsbfof = log(sbf)*log(of); lsbfsi = log(sbf)*log(si); lsbfoi =
log(sbf)*log(oi); lofsi = log(of)*log(si); lofoi = Log(of)*log(oi); lsioi = Log(si)*log(oi); lpfcplc =
log(pf/pc)*log(pl/pc); lsbfpfpc = log(sbf)*log(pf/pc); lsbfplpc = log(sbf)*log(pl/pc); lofpfpc = log(of)*log(pf/pc);
lofplpc = log(of)*log(pl/pc); lsipfpc = log(si)*log(pf/pc); lsiplpc = log(si)*log(pl/pc); loipfpc =
log(oi)*log(pf/pc); loiplpc = log(oi)*log(pl/pc)
228
Appendix 21. Table 6.20. Coefficient of Cost Efficiency Estimation Variables –
BC92 Model 2 -----------------------------------------------------------------------------
Limited Dependent Variable Model - FRONTIER
Dependent variable LTC
Log likelihood function 228.18340
Estimation based on N = 999, K = 31
Inf.Cr.AIC = -394.4 AIC/N = -.395
Model estimated: Sep 10, 2013, 08:09:25
Stochastic frontier based on panel data
Estimation based on 111 individuals
Variances: Sigma-squared(v)= .02955
Sigma-squared(u)= .03594
Sigma(v) = .17189
Sigma(u) = .18957
Sigma = Sqr[(s^2(u)+s^2(v)]= .25589
Gamma = sigma(u)^2/sigma^2 = .54879
Var[u]/{Var[u]+Var[v]} = .30650
Stochastic Cost Frontier Model, e = v+u
Battese-Coelli Models: Time Varying uit
Time dependent uit=exp[-eta(t-T)]*|U(i)|
LR test for inefficiency vs. OLS v only
Deg. freedom for sigma-squared(u): 1
Deg. freedom for heteroscedasticity: 0
Deg. freedom for truncation mean: 0
Deg. freedom for inefficiency model: 1
LogL when sigma(u)=0 101.43897
Chi-sq=2*[LogL(SF)-LogL(LS)] = 253.489
Kodde-Palm C*: 95%: 2.706, 99%: 5.412
--------+--------------------------------------------------------------------
| Standard Prob. 95% Confidence
LTC| Coefficient Error z |z|>Z* Interval
--------+--------------------------------------------------------------------
|Deterministic Component of Stochastic Frontier Model
Constant| 3.29788*** .76292 4.32 .0000 1.80258 4.79318
LSBF| .15162*** .01817 8.35 .0000 .11602 .18723
LOF| .36598*** .10582 3.46 .0005 .15857 .57339
LSI| -.16410 .12004 -1.37 .1716 -.39937 .07117
LOI| .56679*** .10672 5.31 .0000 .35763 .77596
LSBFSQ| .00909*** .00087 10.45 .0000 .00739 .01079
LOFSQ| .13672*** .00603 22.68 .0000 .12491 .14853
LSISQ| .20059*** .02083 9.63 .0000 .15976 .24142
LOISQ| .11771*** .00825 14.26 .0000 .10153 .13388
LPFPC| .02381 .18162 .13 .8957 -.33217 .37979
LPLPC| .81010*** .13924 5.82 .0000 .53720 1.08301
LPFPCSQ| .11784*** .03061 3.85 .0001 .05784 .17784
LPLPCSQ| .09303*** .02367 3.93 .0001 .04664 .13942
LSBFOF| -.00799*** .00116 -6.91 .0000 -.01025 -.00572
LSBFSI| -.00400** .00190 -2.10 .0353 -.00772 -.00028
LSBFOI| .00294** .00131 2.24 .0248 .00037 .00551
LOFSI| -.08769*** .01301 -6.74 .0000 -.11318 -.06219
LOFOI| -.04108*** .00516 -7.96 .0000 -.05120 -.03096
LSIOI| -.09446*** .01366 -6.92 .0000 -.12123 -.06769
LPFCPLC| -.10602*** .02345 -4.52 .0000 -.15198 -.06005
LSBFPFPC| .00637** .00274 2.32 .0203 .00099 .01175
LSBFPLPC| -.00267 .00282 -.95 .3434 -.00820 .00286
LOFPFPC| .04736*** .01619 2.92 .0034 .01562 .07910
LOFPLPC| -.01698 .01500 -1.13 .2578 -.04638 .01243
LSIPFPC| .00321 .01689 .19 .8494 -.02990 .03631
LSIPLPC| .00526 .01658 .32 .7511 -.02724 .03776
LOIPFPC| -.03323** .01382 -2.41 .0162 -.06031 -.00615
LOIPLPC| -.01593 .01314 -1.21 .2253 -.04167 .00982
|Variance parameters for compound error
Lambda| 1.10283*** .07160 15.40 .0000 .96249 1.24317
Sigma(u)| .18957*** .00170 111.43 .0000 .18623 .19290
|Eta parameter for time varying inefficiency
Eta| .08354*** .01172 7.13 .0000 .06057 .10650
--------+--------------------------------------------------------------------
Note: ***, **, * ==> Significance at 1%, 5%, 10% level.
----------------------------------------------------------------------------- ltc = Log (tc/pc); lsbf = Log(sbf); lof = Log(of); lsi = Log(si); loi = Log(oi); lsbfsq = .5*log(sbf)^2; lofsq =
.5*log(of)^2; lsisq = .5*log(si)^2; loisq = .5*log(oi)^2; lpfpc = Log(pf/pc); lpfpc = Log(pf/pc); lplpc = Log(pl/pc);
lpfpcsq = .5*log(pf/pc)^2; lplpcsq =.5*log(pl/pc)^2; lsbfof = log(sbf)*log(of); lsbfsi = log(sbf)*log(si); lsbfoi =
log(sbf)*log(oi); lofsi = log(of)*log(si); lofoi = Log(of)*log(oi); lsioi = Log(si)*log(oi); lpfcplc =
log(pf/pc)*log(pl/pc); lsbfpfpc = log(sbf)*log(pf/pc); lsbfplpc = log(sbf)*log(pl/pc); lofpfpc = log(of)*log(pf/pc);
lofplpc = log(of)*log(pl/pc); lsipfpc = log(si)*log(pf/pc); lsiplpc = log(si)*log(pl/pc); loipfpc =
log(oi)*log(pf/pc); loiplpc = log(oi)*log(pl/pc)
229
Appendix 22. Table 6.21. Coefficient of Cost Efficiency Estimation Variables –
Pooled Model 3
-----------------------------------------------------------------------------
Limited Dependent Variable Model - FRONTIER
Dependent variable LTC
Log likelihood function 55.47266
Estimation based on N = 999, K = 23
Inf.Cr.AIC = -64.9 AIC/N = -.065
Model estimated: Sep 10, 2013, 08:34:17
Stochastic frontier based on panel data
Estimation based on 111 individuals
Variances: Sigma-squared(v)= .03653
Sigma-squared(u)= .36685
Sigma(v) = .19112
Sigma(u) = .60568
Sigma = Sqr[(s^2(u)+s^2(v)]= .63512
Gamma = sigma(u)^2/sigma^2 = .90945
Var[u]/{Var[u]+Var[v]} = .78493
Stochastic Cost Frontier Model, e = v+u
LR test for inefficiency vs. OLS v only
Deg. freedom for sigma-squared(u): 1
Deg. freedom for heteroscedasticity: 0
Deg. freedom for truncation mean: 0
Deg. freedom for inefficiency model: 1
LogL when sigma(u)=0 -157.81325
Chi-sq=2*[LogL(SF)-LogL(LS)] = 426.572
Kodde-Palm C*: 95%: 2.706, 99%: 5.412
--------+--------------------------------------------------------------------
| Standard Prob. 95% Confidence
LTC| Coefficient Error z |z|>Z* Interval
--------+--------------------------------------------------------------------
|Deterministic Component of Stochastic Frontier Model
Constant| 4.38918*** .52530 8.36 .0000 3.35962 5.41874
LOF| .51971*** .08301 6.26 .0000 .35702 .68241
LSI| -.12576 .10187 -1.23 .2170 -.32543 .07390
LOI| .31507*** .10090 3.12 .0018 .11731 .51284
LOFSQ| .12751*** .00748 17.05 .0000 .11285 .14216
LSISQ| .20676*** .01560 13.25 .0000 .17618 .23733
LOISQ| .08634*** .00716 12.07 .0000 .07232 .10037
LPFPC| .19458 .13804 1.41 .1587 -.07597 .46514
LPLPC| .80901*** .12921 6.26 .0000 .55576 1.06225
LPFPCSQ| .12640*** .02881 4.39 .0000 .06994 .18285
LPLPCSQ| .03854* .02210 1.74 .0812 -.00478 .08186
LOFSI| -.11462*** .00937 -12.23 .0000 -.13299 -.09626
LOFOI| -.01935*** .00573 -3.38 .0007 -.03057 -.00813
LSIOI| -.07214*** .01138 -6.34 .0000 -.09444 -.04985
LPFPCPLP| -.07568*** .02476 -3.06 .0022 -.12421 -.02715
LOFPFPC| .05602*** .01077 5.20 .0000 .03492 .07713
LOFPLPC| -.03571*** .01052 -3.39 .0007 -.05633 -.01509
LSIPFPC| .00258 .01441 .18 .8578 -.02566 .03082
LSIPLPC| .00401 .01360 .30 .7679 -.02263 .03066
LOIPFPC| -.03727*** .01220 -3.05 .0023 -.06117 -.01336
LOIPLPC| -.00727 .01174 -.62 .5356 -.03027 .01573
|Variance parameters for compound error
Lambda| 3.16917*** .91135 3.48 .0005 1.38296 4.95538
Sigma(u)| .60568*** .09762 6.20 .0000 .41436 .79700
--------+--------------------------------------------------------------------
Note: ***, **, * ==> Significance at 1%, 5%, 10% level.
-----------------------------------------------------------------------------
ltc = Log(tc/pc); lof = Log(of); lsi = Log(si); loi = Log(oi); lofsq = .5*Log(of)^2; lsisq = .5*Log(si)^2; loisq =
.5*Log(oi)^2; lpfpc = Log(pf/pc); lplpc = Log(pl/pc); lpfpcsq = .5*Log(pf/pc)^2; lplpcsq = .5*Log(pl/pc)^2; lofsi =
Log(of)*Log(si); lofoi = Log(of)*Log(oi); lsioi = Log(si)*log(oi); lpfpcplp = Log(pf/pc)*Log(pl/pc); lofpfpc =
Log(of)*Log(pf/pc); lofplpc = Log(of)*Log(pl/pc); lsipfpc = Log(si)*Log(pf/pc); lsiplpc = Log(si)*Log(pl/pc); loipfpc
= Log(oi)*Log(pf/pc; loiplpc = Log(oi)*Log(pl/pc)
230
Appendix 23. Table 6.22. Coefficient of Cost Efficiency Estimation Variables –
BC92 Model 3 -----------------------------------------------------------------------------
Limited Dependent Variable Model - FRONTIER
Dependent variable LTC
Log likelihood function 88.80593
Estimation based on N = 999, K = 24
Inf.Cr.AIC = -129.6 AIC/N = -.130
Model estimated: Sep 10, 2013, 08:35:17
Stochastic frontier based on panel data
Estimation based on 111 individuals
Variances: Sigma-squared(v)= .03475
Sigma-squared(u)= .18088
Sigma(v) = .18641
Sigma(u) = .42530
Sigma = Sqr[(s^2(u)+s^2(v)]= .46436
Gamma = sigma(u)^2/sigma^2 = .83885
Var[u]/{Var[u]+Var[v]} = .65416
Stochastic Cost Frontier Model, e = v+u
Battese-Coelli Models: Time Varying uit
Time dependent uit=exp[-eta(t-T)]*|U(i)|
LR test for inefficiency vs. OLS v only
Deg. freedom for sigma-squared(u): 1
Deg. freedom for heteroscedasticity: 0
Deg. freedom for truncation mean: 0
Deg. freedom for inefficiency model: 1
LogL when sigma(u)=0 -157.81325
Chi-sq=2*[LogL(SF)-LogL(LS)] = 493.238
Kodde-Palm C*: 95%: 2.706, 99%: 5.412
--------+--------------------------------------------------------------------
| Standard Prob. 95% Confidence
LTC| Coefficient Error z |z|>Z* Interval
--------+--------------------------------------------------------------------
|Deterministic Component of Stochastic Frontier Model
Constant| 3.95999*** .75833 5.22 .0000 2.47369 5.44630
LOF| .40880*** .09808 4.17 .0000 .21657 .60104
LSI| -.01354 .12656 -.11 .9148 -.26160 .23452
LOI| .33506*** .10910 3.07 .0021 .12122 .54890
LOFSQ| .13934*** .00881 15.82 .0000 .12208 .15660
LSISQ| .19508*** .01898 10.28 .0000 .15789 .23227
LOISQ| .09064*** .00703 12.89 .0000 .07686 .10442
LPFPC| .03145 .14309 .22 .8260 -.24899 .31189
LPLPC| .92927*** .12787 7.27 .0000 .67865 1.17990
LPFPCSQ| .10420*** .03136 3.32 .0009 .04274 .16566
LPLPCSQ| .03500 .02271 1.54 .1232 -.00950 .07950
LOFSI| -.11067*** .01000 -11.07 .0000 -.13027 -.09106
LOFOI| -.02173*** .00662 -3.28 .0010 -.03469 -.00876
LSIOI| -.07403*** .01210 -6.12 .0000 -.09775 -.05031
LPFPCPLP| -.06411** .02605 -2.46 .0139 -.11517 -.01304
LOFPFPC| .06151*** .01140 5.40 .0000 .03917 .08385
LOFPLPC| -.03478*** .01071 -3.25 .0012 -.05577 -.01380
LSIPFPC| .00349 .01493 .23 .8151 -.02577 .03275
LSIPLPC| -.00065 .01338 -.05 .9611 -.02688 .02558
LOIPFPC| -.03572*** .01326 -2.69 .0071 -.06170 -.00973
LOIPLPC| -.00670 .01225 -.55 .5844 -.03072 .01732
|Variance parameters for compound error
Lambda| 2.28151*** .05544 41.15 .0000 2.17285 2.39018
Sigma(u)| .42530*** .01355 31.38 .0000 .39874 .45186
|Eta parameter for time varying inefficiency
Eta| .05400*** .00520 10.38 .0000 .04380 .06420
--------+--------------------------------------------------------------------
Note: ***, **, * ==> Significance at 1%, 5%, 10% level.
----------------------------------------------------------------------------- ltc = Log(tc/pc); lof = Log(of); lsi = Log(si); loi = Log(oi); lofsq = .5*Log(of)^2; lsisq = .5*Log(si)^2; loisq =
.5*Log(oi)^2; lpfpc = Log(pf/pc); lplpc = Log(pl/pc); lpfpcsq = .5*Log(pf/pc)^2; lplpcsq = .5*Log(pl/pc)^2; lofsi =
Log(of)*Log(si); lofoi = Log(of)*Log(oi); lsioi = Log(si)*log(oi); lpfpcplp = Log(pf/pc)*Log(pl/pc); lofpfpc =
Log(of)*Log(pf/pc); lofplpc = Log(of)*Log(pl/pc); lsipfpc = Log(si)*Log(pf/pc); lsiplpc = Log(si)*Log(pl/pc); loipfpc
= Log(oi)*Log(pf/pc; loiplpc = Log(oi)*Log(pl/pc)
231
Appendix 24. FIRST REGRESSIONS IN COMPLETION TO TABLE 7.3 (THE
TSLS REGRESSION RESULTS 1)
xtivreg lntl lnta capratio lndep ( p= gdpgr) , fe small first
xtivreg lntl lnta capratio lndep ( p= infl ) , fe small first
xtivreg lntl lnta capratio lndep ( p= unemp ) , fe small first
F test that all u_i=0: F(108, 868) = 4.79 Prob > F = 0.0000 rho .70754853 (fraction of variance due to u_i) sigma_e .01575868 sigma_u .02451156 _cons .3941417 .0138213 28.52 0.000 .3670145 .4212688 gdpgr -.4414485 .0903354 -4.89 0.000 -.6187498 -.2641472 lndep -.0027363 .0029095 -0.94 0.347 -.0084468 .0029742 capratio -.051727 .0091682 -5.64 0.000 -.0697215 -.0337325 lnta -.0119081 .0030424 -3.91 0.000 -.0178794 -.0059368 p Coef. Std. Err. t P>|t| [95% Conf. Interval]
corr(u_i, Xb) = -0.8219 Prob > F = 0.0000 F(4,868) = 104.84
overall = 0.0731 max = 9 between = 0.0250 avg = 9.0R-sq: within = 0.3257 Obs per group: min = 9
Group variable: no Number of groups = 109Fixed-effects (within) regression Number of obs = 981
First-stage within regression
F test that all u_i=0: F(108, 868) = 6.66 Prob > F = 0.0000 rho .76261529 (fraction of variance due to u_i) sigma_e .01396157 sigma_u .02502423 _cons .3574351 .0123695 28.90 0.000 .3331574 .3817127 infl .1890166 .0115404 16.38 0.000 .1663662 .2116671 lndep -.0033422 .0025769 -1.30 0.195 -.0083999 .0017155 capratio -.048485 .0080627 -6.01 0.000 -.0643097 -.0326603 lnta -.011505 .0026721 -4.31 0.000 -.0167495 -.0062605 p Coef. Std. Err. t P>|t| [95% Conf. Interval]
corr(u_i, Xb) = -0.8013 Prob > F = 0.0000 F(4,868) = 193.02
overall = 0.1143 max = 9 between = 0.0229 avg = 9.0R-sq: within = 0.4708 Obs per group: min = 9
Group variable: no Number of groups = 109Fixed-effects (within) regression Number of obs = 981
First-stage within regression
F test that all u_i=0: F(108, 868) = 4.69 Prob > F = 0.0000 rho .6930807 (fraction of variance due to u_i) sigma_e .01572521 sigma_u .02363068 _cons .3330471 .0196571 16.94 0.000 .294466 .3716282 unemp .2758586 .0524321 5.26 0.000 .17295 .3787673 lndep -.0040865 .0029084 -1.41 0.160 -.0097949 .0016218 capratio -.0480258 .0092554 -5.19 0.000 -.0661913 -.0298603 lnta -.0098214 .0031097 -3.16 0.002 -.0159248 -.003718 p Coef. Std. Err. t P>|t| [95% Conf. Interval]
corr(u_i, Xb) = -0.8068 Prob > F = 0.0000 F(4,868) = 106.21
overall = 0.0753 max = 9 between = 0.0219 avg = 9.0R-sq: within = 0.3286 Obs per group: min = 9
Group variable: no Number of groups = 109Fixed-effects (within) regression Number of obs = 981
First-stage within regression
232
xtivreg tlta lnta capratio lndep ( p= gdpgr) , fe small first
xtivreg tlta lnta capratio lndep ( p= infl ) , fe small first
xtivreg tlta lnta capratio lndep ( p= unemp ) , fe small first
F test that all u_i=0: F(108, 868) = 4.79 Prob > F = 0.0000 rho .70754853 (fraction of variance due to u_i) sigma_e .01575868 sigma_u .02451156 _cons .3941417 .0138213 28.52 0.000 .3670145 .4212688 gdpgr -.4414485 .0903354 -4.89 0.000 -.6187498 -.2641472 lndep -.0027363 .0029095 -0.94 0.347 -.0084468 .0029742 capratio -.051727 .0091682 -5.64 0.000 -.0697215 -.0337325 lnta -.0119081 .0030424 -3.91 0.000 -.0178794 -.0059368 p Coef. Std. Err. t P>|t| [95% Conf. Interval]
corr(u_i, Xb) = -0.8219 Prob > F = 0.0000 F(4,868) = 104.84
overall = 0.0731 max = 9 between = 0.0250 avg = 9.0R-sq: within = 0.3257 Obs per group: min = 9
Group variable: no Number of groups = 109Fixed-effects (within) regression Number of obs = 981
First-stage within regression
F test that all u_i=0: F(108, 868) = 6.66 Prob > F = 0.0000 rho .76261529 (fraction of variance due to u_i) sigma_e .01396157 sigma_u .02502423 _cons .3574351 .0123695 28.90 0.000 .3331574 .3817127 infl .1890166 .0115404 16.38 0.000 .1663662 .2116671 lndep -.0033422 .0025769 -1.30 0.195 -.0083999 .0017155 capratio -.048485 .0080627 -6.01 0.000 -.0643097 -.0326603 lnta -.011505 .0026721 -4.31 0.000 -.0167495 -.0062605 p Coef. Std. Err. t P>|t| [95% Conf. Interval]
corr(u_i, Xb) = -0.8013 Prob > F = 0.0000 F(4,868) = 193.02
overall = 0.1143 max = 9 between = 0.0229 avg = 9.0R-sq: within = 0.4708 Obs per group: min = 9
Group variable: no Number of groups = 109Fixed-effects (within) regression Number of obs = 981
First-stage within regression
F test that all u_i=0: F(108, 868) = 4.69 Prob > F = 0.0000 rho .6930807 (fraction of variance due to u_i) sigma_e .01572521 sigma_u .02363068 _cons .3330471 .0196571 16.94 0.000 .294466 .3716282 unemp .2758586 .0524321 5.26 0.000 .17295 .3787673 lndep -.0040865 .0029084 -1.41 0.160 -.0097949 .0016218 capratio -.0480258 .0092554 -5.19 0.000 -.0661913 -.0298603 lnta -.0098214 .0031097 -3.16 0.002 -.0159248 -.003718 p Coef. Std. Err. t P>|t| [95% Conf. Interval]
corr(u_i, Xb) = -0.8068 Prob > F = 0.0000 F(4,868) = 106.21
overall = 0.0753 max = 9 between = 0.0219 avg = 9.0R-sq: within = 0.3286 Obs per group: min = 9
Group variable: no Number of groups = 109Fixed-effects (within) regression Number of obs = 981
First-stage within regression
233
xtivreg tltd lnta capratio lndep ( p= gdpgr) , fe small first
xtivreg tltd lnta capratio lndep ( p= infl ) , fe small first
xtivreg tltd lnta capratio lndep ( p= unemp ) , fe small first
F test that all u_i=0: F(108, 868) = 4.79 Prob > F = 0.0000 rho .70754853 (fraction of variance due to u_i) sigma_e .01575868 sigma_u .02451156 _cons .3941417 .0138213 28.52 0.000 .3670145 .4212688 gdpgr -.4414485 .0903354 -4.89 0.000 -.6187498 -.2641472 lndep -.0027363 .0029095 -0.94 0.347 -.0084468 .0029742 capratio -.051727 .0091682 -5.64 0.000 -.0697215 -.0337325 lnta -.0119081 .0030424 -3.91 0.000 -.0178794 -.0059368 p Coef. Std. Err. t P>|t| [95% Conf. Interval]
corr(u_i, Xb) = -0.8219 Prob > F = 0.0000 F(4,868) = 104.84
overall = 0.0731 max = 9 between = 0.0250 avg = 9.0R-sq: within = 0.3257 Obs per group: min = 9
Group variable: no Number of groups = 109Fixed-effects (within) regression Number of obs = 981
First-stage within regression
F test that all u_i=0: F(108, 868) = 6.66 Prob > F = 0.0000 rho .76261529 (fraction of variance due to u_i) sigma_e .01396157 sigma_u .02502423 _cons .3574351 .0123695 28.90 0.000 .3331574 .3817127 infl .1890166 .0115404 16.38 0.000 .1663662 .2116671 lndep -.0033422 .0025769 -1.30 0.195 -.0083999 .0017155 capratio -.048485 .0080627 -6.01 0.000 -.0643097 -.0326603 lnta -.011505 .0026721 -4.31 0.000 -.0167495 -.0062605 p Coef. Std. Err. t P>|t| [95% Conf. Interval]
corr(u_i, Xb) = -0.8013 Prob > F = 0.0000 F(4,868) = 193.02
overall = 0.1143 max = 9 between = 0.0229 avg = 9.0R-sq: within = 0.4708 Obs per group: min = 9
Group variable: no Number of groups = 109Fixed-effects (within) regression Number of obs = 981
First-stage within regression
F test that all u_i=0: F(108, 868) = 4.69 Prob > F = 0.0000 rho .6930807 (fraction of variance due to u_i) sigma_e .01572521 sigma_u .02363068 _cons .3330471 .0196571 16.94 0.000 .294466 .3716282 unemp .2758586 .0524321 5.26 0.000 .17295 .3787673 lndep -.0040865 .0029084 -1.41 0.160 -.0097949 .0016218 capratio -.0480258 .0092554 -5.19 0.000 -.0661913 -.0298603 lnta -.0098214 .0031097 -3.16 0.002 -.0159248 -.003718 p Coef. Std. Err. t P>|t| [95% Conf. Interval]
corr(u_i, Xb) = -0.8068 Prob > F = 0.0000 F(4,868) = 106.21
overall = 0.0753 max = 9 between = 0.0219 avg = 9.0R-sq: within = 0.3286 Obs per group: min = 9
Group variable: no Number of groups = 109Fixed-effects (within) regression Number of obs = 981
First-stage within regression
234
Appendix 25. FIRST REGRESSIONS IN COMPLETION TO TABLE 7.4 (THE
TSLS REGRESSION RESULTS 2)
xtivreg lnsbl lnta capratio lndep ( p= gdpgr) , fe small first
xtivreg lnsbl lnta capratio lndep ( p= infl ) , fe small first
xtivreg lnsbl lnta capratio lndep ( p= unemp ) , fe small first
F test that all u_i=0: F(108, 868) = 4.79 Prob > F = 0.0000 rho .70754853 (fraction of variance due to u_i) sigma_e .01575868 sigma_u .02451156 _cons .3941417 .0138213 28.52 0.000 .3670145 .4212688 gdpgr -.4414485 .0903354 -4.89 0.000 -.6187498 -.2641472 lndep -.0027363 .0029095 -0.94 0.347 -.0084468 .0029742 capratio -.051727 .0091682 -5.64 0.000 -.0697215 -.0337325 lnta -.0119081 .0030424 -3.91 0.000 -.0178794 -.0059368 p Coef. Std. Err. t P>|t| [95% Conf. Interval]
corr(u_i, Xb) = -0.8219 Prob > F = 0.0000 F(4,868) = 104.84
overall = 0.0731 max = 9 between = 0.0250 avg = 9.0R-sq: within = 0.3257 Obs per group: min = 9
Group variable: no Number of groups = 109Fixed-effects (within) regression Number of obs = 981
First-stage within regression
F test that all u_i=0: F(108, 868) = 6.66 Prob > F = 0.0000 rho .76261529 (fraction of variance due to u_i) sigma_e .01396157 sigma_u .02502423 _cons .3574351 .0123695 28.90 0.000 .3331574 .3817127 infl .1890166 .0115404 16.38 0.000 .1663662 .2116671 lndep -.0033422 .0025769 -1.30 0.195 -.0083999 .0017155 capratio -.048485 .0080627 -6.01 0.000 -.0643097 -.0326603 lnta -.011505 .0026721 -4.31 0.000 -.0167495 -.0062605 p Coef. Std. Err. t P>|t| [95% Conf. Interval]
corr(u_i, Xb) = -0.8013 Prob > F = 0.0000 F(4,868) = 193.02
overall = 0.1143 max = 9 between = 0.0229 avg = 9.0R-sq: within = 0.4708 Obs per group: min = 9
Group variable: no Number of groups = 109Fixed-effects (within) regression Number of obs = 981
F test that all u_i=0: F(108, 868) = 4.69 Prob > F = 0.0000 rho .6930807 (fraction of variance due to u_i) sigma_e .01572521 sigma_u .02363068 _cons .3330471 .0196571 16.94 0.000 .294466 .3716282 unemp .2758586 .0524321 5.26 0.000 .17295 .3787673 lndep -.0040865 .0029084 -1.41 0.160 -.0097949 .0016218 capratio -.0480258 .0092554 -5.19 0.000 -.0661913 -.0298603 lnta -.0098214 .0031097 -3.16 0.002 -.0159248 -.003718 p Coef. Std. Err. t P>|t| [95% Conf. Interval]
corr(u_i, Xb) = -0.8068 Prob > F = 0.0000 F(4,868) = 106.21
overall = 0.0753 max = 9 between = 0.0219 avg = 9.0R-sq: within = 0.3286 Obs per group: min = 9
Group variable: no Number of groups = 109Fixed-effects (within) regression Number of obs = 981
First-stage within regression
235
xtivreg sblta lnta capratio lndep ( p= gdpgr) , fe small first
xtivreg sblta lnta capratio lndep ( p= infl ) , fe small first
xtivreg sblta lnta capratio lndep ( p= unemp ) , fe small first
F test that all u_i=0: F(108, 868) = 4.79 Prob > F = 0.0000 rho .70754853 (fraction of variance due to u_i) sigma_e .01575868 sigma_u .02451156 _cons .3941417 .0138213 28.52 0.000 .3670145 .4212688 gdpgr -.4414485 .0903354 -4.89 0.000 -.6187498 -.2641472 lndep -.0027363 .0029095 -0.94 0.347 -.0084468 .0029742 capratio -.051727 .0091682 -5.64 0.000 -.0697215 -.0337325 lnta -.0119081 .0030424 -3.91 0.000 -.0178794 -.0059368 p Coef. Std. Err. t P>|t| [95% Conf. Interval]
corr(u_i, Xb) = -0.8219 Prob > F = 0.0000 F(4,868) = 104.84
overall = 0.0731 max = 9 between = 0.0250 avg = 9.0R-sq: within = 0.3257 Obs per group: min = 9
Group variable: no Number of groups = 109Fixed-effects (within) regression Number of obs = 981
First-stage within regression
F test that all u_i=0: F(108, 868) = 6.66 Prob > F = 0.0000 rho .76261529 (fraction of variance due to u_i) sigma_e .01396157 sigma_u .02502423 _cons .3574351 .0123695 28.90 0.000 .3331574 .3817127 infl .1890166 .0115404 16.38 0.000 .1663662 .2116671 lndep -.0033422 .0025769 -1.30 0.195 -.0083999 .0017155 capratio -.048485 .0080627 -6.01 0.000 -.0643097 -.0326603 lnta -.011505 .0026721 -4.31 0.000 -.0167495 -.0062605 p Coef. Std. Err. t P>|t| [95% Conf. Interval]
corr(u_i, Xb) = -0.8013 Prob > F = 0.0000 F(4,868) = 193.02
overall = 0.1143 max = 9 between = 0.0229 avg = 9.0R-sq: within = 0.4708 Obs per group: min = 9
Group variable: no Number of groups = 109Fixed-effects (within) regression Number of obs = 981
First-stage within regression
F test that all u_i=0: F(108, 868) = 4.69 Prob > F = 0.0000 rho .6930807 (fraction of variance due to u_i) sigma_e .01572521 sigma_u .02363068 _cons .3330471 .0196571 16.94 0.000 .294466 .3716282 unemp .2758586 .0524321 5.26 0.000 .17295 .3787673 lndep -.0040865 .0029084 -1.41 0.160 -.0097949 .0016218 capratio -.0480258 .0092554 -5.19 0.000 -.0661913 -.0298603 lnta -.0098214 .0031097 -3.16 0.002 -.0159248 -.003718 p Coef. Std. Err. t P>|t| [95% Conf. Interval]
corr(u_i, Xb) = -0.8068 Prob > F = 0.0000 F(4,868) = 106.21
overall = 0.0753 max = 9 between = 0.0219 avg = 9.0R-sq: within = 0.3286 Obs per group: min = 9
Group variable: no Number of groups = 109Fixed-effects (within) regression Number of obs = 981
First-stage within regression
236
xtivreg sbltl lnta capratio lndep ( p= gdpgr) , fe small first
xtivreg sbltl lnta capratio lndep ( p= infl ) , fe small first
xtivreg sbltl lnta capratio lndep ( p= unemp ) , fe small first
F test that all u_i=0: F(108, 868) = 4.79 Prob > F = 0.0000 rho .70754853 (fraction of variance due to u_i) sigma_e .01575868 sigma_u .02451156 _cons .3941417 .0138213 28.52 0.000 .3670145 .4212688 gdpgr -.4414485 .0903354 -4.89 0.000 -.6187498 -.2641472 lndep -.0027363 .0029095 -0.94 0.347 -.0084468 .0029742 capratio -.051727 .0091682 -5.64 0.000 -.0697215 -.0337325 lnta -.0119081 .0030424 -3.91 0.000 -.0178794 -.0059368 p Coef. Std. Err. t P>|t| [95% Conf. Interval]
corr(u_i, Xb) = -0.8219 Prob > F = 0.0000 F(4,868) = 104.84
overall = 0.0731 max = 9 between = 0.0250 avg = 9.0R-sq: within = 0.3257 Obs per group: min = 9
Group variable: no Number of groups = 109Fixed-effects (within) regression Number of obs = 981
First-stage within regression
F test that all u_i=0: F(108, 868) = 6.66 Prob > F = 0.0000 rho .76261529 (fraction of variance due to u_i) sigma_e .01396157 sigma_u .02502423 _cons .3574351 .0123695 28.90 0.000 .3331574 .3817127 infl .1890166 .0115404 16.38 0.000 .1663662 .2116671 lndep -.0033422 .0025769 -1.30 0.195 -.0083999 .0017155 capratio -.048485 .0080627 -6.01 0.000 -.0643097 -.0326603 lnta -.011505 .0026721 -4.31 0.000 -.0167495 -.0062605 p Coef. Std. Err. t P>|t| [95% Conf. Interval]
corr(u_i, Xb) = -0.8013 Prob > F = 0.0000 F(4,868) = 193.02
overall = 0.1143 max = 9 between = 0.0229 avg = 9.0R-sq: within = 0.4708 Obs per group: min = 9
Group variable: no Number of groups = 109Fixed-effects (within) regression Number of obs = 981
First-stage within regression
F test that all u_i=0: F(108, 868) = 4.69 Prob > F = 0.0000 rho .6930807 (fraction of variance due to u_i) sigma_e .01572521 sigma_u .02363068 _cons .3330471 .0196571 16.94 0.000 .294466 .3716282 unemp .2758586 .0524321 5.26 0.000 .17295 .3787673 lndep -.0040865 .0029084 -1.41 0.160 -.0097949 .0016218 capratio -.0480258 .0092554 -5.19 0.000 -.0661913 -.0298603 lnta -.0098214 .0031097 -3.16 0.002 -.0159248 -.003718 p Coef. Std. Err. t P>|t| [95% Conf. Interval]
corr(u_i, Xb) = -0.8068 Prob > F = 0.0000 F(4,868) = 106.21
overall = 0.0753 max = 9 between = 0.0219 avg = 9.0R-sq: within = 0.3286 Obs per group: min = 9
Group variable: no Number of groups = 109Fixed-effects (within) regression Number of obs = 981
First-stage within regression
237
Appendix 26. Correlation Test
0.0000 mktindex 0.5662 1.0000 gdpgr 1.0000 gdpgr mktindex
. pwcorr gdpgr mktindex,sig
0.0000 mktindex -0.6426 1.0000 unemp 1.0000 unemp mktindex
. pwcorr unemp mktindex,sig