technical efficiency and manufacturing ......exploring the determinants of export performance as...
TRANSCRIPT
TECHNICAL EFFICIENCY AND MANUFACTURING EXPORT
PERFORMANCE IN CAMEROON
A Firm Level Analysis
Ngeh Ernest Tingum
Ph.D. (Economics) Dissertation University of Dar es Salaam
October, 2014
TECHNICAL EFFICIENCY AND MANUFACTURING EXPORT
PERFORMANCE IN CAMEROON
A Firm Level Analysis
By
Ngeh Ernest Tingum
A Dissertation Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy (Economics) of the University of Dar es Salaam
University of Dar es Salaam October, 2014
i
CERTIFICATION
The undersigned certify that they have read and hereby recommend for acceptance by the
University of Dar es Salaam a dissertation titled: Technical Efficiency and
Manufacturing Export Performance in Cameroon: A Firm Level Analysis, in partial
fulfilment of the requirements for the degree of Doctor of Philosophy (Economics) of the
University of Dar es Salaam.
………………………………………
Prof. Kidane Asmerom
(Supervisor)
Date…………………………………
.....................................................
Prof. Mbelle Ammon
(Supervisor)
Date………………………………
ii
DECLARATION
AND
COPYRIGHT
I, Ngeh Ernest Tingum, declare that this dissertation is my own original work and that it
has not been presented and will not be presented to any other University for a similar or
any other degree award.
Signature…………………………………
This dissertation is copyright material protected under the Berne Convention, the
Copyright Act 1999 and other international and national enactments, in that behalf, on
intellectual property. It may not be reproduced by any means, in full or in part, except for
short extracts in fair dealings, for research or private study, critical scholarly review or
discourse with an acknowledgement, without the written permission of the School of
Graduate Studies, on behalf of both the author and the University of Dar es Salaam.
iii
ACKNOWLEDGEMENT
“When someone is in the crowd and certainly stands out from the crowd, it is usually
because he/she has been carried on the shoulders of others.” A number of individuals and
organizations have immensely contributed to the completion of my graduate studies and
all deserve my tribute. Above all, I am truly indebted to God the Almighty for His grace
which has enabled me to complete the entire course.
I owe special appreciation to my supervisors, Prof. Kidane Asmerom and Prof. Mbelle
Ammon for their guidance, constructive comments and support throughout the course of
my Ph.D studies. I acknowledge them with great humility.
I express profound thanks to the African Economic Research Consortium (AERC) for
awarding me the scholarship and providing the necessary financial support that enabled me
to persue my Ph.D program. My warm thanks go to the group of resource persons and
researchers who gave me the constructive comments during my presentations.
I am indebted to the Department of Economics of the University of Dar es Salaam for
admitting me to pursue my Ph.D studies. My special thanks to the Head of Department
and all the staff members for the support and constructive comments during my
presentations at the departmental seminars. My profound thanks to the course work
lecturers both at the University of Dar es Salaam and at Joint Facility for Electives (2010)
iv
for laying the necessary theoretical and technical foundation that I will draw from for the
rest of my life. To all my class mates, I say thanks.
I express my heartfelt thanks to the Wage Indicator Foundation especially the Director
Osse Pauline and others such as Dani Ceccon, Iftikar, Arcade, Prof Kea, Tendayi and Prof
Kahyarara for also being the discussant during the advanced seminar presentation.
This work would not have been accomplished without the support from my family and
friends. I am truly proud of my father, mother and brother who have always been there for
me. I am grateful to my uncles and aunties, all my cousins, all my nephews and nieces.
I am thankful to my fellow class mates in Dar es Salaam and Ph.D students in the continent
(CPP class of 2010) for their support during the program. I thank Prof. Tafah Edokat, Prof.
Sondengam and family, Dr Njong, Dr Tabi, Akwi Tafah, Rene Oteh, Sakwe Gervis, Dr
Eno, Ngwi, Nebah Cletus and Nicoline Enugisawnyoh for their ever encouraging words.
Last but not the least, I am grateful to all the people including the friends I interacted with
and from whom I benefited in one way or the other during my study. I owe special
gratitude to the members of the Cameroon Community in Dar es Salaam for the moral and
material support that made my stay in Dar es Salaam both comfortable and memorable.
Nevertheless, while I do acknowledge the contribution of all who assisted me in one way
or the other, I remain solely responsible for the content of this study.
v
DEDICATION
This dissertation is dedicated to my family: my wife – Lizette Neng Sala, my son -Lemuel
Afumbom Tingum, my Dad – Ngeh Emmanuel Nkwain, my Mom – Regina Nsengoin and
my Brother – Ngeh Godlove Nto’oh.
vi
ABBREVIATIONS AND ACCRONYMS
AE Allocative Efficiency
AERC African Economic Research Consortium
CD Cobb-Douglas
CE Cost Efficiency
CEMAC Central African Economic and Monetary Community
CFA Communauté Financière de l’Afrique
COLS Corrected Ordinary Least Squares
DEA Data Envelopment Analysis
DMU Decision Making Units
EE Economic Efficiency
EVA Economic Value Added
FDI Foreign Direct Investment
GDP Gross Domestic Product
GLS Generalized Least Squares
GQBC Generalized Quadratic Box-Cox
ISO International Standards Organization
LDC Least Developed Countries
LT Low Technology
MDG Millennium Development Goals
ML Maximum Likelihood
MLE Maximum Likelihood Estimation
MOLS Modified Ordinary Least Squares
MVA Manufacturing Value Added
NIS National Institute of Statistics
OECD Organization for Economic Co-operation and Development
OLS Ordinary Least Squares
PTA Preferential Trade Agreement
vii
QR Quantitative restrictions
R&D Research and Development
RGE Recensement General des Entrprises
ROW Rest of the World
RPED Regional Programme on Enterprise Development
SAP Structural Adjustment Programme
SFA Stochastic Frontier Analysis
SSA Sub Saharan Africa
STT Standard Time Trend
TE Technical Efficiency
TFP Total Factor Productivity
TP Total Product
UNDP United Nations Development Program
UNIDO United Nations Industrial Development Organization
US United States
WB World Bank
WDI World Development Indicators
WTO World Trade Organization
viii
ABSTRACT
This dissertation addresses the following related issues: efficiency of manufacturing firms,
determinants of technical efficiency of Cameroonian manufacturing firms. It further
investigates the relationship between technical efficiency and export performance while
exploring the determinants of export performance as well. The study employs Stochastic
Frontier Analysis (SFA) to study the technical efficiency of the manufacturing firms and
Probit and Tobit models to examine the determinants of export performance of firms. The
main finding of the study is that most manufacturing firms in Cameroon were technically
inefficient. The most efficient firms are from the food processing sector, followed by wood
and furniture. Firms with 5 to 20 years of operation experience in Cameroon were found to be
more efficient. With regards to the determinants of manufacturing export performance, the
Probit and Tobit models of manufacturing export performance are estimated. The results show
that higher level of efficiency, firm size, foreign ownership, lower tax rates, producing in the
industrial zone, and being in the food processing and textile sectors are the major determinants
of propensity to export and decision to export or not. The policy recommendation is that, there
is still room for technical efficiency improvements with existing firm technologies. In the near
future, however, new technologies must be introduced to sustain higher efficiency levels and
reduce related production costs. More so, in order to promote efficiency and export
performance, polices should be designed at attracting FDIs more especially in the food
processing and textile sectors. The government should as well design strategies to provide
incentives, credit to small and medium sized firms in order to increase output.
ix
TABLE OF CONTENTS
Certification......................................................................................................................... i
Declaration and Copyright ................................................................................................. ii
Acknowledgement............................................................................................................. iii
Dedication .......................................................................................................................... v
Abbreviations and Accronyms .......................................................................................... vi
Abstract ……………………………………………………………………………..…...vi
List of Tables................................................................................................................... xiv
List of Figures ................................................................................................................. xvi
CHAPTER ONE: INTRODUCTION ............................................................................ 1
1.1 Background ......................................................................................................... 1
1.2 Research problem ................................................................................................ 4
1.3 Research Questions ............................................................................................. 7
1.4 Objectives of the study ........................................................................................ 8
1.5 Motivation and Significance of Study ................................................................. 8
1.6 Data ..................................................................................................................... 9
1.7 Organization and Methodology of the study ..................................................... 10
CHAPTER TWO: OVERVIEW OF CAMEROON’S EXPORT AND
INDUSTRIAL SECTOR PERFORMANCE ................................................ 12
x
2.1 Introduction ....................................................................................................... 12
2.2 Overview of Cameroon’s Economic Growth .................................................... 12
2.2.1 Pre-Oil Period: 1963-1977 ................................................................................ 13
2.2.2 The Oil Boom Period: 1978-1986 ..................................................................... 14
2.2.3 The Recession Period, 1987-1993..................................................................... 17
2.2.4 The post-Devaluation, 1994-1999 ..................................................................... 19
2.2.4 The Post HIPC Completion, 2000-2012 ........................................................... 21
2.3 Industrialization and evolution of export performance in Cameroon ............... 27
2.3.1 Rapid growth period, 1960-1986 ...................................................................... 27
2.3.2 Recession period, 1987-1993 ............................................................................ 28
2.3.2 Continuous growth recovery period, 1994-2011 .............................................. 28
2.4 Overview of manufacturing export strategies in Cameroon.............................. 33
2.4.1 Import Substitution Industrialization/inward looking strategy ......................... 33
2.4.2 Industrialization by substitution of exports ....................................................... 34
2.4.3 Industrializing strategy ...................................................................................... 35
2.5 Conclusion ......................................................................................................... 36
CHAPTER THREE: EFFICIENCY AND EXPORT PERFORMANCE: A
CONCEPTUAL FRAMEWORK ................................................................... 37
3.1 Introduction ....................................................................................................... 37
3.2 Definition of Efficiency ..................................................................................... 37
3.3 Types and illustrations of Efficiency ................................................................. 41
3.3.1 Technical Efficiency ......................................................................................... 41
xi
3.3.2 Allocative Efficiency, Profit and Cost Efficiency............................................. 45
3.4 Theoretical basis of Technical Efficiency ......................................................... 49
3.5 Methods of measuring Technical Efficiency ..................................................... 51
3.5.1 Deterministic non-parametric frontiers ............................................................. 53
3.5.2 Deterministic parametric frontiers .................................................................... 56
3.5.3 Parametric Stochastic frontiers ......................................................................... 58
3.6 Empirical studies on efficiency and performance of manufacturing firms ....... 61
3.6.1 Studies on Developing Countries ...................................................................... 61
3.6.2 Studies on Cameroon manufacturing firms....................................................... 70
CHAPTER FOUR: TECHNICAL EFFICIENCY IN CAMEROONIAN
MANUFACTURING FIRMS: A STOCHASTIC FRONTIER
ANALYSIS ....................................................................................................... 74
4.1 Introduction ....................................................................................................... 74
4.2 Purpose and Motivation ..................................................................................... 75
4.3 Production Efficiency and Stochastic Frontier Analysis ................................... 76
4.4 Methodology and data ....................................................................................... 80
4.4.1 Analytical Framework ....................................................................................... 80
4.4.2 Early Developments in the Frontier Analysis ................................................... 82
4.4.3 The Stochastic Frontier Models ........................................................................ 84
4.5 The sample of Cameroonian manufacturing firms and variables ...................... 90
4.6 Definition of variables and the empirical analysis ............................................ 94
4.6.1 Variables of Production Technology ................................................................ 94
xii
4.6.2 Determinants of Manufacturing Efficiency ...................................................... 96
4.7 Pair-wise matrix of correlation coefficients .................................................... 102
4.8 Estimation Procedures and Functional Forms ................................................. 103
4.8.1 Estimation procedures ..................................................................................... 103
4.8.2 Functional Forms ............................................................................................ 104
4.9 Results and discussion ..................................................................................... 108
4.9.1 Production Frontier and Technical Efficiency Estimates................................ 108
4.9.2 The Stochastic Frontier Analysis of Technical Efficiency ............................. 116
4.10 Determinants of Inefficiency ........................................................................... 121
4.11 Mean Technical Efficiency and Inefficiency Scores ....................................... 125
4.12 Conclusion ....................................................................................................... 128
CHAPTER 5: FROM TECHNICAL EFFICIENCY TO EXPORT
PERFORMANCE: EVIDENCE FROM CAMEROON FIRMS .............. 130
5.1 Introduction ..................................................................................................... 130
5.2 Theoretical and Empirical Background ........................................................... 133
5.2.1 Theoretical literature ....................................................................................... 133
5.2.2 Empirical Literature ........................................................................................ 142
5.3 Methodology, Variable specification and data ................................................ 147
5.3.1 Variable specification and Determinants of firm export performance............ 148
5.3.2 Model Specification ........................................................................................ 152
5.3.3 The Data .......................................................................................................... 160
5.4 Empirical Analysis .......................................................................................... 161
xiii
5.4.1 Choice of specification .................................................................................... 162
5.4.2 The probability of Exporting ........................................................................... 164
5.4.3 The propensity to export ................................................................................. 166
5.4.4 The effect of export orientation on Technical Efficiency ............................... 172
5.5 Conclusion ....................................................................................................... 178
CHAPTER SIX: CONCLUSION, POLICY IMPLICATIONS AND
LIMITATIONS OF THE STUDY ............................................................... 180
6.1 Introduction ..................................................................................................... 180
6.2 Summary of findings ....................................................................................... 181
6.3 Policy Implications .......................................................................................... 184
6.4 Limitations of the study and areas of further Research ................................... 186
6.4.1 Limitations ...................................................................................................... 186
6.4.2 Areas for further Research .............................................................................. 187
REFERENCES ............................................................................................................. 188
APPENDICES .............................................................................................................. 198
xiv
LIST OF TABLES
Table 2.1 Export Performance in Cameroon: 1970 - 2011 ......................................... .30
Table 3.1: Selected Stochastic Frontier Studies on the Manufacturing Sector in
Developing Countries. ................................................................................. 67
Table 4.1 Distribution of firms according to size, age and sector of activity……….. 93
Table 4.2: Distribution of firms according to size and regions in Cameroon……......93
Table 4.3: Summary statistics of Variables in different sectors.................................. 101
Table 4.4: Pair-wise Correlation Matrix ..................................................................... 102
Table 4.5: OLS results of the Cobb-Douglas production function with Location and
Industry Dummies ..................................................................................... 111
Table 4.6: Test of hypothesis for Technical Efficiency .............................................. 115
Table 4.7: Cobb-Douglas and Trans-log Stochastic Frontier Estimation of TE ......... 117
Table 4.8: Maximum Likelihood Estimation of Cobb-Douglas and Stochastic frontier models
accounting for Heteroscedasticity (Half-normal MLE) ................................... 120
Table 4.9: Inefficiency effect model ........................................................................... 124
Table 4.10: Mean Technical inefficiency by Size and Sector ...................................... 125
Table 4.11: Mean Technical inefficiency by Ownership and Age for overall sample . 127
Table 5.1: Description and Summary Statistics of the Variables................................ 160
Table 5.2: Probit Estimates of the determinants of the decision to Export ................. 164
Table 5 3: Probit estimates of Determinants of propensity to export to different
regions ....................................................................................................... 165
xv
Table 5.4: Tobit estimates of propensity to export ……..…………………..……….167
Table 5.5: Tobit Estimates of propensity to export controlling for firm size ............. 169
Table 5.6: Tobit Model with Interaction effect ........................................................... 171
Table 5.7: Estimates of the effects of Export orientation and the control variables on
Technical Efficiency ................................................................................... 172
Table 5.8: Group-wise Technical Efficiency comparisons .......................................... 175
xvi
LIST OF FIGURES
Figure 2.1: Cameroon: Sectors Contributions to GDP 1966 – 1976 (%)...................... 14
Figure 2.2: Cameroon: Sector Contribution GDP 1977-1985 (%) ............................... 16
Figure 2.3: Cameroon: Sectors' Contributions to GDP: 2006 – 2009 (%) .................... 22
Figure 2.4: Trends of Cameroon’s GDP and MVA growth rates, 1970 – 2010 ........... 24
Figure 2.5: Sector Contribution to GDP of Cameroon, 2009 ....................................... 27
Figure 2.6: Cameroon Export destinations: Average 2009 - 2011 ................................ 31
Figure 2.8: Exports of Cameroon from 1970 to 2011. .................................................. 32
Figure 3.1: Technical efficiency in outputs .................................................................. 44
Figure 3.2: Technical Efficiency in Inputs .................................................................... 45
Figure 3.3: Allocative and Profit Efficiency ................................................................. 47
Figure 3.4: Cost Efficiency ........................................................................................... 48
Figure 3.5: Illustration of Technical efficiency............................................................. 54
Figure 4.1: Conceptual model of manufacturing firms’ Technical Efficiency ........... 100
Figure 5.1: Export Behavior as Determinant of Change in Productivity Growth ....... 141
Figure 5.2: Export Behavior as non Determinants of Change in Productivity
Growth: ..................................................................................................... 142
Figure 5.3: Percentage of Exporters, By Industry ....................................................... 162
1
CHAPTER ONE
INTRODUCTION
This study examines the technical efficiency of manufacturing firms in Cameroon. It also
analyses technical efficiency as a firm-specific determinant of export performance and
growth. These issues are addressed separately in the study using different methodologies.
The first chapter of the study presents the background of the study and systematically
addresses the problem statement, research questions, objectives, significance of the study
and a brief overview of the research methodology and data used.
1.1 Background
Manufacturing firms play an important role in modern economies and firm output represents
a potential engine for growth in Least Developed Countries (LDC) (Tybout, 2000).
Productivity enhancement therefore, remains crucial to the drive for rapid industrialization
and economic growth in LDCs (Ndulu and O’Connell, 1999). In public policy debates, an
often heard claim is that lack of growth-oriented firms presents the main obstacle to
economic growth and prosperity in a society. It is also often argued that new firms and new
entrepreneurs contribute to the fall in the unemployment rate by creating employment
opportunities. Yet, much still has to be known about the growth processes of individual firms
and the performance of firms in many Sub-Saharan African economies. It is in this context
that manufacturing firms in general, and the efficiency and export performance of such firms
2
in particular, remain topics of great interest in Least Developed Countries, Cameroon in
particular.
Measurement and determinants of efficiency and export performance of manufacturing firms
have invited significant investigations in developing countries especially in Sub-Saharan
Africa (Lundvall and Battese, 2000; Tybout, 2000; Chapelle and Plane, 2005; and Faruq and
Yi, 2010). Early research concentrated more on measuring how efficient the manufacturing
sector was. Most recent research on Cameroon has focused attention on differences in
efficiency, rather than measuring it. Efficiency and productivity provide a criterion for
measuring and improving the manufacturing sector in Cameroon. Research on
manufacturing firms’ efficiency and export performance in Cameroon is still growing and
far from being complete. Much still has to be known about factors that cause differences in
efficiency and export performance among manufacturing firms.
Cameroon a Sub-Saharan African country has experienced increasing concern on the state
of manufacturing productivity. The manufacturing sector is of great importance to the
economy. It employs around 9.2 percent of the total labor force, supplies its output both in
domestic and foreign markets, generates foreign exchange receipts (up to 35 per cent of
export receipts) and contributes up to 17.5 percent to the Gross Domestic Product (GDP) at
current prices1. Moreover, manufacturing induces most of the linkage effects on the other
1 World Trade Organization (2007), Trade policy review: Cameroon and Gabon.
3
sectors of the economy, thus contributing to export diversification, job creation, and poverty
reduction (National Institute of Statistics (NIS), 2009).
The performance of this sector has been declining in recent years, largely because of a
decline in the number of firms as well as a continuous decline in output (NIS, 2009). The
rate of decline in manufacturing output was -0.44 per cent on average between 1995/96 and
2005/06, thus reflecting serious slumps in producer income during the period (WTO, 2007).
Evidence from literature points to the decline in manufactured commodity prices,
appreciation of the Communauté Financière de l’Afrique (CFA) franc relative to the US
dollar, and certain domestic distortions such as high cost of inputs, a cumbersome
administrative machinery, poor management of public enterprises, poor macroeconomic
policy, and cutbacks in government subsidies to firms as the main causes of the fall in
manufactured output (Njikam et al., 2008). Hence, policy makers are concerned about firms
producing low levels of output and that the output is produced inefficiently. As noted by
Faruq and Yi (2010), the key component of the manufacturing sector for improving
efficiency and export performance has to do with making the best use of inputs.
Cameroon’s manufacturing sector increasingly faces critical resource constraints in its
efforts to deliver output to acceptable quantity and quality to satisfy markets. Faced with the
continued deterioration and stagnation of the manufacturing sector, and a significant
disinvestment in manufacturing firms, the government of Cameroon undertook a series of
reforms to attenuate the effects of the crises and safeguard the country’s manufacturing
4
production potential. These measures were in line with the general pattern of the first
structural adjustment program (SAP) adopted in September 1988. The major goals were to:
Liberalize trade in manufacturing exports;
Eliminate input subsidies to firms;
Privatize public enterprises and parastatals, in order to promote firms’
accountability for cost recovery; and,
Restructure manufacturing sector public enterprises and parastatals in order
to achieve a better balance in their financial position and broader autonomy
in internal management.
The overall objective of these measures was to create a sectoral environment likely to
improve firm productivity, reduce production costs to make manufacturing products more
competitive and increase producer income.
1.2 Research problem
The manufacturing sector is one of Cameroon’s most important sectors after agriculture and
oil sectors. Empirical evidence world-wide points to the importance of manufacturing
performance for sustained growth with manufacturing performance contributing to poverty
reduction (Tybout, 2000, Amos, 2007). This is especially relevant to many Sub-Saharan
African countries in view of their heavy reliance on primary exports. Cameroon being one
of such countries, recorded good economic performance during the period 1961–1985, with
agriculture supporting the economy during 1961–1977 and petroleum production taking over
5
the lead during 1978–1985. For these periods, the economy was well managed and the
country had one of the highest per capita incomes in sub-Saharan Africa (Amin, 2002).
However, while the economy is mainly driven by agriculture, output in the manufacturing
sector has been constantly declining over the years (WTO, 2007). The need for rapid output
growth in the sector presents a serious dilemma; whether to concentrate on expanding the
sector in order to achieve a higher economic growth or whether to put greater weight on
protecting existing firms. These challenges call for urgent and thoughtful interventions
because although the manufacturing sector is not a key contributor to GDP, it remains one
of the most important sectors in the economy (see Figures 2.1, 2.2, and 2.3 for the relative
shares of the manufacturing sector in the GDP).
During the 1970s, the share of manufacturing exports was at around 10 percent of total
exports with the growth rate of the manufacturing value added reaching almost 15.3 percent.
In the 1980s, the percentage of manufacturing exports in the total exports doubled from 10
percent to 20 percent, while the share of manufacturing firms in total production remained
weak, at a rate of less than 20 percent (Njikam et al; 2008).
Following strong intervention of the government, numerous industrial activities in
Cameroon flourished from the economic reform. Cameroon implemented policies such as
privatization of public enterprises and relaxation of government controls. Despite the good
6
industrial policies pursued, firms’ outputs remained lackluster regardless of the shift in
strategy from protectionism to liberalization.
In recent years, however, the prices for Cameroon’s manufactured products have been very
low, and production (supply) has fallen. Faced with a fall in prices and quantity produced,
the Cameroon government, in the context of its poverty alleviation program, decided to
increase firm production (through subsidies) to improve producers’ income and profitability.
In order to revive production, the main solution proposed was to improve the productive and
export performance of manufacturing firms (Nchare, 2007).
In the context of the present economic circumstances, characterized in the aftermath of
economic liberalization, by public finance imbalances and significant external debt service
payments, this solution seems to be more appropriate since it is easier to implement and
appears relatively less expensive for Cameroon. However, this solution can be realized if the
sources of inefficiency in the manufacturing firms are identified.
More so, this objective can also be achieved by improving technical efficiency and export
performance of firms, that is, ability to derive the greatest amount of output possible from a
given quantity of inputs. In fact, the presence of shortfalls in efficiency means that output
can be increased without requiring additional conventional inputs or new technologies
(Nchare, 2007). If this is the case, then empirical measures of efficiency and export
7
performance are necessary in order to determine the magnitude of the gain that could be
obtained by improving performance in production with a given technology.
In the first case which is addressed in Chapter four, firm’s performance is indicated by the
firm’s technical efficiency. In the second case which is addressed in Chapter five, firm’s
export performance is indicated by the propensity to export.
1.3 Research Questions
In order to address the research problem, the main research questions of this study can be
stated as follows:
1) How efficient are the manufacturing firms in different industries
2) What are the determinants of firms’ efficiency in Cameroon? How can the
efficiency and export performance of firms be improved in order to increase their
output?
3) How has technical efficiency affected the export performance of manufacturing
firms and what are the observable characteristics of exporting firm that are
closely related to success in international markets?
These questions form the subject matter of the empirical Chapters of this study.
8
1.4 Objectives of the study
Following the research problem outlined above, the main objective of this study is to analyze
the efficiency of manufacturing firms in Cameroon and identify the factors that explain
increases in efficiency level and export performance of firms.
More specifically, the study aims at:
Estimating the level of technical efficiency of manufacturing firms by sectors;
Identifying and analyzing the variables affecting export performance of the firms;
Correlating technical efficiency in manufacturing firms with export performance.
1.5 Motivation and Significance of Study
A number of factors of both practical and theoretical importance motivated this study. At
the practical level, measuring technical efficiency of manufacturing firms, and identifying
the factors that influence export performance will provide useful information for the
formulation of economic policies likely to improve technical efficiency and firm
productivity. Moreover, from the microeconomic perspective, identifying the factors that
improve firm profitability is of major significance, since, by using information derived from
such studies, firms may improve their efficiency and hence profitability. Efficient allocation
of resources at the firm level has great implication for overall economic development as it
leads to a rise in Gross National Product (GNP) and per capita incomes. This study will
provide both qualitative and quantitative analysis of manufacturing firms with special focus
9
on efficiency. In this regard, the findings will greatly inform policy making in general and
industrial policy in particular.
At the theoretical level, the study aims at contributing to the understanding of firms’
technical performance in a developing country. There are few firm level studies on efficiency
on Cameroon. These studies such as (Sjoberg, 1999; Soderling, 1999, Amin, 2002; Njikam,
2003; Njikam et al., 2008) measured Total Factor Productivity (TFP) as a residual of
“Solow” growth accounting; and did not capture mean technical inefficiency of firms, which
has considerable effects on productivity.
This study will examine firms’ performance using micro data that have not been largely used
by previous studies. More so, the few studies that exist have not been able to estimate the
frontiers of the manufacturing firms. This study will therefore provide additional insights
into understanding of technical efficiency as the determinant of export performance. There
are very few empirical studies which investigate this linkage. The main contribution of the
second empirical Chapter is to fill this gap in literature using firm-level data for Cameroon.
1.6 Data
The data used were obtained from Regional Program Enterprise Development (RPED)
dataset for Cameroon’s manufacturing firms for the year 2009 captured by the World Bank’s
RPED survey of 2010. It is also enriched for other variables from the “Récensement General
10
des Entrprises (RGE)” database collected by the National Institute of Statistics (NIS)
Cameroon in 2009. Data on macro-economic variables such as GDP, manufacturing value
added (MVA) were obtained from World Development Indicators (WDI) for Cameroon in
2010 and Kusknir (2013).
1.7 Organization and Methodology of the study
This dissertation is organized along six chapters. Chapter one provided a general
introduction which included the background to the study, problem statement, research
questions, objectives and significance of the study.
Chapter two gives an overview of industrial policies, the sectors’ contributions to GDP,
evolution of GDP and MVA, as well as export performance in Cameroon.
Chapter 3 discusses the concepts of technical efficiency and productive performance from a
conceptual and theoretical perspective. The definitions of efficiency and performance are
provided, followed by a review of the theoretical literature. The chapter ends by providing a
review of empirical studies on Cameroon and other LDCs.
Chapter four provides evaluation of technical efficiency of manufacturing firms in
Cameroon. A stochastic production model is employed in order to estimate firms’ technical
efficiency. This approach is used because the efficiency estimation in stochastic frontier
11
models hinges on the assumption that firms in different industries use different production
technologies. It also seeks to determine the factors responsible for variations in technical
efficiency among firms.
Chapter five explores the link between export performance and technical efficiency for
Cameroonian manufacturing firms. The Chapter evaluates the link using a Probit model. The
Chapter also outlines the determinants of the decision to export or not for Cameroon’s
manufacturing firms using a Tobit model.
Chapter six summarizes the findings and draws conclusions and recommendations based on
the findings of the study.
12
CHAPTER TWO
OVERVIEW OF CAMEROON’S EXPORT AND INDUSTRIAL SECTOR
PERFORMANCE
2.1 Introduction
This chapter provides an overview of industrialization and export performance in Cameroon.
After a brief introduction, the Chapter, first discusses economic growth trends in Cameroon,
followed by an overview of the manufacturing sector and export performance and their
importance to economic growth. The fourth section focuses on manufacturing export
strategies and policies in Cameroon since the 1960s.
2.2 Overview of Cameroon’s Economic Growth
Cameroon recorded good growth performance between 1960 (year of independence) and
1985. In the mid-1970s and early 1980s, economic growth averaged 8 percent per annum.
The country's petroleum production and a rich and diverse agricultural base contributed
much to the growth. Starting in 1986, prospects darkened when the collapse of world prices
for Cameroon's major export commodities - petroleum, coffee, and cocoa – resulted in a
trade shock. From an African economic success story in the early 1980s, Cameroon was in
crisis by the last half of the decade, marked by a shrinking economy and serious deflation
(Njikam, 2003).
13
Cameroon's recent economic trends and performance may be subdivided into five distinct
sub-periods: 1963 to 1977, or the pre-oil era; 1978 to 1986, during which the oil sector
played an important role; 1987-1993, or the economic recession period; 1994-1999, after the
CFA franc devaluation, and the post Heavily Indebted Poor Countries (HIPC) Initiative
decision and completion points. The rest of this section discusses evolution of GDP, MVA
and other indicators of performance over these sub periods.
2.2.1 Pre-Oil Period: 1963-1977
Agriculture played a dominant role until 1978, when oil production started. The primary
sector (including agriculture, forestry, and fishing) accounted for 34 percent of total value
added on average during 1963 - 1977, employed a large section of the labor force, and was
the main source of economic growth and foreign exchange earnings. Real GDP grew, on
average, by 4.6 percent per annum during this period (see figure 2.4). The private
investment-GDP ratio rose from 11 percent in 1963 to about 19 percent in 1977;
government investments, however, remained low as a share of GDP, averaging 2 percent
during 1963 - 1977. Government revenue averaged 17 percent of GDP during the period,
and with total government expenditure averaging at about 18 percent of GDP, the average
overall budget deficit remained low, at 1 percent of GDP (Aerts et al., 2000).
14
Figure 2.1: Cameroon: Sectors Contributions to GDP 1966 – 1976 (%)
Source: Aerts, Cogneau, Herrera, de Monchy, and Roubaud (2000)
The heart of Cameroon’s economic boom came in the early half of the 1970s, an era within
which the service sector supplied half of the country’s GDP. At the time, the country’s
agricultural sector contributed 30 percent of the country’s GDP, while the manufacturing
sector contributed 20 percent of the economy’s GDP (see Figure 2.1). The service sector
was the highest contributor to the GDP of the economy.
2.2.2 The Oil Boom Period: 1978-1986
Beginning in 1978, Cameroon's economy experienced a structural change when oil became
the main source of foreign exchange earnings. The share in GDP of the secondary sector
(including manufacturing) rose from 19 percent on average during 1965-1977 to an average
of 28 percent during 1978 - 1986. Real GDP grew by about 8.8 percent a year during this
Agriculture, 30%
Manufacturing, 20%
Service, 50%
15
period, reflecting in part the oil sector's rising output. Oil production increased from less
than 5 million barrels in 1978 to more than 66 million barrels in 1986. Per capita real GDP
rose by 52 percent from 1978 to 1986. The oil sector also contributed significantly to the
government's budget, with oil revenue growing from less than CFAF 20 billion (1.4 percent
of GDP and 9 percent of total revenue) in 1980 to CFAF 330 billion in 1985 (9 percent of
GDP and 41 percent of total revenue). Total government revenue increased from an average
of about 17 percent of GDP during 1965 - 1977 to an average of 21 percent during 1978 –
1986. Rising government outlays however kept the budget broadly in balance (Ghura, 1997).
With booming economic conditions during 1978-86, the government adopted a
development strategy that centered on expanding the public sector in three ways: first, it
shifted its expenditure priorities by expanding the capital budget from an average of 2
percent of GDP during 1965 - 1977 to an average of 9 percent during 1978 - 1986, while
reducing current outlays from an average of 16 percent of GDP to 12 percent. Thus, the
total investment-GDP ratio increased significantly, but the private investment-GDP ratio
remained broadly unchanged. Second, a large number of public agencies, marketing
boards, public enterprises and industries were set up or expanded in all sectors of the
economy, often supported by government subsidies. Third, the transport sector suffered
from heavy government intervention and was dominated by public enterprises in
railways, urban transport, domestic air travel, merchant shipping, port management, and
road maintenance. Finally, a complex system of regulations on prices, including interest
rates, was put in place. External trade was regulated through import licensing and
16
marketing boards, while quantitative import restrictions were imposed on goods that
competed with domestic production (Njikam et al., 2007)
In principle, the oil boom experienced by Cameroon during 1978-86 should have given rise
to the "Dutch disease" problem, characterized by a rise in the relative price of non-traded to
traded goods. However, the Dutch disease was largely averted, as the real exchange rate
depreciated by about 20 percent between 1979 and 1984, reflecting largely the depreciation
of the French franc. In addition, Benjamin et al., (1989) note that the government saved a
large portion of the windfall income from oil since it perceived the oil boom as temporary,
thus avoiding a spending boom.
Figure 2.2: Cameroon: Sector Contribution GDP 1977-1985 (%)
Source: Ghura (1997)
Agriculture, 20%,
Manufacturing, 35%
Service, 45%
17
The discovery of petroleum in the country’s South West Coast line in 1970 influenced the
contribution of each sector to the country’s GDP as shown in Figure 2.2. The effect of this
discovery was felt between the late 1970s and the first half of the 1980s. The agricultural
sector and the service sector both lost 10 percent and 5 percent respectively to the
manufacturing sector whose contribution to GDP had grown from 20 percent to 35 percent.
This growth arose from the annual 32 percent rise in petroleum earnings realized between
1980 and 1985. After the petroleum discovery, until the economic crisis, only the service
sector faced a relatively stable growth rate, as the manufacturing and agricultural sectors
experienced significant declines (Aerts et al, 2000).
With this GDP structure, the economy fell into a structural crisis (1985-1994) as it depended
on unstable oil revenues to finance its growing recurrent expenses. This led to the country’s
adoption of the Structural Adjustment Program (SAP).
2.2.3 The Recession Period, 1987-1993
The period 1987- 1993 was marked by a severe economic crisis that manifested itself in a
40 percent drop in per capita real GDP. Economic activities shrank in most areas,
particularly in construction and public works, and also in the production of cash crops, retail
trade, and the petroleum sector. The deterioration in Cameroon's economic and financial
situation during this period can be explained by three main factors: a significant
deterioration in world market prices of its main export commodities; an appreciation of its
real effective exchange rate and a decline in oil output. Between 1986 and 1988,
18
international price of crude oil fell by two thirds, while prices of coffee and cocoa dropped
by one-half and one- third, respectively. Overall during 1987 – 1993, the Terms of Trade
declined by nearly 40 percent. Meanwhile, the real effective exchange rate appreciated by
some 40 percent on a cumulative basis between 1985 and 1992, owing to not only the
appreciation of the French Franc (FF) but also an increase in inflation triggered by
expansionary fiscal policies (Njikam, 2003).
Fiscal balance deficit averaged seven percent of GDP during 1987- 1993, compared with an
average surplus of one percent during 1978 - 1986, as the government attempted to jump-
start the economy by expansionary fiscal policy reflected in an increase in total expenditure
by 2.5 percent of GDP between these two sub-periods in the face of a decline in total revenue
by 5.5 percent of GDP. The deficit was financed from two main sources: external borrowing
and the accumulation of domestic and external arrears. External debt rose to 49 percent of
GDP during 1987 - 1993, from 31 percent during 1978 - 1986. Sizable stocks of arrears were
accumulated to external creditors, as well as to domestic suppliers, which prompted several
local companies to halt work and default on their obligations to domestic banks, as well as
on their tax obligations. The deteriorating financial conditions during 1987-1993 exposed
the problems of several local banks, which were undercapitalized, poorly managed, and
marginally profitable. Reflecting the lack of confidence in the domestic banking sector,
money demand fell sharply starting in 1986, as currency rose from 17 percent of broad
money in 1985 to 22 percent by 1993 (Doe, 1995)
19
In order to reverse the declining trend, the government attempted in the late 1980s and early
1990s to jump-start the economy, following a strategy that was based solely on internal
adjustment measures. This strategy consisted mainly of maintaining fixed common peg,
reducing fiscal deficit through increases in tax rates, cuts in wage bill and public enterprise
subsidies, and restoring external competitiveness by reducing domestic cost and
restructuring public enterprises. Given the magnitude of macroeconomic imbalances, it
became clear by end of 1993 that strategies based exclusively on internal adjustment would
not be sufficient to put the economy back on a sustainable economic recovery track. The
internal adjustment strategy alone was unable to restore external competitiveness, as
nominal domestic prices (including wages and producer prices) showed considerable
downward rigidity. In addition, owing to declining government revenue, fiscal adjustment
consisted mainly of cuts in the investment budget and in outlays on nonwage maintenance
and other essential services, a policy that was harmful to growth (Njikam, 2003).
2.2.4 The post-Devaluation, 1994-1999
Given the inability of internal adjustment strategies alone to revive economic performance,
Cameroon, in collaboration with other member countries of the CFA franc zone, devalued
its currency by 50 percent in January 1994.
The devaluation of the country’s currency can be perceived as an effort to re-launch the
economy which had earlier experienced rapid growing petroleum earnings from the oil boom
which occurred in the 1970s. This re-launch followed an over-valuation of the local currency
20
which made the country’s non-petroleum exports less attractive in the global market. After
devaluation, the country’s non-petroleum exports performed better in the global market. This
helped in the reduction of the economic crisis, and the eventual emerging positive GDP
growth (Yang and Nyberg, 2009).
It can be argued that the devaluation had a by-product of evading the Dutch disease through
empowerment of non-petroleum sectors which seemed less competitive earlier. The post-
devaluation recovery program consisted of the Enhanced Structural Adjustment Facility
(ESAF) under the supervision of the World Bank and related multilateral organizations. It
sought to stabilize the newly achieved positive GDP growth, and strengthen the positive
relationship between foreign debt and GDP growth; contrary to the negative foreign debt
to GDP growth relationship witnessed with implementation of SAP during the economic
crisis period.
Besides exchange rate change, the government’s program also consisted of internal
adjustment measures, including further fiscal tightening, as well as implementation of
structural reforms related to the reorganization and downsizing of the civil service,
privatization of public enterprises and industries, bank restructuring, and liberalization of
domestic prices and interest rates. Cameroon's external competitiveness was largely
restored since the devaluation in early 1994, and most exports, including coffee, cocoa,
cotton, timber, aluminum, and manufacturing exports recorded strong gains. Activities in
domestically oriented industries, which had contracted in the wake of the devaluation, also
21
expanded in 1995. This happened particularly with beverages and tobacco. Overall real
GDP turned around from an average decline of four percent during 1987- 1993 to an
average growth of about 2.9 percent during 1994 – 2000, and approximately 3.7 from 2001
to 2011 on average; accompanied by a rise in MVA from 19.1 percent in 1987 to 21.1
percent in 2001/2002, but followed by a decline in MVA from 2003 to 2011 (16 percent in
2011) (Kusknir, 2013).
2.2.4 The Post HIPC Completion, 2000-2012
Achievement of completion point of HIPC initiative had no significant impact on
Cameroon’s Gross Capital formation. Following the decision point of HIPC initiative in
2000, funds were made available for investment in infrastructure. This led to a five percent
rise in the country’s Gross Capital Formation to GDP ratio. This stayed stable until after the
completion point in 2006. Within this period, the economy experienced a slight fall in its
GDP growth.
After the decision point of HIPC initiative, the contribution of the country’s fiscal revenue
to GDP continuously fell until 2004, when the IMF recommended increased government
attention to the implementation of fiscal targets with medium term perspectives. The HIPC
funding in this case acted as a substitute to fiscal earnings which dropped by five percent of
GDP between 2000 and 2004. In this light, Yang and Nyberg (2009) argued that HIPC’s
decision point, through the easily accessible funding allocations, led to an amelioration of
22
the country’s monitoring mechanism which controlled the objectives and achievements of
the state budget, local and foreign debt, as well as the performance of state owned companies
and industries to ensure their growth enhancing potential.
Yang and Nyberg (2009), argue that the majority of countries that attained the completion
point of the HIPC initiative still depend to a great extent on a single export product for a
large percentage of their export revenue. Thus the degree of exposure to external shocks,
which could arise in these economies, following changes in the prices of these products has
not been mitigated. Also, it is noticed, using the revenue to GDP ratio, that an average of
less than 20 percent of the HIPCs GDP is earned from the countries’ fiscal revenues, thus
suggesting that their degree of dependence on foreign revenues was not improved after the
cancellation of their foreign loans.
Figure 2.3: Cameroon: Sectors' Contributions to GDP: 2006 – 2009 (%)
Source: World Bank (2012)
Agriculture, 25%
Manufacturing,31%
Service, 44%
23
After the SAP, the sectors which contributed to Cameroon’s GDP were service sector, 44
percent; Manufacturing, Oil and mining, 31 percent and Agriculture, Forestry and livestock,
25 percent.
The country’s technological base is relatively weak, as is the case with other low income
less developed countries. The trade liberalization which followed SAP opened the country’s
markets to competition from foreign manufactured products. It is therefore important to
understand how an economy whose GDP arises predominantly from the service sector can
improve its industrial output, considering the need for externally earned income to finance
maturing foreign loans. Also, considering the relative instability of externally earned income
due to export price fluctuations, it is important to assess the means by which such a small
economy can generate GDP growth from its export performance.
24
Figure 2.4: Trends of Cameroon’s GDP and MVA growth rates, 1970 – 2010
Source: Plotted using Kusknir (2013) data
From Figure 2.4, we can discern the following sub periods:
Crude oil boom (Period I) which began after 1972 when the economy experienced
a wave of erratic growth until the end of 1982 when growth stabilized.
Economic Crisis (Period II) which extended from 1985 to 1994, when the country
experienced a transition from steady GDP growth at 8 percent to negative GDP
growth.
Economic Recovery (Period III) between 1995 and 2005, within which the economy
regained positive growth and a relative re-stabilization of the country’s industrial and
export performance.
-10
-5
0
5
10
15
20
25
19
70
19
72
19
74
19
76
19
78
19
80
19
82
19
84
19
86
19
88
19
90
19
92
19
94
19
96
19
98
20
00
20
02
20
04
20
06
20
08
20
10
Gro
wth
rat
es
GDP
MVA
25
Post HIPC (Period IV) within which the economy’s GDP growth rate remained
stable despite the fact that the stock of the country’s foreign debt had fallen to close
to one fifth of the level at which it was four years earlier.
Overall industrial output, expanded over the period 1960/1 - 1984/5 by 9 percent on average,
much higher than GDP growth. Within the industrial aggregate, it was manufacturing which
was the best performer. The sector's average annual growth rate of 17 percent during the first
15 years after independence was double that of GDP. However, in the nine years that
followed its expansion slowed down considerably to 7.5 percent close to the then oil
propelled domestic gross output (Benjamin et al., 1989).
As shown in Figure 2.4, the relative instability in the country’s GDP growth rate between
1970 and 1980 arose from the discovery of petroleum resources in the country’s coast line
in the early 1970s; (production began in 1978) and price hikes in petroleum products during
the 1980s “oil boom”. This had a strong impact on the economy, as it led to investment
choices which prioritized petroleum, as well as other non-tradable resource sectors over
agriculture and other tradable resources. This investment policy failed to consider the fact
that the agricultural sector supported a high percentage of the country’s labor force. The
instability resulted from the need to return to the country’s initial GDP contribution structure,
focusing on agriculture, after the drop in fuel prices (Benjamin and Devarajan, 1985).
26
According to Benjamin and Devarajan (1985), by injecting oil revenues into the economy,
inflation levels rose. This increased the prices of locally manufactured agricultural products,
making them less competitive in both local and foreign markets, following exchange rate
appreciation. This was a challenge to the economy’s quest to use import substitution policy
as a driver of growth in the late 1970s. Therefore, allowing the contraction of the country’s
agricultural sector in the face of oil discovery was not a proper orientation of the economic
policy. This was responsible for the unstable growth of the economy after the “oil boom”.
For many years, the country faced the huge challenge of stabilizing its GDP. This goal could
be achieved by improving on export diversification in order to reduce dependence on oil
revenues, as well as fight countering falling commodity prices through sufficient processing
of raw material exports (World Bank, 2012).
Real GDP growth averaged around 3.4 percent a year between 2002 and 2007. From 2007,
economic performance was affected negatively by the global economic and financial crisis,
which led to the disruption in manufacturing, mining and energy sectors. The global demand
and prices for the country’s main exports (particularly oil, timber and rubber) fell. As a
result, GDP growth decreased from 3.4 percent in 2007 to 2.6 percent in 2008 and 2.4 percent
in 2009. However, due to an increase in external demand, economic activities picked-up as
real GDP increased to 3 percent in 2010 and 3.5 percent in 2011 (World Bank, 2012). Import-
substitution industrialization policies led to the creation of excess industrial capacity and low
capacity utilization (Njikam et al., 2008).
27
Figure 2.5: Sector Contribution to GDP of Cameroon, 2009
Source: World Bank, 2012
2.3 Industrialization and evolution of export performance in Cameroon
Three main phases in the evolution of Cameroon’s exports can be distinguished from
independence (1960) up to 2012: (I) rapid growth from 1960 to 1986, (II) a fall in growth
from 1987 to 1993, and (III) continuous growth recovery since 1994.
2.3.1 Rapid growth period, 1960-1986
The first phase lasted for more than 20 years and was characterized by rapid growth at an
annual rate of 106 percent. Spurred by the good performance of primary agricultural
products (coffee, cocoa, cotton, timber, etc.) during the first 15 years, growth was further
supported by oil exports. Behind this global good performance, however, there were great
28
sectoral imbalances: a very limited number of agricultural products represented over 75
percent of total exports before oil started being exported. Industrial export products were
dominated by mineral derivatives, mostly aluminum (Benjamin et al., 1989).
2.3.2 Recession period, 1987-1993
During the second phase, sectoral imbalance worsened, with a fall in both agricultural and
industrial contributions and a boom in oil contribution to total exports, despite the decline in
export revenue due to both world economic recession and depreciation of the US dollar,
which was the main currency for the receipt of exports. In spite of the poor performance of
exports in this phase, there was a relative diversification of industrial exports. Chemical
industry and timber products declined while mineral derivatives and agricultural food
products increased. This phase coincided with implementation of the first SAP leading to
the gradual abandonment of the import substitution policy which was in place since
independence. Quantitative restrictions (QRs) as well as price controls and other nontariff
barriers were gradually abandoned from 1989 (Yang and Nyberg, 2009).
2.3.2 Continuous growth recovery period, 1994-2011
The third phase began with some major changes in the country’s trade policy: a fiscal reform
was implemented and the local currency (CFAF) was devalued by 50 percent relative to the
French franc (FF). Export growth in this phase was also accompanied by a relative
harmonization of contributions to total exports, especially in the industrial sector. Although
29
primary export products still fetched over 80 percent of total export earnings, agricultural
exports gradually reclaimed the front position they had occupied in the preceding years. That
was an indication that the Dutch disease was avoided. Growth in the contribution of both
chemical industry and agricultural food products was closer to 45 percent of the industrial
export earnings (World Bank, 2012).
30
Table 2.1: Export Performance in Cameroon: 1970 - 2011
Year
Export, Billion dollars
Share in the World Export, 0/00
Export Share in GDP, %
Export per Capita, US$
Growth rate of Export, %
1970 0.25 0.66 20.8 37 1971 0.27 0.63 20.8 38 108 1972 0.41 0.81 24.1 57 151.9 1973 0.57 0.82 24.8 77 139 1974 0.56 0.59 23.3 74 98.2 1975 0.74 0.73 23.1 94 132.1 1976 0.85 0.75 24.3 105 114.9 1977 0.94 0.74 22.9 113 110.6 1978 1.3 0.87 24.5 152 138.3 1979 1.6 0.85 22.5 181 123.1 1980 1.8 0.8 20.2 198 112.5 1981 1.6 0.7 19.3 171 88.9 1982 1.6 0.75 19 166 100 1983 1.7 0.81 19.5 171 106.3 1984 1.7 0.76 19.8 166 100 1985 1.5 0.67 17.9 143 88.2 1986 1.6 0.65 16 148 106.7 1987 1.9 0.65 17.3 170 118.8 1988 2.1 0.62 19.1 183 110.5 1989 2.1 0.58 21 177 100 1990 2.4 0.55 20 197 114.3 1991 2.4 0.53 21.8 192 100 1992 2.4 0.47 20 186 100 1993 2.2 0.44 18.3 166 91.7 1994 1.5 0.27 21.1 110 68.2 1995 2.1 0.32 23.6 151 140 1996 2.2 0.32 23.2 154 104.8 1997 2 0.28 22 137 90.9 1998 2.1 0.3 21.4 140 105 1999 2.2 0.3 22 144 104.8 2000 2.2 0.27 23.7 140 100 2001 2.1 0.27 21.9 131 95.5 2002 2.2 0.27 20 134 104.8 2003 2.8 0.3 20 167 127.3 2004 3.1 0.27 19.4 181 110.7 2005 3.4 0.26 20 194 109.7 2006 4.1 0.27 22.8 228 120.6 2007 4.9 0.28 24.5 167 119.5 2008 5.6 0.28 24.3 199 114.3 2009 3.7 0.23 16.1 193 66.1 2010 4.1 0.22 17.1 209 110.8
2011 4.8 0.21 18.5 240 117.1
Source: World Bank (2012)
31
Export diversification was accompanied by a diversification of the market, representing a
drop in exports to France, Cameroon’s traditional export destination since independence.
France accounted for 70 percent of the 85 percent of total exports to Europe. In 1997, this
share dropped to only 25 percent of the 78 percent exports to Europe and to a further 8
percent between 2009 to 2011 (World Bank, 2012).
Figure 2.6: Cameroon Export destinations: Average 2009 - 2011
Source: World Bank (2012)
Europe still remains the main outlet for Cameroonian goods, thanks to the preferential trade
agreements (PTAs) between the European Union and Cameroon in the Lomé conventions.
Nevertheless, there are some openings in America, Africa and Asia as well. This opening
towards Asia became remarkable in 1991, and that in Africa was timid due to the drop in the
Maghreb market, which somehow counter-balanced the upsurge of the SSA market. The
rather timid opening on the American market was due to the limited number of partners.
ROW, 25%
Series1, Spain, 15%, 15%
Netherlands, 12%
SSA, 10%
Italy, 9%
China, 8%
France, 8%
Tchad, 7%,
USA, 6%
32
Almost all exports to this market were destined to North America, US in particular, which
accounts for more than 95 percent.
In spite of the relatively high diversification of industrial exports, primary products still
remain preponderant in the country’s export earnings. Primary agricultural products, which
represented 76 percent of total export earnings in 1959, still stood at 40 percent in 1996/97,
and almost reached 82 percent if crude oil exports are added. This predominance of the
primary sector showed that in spite of advancements in the industrial sector, industrialization
was geared more towards import-substitution (Bamou, 1998).
Figure 2.7: Exports of Cameroon from 1970 to 2011.
Source: Adapted from Kusknir (2013)
33
The global downward trend in both volume and value of non-traditional exports, reversed
only by the CFAF devaluation and commercial liberalization policies of early 1994, as well
as the relatively continuous drop in their contribution to total export earnings, indicates that
there is need for urgent government action in export promotion. Economic rationale demands
that priority be given to products with good prospects.
2.4 Overview of manufacturing export strategies in Cameroon
The manufacturing export or industrialization strategies in Cameroon started immediately
after achieving independence in 1960. We distinguish three strategies:
2.4.1 Import Substitution Industrialization/inward looking strategy
After achieving independence in 1960, Cameroon embarked on an industrialization strategy
based on import substitution. This strategy was marked by extensive use of quantitative
restrictions and controls, high levels of tariffs, widespread rent-seeking activities. The
strategy had an objective to provide the internal market with foodstuffs, clothes and drinks.
It implied the substitution of imported goods by locally produced goods to reduce the
dependency on imported products and to diversify the productive capacity step by step.
Industrialization was done in a gradual way and passed through four main stages. In the first
stage, emphasis was put on “light” industry which requires low technology. In the second
stage, industries constructed were textile and chemical products. In the third stage, the
authorities developed equipment sectors such as electrical manufacturing. In the last stage,
34
resources were to be mobilized towards the sector that produces durable consumption goods
such as vehicles (Benjamin and Deverajan, 1985). The policy in the 1960s and 1970s
shielded state-owned enterprises from foreign competitors, thus compromising their
efficiency and their competitiveness in international market (Benjamin et al., 1989).
However, Cameroon failed to industrialize using inward-looking strategies. Various
hypotheses have been advanced to explain Cameroon’s disappointing industrial performance
record. The poor performance of Cameroon industries during the last decades was mostly
explained by inappropriate domestic policies e.g. the inward-orientation of the trade regime
and the subsequent distortions due to industrial licenses. According to Njikam (2003), other
factors such as; import – substitution of consumer goods, exchange rate overvaluation, and
high tariffs also contributed to the failure of this policy.
2.4.2 Industrialization by substitution of exports
Since the late 1980s and early 1990s, policies that reduced openness to foreign trade were
largely reversed. The policy reform started in Cameroon in 1988 when the government
accepted a stabilization program supported by an 18 month IMF standby agreement,
followed one year latter with the adoption of SAP financed by the World Bank and bilateral
donors. Between 1990 and 1992, trade reform was marked by the elimination of non-tariff
barriers. In 1993-94, the trade reform gained momentum; firstly, through the consolidation
of existing regional trading arrangements i.e. the Communauté Economique et Monétaire
35
de l‟Afrique Centrale (CEMAC) - member states succeeded in establishing a custom union
and lowering their external tariff, and secondly, the devaluation of the CFA–Communauté
Financière Africaine- franc by 50 percent against the French franc.
Actually, this strategy consisted of gradually substituting traditional exports by non-
traditional exports, for example, it transformed primary products, semi-manufactured goods,
and industrial products. The strategy presented many advantages such as exports as a means
of securing foreign exchange for economic development, introduction of export incentives,
financial credits and tax incentives. The strategy resulted in a specialization plan that led to
an effective allocation of resources and gave domestic firms the opportunity to benefit from
effects of scale in their production (Soderling, 1999).
2.4.3 Industrializing strategy
This consisted of developing some industries which will have a strong effect on the
formation of other industries. Priority was given to heavy industries which take advantage
of downstream relations. The aim of this model was to focus more on inter-sector-based
production strategies than intra-industrial, as in the previous case (Njikam, 2003).
36
2.5 Conclusion
From the foregoing discussion, manufacturing dominates industrial output. It is a secondary
sector of the economy which transforms the outputs of primary, agriculture and mining
sectors into semi finished and finished products. It is a foreign exchange generating sector
as products are exported to different countries and hence contributing highly to GDP of
Cameroon. Manufacturing is thus an important activity in promoting economic growth and
development.
Given the role that the manufacturing sector plays in the overall growth of the economy, it
is no doubt that accelerating the growth of the manufacturing sector will boost the growth of
the overall economy. The next chapter discusses the theoretical and empirical foundations
linked to technical efficiency and manufacturing export performance.
37
CHAPTER THREE
EFFICIENCY AND EXPORT PERFORMANCE:
A CONCEPTUAL FRAMEWORK AND LITERATURE REVIEW
3.1 Introduction
This Chapter explores the concepts of efficiency and export performance and gives an
overview of the methods of estimation of these concepts. Moreover, it describes the choices
that have to be made to estimate efficiency and export performance for the manufacturing
sector. Since the main methodology of this thesis is related to these concepts, it is important
to make clear what efficiency means and what types of efficiency measures are developed.
This chapter begins with the theoretical concept of efficiency and ends with an empirical
review of literature on this concept.
3.2 Definition of Efficiency
Efficiency of a firm is the ability to produce the greatest amount of output possible from a
fixed amount of inputs. Another way of putting this is to say that an efficient firm is one that,
given a state of technical know-how, can produce a given quantity of goods by using the
least possible quantity of inputs. In fact, the concept of efficiency is derived from a particular
interpretation of the notion of production frontier, which in the classical sense is the
relationship between output, on the one hand, and the quantity of inputs used in the
38
production process to obtain that output, on the other hand. In estimation methods of
efficiency frontiers, the production function becomes the production frontier.
The concept of efficiency was introduced in the 1950s by Koopmans (1951). In a rather
technical monograph Koopmans gives the definition of an efficient point: “A possible point
[…] in the commodity space is called efficient whenever an increase in one of its coordinates
(the net output of one good) can be achieved only at the cost of a decrease in some other
coordinates (the net output of another good).” In other words, a point is efficient if the output
is maximized given a set of inputs. The distance function measures the distance between the
points in what Koopmans called the commodity space that represents the achieved output of
a firm and the point that it might have achieved if none of the inputs had been wasted. The
greater the distance, the less efficient is the producer.
Debreu (1951) used this definition to develop a measure of efficiency: “a numerical
evaluation of the ‘deadweight loss’ associated with a non-optimal situation (in Pareto sense)
of an economic system.” The general idea of this measure is to determine the distance
between the produced output and the output that could have been produced given the inputs.
Debreu (op.cit) showed that the distance function is well suited for analyzing efficiency.
Shepard (1953) used the same concept of distance functions, but stating it as a problem that
a producer uses too many inputs to produce a certain amount of output. Shepard has an input
oriented approach while Koopmans has an output approach.
39
The idea of measuring efficiency with a distance function has not only been restricted to
theory but is also feasible in practice. Farrell (1957) used the works of Koopmans (1951)
and Debreu (1951) to show how the distance function can be used in a practical way. To
illustrate this practical way, Farrell used an empirical example of the efficiency in the
agricultural sector. Although Farrell showed that the concept was feasible, he also mentioned
the associated computational intensity. For his sample of 48 observations, he needed two
hours of computation time to calculate the measures, and the time would have increased
dramatically if the number of data points grew larger. Coelli (1995) noted that this
computational time no longer constitutes a problem because large data sets can be estimated
with the benefits of modern computers and new algorithms. The feasibility of distance
functions for measuring efficiency also appears in the vast number of applications for which
they are used. It has been used to evaluate farmers, electricity plants, banks and micro
finance institutions, manufacturing firms among others.
Farell (1957) disaggregated efficiency of a firm into two components: technical efficiency
and allocative efficiency. According to the author, technical efficiency reflects the ability of
a firm to obtain maximum output from a given set of inputs. In this case, technical
inefficiency refers to the inability of a firm to use a set of inputs to generate the highest
attainable output level from the same inputs. Hence a firm fails to produce at the outer bound
of its production function (Forsund and Hjalmarsson, 1987).
40
Allocative (price) efficiency on the other hand measures the ability of a firm to use inputs in
optimal proportions, given their respective prices and the production technology. Given the
prevailing price ratios of inputs, the allocative efficiency is represented by only one point
out of the several points on the technically efficient isoquant. This is the point at which the
price ratio line is tangent to the technically efficient isoquant. It is the least-cost point at
which the amount of each input required to produce the specified output level is the
minimum possible at the given prices of inputs. Thus, allocative inefficiency arises when a
firm fails to use substitutable cheaper inputs to incur the minimum cost of production.
According to Forsund et al. (1987) firm efficiency may be a combined effect of technical
and allocative efficiency, with the combined effect known as economic efficiency.
The measures of efficiency are bounded by zero and one. They are measured along a ray
from the origin to the observed production point. Hence, they hold the relative proportion of
inputs (outputs) constant. The main advantage of these efficiency measures is that they are
units invariant (Coelli, 1996). This means that changing the units of measurement (for
example, measuring the quantity of labor either in person hours as against person years) will
not change the value of the efficiency measure.
Since the seminal works of Debreu (1951), Koopman (1951), and Farrell (1957), firm
efficiency has been defined and studied in different dimensions which include: scale
efficiency, referring to the relationship between the level of output and the average cost;
scope efficiency, defining the relationship between average cost and production of
41
diversified output varieties; and operational efficiency, also known as X-efficiency which
measures deviations from the cost efficient frontier that represents the maximum attainable
output for the given levels of inputs. Concerning scale efficiency, it is achieved from firms’
output expansion resulting in an increase in the industry’s output which reduces costs of
production owing to a strong technological economy. Thus, a production unit is scale
efficient when its size of operation is optimal. At the optimal scale, when the size of
operation is either reduced or increased, its efficiency will drop. A scale efficient unit is one
that operates at optimal returns to scale. As noted by Coelli (1996), based on the various
definitions, inefficiency is therefore regarded as a multifaceted concept depending on the
context in which it is employed. The next section discusses several types of efficiency and
how the distance between produced output and optimal output can be measured.
3.3 Types and illustrations of Efficiency
3.3.1 Technical Efficiency
Considering the definition of Koopmans (1951) a firm is efficient if it maximizes output
given the inputs it uses in a production function. Hence, the production function is the
technical relation which connects factor inputs and outputs given existing technology at any
particular time period. If technology changes, then technological improvement is considered
to have taken place. Since this type of efficiency deals solely with technology, it is referred
to as technical efficiency (TE). The production frontier is simply the maximum output
possible for each combination of inputs given the existing technology (Forsund and
42
Hjalmarsson; 1987). Burki and Dek (1998) noted that firms producing on the frontier are
efficient, while firms inside the frontier are inefficient. Therefore, any output deviation from
the production is assumed to be as a result of technical inefficiency.
In order to illustrate the concept of technical efficiency, we assume a firm which uses a single
input and where one unit of the input can be converted to a maximum output. Considering
this, overall technical efficiency will be the ratio of the quantity an efficient firm would have
used to produce a unit of output to the quantity used by the firm being evaluated. Thus, a
firm using two units of inputs to produce two units of output has an overall technical
efficiency score of 1. A firm using four units of inputs to produce two units of output has an
overall technical efficiency of 0.5. According to Burki and Dek (1998), the second firm’s
efficiency score implies that the firm could produce the same output with half the units of
the input it currently uses or equivalently that the firm could double output using the same
amount of the input.
Technical efficiency measurement is illustrated in Figures 3.1 and 3.2. Suppose that one
input ( X ) is needed to produce two outputs ( 21 YandY ) by a certain technology. The
simplest way to describe technology is by the use of a production function (Varian, 1992).
Figure 3.1 represents a production possibilities function (1SS ). The curve
1SS denotes the
possibilities of output given an amount of X . In an ideal situation, every producer who has
that specific amount of input X will produce somewhere on the curve 1SS . In a less ideal
situation, however, it is possible that a producer produces less than the outputs represented
43
by 1SS , for example, producing at
21 YandY (represented by point P in Figure 3.1). To
determine technical efficiency, a fully efficient point is necessary. Such a point is located on
the curve 1SS . Although it is possible to calculate the distance from point P to each point
on the curve 1SS , it is more sensible to choose a point with the same ratio of 21 YtoY as
21 YtoY . This point is represented in Figure 3.1 by point Q. The distance between points P
and Q is therefore a measure for efficiency. This measure has one drawback because it is an
absolute measure and does not take into account the amount of output that could have been
produced. To overcome this problem, efficiency as a relative measure can be determined by
the ratio of distance OP to OQ . This measure gives the value of 1 if P is equal to Q. This
is the case where the amount of output produced lies on the 1SS curve and thus is fully
efficient. The measure gets a value of 0 if P is equal to zero. This implies that although
inputs are used, no outputs are produced.
44
Figure 3.1: Technical efficiency in outputs
Source: Adopted from Burki and Dek (1998)
On the other hand, there exists an input oriented measure of technical efficiency. This
measure assumes that output is given and the firm minimizes inputs. From Figure 3.2,
suppose that two inputs ( )21 XandX are needed to produce one output (Y ) by a certain
production process. The curve 1SS represents the amount of 21 XandX that can be used to
produce an identical amount of Y . In an ideal situation, every producer who wants to
produce a certain amount of output Y needs the amount of inputs represented by the frontier.
In a less ideal situation, it is possible that a producer needs 21 XandX for the production
of the amountY . This is represented by point P in Figure 3.2. To determine efficiency, a full
efficient point is necessary located on the curve SS . The point Q represents a case where
the proportion of 21 XandX should be equal to the proportion of 21 XandX . With the
Q S
P
O
45
use of point Q, efficiency can be determined by the ratio of distance OPtoOQ . This
measure gives the value of ‘1’ if P is equal to Q. This is the case if the amount of inputs
needed for the production lies on the SS curve and thus is fully efficient. The measure
assumes a value of ‘0’ if P is equal to infinity.
Figure 3.2: Technical Efficiency in Inputs
Source: Adopted from Burki and Dek (1998)
3.3.2 Allocative Efficiency, Profit and Cost Efficiency
The discussion of technical efficiency shows a situation where only a production function is
used to measure efficiency. A producer however, does not only deal with a production
function. Part of the profits and costs are determined by the prices of the inputs used and
outputs produced. If prices are taken into account, not every point on the production function
in Figures 3.1 and 3.2 is efficient. Only points that maximize profit or minimize costs are
most efficient.
P
S
Q
O
46
For profit efficiency, a producer has to maximize profits given the amount of inputs and their
prices (Meesters, 2009). Assume that one input ( X ) is available to produce two outputs
1 2( )Y and Y . The prices of these outputs 1P and 2P (the proportion of the prices) is
represented by line 1AA . Since the price ratio is used to draw the line 1AA , every point on
the line should generate the same amount of profit. The producer maximizes profit if line
1AA is shifted as far to the right as possible. This implies that a firm is profit efficient if it
produces the amount where 1AA is tangent to the production curve
1SS (point Q).
Supposing that the firm fails in setting the production to 1Q but produces at point Q, the
firm is still technically efficient yet the allocation of the outputs is inefficient. The allocation
mismatch can be measured with the use of allocative efficiency (AE). For this measure, a
point ( )R on the 1AA line is needed that can be compared with the point
1Q . The point R
has the same proportion of 1Y and 2Y as point Q but is still located on the line 1AA . The
ratio of OQ to OR is then the measure of allocative efficiency. The ratio ranges between
‘0’ and ‘1’. A value ‘1’ represents the most allocative efficient producer. This can only be
achieved if the producer produces at the point where the profit line is tangent to the
production curve, implying that the producer chooses the right output mix. This measure
assumes a value ‘0’ for the completely inefficient producer, and this can only happen if the
distance between the profit line and the production curve is infinite.
47
Figure 3.3: Allocative and Profit Efficiency
Source: Adopted from Meesters (2009)
Profit efficiency is measured by combining technical efficiency and allocative efficiency
(Battese and Coelli, 1988; 1993). Suppose that a producer generates outputs represented by
point p . Figure 3.3 shows that if a firm maximizes profits, it should produce on the 1AA
line. The point R on the graph is best suited for evaluation because it has the same
proportion of 1Y and 2Y as point .P Thus, profit efficiency can be calculated by the ratio
.OP to OR This measure will be equal to ‘1’ if the producer is most profit efficient and ‘0’
if no output is produced from the inputs used. Therefore, profit efficiency is a product of
technical efficiency and allocative efficiency. Using the same analogy, cost efficiency is
determined by using input oriented technical efficiency and allocative efficiency. Figure 3.4
gives a graphical representation of allocative and cost efficiency.
A
A1
Y1 S1
O
S
Y2
R
Q
P
48
Figure 3.4: Cost Efficiency
Source: Adopted from Meesters (2009)
Assume that a producer needs two inputs 21 XandX with prices 21 PandP to produce a
certain amount of output .Y The line 1AA represents the proportion of input prices 21 PandP
. To minimize cost, a firm has to shift this line as low as possible by setting the input level
to the point where 1AA is tangent to the production curve, represented by point Q in Figure
3.4. Supposing that a producer uses inputs represented by point Q, then input allocative
efficiency can be calculated as OQ divided by OR where R is a point that lies on the line
1AA and has the same proportion of inputs as Q. As indicated above, cost efficiency is
calculated as a product of input oriented technical efficiency and allocative efficiency. If a
producer uses inputs 21 XandX as denoted by point P , then the measure for cost efficiency
will become the ratio of .OP over OR Since OP is smaller or equal to ,OR the ratio is
smaller or equal to ‘1’. A score of ‘1’ is only achieved if the producer is fully efficient.
P
S
Q
O
Q1
R
A1
A
49
3.4 Theoretical basis of Technical Efficiency
The theoretical foundation of efficiency is an extension of basic microeconomics of the firm
and production/cost functions. Pareto established the basis of modern “welfare economics”,
by setting positive and normative underlying principles for deriving efficiency analysis to
enhance tangible value and to obtain useful policy information respectively. This welfare
principle evaluates public policies based on efficiency. According to Pareto efficiency
criterion, a social policy could be justified if it makes some persons better off without making
others worse off. Hence Pareto optimality requires that resources be allocated efficiently. If
an allocation is not efficient, there is wastage of resources and therefore room for
improvement so that at least one agent is better off without making the others worse off
under given resources (Schenk, 2004). In economic theory, Varian (1992) stipulates that a
production vectorY is efficient if there is no other feasible production vector 'Y that
generates as much output as Y using no additional inputs.
Based on the definitions and measures of technical efficiency provided by Debreu (1951)
and Koopman (1951), Farrell (1957) developed actual measures of efficiency following the
failure of previous studies to combine these measures into satisfactory measures of
efficiency. This failure was due to the fact that previous studies considered average
productivity of labor as a measure of efficiency, consequently, ignoring all other inputs.
Farrell’s concern was that ‘if any economic planning is to concern itself with particular
industries, it is important to know how far a given industry can be expected to increase its
output by simply increasing its efficiency, without absorbing further resources.” Farrell then
50
estimated the production frontier for ‘fully efficient’ firms where they are producing a
maximum output from a given amount of inputs.
Broeck et al. (1980) specified a production frontier by considering a firm which uses inputs
),...( 1'
nxxX to produce its output Y , on the production plan YX ,' . The efficient
transformation of inputs into output is characterized by the production function ),...( 1 nxxf
which shows the maximum output obtainable from various inputs vectors. In econometric
literature, ),...( 1 nxxf is typically referred to as a frontier since it characterizes optimizing
behavior on the part of an efficient firm and thus places limits on the possible value of its
dependent variable. A firm could be considered to be technically efficient or technically
inefficient depending on the following conditions:
If ),,...( 1 nxxfY then the firm is considered to be technically efficient. On the other hand,
if the production plan is such that, ),,...,( 1 nxxfY then it will be technically inefficient
(firms producing less than maximal possible output). Forsund et al., (1987) assume
),...,( 1 nxxfY
to be impossible since no points can lie above the frontier.
Battese and Coelli (1988) defined technical efficiency as the ratio of a firm’s mean output to
the corresponding mean potential output, conditional on both the level of factor inputs being
used and inefficiency effects. Based on this definition, technical efficiency would be
measured theoretically and simply stated as the ratio of the observed output for the firm,
relative to the potential output:
51
1
0 1 3.1( ,..., )n
Y
f x x
In this case, technical inefficiency is due to excessive input usage, which is costly (failure to
minimize cost) and consequently, profit is not maximized.
In a cross-section data theoretical approach, the technical efficiency of a given firm is
defined as the ratio of its mean production (in original units), given its realized firm effect,
to the corresponding mean production if the firm effect is zero (Battese and Coelli, 1988).
Thus, the technical efficiency (TE) of the thi firm is defined by:
( | , , 1,2,..., )
3.2( | 0, , 1,2,..., )
i i ii
i i i
E Y U X i nTE
E Y U X i n
where iY denotes the value of production (in original units) for the
thi firm. This measure
necessarily has values between ‘0’ and ‘1’. If a firm’s technical efficiency is closer to one,
this implies that the firm realizes, on average, a higher percent of the production possible for
a fully efficient firm having comparable input values.
3.5 Methods of measuring Technical Efficiency
After discussing the concept of efficiency, it is useful to show how it can be estimated. Many
studies (using both panel and cross-section data) have applied, extended as well as modified
frontier modeling for measuring efficiency since the works of Debreu (1951), Koopmans
(1951), and Farrell (1957). Broeck et al. (1980), Battese (1992), Coelli (1995) attributed the
widespread use of frontier modeling to many reasons among which Bauer (1990) had
52
outlined three main factors: first, the notion of a frontier is consistent with the underlying
economic theory of optimizing behavior. Second, deviations from a frontier have a natural
interpretation as a measure of the efficiency with which economic units pursue their
technical objectives. Finally, information about the structure of the frontier and about the
relative efficiency of economic units has many policy applications.
The evaluation of a firm’s technical efficiency level results from the estimation of a frontier
production function. Studies on frontier technology and efficiency measurement can be
classified according to the way the frontier is specified and estimated. First, researchers have
specified frontiers as parametric or non-parametric functions. Second, an explicit statistical
model of the relationship between observed output and the frontier may be specified or not.
Finally, the frontiers may be specified to be either deterministic or stochastic (random).
Parametric and non-parametric approaches may be distinguished in many aspects: first, the
non-parametric approach does not impose any functional form on the data. Second, it does
not make assumptions about the distribution of the error term that represents inefficiency.
Lastly, the estimated non-parametric frontiers have no statistical properties on which to be
gauged (Bauer, 1990; Coelli, 1996).
With regards to the deterministic and random specifications, the deterministic specifications
assume any deviation from the frontiers to be resulting solely from inefficiency (Broeck et
al., 1980).The frontier is called deterministic if all the observations must lie on or below the
53
frontier. Inefficiency in this case is therefore defined as the proportion by which the level of
production is less than the estimated frontier output. One obvious weakness of constraining
the observations to be on or below the frontier is when measurement errors are present.
Therefore, in failing to account for the possibility of random influence, the deterministic
specification is particularly sensitive to outliers and measurement errors. Aigner et al. (1977)
addressed this problem by allowing observations to be above the frontier, but putting
different weights on positive and negative disturbances. This approach was more
satisfactorily developed in Broeck et al., (1980) by introducing two separate disturbance
terms: One variable, capturing the efficiency differences between units, distributed over the
natural interval ),1,0( and another variable, reflecting true random differences, such as
measurement errors, distributed over the interval ).,0( Conversely, the stochastic
estimation methods involve a specification of a probabilistic frontier that takes into account
the possibility of variations in output due to factors not under the control of the firm. The
frontier is called stochastic if observations can be above the frontier due to random events
(corruption within the firms, existence of trade unions, size of the firms, location of the firms,
measurement errors, etc.). The next subsection discusses the frontier specifications.
3.5.1 Deterministic non-parametric frontiers
Farrell (1957) proposed specific measures of technical efficiency which are valid for
restrictive technologies. This approach is deterministic and non-parametric, and provides
definitions and a computational framework for technical inefficiency. Figure 3.5 shows a
54
situation where a firm is using two inputs 1x and 2x to produce an output ,y and assuming
that the firm’s production (frontier) is ).,( 21 xxfy
Figure 3.5: Illustration of Technical efficiency
Source: Burki and Dek (1998)
If frontier technology is characterized by constant returns to scale, then it can be represented
as .),(1 21 yxyxf The line 1pp represents the ratio of input prices or the iso-cost line
which shows all combinations of inputs 1x and 2x such that input costs sum to the same
total cost of production. The curve 1UU denotes a unit isoquant, representing technically
efficient combination of inputs 1x and 2x used in producing output y .
U
0
P
P1
A
B
C D
55
If the firm is observed using ),( 02
01 xx to produce ,0y assuming point A to be represented by
),( 02
01 yxyx , then the measure of technical efficiency at this point is given by the ratio
OAOB : which is the ratio of inputs needed to produce 0y to the inputs actually used to
produce 0y , given the inputs mix used (Broeck et al., 1980). Thus, the distance between B
and A represents the proportional reduction in all inputs used in production that could
theoretically be achieved without any reduction in output.
This approach is non-parametric in the sense that it simply constructed the free disposal
convex hull (FDH) of the observed input-output ratios by linear programming techniques;
not based on any explicit model of the frontier or of the relationship of the observations to
the frontier (other than the fact that observations cannot lie below the production frontier
(see Farrell, 1957). Farrell’s (1957) approach was extended and applied by Farrell and
Fieldhouse (1962), Forsund and Hjalmarsson (1987). More especially, Charnes et al., (1978)
extended and refined this approach into the Data Envelopment Analysis (DEA). DEA is
based on linear programming and consists of estimating a production frontier through a
convex envelope curve formed by line segments joining observed efficient production units.
No functional form is imposed on the production frontier and no assumption is made on the
error term. Nevertheless, this method is limited because of the following reasons; firstly, it
lacks the statistical procedure for hypothesis testing. Secondly, it does not take measurement
errors and random effects into account. In fact, it supposes that every deviation from the
56
frontier is due to the firm’s inefficiency, and thirdly it is very sensitive to extreme values and
outliers.
Coelli (1995) observed that these methods, in addition to having the advantages of the non-
parametric approach, enable one to estimate efficiency for multiple-input multiple-output
technologies.
3.5.2 Deterministic parametric frontiers
Farrell (1957) proposed the second approach (parametric approach) to the non-parametric
approach. The Cobb-Douglas form was recommended since the selection of the functional
form was limited. Aigner and Chu (1968) followed Farrell’s suggestion by specifying a
homogenous Cobb-Douglas production frontier, with all observations required to be on or
beneath the frontier. The model is specified as follows:
01
( )
, 0, (3.3)n
i ii
In y In f x
Inx
where the one-sided error term forces ).(xfy
The elements of the parameter vector '10 ),...,,( n may either be estimated by linear
programming (minimizing the sum of the absolute values of the residuals, subject to the
constraint that each residual be non-positive) or by quadratic programming (minimizing the
57
sum of squared residuals, subject to the same constraints). Hence, technical efficiency of
each observation can be computed directly from the vector of residuals, with representing
the technical inefficiency.
The main advantages of the parametric approach vis-à-vis the non-parametric approach are
the ability to characterize frontier technology in a simple mathematical form, and the ability
to accommodate non-constant returns to scale. However, the parametric approach often
imposes a limitation on the number of observations that can be technically efficient (Forsund
and Hjalmarsson, 1987). In the homogenous Cobb-Douglas case, for example, when the
linear programming algorithm is used, there will, in general, be only as many technically
efficient observations as there are parameters to be estimated.
As was the case with the non-parametric approach, the estimated frontier is supported by the
subset of data and is therefore extremely sensitive to outliers. Aigner and Chu (1968)
suggested one possibility which was to essentially just discard a few observations. If the rate
of change of the estimates with respect to succeeding deletions of observations diminishes
rapidly, then the suggestion will be useful. The main problem with this approach is that the
estimates which it produces have no statistical properties. That is, mathematical
programming procedures produce estimates without standard errors, t-ratios, etc. This is
basically because no assumptions are made about the disturbance or the regressors and
without statistical assumptions inferential results cannot be obtained (Broeck et al., 1980).
58
3.5.3 Parametric Stochastic frontiers
The preceding frontiers discussed are all deterministic in which all firms share a common
family of production, cost and profit frontiers, and all variations in firm performance are
attributed to variation in firm efficiencies relative to the common family of frontiers. Broeck
et al. (1980) noted that this scenario proves difficult to justify empirically although it
conforms with theoretical underpinnings. Hence, the notion of deterministic frontiers shared
by all firms ignores the very real possibility that a firm’s performance may be affected by
factors entirely outside its control (such as poor machine performance, input supply
breakdowns, weather and so on) as well as by factors under its control (inefficiency). Many
authors have indicated that lumping the effects of exogenous shocks together with the effects
of measurement error and inefficiency into a single one sided error term, and to label the
mixture ‘inefficiency’ is somewhat questionable.
The essential idea behind the stochastic frontier model is that the error term is composed of
two parts. A symmetric component permits random variations of the frontier across firms
and captures the effects of measurement error, and other statistical ‘noise’, and random
shocks outside the firm’s control. A one-sided component captures the effect of inefficiency
relative to the stochastic frontier. A parametric stochastic frontier model may be specified
as:
( )exp( ) 3.4y f x v u
Therefore; uvii eexfy .).,(
59
which in log form gives; iiii uvxfy )),(log()log(
where the stochastic frontier production frontier is ),exp()( vxf v having some symmetric
distribution to capture the random effects of measurement error and exogenous shocks which
cause the placement of the function )(xf to vary across firms; iv is considered as a normal
error ).,( 2vvi Nv Technical inefficiency relative to the stochastic production frontier is
then captured by the one-sided error component exp( ), 0.u u According to Broeck et al.
(1980) the condition 0u ensures that all observations lie on or beneath the stochastic
production frontier.
Assuming a Cobb-Douglas production function;
0
1
( , ) 3.5i
K
i ii
f x e x
which in log form gives:
01
log ( , ) log 3.6K
i i ii
f x x
So the model becomes;
01
log( ) log 3.7K
i i i i ii
y x v u
This leads to firm-specific efficiency scores in the Cobb-Douglas case
( , ). .
3.8( , ).
v uui
i vi
f x e eTE e
f x e
60
Direct estimates of the stochastic production frontier models may be obtained by either
Corrected Ordinary Least Squares (COLS) or by Maximum Likelihood Methods. Broeck et
al. (1980) noted that whether the model is estimated by COLS or by maximum likelihood,
the distribution of u must be specified. There is no consensus about the assumptions for the
distribution of the efficiency term .iu
Aigner et al. (1977) and Meeusen and Broeck (1977) considered two types of distributions
that can be used for ,u that is, an exponential and half-normal distribution for .u Both of these
distributions have a mode of ‘0’. Cummins and Zi (1998) noted that in general the
assumption about the distribution of the inefficiency term does not affect the estimated
inefficiencies. Most often the half normal assumption is applied (Behr and Tente, 2008), but
the exponential and truncated normal cases are also discussed in specific literature2.
While the two parameter distributions (the truncated normal and the gamma) potentially
increase the flexibility of the model, in practical applications problems of identification seem
to outweigh the potential gains for either distribution (Greene, 1997; Ritter and Simar, 1997).
To make sure that the model is properly identified and that the noise really measures noise
and not efficiency, it also has to be assumed that iv and iu are independent from each other.
If assumptions are made about iv and iu terms, the model represented above can be
estimated by using Maximum Likelihood (ML) technique3. The next section reviews the
2 See Greene, 1997; Ritter and Simar, 1997, Olson et al, 1980. 3 See appendix for the derivation of the ML functions and their log-likelihood for half-normal and exponential models.
61
empirical literature related to technical efficiency and export performance in selected LDCs
and Cameroon in particular.
3.6 Empirical studies on efficiency and performance of manufacturing firms
This section reviews empirical studies of efficiency and productivity for manufacturing
firms in other less developed countries and Cameroon in particular. The reviewed studies
bring out differences in issues such as definition of efficiency, methodologies employed to
estimate the various types of efficiency, choice of functional form or the structure of the
error term.
3.6.1 Studies on Developing Countries
Burki and Dek (1998) investigated technical and scale efficiencies of small manufacturing
firms in nine small manufacturing industries in Gujranwala, Pakistan using the
nonparametric DEA approach. The authors defined production efficiency as producing the
maximum quantity of output possible with a given set of inputs. Efficiency was studied in
two steps. The first step consisted of using data envelopment analysis to calculate measures
of efficiency for each firm in the sample. In the second step, the authors used the Tobit
regressions of the efficiency measures on attributes of the firm to analyze the sources of
efficiency. They found that on average the sampled firms could raise output by 6 percent to
29 percent by improving their overall technical efficiency. About 46 percent of the firms
62
exhibited increasing returns to scale while only 16 percent of the firms operated at decreasing
returns. This shows that a primary source of scale efficiency is operating at less than the
optimal level of production. Concerning the second step, the study indicated that functional
literacy of firm owners and their experience positively affects technical efficiency of firms.
Bigsten et al. (1999) used panel data (1992 to 1995) to investigate manufacturing investment
in four African countries – Cameroon, Ghana, Kenya and Zimbabwe – in which the financial
market had been heavily controlled.
Lundvall and Battese (2000) used an unbalanced panel of 235 Kenyan manufacturing firms
to estimate trans-log stochastic frontier functions of firms in the food, wood, textile and
metal sectors. The frontier was used to investigate the relationship between age, size and
technical efficiency and to simultaneously estimate the parameters of the production
function with those of the inefficiency model. The sectors were estimated individually in
order to assess whether technical efficiency is systematically related to the size and age of
firms. The authors found that size, and not age, often had a strong positive association with
technical efficiency. The size effects were positive for an overwhelming majority of firms
in all sectors with a significant parameter in the wood and textile sectors. As well, the
marginal effects of firm size on technical efficiency tended to be positive especially for firms
over five years of age. The age effect was less systematic, and insignificant in all sectors,
except wood.
63
Mahadevan (2000) studied the technical efficiency of 28 industries in Singapore from 1975-
94 using a Cobb-Douglas production function and stochastic production frontier approach.
The author showed that on average, Singapore’s manufacturing industries were operating at
73 percent of their potential output level and showed that capital intensity and labor quality
were important factors in determining the efficiency levels. In their study of technical
efficiencies of firms in the Indonesian garment industry, Battese et al. (2001) used stochastic
frontier models for firms in five different regions of Indonesia for the period 1990 to 1995
and found that there are substantial efficiency differences among garment industry firms
across the five regions.
Oczkowski and Sharma (2005) studied the determinants of efficiency in Least Developed
Countries using the evidence of Nepalese manufacturing firms, using a trans-log stochastic
production frontier and maximum likelihood econometric methods. The paper estimated and
modeled the determinants of firm efficiency in the Nepalese framework, with results broadly
in line with theoretical expectations. The results showed that large firms were more efficient
and that higher capacity intensity leads to inefficiency. The results also showed no statistical
evidence that foreign participation leads to efficiency improvement. As well, no observation
was made to establish the link between export intensity and efficiency improvement. Hence
higher protection leads to inefficiency. The overall results suggested that an outward looking
industrial strategy, which relies on less intervention and permits the development of large
scale industries, is conducive to efficiency improvement in developing economies.
64
Niringiye and Luvanda (2010) established the relationship between firm size and technical
efficiency in East African manufacturing firms. Panel data for 403 firms in Kenya, Tanzania
and Uganda were used. Two-step methodology was applied: In the first step, technical
efficiency measures were calculated using DEA approach. In the second step, the study used
a Generalized Least Squares (GLS) technique, where a technical efficiency equation was
estimated to investigate whether technical efficiency was increasing with firm size. The
results showed a negative association between firm size and technical efficiency in both
Ugandan and Tanzanian manufacturing firms. The study found a positive relation between
size squared and technical efficiency as well as a negative association between firm size and
technical efficiency in Uganda and Tanzania manufacturing firms suggesting an inverted U-
relationship existing between firm size and technical efficiency in these Countries. The
relationship between firm size and technical efficiency was not statistically significant for
Kenyan manufacturing firms.
Roudant and Vanhems (2011) used a sample of 195 and 174 firms in 1994 and 1995
respectively to explain firm efficiency in the Ivorian manufacturing sector using a robust
non-parametric approach. The authors adapted the one-step nonparametric robust
methodology of Daraio and Simar (2005) to take in account qualitative environment factors
and also compare the difference in behavior among two sub groups of firms characterized
by different levels of technology (high technology and low technology sectors).
Accordingly, efficiency in production measurement consists in analyzing how firms
combine their inputs to produce their output in an efficient way. The choice of the
65
nonparametric deterministic method was in order not to restrict the shape of the frontier to
be parametric, contrary to the parametric stochastic frontier approach (Kumbhakar and
Lovell, 2000). The results suggested that there is no strong effect of environmental factors
on firm efficiency levels in the high technology (HT) sector. The HT sector was quite
homogenous with respect to exogenous variables used. Contrary to expectation, the results
showed that environmental factors had a positive impact on efficiency for a low technology
(LT) sector except for the variable age. The results also indicated that among the LT sector,
firms from the formal sector were more efficient than firms from the informal economy.
Two other studies on Ivorian manufacturing had used the same database (RPED) as Roudant
and Vanhems (2011) to estimate efficiency scores. Chapelle and Plane (2005) investigated
the technical efficiency of Ivorian manufacturing firms in four sectors: textile and garments,
metal products, food processing, and wood and furniture, applying nonparametric DEA
techniques and using a methodology in four steps to capture three effects: the purely
managerial, impact of the scale of production, and technical effects capturing the potential
gain that could result from the adoption of modern technology by small informal
organizations. Roudant (2006) studied the impact of business environment on technical
efficiency using an unbalanced panel of Ivorian manufacturing firms. A stochastic frontier
production model with non-neutral effects on technical efficiency of the business
environment variables was specified. This specification allows evaluation and comparison
of technical efficiencies and efficiency levels net of business environment influences.
Roudant (op.cit) proposed a practical method based on the definition of an artificial
66
environment in order to determine net efficiency levels in the non-neutral case. The results
suggested that informal firms were less technically efficient than formal firms whereas their
managerial performances were close to those of formal firms.
Faruq and Yi (2011) used the non-parametric linear programming technique (Data
Envelopment Analysis (DEA) to estimate the technical efficiency of firms in Ghana across
six manufacturing industries using data from 1991 – 2002. This technique was used because
it does not require the specification of the functional form of the production function or make
any assumptions about the probability distribution of the errors. The DEA approach
measures the efficiency of a firm relative to other firms in a comparable environment (i.e.
within the same industry and/or country), and the method has the advantage that the
efficiency measurements are similar regardless of whether the efficiency estimates are
‘input-oriented’ (whether firms can reduce their inputs usage to produce a given level of
output) or ‘output-oriented’ (whether firms can increase their output level for a given set of
inputs). The authors found that the overall mean efficiency of manufacturing firms in Ghana
ranged between 54 and 55 percent. The results showed that among the six industries, the
textile and garment industries seemed to be relatively more efficient, while furniture industry
appeared to be relatively less efficient. The authors also observed that manufacturing firms
in Ghana were significantly less efficient than their counterparts in other countries. In
addition, they found that firm characteristics such as size, age foreign ownership, and the
mix of labor and capital used in the production process had positive effects on efficiency.
67
Table 3.1: Selected Stochastic Frontier Studies on the Manufacturing Sector in Developing Countries.
Study Country and Sectors,
No. of firms (N), Years
(T)
Methodology SFA: Variables Used Inefficiency Variables
and firm
characteristics model
Level of TE for
selected sectors
Radam et
al. (2010)
Country: Malaysia
Sector: Wood furniture
N=? and T=?
Cobb-Douglas
SFA
Output = Value Added in RM; Capital =
Value added in RM (+); Labor = No. of
workers (+); Energy = Expenditure (+)
n.a
Wood= 45.5%
Kinda et al
(2008)
Cross-country Analysis,
Sector: Steel and Iron
N =52 and T = 20
Cobb-Douglas
SFA
Output = Value added
Labor = No. of permanent workers
Capital = Gross Value of property, plants
and equipment.
n.a
MENA NON
Food = 43% 45%
Wood = 46% 48%
Metal = 53% 62%
Textile = 42% 44%
Natarajan
and Raj
(2007)
Country: India
Sector: food, textile,
wood, minerals, others
N = 52 and T =20
Trans-log SFA
Output = Gross value added; Capital =
value of total equipment (+); Labor =
number of workers (+)
Firm size, ownership,
location and nature of
seasonality of operation
Overall = 48%
Food = 53%
Wood = 36%
Metal = 53%
Textile = 47%
Bhandari
and Maiti
(2007)
Country: India
Sector: Textile
N =? and T =5
Trans-log SFA
Output = value products and by products;
Capital = net value of fixed assets (+);
Labor = total number of man days
worked (+); intermediate inputs =
nominal value of inputs (+)
Firm size-intermediate
inputs (+), ownership,
location, age (-)
Textile = 68% - 84%
68
Study Country and Sectors,
No. of firms (N), Years
(T)
Methodology SFA: Variables Used Inefficiency Variables
and firm characteristics
model
Level of TE for
selected sectors
Kim et al.
(2005)
Cross Country
Sector: Steel and Iron
N = 52 and T = 20
Trans-log SFA
Output = Crude steel production is
millions of tons; Labor = total number
of employees; Capital = productive
capacity of equipment (millions of tons),
and material inputs (U.S dollars)
Firm ownership, Age,
Scale – firm’s production
as a share of the total
production in all non-
communist countries
Average > 90%
Nikaido,
Y (2004)
Country: India
Sector: all sectors
N = 505 and T = 1
Cobb-Douglas
SFA
Tobit model
Output = output per employee; capital
per employee (+) and Labor (+)
Firm size – employees per
unit (-) and location
Food = 82%
Wood = 81%
Metal = 81%
Textile = 81%
Margono
and
Sharma
(2004)
Country: Indonesia
Sector: food, textile,
chemical and metal
products
N = 733 and T = 8
Trans-log SFA
Output = total value of output; capital =
total cost of capital depreciation and
interest paid by the firm (+), Labor = the
total number of employees (+); Material
(m) = the total value of the material used
by the firm.
Firm size = output;
maturity = year; location;
ownership
Overall = 55.8%
Food = 50.8%
Wood = n.a
Metal = 68.9%
Textile = 47.9%
Lundvall
and
Battese
(2000)
Country: Kenya
Sector: food, wood,
textile and metal
N =235 and T =3
Trans-log SFA
Output = value of output; capital =
replacement cost corrected by capacity
utilization (+ in food; - for others);
wages = total wage and allowances (- in
food; + for others); intermediate inputs
= costs of raw materials including solid
and liquid fuel, electricity and water (+).
Firm size = intermediate
inputs (-in food;+ for
others); firm age = years in
operation (+ for food and
textile; - for wood and
metal)
Food = 77%
Wood = 68%
Metal = 80%
Textile = 76%
Bigsten et
al. (1998)
Country: Cameroon,
Ghana, Kenya and
Zimbabwe
Sector: food, wood,
textile and metal
Cobb- Douglas
SFA
Output = value added
Capital = replacement value of
equipment
Labor = number of employees
Export
ownership
Average = 22.1%
Food = 20%
Wood = 34.9%
Metal = 12.1%
Textile = 18.5%
69
Cameroon: N=? T=
Study Country and Sectors,
No. of firms (N), Years
(T)
Methodology SFA: Variables Used Inefficiency Variables and
firm characteristics model
Level of TE for
selected sectors
Brada et
al. (1997)
Country:
Czechoslovakia and
Hungary
Sector: all in the
Industry
N = (800 in Czech and
1121 in Hung) T = 1
Cobb-Douglas
SFA
Output = value added, capital = average
(annual) stock of capital (+)
Labor = hour worked (+) (Czech) and
average number of employees (+)
Hungary
Firm size = value added
(+)
Profitability (+)
Others
Czech Hung
Food = 53% 64%
Wood = 51%
47%
Metal = 52%
74%
Textile = 52%
45%
Biggs et
al. (1995)
Country: Cameroon,
Kenya, Ghana and
Zimbabwe
Sector: food, wood,
textile and Metal
N = 563 and T = 3
Cobb-Douglas
SFA
Output = value added
Labor = total number of employees (+)
Capital = capital stock measured by
replacement cost (+)
Additional variables – ratio of non
manual workers to total workers (+) and
the rate of capital utilization (+).
n.a
Average = 41%
Food = 68%
Wood = 48%
Metal = 56%
Textile = 43%
70
While efficiency is a key to sustaining growth and alleviating poverty, several existing
studies on the efficiency of manufacturing in Less Developed Countries (LDCs) are based
on highly aggregated data. Among the studies, Lundvall and Battese (2000) examined the
effect of firm size and age on technical efficiency of Kenyan manufacturing firms. Njikam
(2003) examined the effect of trade reforms on efficiency of manufacturing firms in
Cameroon by adapting a trans-log production function. Oczkowski and Sharma (2005)
investigate the effect of firm size and other firm characteristics on technical efficiency of
manufacturing firms in Nepal using a parametric frontier analysis. Niringiye et al. (2010)
examined the relationship between firm size and technical efficiency in East African
manufacturing firms using a two-step methodology. They used the non-parametric approach
in the first step to calculate technical efficiency measures, and in the second step, the
Generalized Least Squares (GLS) technique was used. Most of these studies used long time
series data. This study uses the parametric approach of stochastic frontier analyses.
3.6.2 Studies on Cameroon manufacturing firms
Soderling (1999) used firm level data covering the period 1980 – 1995 to present main
developments in the manufacturing industry in Cameroon. The study laid more emphasis on
structural factors of competitiveness. A production function and an export function were
estimated in order to study the determinants of total factor productivity (TFP) and export
performance. The results provided evidence indicating that openness to trade, development
of skilled labor and adequate management of the real exchange rate were crucial factors in
the enhancement of productivity and exports. The simple model used to quantify these
71
impacts revealed that the devaluation of the CFA franc in 1994 had some appreciably
beneficial effects on manufacturing productivity and exports. More so, Soderling (op.cit)
demonstrated a mutually reinforcing relationship between productivity and export
performance and constructed a model to assess the cost of Real Effective Exchange Rate
(REER) evaluation, both in terms of productivity and exports. The study showed that
performance of the manufacturing sector in Cameroon deteriorated considerably after the
mid-1980s. The decline was to a large extent explained by in-ward looking policies in the
manufacturing sector.
Njikam (2003) using firm-level data to establish the trade reform efficiency on Cameroonian
manufacturing firms reported a positive (but statistically insignificant) association between
the official tariff rates and the level of average technical efficiency achieved by these firms.
The author also found the association between effective protection rate and the level of mean
technical efficiency in the manufacturing firms to be positive but statistically insignificant.
Further, the study observed a strong positive association between import penetration ratio
and the level of mean technical efficiency achieved in the manufacturing industry. Even
though the results obtained by Njikam conformed to the a priori expectation of a positive
relationship between the two variables. However, the results were obtained from a
correlation analysis which does not provide a basis for measuring the impact of one variable
on the other.
72
The results of Njikam (2000) indicated a positive and significant correlation between
manufacturing share of exports and average technical efficiency achieved in the
Cameroonian manufacturing sector. The results showed that the higher the share of
manufacturing in total exports, the higher the mean technical efficiency achieved in the
manufacturing sector. The study also reported a positive and significant association between
changes in import penetration rate, export share, effective rate of protection and intra-
industry trade index and the mean technical efficiency achieved in the firms. Further, a
negative and insignificant correlation between changes in official tariff rates and the mean
technical efficiency were found. Moreover, the results indicated that, while macroeconomic
instability (inflation) had a negative and statistically significant impact on average technical
efficiency achieved in this sector, the impact of political instability on the mean technical
efficiency was also negative but statistically insignificant. The author also revealed that the
impact of property right protection on mean technical efficiency is positive and statistically
significant. These results imply that political and macroeconomic instability hindered
efficiency of manufacturing sector while property rights protection promoted manufacturing
sector’s efficiency in Cameroon.
Njikam and Cockburn (2007) used pooled pre and post reform period data (from 1988/89 to
1991/92 and from 1994/95 to 1997/98) for Cameroon manufacturing firms to estimate a
single stochastic production frontier for each industrial sector. This frontier was used to
assess the effects of trade reforms in manufacturing firm-level technical efficiency. A Cobb-
Douglas production function was specified and estimated for the production frontier. The
73
link between trade reforms and firm-level technical efficiency was established using a two-
stage procedure. In the first stage, the production frontier parameters were estimated and
firm-level technical efficiencies derived. In the second stage, the derived firm-level technical
efficiencies were regressed on trade policy and macroeconomic variables to assess the
impact of trade reform and macroeconomic variables.
The results suggested that trade reform provided an enabling environment for improving
firm-level technical efficiency. Average technical efficiency increased in six of the eight
sectors following trade reforms. The pre-reform firm-specific technical efficiencies
decreased on average at an annual rate of 0.76 percent, while the post-reform firm-specific
technical efficiency increased on average at an annual rate of 1.4 percentt. Lastly, factors
that characterize firm-level technical inefficiency prior to trade liberalization, as showed by
the Tobit and fixed effects results were macroeconomic instability and political instability
of the early 1990s, coupled with restricted trade regime. After the trade reforms, the potential
determinants of firms’ technical efficiency were export share and import penetration rate
(Njikam et al., 2008).
74
CHAPTER FOUR
TECHNICAL EFFICIENCY IN THE CAMEROONIAN MANUFACTURING
FIRMS: A STOCHASTIC FRONTIER ANALYSIS
4.1 Introduction
This Chapter presents an analysis of the factors that influence technical efficiency of
manufacturing firms in Cameroon. This analysis is motivated by increasing concern on the
need for firms to increase their output in order to achieve higher performance. This
necessitates that efficiency of a firm be defined. Efficiency measurement in production
basically consists of analyzing how firms combine their inputs to produce a level of output
in an efficient way (Coelli et al., 2005). Analysis of the efficiency in manufacturing firms in
Cameroon is further motivated by the fact that the manufacturing sector has been a driver of
economic expansion and accounted for one fifth of the GDP in 2000, contributed
approximately 18.7% to GDP in 2005 and rising to 20% in 2009 (World Bank, 2012). It also
contributes to aggregate export earnings. It generates income to the manufacturers and also
serves as a source of employment
Even though the contribution of the sector has been increasing over the years, low output
growth over the last two decades is causing concerns in Cameroon. To identify the
underlying causes, studies at the micro level are highly relevant given that efficiency studies
at the micro level of the Cameroonian manufacturing sector are very limited.
75
Cameroon industrial policies raise different questions which this study sheds light on. In
explaining this gap, the following questions will be addressed: What are the determinants of
Cameroon’s manufacturing efficiency, and using these determinants, how efficient are the
manufacturing firms in Cameroon across different industries?
4.2 Purpose and Motivation
The main purpose of this chapter is to examine technical efficiency of manufacturing firms
and investigate the determinants for inefficient operation, using firm level data. The chapter
also examines whether foreign-dominated firms are more technical efficient than those
dominated by nationals, given the significant amount of Foreign Direct Investment (FDI) in
Cameroon. Also, the mean technical efficiency of the manufacturing firms is measured by
sectors. Reliable estimates of technical efficiency are important from a policy perspective as
they play a vital role in production economics especially in LDCs.
From the review of studies, this chapter will employ the stochastic frontier production
function to estimate the efficiency of manufacturing firms in Cameroon4. There are of
course, other ways of evaluating manufacturing firms such as profitability, measured as
return on assets or operating profit margin, or efficiency ratios such as total assets turnover.
These approaches have inherent problems because of their failure to incorporate the
environment in which the firm operates. Hence, the concept of efficiency can overcome such
problems and thereby indicate which firms are the least efficient. Using the stochastic
4 See chapter 3 for a detailed empirical literature review
76
frontier not only provides insight about the most efficient firms but also about the
determination of the sources of the best practices.
In terms of contribution to knowledge, the chapter contributes positively to the ongoing
efficiency debates in academic circles. Firms are being encouraged to adopt efficient
production processes by using fewer inputs in order to maximize output. It will also inform
managers if there is any benefit of adopting a frontier production technology and if it has an
impact on a firm’s output. For policy makers, this chapter will inform them on the benefits
of providing incentives to firms to promote efficient production technology in order to
maximize output.
For meaningful results, the following hypotheses are tested.
The Cameroonian manufacturing firms are technically efficient. In other words, the
hypothesis stipulates that no productivity gains linked to the improvement of
technical efficiency may be realized in the manufacturing sector.
The firm age, size, ownership structure, tax rates, labor regulations, access to finance
and corruption do not significantly influence the firm’s technical efficiency.
4.3 Production Efficiency and Stochastic Frontier Analysis
Frontier methodologies have emerged as an important development for estimating efficiency
and productivity which originated from the theoretical contribution by Farell in 1957. Farell
77
(1957) created a framework to analyze firms that are not fully efficient. He suggested that
efficiency could be evaluated by comparing firms to “best practice” efficient frontiers
formed by the dominant firms in an industry. Empirically, efficiency is measured by
estimating best practice efficient frontier based on a relevant sample of firms. The firms on
the frontier are considered to be the best practice firms in the industry in the sense that their
performance is at least as good as that of other firms with similar characteristics. The
efficiencies of other firms in the market are measured in comparison to the efficient frontier.
There are two major classes of efficiency estimation methodologies, and they are the
econometric, which is also known as the parametric approach and the mathematical
programming approach also known as the non-parametric approach.
The stochastic frontier technique ( Aigner, Lovell and Schmidt, 1977; Meeusen and Van den
Broeck, 1977) or the parametric approach can be formulated in two steps: firstly, an
appropriate function such as a production, cost, revenue or profit function is estimated using
an econometric method such as the OLS, non-linear least squares or maximum likelihood;
then secondly, the estimated regression error terms are separated into two components,
usually a two sided random error component and a one-sided inefficiency component. This
produces an estimate of efficiency for every firm in the estimated sample. In the
mathematical programming approach, the implementation that is used most frequently is
data envelopment analysis (DEA), which was originated by Charnes et al (1978). The
method can be used to estimate production, cost, and profit frontiers and provides a
particularly convenient way for decomposing efficiency into its components.
78
Lovell (1993) noted the advantages pertaining to stochastic frontier analysis. The major
benefit of the econometric approach is that it allows firms to be off the frontier due to random
error as well as inefficiency and separates purely random error from inefficiency effects. It
also requires distributional assumptions for the error term in order to recover the efficiency
estimates, the selection of which may be arbitrary (Coelli, 1995). The stochastic frontier
models also have the advantage of controlling for random events and of distinguishing the
statistical noise effects from technical inefficiency. The technique assumes that producers
may deviate from the frontier not only because of measurement errors, statistical noise or
any non-systematic influence but also because technical inefficiency. Based on the
econometric estimation of the production frontier, the efficiency of each firm is measured as
the deviation from the best practice technology.
More so, one of the most important issues in stochastic frontier models is to take account of
the unobserved heterogeneity among firms operating in different production environments
(Greene, 2005). Individual firms carry out their production in different environments
characterized by external factors which can influence their technology but are not under their
control. Hence, production possibilities are expected to differ in a cross-section of firms, and
a set of different technologies may simultaneously coexist at any given time. If that is the
case, the evaluation of technical efficiency cannot be performed by considering a common
technology.
79
Although they have a few particular problems of their own, stochastic frontiers are less
subject to the aforementioned pitfalls. In addition, their statistical nature allows hypotheses
to be tested regarding the existence of inefficiency and the structure of the production
technology (Coelli, Rao and Battese, 1998).
The mathematical programming approach, DEA, is non parametric which implies that it is
not necessary to specify a functional form or distributional assumptions; hence it is not prone
to specification errors. However, the fact that it is not stochastic renders it impossible to
isolate technical efficiency from random noise (Lovell, 1993). Therefore, any departure from
the frontier is measured as inefficiency. One other advantage of the DEA is that it solves the
optimization problem separately for each decision making unit (Charnes et al. 1994) unlike
the econometric models which optimize over a sample as a whole, and the estimated function
is assumed to apply to all units in the sample, with all of the differences among firms
captured through the estimated residuals (Cummins and Zi, 1998).
The choice of the methodology is based on the advantages and disadvantages of each of
these approaches and on the data sets used. Cummins and Zi (1998) suggest that in datasets
that are known to be noisy, the econometric approach or stochastic frontier method is more
appropriate because it is capable of filtering out random errors from inefficiency. When the
objective is to study the performance of specific units of firms, mathematical programming
is more appropriate because the optimization is conducted separately for each unit.
80
From the foregoing, it is evident that both stochastic frontier analysis and data envelopment
analysis are useful in measuring the inefficiency of firms. However, the stochastic frontier
is more appropriate for this study because although, unlike data envelopment analysis, there
is need to specify the functional form and distribution assumption, it has an advantage of
isolating technical inefficiencies from random noise, which is the main methodology of this
chapter. The next section presents the methodology and data used for empirical analysis of
technical efficiency.
4.4 Methodology and data
This section presents the methodology used in the empirical analysis of technical efficiency
of the firms in the sample. It begins with the analytical framework of the concept of technical
efficiency. This is followed by the earlier development in the frontier analysis. Then, the
model for technical efficiency is specified as well as the data used in the analysis is
discussed.
4.4.1 Analytical Framework
Typical models of frontier function analysis start with a production function and in these
models producers are assumed to operate on their frontier production functions, maximizing
output using the available inputs. Different least square techniques have long been used in
empirical analyses in which error terms were assumed to be symmetrically distributed with
zero means and the only source of departure from the estimated function was assumed to be
81
statistical noise. These analyses considered productivity only and did not deal with technical
efficiency. The pioneering work of Koopmans (1951) provided a definition of technical
efficiency suggesting that not all producers were technically efficient and since then, there
are increasing number of studies modeling production functions with the assumption that
not all firms might be operating efficiently.
Chen et al. (2003) describe two measures of technical efficiency; an output oriented measure
and an input-oriented measure. An output-oriented measure of is measured as the ratio
observed to maximum feasible output, conditional on technology and observed input usage
and is defined formally as the function:
1
0 ( , ) max : : 4.1TE y x y y xcan produce y
Where y is the level of output and x is a vector of inputs.
An input-oriented measure of technical efficiency is measured as the ratio of minimum
feasible input use to observed input use, conditional on technology and observed output
production and is defined as the function:
1( , ) min : : 4.2TE y x x x xcan produce y
where the variables are defined above.
82
The two measure yield the same results if and only if the technology is constant return to
scale (Chan et al. 2003). This chapter measures technical efficiency using an output
expanding orientation specified as:
1`
max : ( , ) 1 4.3TE Y F X Z
where (.)F is the production frontier, Y is the output and X is a vector of inputs.
4.4.2 Early Developments in the Frontier Analysis
Farrell (1957) pioneered measurement of productive efficiency empirically. Using data on
US agriculture, he defined cost efficiency and decomposed it into its technical and allocative
parts using linear programming techniques rather than econometric methods. His work
eventually led to the development of data envelopemnt analysis (DEA) and this method is
widely used in the literature as a non-parametric non-stochastic technique.
Farrell’s work also led to the development of stochastic frontier analysis which involved
estimating deterministic production frontiers, either by means of linear programming
techniques or by modification to the least squares techniques. Initial studies on efficiency
using deterministic production frontier models assumed the error term was not affected in
any way by statistical noise and thus represented inefficiency.
Following Farrell (1957), Aigner and Chu (1968) considered the idea of a deterministic
production frontier using a parametric frontier function of a C-D form defined as:
83
1, 2,3..., 4.4i i iIn y x u i N
where ty is the output for the thi firm, ix is a vector of inputs, is a vector of unknown
parameters of the intercept and the slope terms and iu is non-negative random variable
associated with technical inefficiency. The measure of efficiency is given as the ratio of the
observed output of the thi firm to the potential output defined by the frontier function and is
outlined as:
expexp 4.5
exp expi ii
i i
i i
x uyTE u
x x
Following Aigner and Chu (1968) there have been other studies that used the same approach
by applying different estimation techniques. Some studies used the Corrected Ordinary Least
Squares (COLS) method to estimate the production frontier, which involved the estimation
of the model in two stages where parameter estimates are obtained in the first stage using
Ordinary Least Squares (OLS) method and the intercept term is corrected by shifting it
upwards until all residuals are non-positive and the largest residual is zero, in the second
stage (Lovell, 1993). These corrected residuals are then used to calculate technical efficiency
for each producer. The corrected least square (COLS) method leads to negative variances.
According to Olson, Schmidt and Waldman (1980) this would affect the efficiency of the
estimators in the model and hence the TE estimates. The main drawback of this method was
84
the implication of both efficient and inefficient producers having the same structure of
frontier technology.
In order to overcome this drawback of the COLS method, an alternative method known as
Modified Ordinary Least Squares (MOLS) was proposed. It involved the assumption that
the error term followed a one-sided distribution.
Schmidt (1976) argued that if the error term associated with the technical inefficiency effects
followed a one side distribution such as exponential or half normal, then linear programming
estimates proposed by Aigner and Chu (1968) were maximum likelihood estimates of the
deterministic frontier model, which led to the wide use of Maximum Likelihood Estimation
(MLE) techniques in stochastic production frontier analysis.
Although these early studies estimated technical inefficiency, their approach was
deterministic because no allowance was made for the possible influence of a measurement
error and other statistical noise on the estimated production frontier. In other words, all the
deviations from the frontier were assumed to be the result of technical efficiency.
4.4.3 The Stochastic Frontier Models
To estimate firm-level technical efficiency and investigate its determinants, the parametric
stochastic frontier production function approach suggested by Battese and Coelli (1995) is
applied. This procedure is a development of Aigner, Lovell and Schmidt (1977) and
85
Meeusen and van den Broeck (1977). Stochastic frontier model analysis technical
inefficiency effects in terms of other explanatory variables. Considerable amount of research
has been conducted thereafter to extend and apply the model (Schmidt, 1986; Bauer, 1990;
Battese, 1992; Greene, 1993; Battese et al., 1995).
The specification of the stochastic frontier model is a production function with an error term
incorporating two components: the output-based unobservable technical inefficiency factor
,iu and a symmetric component ,iv capturing random variations across production units and
random shocks that are external to its control. The model is specified as;
,( ) ; 1, 2, ..., 4.6i i iY f X e i N
Where iY represents the potential output level on the frontier for firm i, given technology
iXf (.), is a )1( k vector of inputs and other explanatory variables associated with the thi
firm. β is a )1( k vector of unknown parameters. The error term ie is composed of two
independent elements, i.e., iii uve , with the iv term being a random (stochastic) error,
which is associated with random factors not under the control of the firm. It is assumed to
be independently and identically distributed as ),0( 2vN , where 2
v stands for the variance
of stochastic disturbance iv . The term iu captures technical efficiency and is a non-negative
one sided component (since realized output is lower than potential output) associated with
86
industry-specific factors. It is distributed independently from and identically to iv . If
industries achieve their maximum output, then they would be technically efficient and this
means that 0.iu iu is associated with the technical inefficiency of the thi firm and defined
by the truncation (at zero) of the normal distribution ),( 2uizN , where iz
is a )1( m vector
of explanatory variables associated with technical inefficiency of firms; and is an )1( m
vector of unknown coefficients.
Following Battese and Coelli (1995), the stochastic frontier production function can be
specified in terms of the original values as follows:
( , )exp( ) 4.7i i i iIn Y f X v u
The model is such that the possible production iY is bounded above by stochastic quantity,
),exp(),( iii uvXf hence the term stochastic frontier.
From equation (3.4), the technical inefficiency effects, ,iu are assumed to be a function of a
set of explanatory variables, ,iz and an unknown vector of coefficients, .i The explanatory
variables in the inefficiency model may include some input variables in the stochastic
frontier, provided the inefficiency effects are stochastic. Battese and Coelli (1992) suggested
that if all the elements of the -vector are equal to zero, then the technical inefficiency
effects of firms are not related to the z -variables and so the half normal distribution
87
originally specified in Aigner, Lovell and Schmidt (1977) is obtained. If interaction between
firms-specific variables and input variables are included as z -variables, then a non-neutral
stochastic production frontier is obtained (Battese and Coelli, 1995). Hence, the technical
inefficiency effect, ,iu in the stochastic frontier model in (3.4) can be specified as:
4.8i i iu z w
where the random variable, iw is defined by the truncation of the normal distribution with
zero mean and variance , ,2 such that the point of truncation is ,iz i.e., .ii zw
According to Battese and Coelli (1993), these assumptions are consistent with iu being a
non-negative truncation of the 2( , )iN z distribution. Hence, the mean, ,iz which is
truncated at zero to obtain the distribution of ,iu is not required to be necessarily positive
for each observation, si ' are inefficiency parameters to be estimated. The assumptions that
iu and iv
are independently distributed for all ,,...,2,1 Ni is obviously a simplifying, but
restrictive condition.
Battese et al. (1993) proposed the method of Maximum Likelihood for simultaneous
estimation of the parameters of the stochastic frontier and the model for the technical
efficiency effects. By following different parameterization such as those of Battese and
Coelli (1988), and Battese (1992), the likelihood function of the model defined in equation
(3.3) can be written:
88
2 2
21 1
1( ) 1 4.9
2 2 2(1 )
N Ni
i i
eNIn L In In In e
With );( iii xfInYe , ie is the residual of Equation 4.9, N is the number of observation,
(.) is the standard normal distribution function, and 222uv and 22
vu are
variance parameters.
The technical efficiency of production for the thi firm from the above can be defined in terms
of the ratio of the observed output to the corresponding frontier output, given the available
technology. The technical efficiency is thus empirically measured by decomposing the
deviation into random component )(u as follows:
*
( , ) exp( )exp( ) 4.10
( , )exp( )
exp( ) exp( )
i i i ii
i i
i i i i
Y f X v uTE u
Y f X v
TE u z w
Where iY is the observed output and *
iY is the frontier output and iw being an error term that
follows a truncated normal distribution. This is such that .10 TE If industries achieve
their maximum output, then they would be technically efficient and this means that 0.iu
89
By assuming a half-normal distribution of iu, mean technical efficiency can be computed
as follows:
2exp( ) 2 1 exp 2 4.11iE u
Moreover, the measurement of technical efficiency (or inefficiency) level of firm i requires
estimating the random term ..iu The prediction of the technical efficiencies is based on its
conditional expectation, taking into consideration the assumptions made on the distribution
of iu and .iv Jondrow et al. (1982) first computed the conditional mean of iu given .ie
Battese et al. (1988) derived the best indicator of firm i technical efficiency, written as
exp( )i iTE u using the formula:
21 /
exp( ) exp / 2 4.121 ( / )
A i Ai i i A
i A
eE u e e
e
where 21 A
The density function for ite where ( )e u v is given by:
2 2
2 2 2
1 2
1( ) 1 ( ) exp( 1 2) (4.13)
(2 )i i v i u i vf e F e e e
Where 2 2 2u v
(.)F is the cumulative distribution function of the standard normal random variable.
90
4.5 The sample of Cameroonian manufacturing firms and variables
The data set used in this chapter is obtained from the Regional Program Enterprise
Development (RPED) dataset for Cameroon’s manufacturing sector for the year 2009
captured by the World Bank’s RPED survey of year 2010.
The main objective of these surveys in African countries is to increase the knowledge of the
creation process of African manufacturing firms and to shed some light on the problems they
face in their development. The RPED defines formal firms as those recorded in the trade
register. They are known to the government tax authorities and are potential taxpayers for
all regular taxes resulting from their commercial activities.
The purpose of the survey in Cameroonian manufacturing was to capture business
perceptions on the main obstacles to enterprise growth, the relative importance of various
constraints to increasing employment and productivity, and the effects of a country’s
business environment on its international competitiveness. The sample consisted of 319
employing at least 5 permanent workers, and covered the following manufacturing sub-
sectors: food processing, textile and garments, chemicals and pharmaceuticals, non-metallic,
machinery and equipment, electronics and wood processing.
An important advantage of this data set is that it enables one to test for inefficiency using
truly microeconomic data. In fact, it has been found that empirical tests which rely on
91
microeconomic data, provide clearer evidence of inefficiency than studies that make use of
more aggregate data, since there is a loss of information in the aggregation process (Schmidt
and Lovell, 1979).
The survey was conducted on firms located in the major industrial regions in Cameroon
which consist of Littoral (Douala), Centre (Yaoundé), West (Bafousam), representing
approximately 92 percent of the total number of firms in the country. Table 4.1 shows the
structure of firms by region and type. Littoral (Douala) which has the industrial zone as well
as the export zone in Cameron has approximately 75.86 per cent of the total firms in the
sample. The dataset contain firm-level information on various aspects which include sales,
capital and labor costs, purchases, energy costs and water costs. This renders the dataset
useful as a basis for analysis of the technical efficiency of the sampled firms.
The five sectors covered in this study represent approximately 76.18 per cent of total
manufacturing in Cameroon (RPED, 2010). The food, wood and textile and garments sectors
are the dominant sectors in terms of output and employment, followed by metals and
machinery, electronics, chemical and pharmaceutical industries among others. During the
years of import substitution, most resources were invested in the food sector, and later,
during the 1980s, in the wood and other sectors. Because some of the investments in food
and wood production were foreign, it has been suggested that these sectors are the most
productive and technologically advanced. Output in the food sector comprises a wide range
of commodities, including grain milling, dairy products, canning and preservation of meat,
92
fruit and vegetables, bakery and confectionery, and production of salt, beverages, food
preservatives and animal feed (Njikam and Cockburn, 2007).
The wood sector makes timber products, furniture, wooden art, and storage and packaging
materials. The products of the metal sector consist of both simple engineering items, based
on sheet metal (containers, utensils, window frames, metal furniture, etc.), and more
sophisticated equipment to serve the needs of the construction industry, railway system and
the agricultural sector. Production in the textile sector consists of the manufacturing of
garments, furnishings and carpets, and industrial goods, including belting, ropes, and sacks
among other products which are exported to most of the CEMAC member countries.
Table 4.1 shows the distribution by size, the sector of activity and the age of the firms.
The greater proportion of medium size firms are 20 years old and above. Generally, there
are more medium size firms in the sample, followed by large firms.
93
Table 4.1: Distribution of firms according to size, age and sector of activity
Sector of activity and size of firm Age of firm
Food Wood Textile Metal Electro
nics Non
Metal Others Total [0,5] (5, 10] (10, 20] (20, +) Total
Small (<20) 15 15 11 16 8 14 19 86 4 21 23 38 86 Medium (20-99) 27 26 20 13 16 11 29 129 7 15 38 69 129 Large (100 and above) 29 14 10 10 13 11 20 104 6 13 30 55 104
Total 71 55 41 39 37 26 68 319 17 49 91 162 319 Source: Cameroonian data base, RPED, World Bank
Table 4.2: Distribution of firms by size and region in Cameroon
Littoral (Douala) Centre (Yaounde) West (Bafoussam) Total
Small (<20) 38(31) 6(9) 2(0) 46(40)
Medium (20 - 99) 58(41) 5(15) 4(6) 67(62)
Large (100 and above) 47(29) 7(11) 3(7) 57(47)
Total 141(101) 18(35) 9(13) 170(149) Source: Cameroonian data base, RPED, World Bank.
94
4.6 Definition of variables and the empirical analysis
4.6.1 Variables of Production Technology
Output
Output (Y), can be measured in physical or value terms. Since firms under consideration in
this study produce a number of products, an aggregated measure of output in value term is
used. Thus, the y-output is measured by total output (sales) at constant prices. Therefore,
Output is the value of all output produced by the firm in the given year.
Inputs
Four categories of inputs are used in this study: Capital (K), Labor (L), Human Capital (H)
and intermediate inputs (R). According to Taymaz and Saatci, (1997), the capital input is
defined theoretically as the services of capital goods in value terms. The data set does not
provide data for capital services and replacement value of fixed assets. Therefore, we use a
proxy variable. There are two alternatives available: the book value of fixed assets
(machinery and vehicles), and the total annual depreciation or depreciation allowances. In
this study, depreciation allowances are used to measure the capital input. Conceptually the
replacement values should reflect the superior quality of capital used in larger firms.
Therefore, Capital is defined as the replacement cost of existing machinery and other
equipment employed in the production process (Lundvall et al., 2002).
95
The labor input (L) is measured as total number of hours worked in the firm. This measure
applies to both temporary and permanent employees, which tend to vary across firms. Thus
it is underestimating the quality of labor which can be expected to increase with firm size.
This part of the problem might be taken care of to some extent by adding a quality dimension
to the labor factor in the production function.
A third input variable was taken into account to capture the specific impact of human
qualifications. The dimensions of the human capital that can be measured from the RPED
survey are numbers of years of education attainment of an employee in the firm. This
variable captures the specific impact of human qualifications.
Production functions with this alternative specification have been estimated by Bigsten et
al., (2000), Chapelle and Plane (2005).The intermediate inputs variable (R) is measured as
the expenditures on inputs (raw materials and supplementary materials such as solid and
liquid fuel, electricity and water costs) adjusted for stock changes. The stochastic frontier
production function to be estimated is:
0 1 2 3 4( ) ( ) ) ( ) 4.15i i i i i i iInY In K In L InH In R v u
Where In denotes natural logarithms, i are unknown parameters to be estimated.
The input choice in the model is based on economic theory in regard to the firm profit
maximization with efficient resource allocation
96
4.6.2 Determinants of Manufacturing Efficiency
Previous studies (Faruq and Yi, 2010; Chapelle and Plane, 2005, Taymaz and Saatci., 1997)
enumerated and examined various attributes that may affect technical efficiency. Drawing
on the existing theoretical and empirical literature on the determinants of efficiency, this
study uses among others, the following variables to study their effect on technical efficiency
in Cameroon’s manufacturing industries.
Firm size: Size is measured in this study by considering three categories defined according
to the number of workers: small, medium and large (Table 4.1). Most studies have used
firm’s total employees as a measure of firm size (Faruq and Yi, 2010; Oczkowski and
Sharma, 2005; Lundvall and Battese, 2000). These studies find significant positive
relationship between firm size and firm efficiency. Larger firms are considered to be more
efficient than smaller firms because they are thought to have superior organization or
technical knowledge, greater market power, better access to important resources and they
enjoy economies of scale. According to Faruq and Yi (2010), the relationship between firm
size and firm efficiency becomes ambiguous because it can be argued that small firms can
be more efficient sine they are more exposed to competition than large firms and have a
strong incentive to address their own weaknesses in order to survive. Lundvall and Battese
(2000) argue that small firms adopt more appropriate technology, are more flexible in
responding to changes in technology, product lines and markets, and foster more competitive
factor and product markets, and thus are able to use resources more efficiently. Hence,
increasing the size of small firms may result in coordination problems within the firms, thus
97
leading to inefficiency. According to Jovanovic (1982), efficient firms grow and survive,
while inefficient firms stagnate or exit the industry. As efficient firms grow, they gain
experience and improve work practices, leading to efficiency improvement.
Firm age: Mixed evidence has been found in empirical studies regarding the relationship
between firm age and firm efficiency. Evidence from the Indonesian weaving industry, as
provided by Pitt and Lee (1981) has shown that age has a consistent positive effect on
efficiency. Older firms are usually considered to be more efficient than younger firms
because they are thought to have gained experience from past operations and thus their
survival may reflect their superior efficiency. Older firms may identify and reject previously
used inefficient production methods (Malerba, 1992). However, Little, Mazumdar and Page
(1987), on a comparative analysis of small manufacturing enterprises in India and other
economies conclude on the possibility that older firms may be less efficient if they fail to
upgrade to new production technology and adapt to changing market conditions. Lundvall
and Battese (2000) suggest that the link between age and efficiency may depend on the
nature of the industry. They find a positive relationship between efficiency and age among
Kenyan firms in the textile sector, but no effect of firm age on efficiency was identified in
the food, wood and metal sectors.
Ownership structure: Efficiency may also be related to ownership of firms. Mahadevan
(2000) puts it that domestic ownership may improve efficiency since foreign owners are
generally less familiar with the local environment; local shareholders can help to improve
98
firm efficiency. According to Benard et al. (2003), foreign ownership is associated with
lower profits due to coordination problem and high cost of learning about the different local
markets. However, as investigated by Faruq and Yi (2010), foreign ownership can help
improve the efficiency of domestic firms by giving them access to foreign technology,
management talents and distribution network. According to Oczkowski Sharma (2005), the
link between foreign ownership/participation and efficiency improvement still remains an
empirical matter. The inclusion of foreign ownership variable in the model is due to the fact
that Cameroon has had a long history since independence as a major recipient of FDI. Hence,
the effect of foreign ownership on the use of resources and given technology is investigated.
More so, the inclusion of foreign ownership variable is based on the assumption that foreign
firms operating in the Cameroon environment share the same technology frontier as the
domestic firms. In this study, a dummy variable is used for industries which have more than
45% of the total number of firms in that industry wholly foreign or joint ventures which are
less than half locally owned. Dunning (1988) explains that FDI often stems from ownership
advantages like specific knowledge on the use of resources due to research and development
experience and/or exposure to international competition. However, empirical evidence of
foreign ownership on the efficient use of resources in the host country is mixed.
Trade Union: The presence of trade union has also been taken into account, the expected
sign being unclear. On the one hand, the existence of strong unions within a firm may
contribute to restricting the set of managerial decisions by reducing the speed of adjustment
of the labor force to the business cycle. However, on the other hand, it can be a source of
99
positive stimulation for the emergence of procedural arrangements encouraging high level
of efforts and loyalty within the firm (chapelle and Plane, 2005).
Corruption (COR): This variable is tested under the hypothesis that it represents a
transaction cost effect, which is potentially more serious for large firms. Qualitative
variables are defined on the basis of the surveyed manager answers concerning the excess
unit cost constituting an obstacle to the current operation of the firm. In this case, an
increasing disturbance is measured by a discrete variable with a value ranging from 1 (no
obstacle) to 5 (very severe obstacle). In brief, the inefficiency effect model is written as:
0 1 2 3 4 5 6
7 8 9 10
Re
exp 4.16
i
i
u Firmsize Age Foreign Union glabor Corruption
Taxrates Acessfin Mngedu Mng
Equation 4.16 is then estimated in logarithms to measure changes in TE. Since TE levels are
bounded between the value of zero and one, in order to comply with the standard normal
assumptions of the error term in a multiple regression equation, the TE values were
transformed to (1 ).InTE In TE Therefore, the regression coefficients have no direct
interpretation but it is possible to calculate the elasticity value from the estimated coefficient.
The variables discussed above are shown in the conceptual model as indicated in Figure 4.1.
100
Figure 4.1: Conceptual model of manufacturing firms’ Technical Efficiency
Source: Author
Table 4.3 presents summary statistics for the variables of production technology as well as
the some of the variables (with the exception of the binary and categorical variables) used in
the efficiency effects model in explaining the differences in the inefficiency levels of firms
in Cameroon.
Inputs
Capital, Labor, Human capital
and intermediate inputs
Production function (process)
Output
Efficien
cy
Corruption
Factors
Influencing
Technical
Efficiency Tax rates
Firm
size
Firm age Ownership
Access to finance
Labor regulations
Manager’s educ
Manager’s exp Firm location
101
Table 4.3: Summary statistics of Variables in different sectors
Obs. Mean Std. Dev Min. Max.
Food Processing Log of Output 71 20.8808 2.5621 15.4949 26.0846 Log of labor 71 18.8404 2.1716 14.5856 24.5945 Log of capital 71 17.8965 2.9496 10.8198 24.5945 Log of human Capital 71 1.14130 0.4482 0 1.6094 Log of Intermediate Inputs 71 19.1531 2.7453 12.4292 24.9159 Firm age 71 26.9437 26.944 1 61 Mng exp.(Manager experience) 71 16.3944 8.5815 2 40
Wood and Furniture Log of Output 55 19.566 2.1665 13.6412 25.1053 Log of labor 55 17.6469 1.7725 13.0815 23.0259 Log of capital 55 17.4585 2.4471 11.9087 23.3623 Log of human Capital 55 1.1955 0.3168 0 1.6094 Log of Intermediate Inputs 55 17.9509 1.1377 11.9184 22.1096 Firm age 55 22.8545 14.7226 4 61 Mng exp. (Manager experience) 55 18.4546 11.1186 3 50
Textile and Garments Log of Output 41 19.5606 2.9291 14.5087 25.3284 Log of labor 41 17.2192 2.3456 13.017 23.0259 Log of capital 41 16.3661 3.2082 9.9688 23.3623 Log of human Capital 41 1.1066 0.4037 0 1.6094 Log of Intermediate Inputs 41 17.5829 2.6484 11.7906 22.7671 Firm age 41 23.4878 11.485 4 47 Mng exp. (Manager experience) 41 19.3659 8.8396 6 40
Metal and Machinery Log of Output 39 192733 1.8827 16.1181 24.2599 Log of labor 39 17.3604 1.8581 13.5924 21.3609 Log of capital 39 16.7419 2.2385 13.3047 23.1212 Log of human Capital 39 1.1470 0.3929 0 1.6094 Log of Intermediate Inputs 39 17.6146 2.1945 14.2209 22.9954 Firm age 39 20.3333 16.1772 2 63 Mng exp. (Manager experience) 39 19.8718 8.7934 3 45
Electronics Log of Output 37 19.1649 2.278 14.7318 14.6353 Log of labor 37 17.2202 1.9963 13.6171 21.8219 Log of capital 37 16.7325 2.3859 10.8198 23.1211 Log of human Capital 37 1.2225 0.3296 0.6931 1.6094 Log of Intermediate Inputs 37 17.9704 2.2241 12.5818 23.719 Firm age 37 23.7027 13.824 5 76 Mng exp. (Manager experience) 37 20.6487 8.1454 9 45
Overall Sample Log of Output 319 19.8927 2.4911 13.6412 26.0845 Log of labor 319 17.839 2.0989 13.017 24.5945 Log of capital 319 17.1322 2.6716 9.9688 24.5945 Log of human Capital 319 1.1476 0.4242 0 1.6094 Log of Intermediate Inputs 319 18.3764 2.5805 11.7906 24.9159 Firm age 319 23.5047 15.0264 1 76 Mng exp. (Manager experience) 319 18.4075 9.2684 2 50
Source: Author’s calculation from the Cameroonian base, RPED, World Bank
102
4.7 Pair-wise matrix of correlation coefficients
Matrices of correlation coefficients between variables are illustrated in Table 4.4.
Table 4.4: Pair-wise Correlation Matrix
Output Labor Capital Human Capital Intermediate Inputs
Food Processing (n=71) Output 1.0000 Labor 0.8932 1.0000 Capital 0.2353 0.2269 1.0000 Human Capital -0.0435 -0.0811 -0.0570 1.0000 Intermediate Inputs 0.8354 0.8311 0.2552 -0.1523 1.0000
Wood and Furniture (n=55) Output 1.0000 Labor 0.7848 1.0000 Capital 0.0322 0.0957 1.0000 Human Capital -0.0870 -0.0397 0.1855 1.0000 Intermediate Inputs 0.7282 0.6431 -0.0996 -0.2631 1.0000
Textile and Garments (n=41) Output 1.0000 Labor 0.7189 1.0000 Capital 0.4569 0.3150 1.0000 Human Capital 0.0576 0.1262 0.0127 1.0000 Intermediate Inputs 0.6038 0.5922 0.4341 0.1402 1.0000
Metals and Machinery (n=39) Output 1.0000 Labor 0.4515 1.0000 Capital -0.2116 -0.0064 1.0000 Human Capital 0.2274 0.1752 0.2590 1.0000 Intermediate Inputs 0.6317 0.6460 0.0448 0.2573 1.0000
Electronics (n=37) Output 1.0000 Labor 0.9201 1.0000 Capital 0.2559 0.2252 1.0000 Human Capital -0.0431 0.0649 0.2694 1.0000 Intermediate Inputs 0.7016 0.6679 0.0992 0.0364 1.0000 Overall Sample (n=319) Output 1.0000 Labor 0.8259 1.0000 Capital 0.2705 0.2637 1.0000 Human Capital 0.0087 0.0201 0.0670 1.0000 Intermediate Inputs 0.7780 0.7454 0.2606 0.0143 1.0000
103
An overview of variables correlations shows that output is significantly correlated with a
number of variables in the Cobb-Douglas production and stochastic frontier functions. The
zero order correlation coefficients table shows that output, labor, capital and intermediate
inputs are correlated at least at ten percent level. Human capital has a weak correlation with
output as well as an unexpected sign in most of the sectors.
4.8 Estimation Procedures and Functional Forms
4.8.1 Estimation procedures
Stochastic frontier production functions can be estimated using either the maximum
likelihood method or using a variant of the COLS (corrected ordinary least squares) method
suggested by Richmond (1974). This study will consider the maximum likelihood estimation
(MLE) method because of the availability of software programs which have automated the
MLE (Coelli, 1996). The MLE method has also been found to be significantly better than
COLS where the contribution of the inefficiency effects of the total variance is large, and is
the preferred estimation technique whenever possible (Coelli, et al. 1998).
This study employed a two-stage approach (Pitt and Lee, 1981; Kalirajan, 1981; 1991). The
first stage consisted of specifying and estimating the stochastic production frontier and
predicting the technical inefficiency effects, under the assumption that these inefficiency
effects are identically distributed. In the second stage, a regression model for the predicted
technical inefficiency effects is specified and estimated. This approach was highly defended
104
by Kaliajan (1991) in which he claimed that the production unit-specific factors exert only
direct influence on production through their association with inefficiency. Coelli (1995)
stipulated the use of inefficiencies as a dependent with the assumption of identically
distributed efficiency effects in the stochastic frontier. In estimating the parameters of both
the stochastic frontier and the model explaining technical inefficiency effects, the chapter
applies MLE of parameters of a variety of stochastic production (Abuka, 2005).
4.8.2 Functional Forms
Defining the production function requires giving (.)f some type of algebraic form based on
economic theory. Production functional forms are characterized by several properties. There
are basically two most common functional forms of production functions used in the
literature in studying technical efficiency using stochastic frontier functions, namely Cobb-
Douglas and the general trans-log functional forms. These functional forms are specified to
both model and data. As indicated in the literature, most efficiency studies focused on
determining the degree of inefficiency and hardly did they examine alternative specifications
of the technology (Chapelle and Plane, 2005). Thus, if an incorrect functional form is
employed, the model will potentially predict responses in biased and inaccurate way (Amos,
2007). Hence, the consequences of this mis-specification of the functional form may include,
among others, misleading policy implications.
105
This study employs both the Cobb-Douglas and trans-log functional forms mainly for two
reasons. In addition to being the most commonly used functional forms, they will also allow
for comparison to be made between the findings of the current study relative to previous
studies that have analyzed the manufacturing firms in Cameroon. Examples of such include
among others, Soderling (1999) who used the Cobb-Douglass form and Njikam (2000; 2003)
who employed the trans-log form.
Since the Cobb-Douglas specification is nested in the trans-log model, we start with the
trans-log specification defined as:
20
1 1 1 1
1 14.17
2 2i j ji T TT jt ji jk ji ki i i
j j j k
In y In x t t In x t Inx In x v u
where the subscript i indicates firm, kandj index inputs, y is output, jx is a vector of
inputs, v random errors, u firm specific technical efficiency effects, and s' parameters to
be estimated. The v random errors are assumed to be independently and identically
distributed.
The trans-log specification is the most commonly used flexible functional form, because it
can provide a second order approximation Therefore, the trans-log does not impose
restrictions on the structure of the technology, such as, restrictions on returns to scale or
elasticity of substitution. It permits the elasticity of substitution to be determined by the data
(Amos, 2007). Coelli (1995) stated that the main weaknesses associated with the trans-log
106
are its susceptibility to multi-collinearity and the potential problem of insufficient degrees
of freedom due to the presence of interaction terms.
In the trans-log specification, if the jk (the second-order terms) are all equal to zero, then
the model reduces to the standard Cobb-Douglas frontier production function specified in
logarithmic form as:
01
4.18i j ji T i ij
In y In x t v u
where all variables and parameters are defined as in the case of trans-log form.
Richards and Jeffrey (1998) noted that the Cobb-Douglas form specified above is used
mainly because of its simplicity and parsimony. More so, Coelli (1995) observed that by
transforming the model into logarithms, one can obtain a model that is linear in inputs and
thus easy and straight forward to estimate.
Zhu et al. (1995) stipulated that the Cobb-Douglas and trans-log production functions are
among the functional forms nested in the generalized quadratic Box-Cox (GQBC) functional
form specified as:
2 2 2 2( ) ( ) ( ) ( ) ( )20
1 1 1 1
14.19
2i
i T TT j ji k ki jk ji ki i ij k j k
y t t x x x x v u
107
KkandJjNi ...,2,1;...,2,1;...,2,1
where kandj stand for the inputs used in producing output, i represents firms. The
21,' ands represent the parameters to be estimated, iv is the random error;
)()( 21 ii xandy are the Box-Cox transformations defined by Giannakas et al. (1998) as:
1
1( )
1
1ii
yy
and
2
2( )
2
14.20i
i
xx
Earlier studies (Zhu et al. 1995, Giannakas et al. 1998) have utilized this form in frontier
analysis to examine the relative performance of different functional forms. Giannakas et al.
observed that the trans-log and Cobb-Douglas forms result from GQBC by applying
appropriate restrictions on the values of .21 and
Case 1: 00 21 and , the GQBC becomes a trans-log production frontier.
Case 2: 021 , the GQBC takes the Cobb-Douglas production form, with the second
order parameters )( jk assumed to be equal to zero for all kj, . However, the GQBC may
sometimes fail to select one of the forms discussed above.
In using the GQBC estimations when measuring technical efficiency of firms, problems of
biases caused by heteroscedasticity and/or autocorrelation as well as data scaling may arise,
108
and these problems can seriously bias the transformation variable and invalidate statistical
tests (Giannakas et al, 1998). As a result of these, the GQBC is not considered in this study.
The analysis of this study focuses on the Cobb-Douglas and trans-log production functions.
This Chapter applies the newly developed techniques for the estimation of cross-sectional
data stochastic frontier models (Belotti, Daidone, Ilardi, 2012). The models are based on the
official frontier capabilities by including additional variables.
4.9 Results and discussion
In this section, the results obtained from data analysis are presented and discussed. The aim
is to discuss the determinants of technical efficiency in Cameroon manufacturing firms. The
mean level of TE is estimated by ownership, sectors as well as for the overall sample. The
sectors will be ranked according to their levels of technical efficiencies/inefficiencies.
Firstly, the analysis begins by estimating the average production function using the OLS.
Second, the hypothesis test about the stochastic frontier model is presented; third, the MLE
and efficiency estimates are presented; and finally, the determinants of efficiency and the
mean scores of technical inefficiencies calculated.
4.9.1 Production Frontier and Technical Efficiency Estimates
The section estimates of OLS based on the Cobb-Douglas production function. Basing on
the significance of the parameter estimates, information is gained from which variables
109
should be included in the stochastic frontier analysis. Four inputs are included in the
production function - capital, labor, human capital and raw material. The choice of the
variables is made because the inputs are conventional inputs used in manufacturing firm in
Cameroon.
The other regression involves the expression of the empirical version of stochastic frontier
model with the decomposed errors after getting necessary information about the inclusion
of variables for the frontier analysis. Table 4.5 shows the results of the OLS estimates of the
Cobb-Douglas stochastic frontier model.
4.9.1.1. Ordinary Least Square Estimation
The OLS estimates of the parameters of the Cobb-Douglas production show the average
performance of the sample firms as presented in Table 3.4. The OLS estimates of the
parameters are used as initial values (to estimate) for the Maximum Likelihood estimates of
the parameters. The OLS as well shows how different variables relate to output. The dummy
variables for location capture the impact of geographical location and localization of the
firms. The industry dummies show the importance of the different sectors’ contribution to
output5.
5 Due to the dummy variable trap, the electronic industry dummy is dropped.
110
From the results, capital, labor and raw material inputs have a positive impact on output of
the firm and they are all significant. However, their contribution to output differ, with the
coefficient of labor being highest followed by that of intermediate input, and then that of
capital. The results indicate that these input variables significantly affect the amount of
output in manufacturing firms. The results show that labor comes as the most important
factor of production. Firms still rely heavily on labor in their production process (Labor-
intensive), hence, the capital used may not be sophisticated6. Human capital is seen to be an
insignificant explanatory factor of the firm’s output. This might be explained by the lack of
on-the-job training and low level of education for the employees.
6 Book value of capital is measured by depreciation.
111
Table 4.5: OLS results of the Cobb-Douglas production function with Location and Industry Dummies
Variable Basic Model (No Dummies)
With Location Dummies
With Loc and Industry Dummies
log of capital 0.030** 0.030** 0.033*
(1.99) (1.97) (1.65)
log of labor 0.651*** 0.651*** 0.646***
(12.69) (12.62) (12.14)
log of Human capital 0.057 0.056 0.040
(0.34) (0.34) (0.24)
log of Intermediate inputs 0.348*** 0.344*** 0.348***
(8.35) (8.19) (8.06)
Expzone 0.142*** 0.192***
(2.69) (2.92)
Induszone 0.142* 0.176*
(1.76) (1.93)
Dfood 0.004**
(2.02)
Dwood - 0.116
(-0.44)
Dtextile 0.201*
(1.70)
Dchemicals -0.134
(-0.45)
Dmetals -0.051**
(-2.18)
Constant 1.425** 1.383* 1.349*
(2.02) (1.93) (1.68)
Obs 319 319 319
Prob>F 0.00*** 0.000** 0.000***
R-Squared 0.742 0.743 0.745
Adjusted R-sq 0.739 0.738 0.735 Skewness of OLS residuals -1.272 -1.275 -1.282
,,, Significance at 1%, 5%, 10% respectively and values in parenthesis are the t-statistics.
According to Nikaido (2004), in the presence of technical inefficiency, the OLS model will
be a wrong specification that could possibly result in biased estimates. Therefore, a further
exploration of the assumption of technical efficiency leads to the appropriate model
112
estimation. Hence, the skewness of OLS residual is used to check the presence of
inefficiency in the model. Coelli (1995), Reinhard et al. (2002) showed that given the
underlying assumption of 0,iu a negatively skewed residual, ,i i iv u implies
inefficiency in the firms. By identifying negative skewness in the residuals with the presence
of an inefficiency term, Coelli (1995) derived a one-sided test for the presence of the
inefficiency term7. A positive skewness of the residual is therefore considered problematic
because it cannot be reconciled with a one-sided distribution of inefficiencies Nikaido (2004)
suggested that when a firm shows positive skewness of the OLS residuals, it is assumed that
there are little if any inefficiencies.
From Table 4.5, the skewness of the OLS residuals is negative for all the regressions. This
actually means that firms in the sample are characterized by inefficiencies. It indicates that
the observed output differs from frontier output due to factors which are within the control
of the industries (inefficiencies). Reinhard et al. (2002) showed that computing the skewness
of the OLS residuals acts as a way of testing the appropriateness of the frontier specification.
Based on Reinhard et al. (2002) and since the skewness of the residuals are all negative, it
implies that the OLS estimation of the production function is not the right estimate in this
case. Therefore, the study will adopt the stochastic frontier function as the appropriate model
for the analysis. Before estimating the stochastic frontier model, there is a need to establish
the functional form (Cobb-Douglas or trans-logarithmic), as well as the presence of technical
inefficiencies in the model. This is implemented using a generalized likelihood ratio test.
7 The results of the test by coelli (1995) are given at the bottom of the frontier output when using STATA.
113
4.9.1.2 Hypothesis Testing and Model Robustness
In estimating the production technology for the overall sample and five sectors of
Cameroon’s manufacturing firms, the Cobb-Douglas and trans-log production functions are
specified for the empirical analysis. However, many studies estimate the Cobb-Douglas
function for two reasons;
Firstly, the selection of the Cobb-Douglas frontier model solves the problem of degrees of
freedom normally encountered in the trans-log production model (Amos, 2007). The
assumption is that the Cobb-Douglas specification is nested in the trans-log model; hence
the Cobb-Douglas frontier is an adequate representation of the data8.
Second, the preference has been based on the generalized-likelihood ratio-test which is
defined by the test statistics, chi-square ( )2 .Various tests of hypotheses of the parameter
in the frontier production function and the inefficiency models are performed using the
generalized likelihood ratio test statistic, defined by the negative of twice the logarithm of
the likelihood ratio as approximately the 2 distribution with degree of freedom equal to the
difference of the estimated parameters between the two nested hypotheses.
20 12[log ( ) log ( ) ] 4.21L H L H
8 A trans-logarithmic stochastic frontier model offers a more flexible form due to the inclusion of second order inputs quantities and cross terms but this leads to the model having fewer degrees of freedom. On the other hand, the Cobb-Douglas is easier to implement.
114
where )( 0HL and )( 1HL denote the values of the likelihood function under the null )( 0H
and the alternative )( 1H hypotheses, respectively. If the null hypothesis is true (accepted),
then the likelihood ratio test statistic has an approximately a chi-square or a mixed chi-square
distribution with degrees of freedom equal to the difference between the number of
parameters in the unrestricted and restricted models. Two tests are performed;
Firstly on the functional form, the form of production function encompasses the Cobb-
Douglas form (since the Cobb-Douglas is nested in the trans-log form). So the test of
preference for one form over the other can be undertaken by analyzing the significance of
the cross terms in the trans-log form. If the cross products have t values less than one or
close to zero, then the Cobb-Douglas will best fit the data and will be more appropriate than
the trans-log model specification.
Secondly, as concerns the inefficiency effects model, the null hypothesis is tested as:
0 0 1 10: ... 0H , which specifies that the technical inefficiency effects are not
present in the model, that is, manufacturing firms in Cameroon are efficient and have no
room for efficiency growth. In addition, a stochastic trans-log production frontier is
estimated as a test of robustness in the choice of functional form.
115
Table 4.6: Test of hypothesis for Technical Efficiency
Food
Processing Wood Textile and Garments
Metal and Machinery Electronics
Overall sample
Critical value
0 0 1...4iH for all i Cobb Douglas function
24.47* 17.58 17.02 16.79 16.54 36.32* 17.67
22.63* 19.65*
16.91*
15.57* 12.85* 32.9* 10.37
Notes: 1. * denotes cases where the null hypothesis is rejected. This happens when the calculated value exceeds the critical value.
2. Critical values are obtained from Kodde and Palm (1982) 3. The critical values are at 5% level of significance
The results from Table 4.6 show that the Cobb-Douglas production function is accepted for
four sectors (Wood and furniture, Textile and Garments, Metals and Machinery, and
Electronics), except for the food processing and the overall sample given the assumption of
the trans-log production function. Therefore, the Cobb-Douglas function will be specified
for the four sectors whereas the trans-log specification is adopted for the food processing
and the overall sample. The null hypothesis of no technical efficiency effects is rejected for
all the sectors including the overall sample. Therefore, there are inefficiencies effects in all
the firms in the sample.
0 0 1 10: ... 0 ( )H No inefficiency effects
116
4.9.2 The Stochastic Frontier Analysis of Technical Efficiency
As earlier discussed, due to its ability to decompose the composite error term into a technical
inefficiency term and a stochastic error term, the stochastic frontier analysis has been widely
used in estimating technical efficiency.
Table 4.7 reports the estimates of the Cobb-Douglas production functions for four sectors
(Wood and furniture, Textile and Garments, Metals and Machinery, and Electronics) and the
trans-log estimates for the food processing sector and the overall sample.
117
4.9.2.1 Maximum Likelihood estimation of stochastic frontier model
Table 4.7: Cobb-Douglas and Trans-log Stochastic Frontier Estimation of TE
Variable Food Processing
Wood & furniture
Textile & Garments
Metals and Machinery Electronics
Overall Sample
Constant 0.603 0.379 1.184 12.743*** 0.046 1.431
(0.35) (0.17) (0.44) (4.26) (0.03) (1.38)
Loglabor(L) 0.748*** 0.645*** 0.660*** 0.056 0.909*** 0.651***
(7.16) (5.31) (4.31) (0.37) (10.11) (12.79)
Logcapital(K) 0.013 0.013 0.165* -0.229** 0.089* 0.03
(0.29) (0.19) (1.61) (-2.28) (1.54) (1.09)
Loghumcap(H) 0.323 0.249* -0.505 0.677 -0.871** -0.056
(1.13) (1.47) (-0.71) (1.14) (-2.12) (-0.34)
Logintermediate(R) 0.292*** 0.405*** 0.271* 0.49*** 0.168** 0.348***
(3.48) (3.91) (1.83) (3.69) (2.13) (8.42)
(1/2)log(k*K) 0.251 0.016
(0.39) (0.14)
(1/2)log(L*L) 0.115*** 0.341***
(6.01) (8.32)
(1/2)log(H*H) 0.333 0.027**
(0.13) (2.02)
(1/2)log(R*R) 0.219** 0.115**
(2.24) (2.35)
log(K*L) 0.271*** 0.239**
(7.32) (1.98)
log(K*H) -0.017 -0.072
(-0.52) (-0.34)
log(K*R) -0.013* 0.299**
(-1.40) (2.47)
log(L*H) 0.422** -0.362
(2.79) (-0.97)
log(L*R) 0.129** 0.422**
(2.33) (2.79)
log(H*L) 0.196* -0.034
(1.92) (-0.39)
Sigma-squared 1.11 1.37 3.28 1.81 0.614 1.59
Lambda 0.017 0.007 0.008 0.013 0.015 0.006
No. Obs. 71 55 41 39 37 319
Wald Chi 344.43 129.91 71.43 35.46 267.16 919.53
Prob>Chi2 0.000*** 0.000*** 0.000*** 0.000*** 0.000*** 0.000***
Mean TE 0.724 0.653 0.555 0.498 0.631 0.619 Log-likelihood -104.32 -86.714 -82.527 -66.897 -43.484 -526.94
118
Notes: ***, **,* show significance level at 1%, 5% and 10% respectively. Values in parenthesis are the z-
values. Sigma squared .
The variables of the production function display the expected positive signs. The coefficients
are generally significant at the conventional statistical level although the coefficient of
expenditure on raw material is not significant for the two sectors. The results show that the
elasticity of output with respect to labor dominates over capital. Similar results are obtained
by Chapelle and Plane (2005) among Ivorian manufacturing sectors. This indicates that for
specific policy formulation in addressing low productivity, there is a possibility of increasing
the number of hours worked in the firms in Cameroon.
More so, an increase in total annual depreciation (K) and average educational attainment (H)
will significantly and positively increase the firms’ output. This shows that technical
efficiency and output should increase with increase in the average educational attainment of
the workers since education and capital replacement were expected to be positively
correlated with technical efficiency. In Table 4.7 above, the negativity of the generalized log
likelihood ratio shows the presence of the inefficiency term across all the sectors.
4.9.2.2 MLE of stochastic frontier model accounting for heteroskedasticity
Problems with efficiency estimation can arise when the variance of the dependent variable
varies across the data, known as heteroscedasticity. Heteroscedasticity affects standard
errors, and thus determinations of significance of a given variable. Standard tests for
heteroscedasticity following a linear regression are not available for frontier maximum
2 2 2( )s v
119
likelihood estimation. However, the Cobb-Douglas function shows that the firms are using
labor, capital, human capital and intermediate inputs in the production process with constant
returns-to-scale technology, but the sizes of the firms differ. This size variation introduces
heteroskedasticity into the idiosyncratic error term (Coelli, 1995). Stata allows for explicit
modeling of variables thought to influence the variance of both ui and vi, but an assumption
of a half-normal inefficiency error term is required.
Therefore, the parameters of the Cobb-Douglas are estimated taking into account the
heteroskedastic effects. To do this, a conditional heteroskedastic half-normal model is used,
with firm size as an explanatory variable in the variance function for the idiosyncratic error.
120
Table 4.8: Maximum Likelihood Estimation of Cobb-Douglas and Stochastic frontier models accounting for Heteroscedasticity (Half-normal Maximum Likelihood Estimation)
Variable Food Processing
Wood & Furniture
Textile & Garments
Metals & Machinery Electronics
Overall Sample
Constant 0.326 -2.099 -0.804 10.799*** 0.097 1.459*
(0.25) (-1.27) (-0.39) (3.48) (0.06) (1.83)
Loglabor (L) 0.741*** 0.846*** 0.676*** -0.041 0.902*** 0.643***
(7.11) (9.16) (4.53) (-0.27) (9.73) (12.14)
Logcapital (K) 0.001 0.033 0.226** -0.148 0.089* 0.031
(0.02) (0.54) (2.14) (-1.27) (1.55) (1.13)
Loghumcap (H) 0.321 0.623 -0.289 0.212 -0.847 -0.071
(1.20) (1.28) (-0.47) (0.30) (-0.20) (-0.42)
Loginterinputs (R) 0.323*** 0.306*** 0.304** 0.644*** 0.171** 0.354***
(3.78) (3.64) (2.48) (3.88) (2.18) (8.29)
(1/2)log(k*K) 0.326 0.704
(0.41) 0.32
(1/2)log(L*L) 0.02*** 0.231***
(6.43) 8.059
(1/2)log(H*H) 0.085* 0.048
(1.55) 0.88
(1/2)log(R*R) 0.100* 0.059*
(1.68) 1.76
log(K*L) 0.107* 0.081
(1.72) 1.27
log(K*H) -0.558 -0.564
(-0.29) -0.23
log(K*R) 0.779* 0.600*
(1.45) 1.69
log(L*H) -0.645 -0.052*
(-1.35) -1.59
log(L*R) 0.116** 0.066*
(1.96) 1.57
log(H*L) -0.535 -0.458
(-0.41) -1.06
Firmsize -0.463 1.037*** -1.288 -0.575 0.125 -0.073
(-1.37) (3.90) (-1.35) (-1.26) (0.28) (-0.57)
Constant 1.072* -2.009*** 3.815*** 1.617* -0.756 0.614**
(1.79) (-3.45) (2.73) (1.85) (-0.77) (2.25)
No. Obs. 71 55 41 39 37 319
Wald Chi 363.54 311.29 103.24 42.17 243.71 923.86
Prob>Chi 0.000*** 0.000*** 0.000*** 0.000*** 0.000*** 0.000***
Log-likelihood -102.7 -79.456 -80.993 -66.179 -43.445 -526.773
Note: and,, show levels of significance at 10%, 5% and 1% respectively. The values in
parenthesis are the z-values.
2uIn
121
The output shown in Table 4.8 indicates that the variance of the idiosyncratic error term (
2v ) is not really a function of firm size in four of the five sectors considered.
Heteroscedasticity only occurs in the wood and furniture industry. However, when the
overall sample is considered, no strong pattern of heteroscedasticity is apparent. Therefore
the results suggest that heteroscedasticity is not a significant problem. The Wald chi tests
and its corresponding probability for all the sectors indicate that the study fails to reject the
hypothesis that the firms use constant returns to scale technology.
4.10 Determinants of Inefficiency
The focus of this section is to provide an empirical analysis of factors that contribute to
technical inefficiency and productivity variability among manufacturing firms in Cameroon.
Hence knowing that firms are technically inefficient (as shown in the Cobb-Douglas and
Stochastic frontier models) might not be useful unless the sources of the inefficiency are
identified. Thus, in the second stage of this analysis, the study investigates the firm-specific
attributes that have impact on technical efficiency. The inefficiency function is written as:
0 1 2 3 4 5
6 7 8 9 10
Re
exp 4.22
i
i
u Firmsize Age Foreign Union glabor
Corruption Taxrates Acessfin Mngedu Mng
The estimated coefficients in the inefficiency model are presented in Table 4.9. The analysis
of the inefficiency model shows that the signs of the estimated coefficients in the model have
122
important implications on the technical efficiency of manufacturing firms. Variables are
included as inefficiency variables; thus a negative coefficient means an increase in efficiency
and a positive effect on firms’ output.
From Table 4.9, firm size is negatively correlated with firm technical inefficiency effects
which imply a positive effect on productivity. The result conforms to a number of theoretical
arguments. The literature of early development economics placed a strong emphasis on large
firms, which were considered as the driving force of economic growth. Hence, small firms
were being perceived as archaic modes of production. According to Chapelle and Plane
(2005), large firms with their managerial know-how would offer a better organizational
framework to reduce transaction cost. Hill and Kalirajan (1993) concluded with respect to
Indonesian garment industry that large firms benefit from more efficient management. Thus
the larger the size of a firm, the more labor is available for firms operations thus increasing
the efficiency of firms.
Firm age is also a major determinant of technical inefficiency of manufacturing firms in
Cameroon as it reduces the efficiency of these firms. This is plausible given that majority of
firms were established in the late 1970s (see Table 4.3 for mean age of firms). The firms are
old and may not be willing to try new innovation and technology due to financial constraints.
A significant relationship was found between the existence of trade union and the technical
inefficiency levels of individual firms in the industries (except in the metal and machinery
123
and electronics sub sectors). However, the variable has positive coefficients for the
significant sub sectors. This shows that the variable explaining the existence of trade unions
contributes significantly to technical inefficiency.
Another important variable which has an effect in determining technical efficiency level is
the foreign variable. It is significant in the food processing and beverages, wood processing,
textile and garments as well as in the overall sample. Hence, it increases technical efficiency
in these the sub-sectors. Finally, the results also show that corruption plays a significant role
in increasing technical inefficiency especially in all the subsectors as indicated by the
positive coefficient of the variable.
124
Table 4.9: Inefficiency effect model
Variable Food Processing
Wood & Furniture
Textile & Garments
Metals & Machinery Electronics
Overall Sample
Firmsize -0.454*** 0.799*** -0.706* 0.280* 0.879** 0.308**
(-3.84) (2.75) (-1.66) (1.45) (2.23) (2.47)
Firmage 0.076*** 0.023* 0.011 0.047*** 0.059*** 0.053***
(8.16) (1.43) (0.41) (3.29) (2.73) (6.76)
Foreign -0.491*** -1.706*** -1.506* -0.455 0.751 -1.276***
(-8.76) (-3.31) (-1.68) (-0.99) (0.77) (-4.95)
Union 1.216* 0.487* 2.394*** -0.166 -0.391 1.068***
(1.75) (0.75) (2.95) (-0.46) (-0.51) (4.67)
Reglabor 0.442*** 0.059 0.369* 0.387** 0.202 0.028
(8.62) (0.38) (1.45) (2.18) (0.73) (0.34)
Corruption 0.245*** 0.333** 0.108 0.151 0.387* 0.007
(8.62) (2.16) (0.28) (1.26) (1.69) (0.10)
Taxrates -0.335*** -0.007 0.007 -0.325** 0.165 0.022
(-5.40) (-0.03) (0.03) (-2.61) (0.62) (0.26)
Acessfin -0.428*** -0.136 -0.042 -0.107 0.600* -0.016
(-8.83) (-0.75) (-0.11) (-0.63) (1.59) (-0.15)
Mngedu -0.235*** 0.104 0.277* 0.167 -0.094 0.156***
(-5.94) (1.06) (1.78) (-1.22) (-0.62) (3.03)
Mngexp 0.028*** -0.0003 -0.003 -0.019 -0.020 -0.019*
(11.05) (-0.02) (-0.07) (-1.04) (-0.68) (-1.73)
Constant 22.061*** 18.686*** 19.987*** 17.642*** 14.872*** 17.745***
(3.25) (6.27) (9.62) (8.71) (4.20) (6.35)
Note ***, **, * shows levels of significance at 1%, 5% and 10% respectively. The values in parenthesis show
the z-statistics.
125
4.11 Mean Technical Efficiency and Inefficiency Scores
Table 4.10 shows the mean technical inefficiency in all the sub sectors and for the overall
sample. Technical efficiency is defined as:*
exp( );ii i
i
yTE u
y where *
iy is the production
frontier – maximum output given the inputs for each firm. Hence, exp( ).i iTE u Therefore,
in all specifications, total average technical efficiency would be: 1
1 ˆ ,I
iiTE TE
I for each firm,
1,2...i I (Coelli et al. 2005). From the technical efficiency equation, average inefficiency is
calculated as; 1 .TE
Table 4.10: Mean Technical inefficiency by Size and Sector
Size/Sector Food Processing
Wood & Furniture
Textile & Garment
Metals & Machinery Electronics
Overall Sample
Small 0.187 0.210 0.204 0.206 0.194 0.184
SD (0.142) (0.169) (0.155) (0.157) (0.136) (0.152)
Medium 0.103 0.181 0.177 0.145 0.113 0.159
SD (0.044) (0.132) (0.129) (0.080) (0.045) (0.157)
Large 0.227 0.236 0.241 0.224 0.240 0.236
SD (0.183) (0.188) (0.186) (0.185) (0.175) (0.183)
Notes: Values in parenthesis are the standard deviations for the mean technical efficiencies. 1) Small shows firms with less than 30 employees 2) Medium represent firms with 30 to 100 employees 3) Large represent firms with over 100 employees
As shown in Table 4.10 above, total average technical inefficiency ranges from 10.3% to
24.1% across the five sectors. For the food processing sector, the average inefficiency varies
126
widely, from 10.3% in medium sized firms to 22.7% in large firms. Inefficiency in the wood
and furniture sector varies across firm sizes from 18.1% to 23.6%. Taking the overall sample,
inefficiency of for-profits varies the firms varies across sizes from 15.9%% to 23.6%. Thus,
the wood and furniture sector is the least efficient almost the five sectors, followed by the
textile and garments sector. The results also show that the food processing sector is the most
efficient sector in the sample. This result could be due to the fact that the food processing
sector have experienced higher technical change than the other sectors in the manufacturing
sector, which could have pushed the production frontier further for some firms in the sector.
It maybe also be as a result of economies of scale due to the high demand for food products.
As concerns firm size, the medium sized firms are found to be most efficient while large
firms are found to be the most inefficient. Although some studies have found a positive
relationship between technical efficiency and firm size (Lundvall and Battese, 2000;
Niringiye et al., 2010), the findings in this present study are in conformity with Biggs et al.
(1995) who found an inverted U-shaped relationship between firm size and efficiency. Biggs
et al (1995) found the size-efficiency relationship to be negative for large firms and positive
for small firms, with the medium-sized firms being the most efficient.
127
Table 4.11: Mean Technical inefficiency by Ownership and Age for overall sample
Variable Inefficiency
Ownership
Domestic firms 31.02%
Foreign firms 28.77%
Firm Age
0 – 5 years 30.15%
6 – 10 years 23.07%
11 – 20 years 27.32%
20 and above 35.97%
Notes: 1) Ownership is measured by number of shares owned in the firm. 2) Firm age has been calculated as 2009 minus the year the firm started operations in Cameroon.
Table 4.11 above shows that foreign owned firms are more efficient than domestic owned
firms. This might be explained by the issue of transfer technology especially as most of the
foreign owned firms in Cameroon export to other countries. Learning by exporting, in which
experience brings about improvements in performance, may be the explanation for this
finding. Concerning firm age, firms between 6 to 10 years are the most efficient while the
much older firms are the least efficient. This might be explained by the fact that at the start
of the operations (0 to 5years), firms might still be adjusting to cover sunk cost and enter the
market where already established firms are operating. More so, in the context of Cameroon,
the high inefficiency of the older firms might be explained by the type of technology used
in the production process. Some of the technology is highly considered to by archaic and out
dated. Therefore, older firms operate 35.97% below their potential frontier production level
with the given inputs and production technology.
128
4.12 Conclusion
The primary objective of this study is to analyze the determinants of efficiency in
manufacturing firms in Cameroon. This is achieved by determining the efficiency of
manufacturing firms in five industries and identifying the determinants of inefficiency. The
study used a stochastic frontier model employing RPED data of 319 firms from different
manufacturing industries. The data are micro-level which is the most adequate type of data
used in the estimation of these models.
The model used is that outlined by Battese and Coelli (1995) which determines the causes
of inefficiency in the manufacturing sector in Cameroon. The estimates of the stochastic
production frontier with inefficiency effects model indicate that Cameroonian firms exhibit
various degrees of technical inefficiency for the sample of firms considered. The results
show that firm size plays an important role in explaining technical efficiency in the sub-
sector of food processing. The results show that large firms reduce technical inefficiency
levels of firms in all the sub sectors. A significant relationship was found between trade
unions existence and the technical inefficiency levels of individual firms in the industries
(except in the wood and furniture and metal and furniture sub sectors). The age of firms also
play an important role in determining inefficiency levels in the industry. This could be
explained by the fact that most of the older firms were established in the post-colonial
periods and still heavily rely on the archaic technology. The firms may be bar from taking
on new technology or trying new innovation by financial constraints.
129
Another important variable which has an effect in determining technical efficiency level is
the foreign variable. It is significant in food processing, wood processing, textile and
garments as well as in the overall sample. Hence, it increases technical efficiency in all the
sub-sectors. The results also show that corruption plays a significant role in increasing
technical inefficiency especially in the food processing sector. Finally, since an increase in
age of firms leads to a reduction in efficiency levels in manufacturing firms, policies should
be adopted to replace the existing capital in the large firms.
130
CHAPTER 5
FROM TECHNICAL EFFICIENCY TO EXPORT PERFORMANCE:
EVIDENCE FROM CAMEROONIAN MANUFACTURING FIRMS
5.1 Introduction
In the era of globalization, many firms have adopted an export-oriented strategy to seek
organic growth through active participation in the international market. Internationalization
allows firms to become more familiar with the activities of their international competitors
and also affords them greater access to new market opportunities (Mok, Yeung, Han and Li,
2010). The pursuit of an export-orientated strategy is considered fundamental for firms’
competitiveness in the long term, as exporters tend to be more productive through learning-
by-exporting than non-exporters (Wagner, 2007).
By following an export-oriented strategy, firms are also able to exploit the possible
advantages of economies of scale and gain cost advantages (Wagner, 2007). Furthermore,
keen competition in the international market obliges firms to meet international standards
and high customer expectations in terms of product quality and choices, driving them to
upgrade their technological capabilities and thus their competitiveness. Among others, two
recent papers have given empirical reviews to demonstrate that export-oriented enterprises
perform better than non-exporters (Farinas and Martin-Marcos, 2007; Wagner, 2007). These
studies produced empirical results supporting the view that higher level of
131
internationalization for U.S. manufacturing multinational enterprises is associated with
improved performance measured by technical efficiency.
The question then arises “Do firms learn from their exporting experience?” In recent years,
research on export as an economic activity and as a driver of productivity and growth has
focused around this question. The underlying idea of these works is that exporting to the
international market should improve firms’ efficiency through two main channels. On the
one hand, the larger international markets allow the exploitation of economies of scale and,
on the other hand, international contacts foster a learning process through technology and
knowledge spillover.
For the past decades, empirical evidence in this field, though not conclusive has been
quantifying the contribution of export to economic growth, designing appropriate trade and
industrial policies and identifying macroeconomic factors that affect trade performance.
Most of these studies used data at the country or industry level to test the impact of exports
on firm level productivity and growth (Giles and Williams, 2000). Focusing on the role of
firms in shaping international competition, a critical observation made is that all firms face
the same macroeconomic conditions but respond and perform differently in their export
activities (Pusnik, 2010). This suggests that there must be firm-specific characteristics that
significantly influence a firm’s capability to export. The relationship between efficiency and
export performance has not been well exploited in economic literature especially in the
Cameroon economy.
132
Due to a shortage of micro data, there is not much empirical evidence on the current topic
for Africa and more specifically in Cameroon. Recent years, however, have seen an
expansion in the availability of such data, primarily through the Regional Programme of
Enterprise Development (RPED) surveys organized by the World Bank in the early and mid
1990s. To date, a handful of studies have used these data to examine various aspects of
exporting behavior9. Using data from manufacturing firms in Cameroon surveyed within the
RPED this study attempts to shed light on the issue whether exporting in Cameroon is more
accurately described by firm-level mechanisms of technical efficiency or standard trade
theory predicting close links between industry and exporting.
Therefore, the main objective of this chapter is to analyze the export behavior of a sample
of Cameroonian manufacturing firms estimating which factors affect export performance.
Specifically, the chapter seeks;
To analyze technical efficiency as firm-specific determinant of export performance.
To identify the determinants of the propensity to export among firms in Cameroon.
In order to achieve these objectives, the following null hypotheses are tested:
Technical efficient firms are more likely to export than less efficient firms.
9 Bigsten et al. 1999, 2000 have used RPED data from the Cameroon, Ghana, Kenya and Zimbabwe to undertake a comparative study of manufacturing exports. Country specific studies have been undertaken by Granér and Isaksson 1998, 1999 (Kenya), Hoogeveen and Mumvuma 1999 (Zimbabwe) and Söderbom and Teal 2000 (Ghana)
133
Firm size, firm age and Foreign-owned have positive influences on export
performance. That is, larger firms are more likely to export than small firms.
The propensity to export is positively influenced by the business environment
factors, factor intensity factors as well as firm specific factors.
5.2 Theoretical and Empirical Background
5.2.1 Theoretical literature
The theoretical foundation on export performance originates from the neo-classical models
on comparative advantage, with labor productivity as a determinant of export. Heckscher
(1991) and Ohlin (1991) developed a trade theory that takes into account the difference in
the location of labor, capital and natural resources as determinants of trade. According to the
Hecksher - Ohlin model, countries export goods whose production is intensive in factors
with which they are abundantly endowed. These are known as the factor intensity theories
which argue that factor–based advantages may be important if the firm has either a natural
monopoly of a particular factor or is located in a particular region where the factor is
plentiful. This model disregards the difference in labor productivity unlike in Ricardo’s
model. This means that even if labor productivity were identical among two countries, there
would be a possibility for competitive advantage due to the differences in the production
factor endowment. The difference in factor supply is the reason for the difference in relative
prices between countries. Hence, capital abundant countries would, therefore, export capital
intensive goods, while labor abundant countries would export labor intensive goods.
134
Therefore, export-led growth literature shows exports induce an increase in country’s output
and productivity. On the other hand, some scholars claim that the direction of causality runs
from economic growth to export. Many theoretical arguments in favor of the export-led
hypothesis have been put forward over the years.
First, the traditional approaches of export-led growth hypothesis (Kaldor, 1970) posited that
external demand would enable firms to exploit economies of scale leading to productivity
growth. Hence, firms move to a lower point on the average cost curve since a rise in output
is accompanied by a less than proportionate rise in average costs. They predicted that firms
could invest in productivity-enhancing technology in anticipation of larger export markets.
Hence, exporting activity is an important component of autonomous demand and determines
a multiplier effect on investment and output both in the exporting (direct effect) and in
related (linkage effect) sectors in the home country (Castellani, 2002). Second, the growth
of the exporting sector promotes a reallocation of resources from nontrade sectors to the
export sector itself which, being relatively more productive; raise the overall productivity of
the country (Wagner, 2007). Third, export is a means to generate foreign currency inflows,
required to finance imports (Wagner, 2007). Finally, outward orientation may result in
efficiency gains for firms, due to the exploitation of economies of scale and to learning
associated with knowledge spillovers from international contacts (Clerides et al. 1998).
Advocates of the alternative view argue that the relevant direction of causality is from
productivity growth to exports (Caves, 1971). In particular, it is claimed that economic
135
growth produces an enhancement of skills and technologies, which creates the basis for the
international competitive advantage and in turn determines exports (Krugman, 1984).
Another theoretical foundation by Krugman (1989) suggested that hysteresis in exports may
be due to sunk costs in entering the export market at firm level. It is argued that exporting
firms incur sunk costs, due to the adaptation of products to foreign standards, which
determine that only the larger and more productive firms will start exporting (Roberts and
Tybout, 1997; Benard et al. 2000). The underlying theory of sunk cost stipulates that there
are fixed costs of exporting that deter those firms operating below the threshold level of
efficiency because their prospective profits from exporting do not compensate for individual
costs. Sunk costs may include expenses related to establishing a distribution channel and
modification of commodities for foreign tastes. According to Graner and Isaksson (2007),
these costs may vary with firm size, firm age (capturing the extent of a firm’s learning
experience), and ownership structure10.
However, in recent years, two hypotheses have dominated the theoretical framework of the
export performance of individual firms (Benard and Wagner, 1997; Benard and Jensen,
1999; Giles and Williams, 2000; Wagner; 2007). The first hypothesis points to “self-
selection” of the more productive firms in the export market. The reason for this is that there
exist additional costs of exporting and selling goods in foreign markets (Wagner, 2007). The
second hypothesis dwells on the argument that in view of limited experience in the export
10 Firm size also serves as a proxy for the magnitude of the firm’s resources that are important for the decision to enter into the export markets (Wagner, 1995; Bernard and Jensen, 1999). According to Berry (1992), foreign-owned firms may have better access to finance, making it easier to bear fixed costs associated with entering the export market.
136
business and the significant technological gaps faced by potential exporters from most of the
developing countries, there is a lot of scope to “learn by exporting”.
5.2.1.1 Self-selection hypothesis
The central idea of this hypothesis is to identify firm specific characteristics that make a firm
more likely to export, therefore searching the dividing line between firms that sell only
domestically and those that export to foreign markets. The framework is drawn from the
models developed by Robert and Tybout (1997), Clerides et al. (1998), Tybout (2003) and
Melitz (2003). Starting from a firm’s static problem of export participation with no sunk
cost, assume iY to denote a dummy variable taking the value 1 if firm i exports in the given
year, and 0 otherwise. The foreign market participation of firm i in the given year will be;
0,1( ) 5.1
i i i i iYMax X Y
where i denotes the profits made by exporting, in excess of those made on the export
market. This depends on the market characteristics iX (which also determine the marginal
production costs), and on an error term .i Firm i will decide to export at the given period
)1( iY if 0)( iiX otherwise it will only serve the domestic market. Using a reduced
form approximation for the determinants of firm profits from exporting, this leads to
following export market participation model:
137
1: 0
5.20 :
i i i
i
if XY
Otherwise
i in the model is approximated as a reduced-form expression in exogenous firm and market
characteristics. The vector iX contains firm size, firm age, foreign ownership, physical
capital, human capital, output and other firm-level characteristics which typically determine
the marginal production costs faced by the firm and consequently the expected profits it is
likely to generate by exporting.
Assuming that a firm has to incur sunk cost, therefore profit will be adjusted for costs of
foreign market entry. The reason for this is that there exist additional costs of selling goods
in the foreign market (Wagner, 2007). The range of extra costs include transportation costs,
establishment of distribution network or marketing cost, production costs in modifying
current domestic products for foreign consumption, personnel with skill to manage foreign
networks, as well as gathering information and dealing with the different legal and economic
environment in the foreign country. These costs provide an entry barrier that less successful
firms cannot overcome. Clerides et al. (1998) presented a model in which incumbent
exporters would choose to export whenever gross operating profit plus expected future
payoff from remaining an exporter is higher than the per-period fixed cost of being an
exporter (that is, costs dealing with customs and other intermediaries), and non-exporters
begin exporting whenever this sum is higher than the per-period cost plus the sunk entry cost
for entering foreign markets (in other words, expenses related to establishing a distribution
channel, or production costs for modifying domestic products to foreign tastes). Since gross
138
profit is positively related to productive efficiency the probability that a firm exports should
increase with its efficiency level.
Furthermore, Tybout (2003) and Wagner (2007) pointed out that the behavior of firms might
be forward-looking in the sense that the desire to export tomorrow leads a firm to improve
performance today to be competitive on the foreign market, too. Cross-section differences
between exporters and non-exporters, therefore, may in part be explained by ex-ante
differences between firms: the more efficient firms become exporters.
5.2.1.2 Learning-by-exporting hypothesis
The hypothesis of learning-by-exporting dwells on the argument that there is a lot of scope
for firms to “learn by exporting”. This is because potential exporters from most of the
developing world have limited experience in the export business and significant
technological gaps as well. The theory postulates that firms gain information when exporting
and that such learning enhances their efficiency (Clerides et al, 1998). Exporters, therefore,
acquire information from their foreign customers on how to improve the manufacturing
process, decrease production costs and upgrade product quality.
Using the classical learning-by-doing model of Arrow (1962), two main characteristics of
learning were suggested. First, “learning is the product of experience. Learning takes place
through an attempt to solve a problem and therefore only takes place during activity”.
Second, “learning associated with repetition of essentially the same problem is subject to
139
sharply diminishing returns. The Arrow’s classical model of learning is applied to domestic
firms breaking into the export markets. The firms need to solve new problems such as
adopting stringent technical standards to satisfy more sophisticated consumers, because
export markets are likely to be more competitive than the domestic markets.
Using the production function approach, the link between efficiency and exporting can be
analyzed. In the Cobb-Douglas production function, output is modeled as a function of
capital, labor and productivity.
5.3i i i iY A K L
where iA is the productivity parameter (or level of total factor productivity), iK and iL are
stocks of capital and labor respectively. Using the intensive form model of the production
function can be rewritten as:
5.4i i iy A k
where iy is output per worker in each firm (measure of productivity) and ik is capital
intensity or capital per worker. Using the learning-by-exporting hypothesis, the parameter
of productivity, iA , is assumed to depend on exporting as follows:
, 5.5i i iA f X Z
140
where iX is a dummy variable for export status; equal to “1” if firm i is an exporter and “0”
elsewhere. iZ is a vector of firm-specific characteristics (control variables) including size,
ownership, location, industry, source of finance at start-up and capacity utilization among
others. The imposition of elasticities on the right hand side of the productivity parameter
gives;
5.6i i iA X Z
Substituting for the productivity parameter in the intensive form model gives;
5.7i i i iy X Z k
Taking the natural logs of both sides, gives an estimable linear function:
5.8i i i i i iIn y In X In Z In k
The disturbance term is composed of two terms. The first, ,i is the firm-specific effect
(unobserved firm heterogeneity) that reflects firm efficiency and managerial skills. The
second ,i is a random disturbance term assumed to be distributed identically and
independently across firms. This may represent factors such as weather conditions and
unpredictable variations in inputs. This theoretical framework has been developed to explain
firms’ learning from exporting. Most empirical investigations build on the idea that if export
behavior determines learning effects, the stochastic process governing productivity should
be changed by the events of exporting. As shown in Figure 5.1, comparing two firms: firm
A which exports at time t and firm B which does not export at time t . One would expect
that the productivity trajectory of A will steepen after exporting, due to the learning process,
while firm B will continue on its trajectory.
141
Figure 5.1: Export Behavior as Determinant of Change in Productivity Growth
Source: Adapted from Castellani (2002).
However, a situation may occur where exporting does not affect the stochastic process
governing the dynamics of productivity but where exporters have a higher productivity
growth before entering the export market (Figure 5.2).
Wagner (2007) used the concepts of self-selection and learning-by-exporting to compare
cross sectional average productivity of firms which have undergone different patterns of
transition in and out of the export market. He identified 4 different status for their sample
firms: stay out (firms which do not export neither in period t , nor in period 1t ), entry
(firms which do not export in period t and export in period 1t ), exit (firms which export
in time t and do not export in time 1t ), stay in (firms which export both in t and 1t ).
0 X T
Non-exporters
B
Exporters
A
Lo
g (
pro
du
ctiv
ity
)
Time
142
Figure 5.2: Export Behavior as non-Determinants of Change in Productivity Growth:
Source: Adapted from Castellani (2002)
5.2.2 Empirical Literature
The association between exports and productivity is ambiguous and not conclusive
(Castellani, 2002; Wagner, 2007). Giles and Williams (2000) review more than one hundred
and fifty empirical papers, using either cross-section and time series data, and do not reach
any conclusion about the direction of causality.
Some previous studies find that firm heterogeneity plays a crucial role in the firm’s decision
to enter foreign market through exporting. The findings are that better-performing firms in
an industry are more likely to be exporters (Wagner, 2007). Early research in this area
X T
Non-exporters
B
Exporters
A
Lo
g (
pro
du
ctiv
ity
)
Time 0
143
investigated firm’s competitive advantages that facilitate its involvement in exporting
activities and was limited to highly industrialized countries.
There are arguments suggesting that increased foreign competition may be injurious to
domestic industries if it leads to a closure of factories. Pusnik (2010) found that foreign
technology adoption may be relatively unimportant. This is because the efficiency difference
between foreign and domestic inputs has only a minor impact on productivity in some cases.
The explanation for the minor impact lies in the fact that foreign technology adoption takes
time due to delays in learning, difficulties with factor complementarities and differences in
production arrangements.
The empirical literature on the firm level export and productivity of less developed country
firms has been documented in cross countries studies but more still has to be done on
individual countries. Among a few others, Roberts and Tybout (1997) and Clerides et al.
(1998) carried out studies in Colombia, Mexico, Morocco, South Africa, Mauritius, and
Ghana. These studies mostly focus on a few variables, for example the effect of firm size
and research and development (R&D) expenditures on export performance. Yet export
performance may be influenced by a multitude of other important variables.
Although the issue about determinants of export activity of individual firms has been
researched, there is a research gap on the relationship between a firm’s technical efficiency
and its export performance. Earlier studies showed that exporting firms are more efficient
than non-exporting firms (Bernard and Jensen, 1995). Recent studies brought forth an
144
alternative explanation, i.e, efficient firms self-select into the export activity because returns
on doing so are high to them (Robert and Tybout, 1997; Clerides et al, 1998). To explain the
self-selection hypothesis, Clerides et al. (1998) present a model in which incumbent
exporters would choose to export whenever gross operating profit plus expected future
payoff from remaining an exporter is higher than the per-period fixed cost of being an
exporter. Similarly, non-exporters begin exporting whenever this sum is higher than the per-
period cost plus the sunk entry cost for entering foreign markets. Since gross profit is
positively related to productive efficiency the probability that a firm exports should increase
with its efficiency level.
Melitz (2003) provided a general equilibrium model showing that firms self-select into
export markets, i.e. only more efficient firms can bear fixed entry costs in the export markets.
In a dynamic industry model based on heterogeneous agents, as opposed to the standard
representative-agent model, Melitz (2003) showed that trade may generate productivity
gains at the aggregate level, however, without necessarily improving the productivity of
individual firms. This can happen because costs associated with export entry alter the
distribution of trade gains across firms. The most efficient firms reap trade gains by
increasing their market share and profit, while less efficient firms lose in terms of both, and
the firms worst off are forced to exit.
145
Bigsten et al. (1998) found that state-owned enterprises and large firms in Tanzania were
less likely to export and found strong evidence that foreign-owned enterprises were more
likely to export. The results also show that foreign-owned enterprises were more likely to
export than similar private domestically-owned firms. On the basis of evidence presented by
Bigsten et al. (1997), this minimum size appears to be firms with 100 employees; these
authors show that for Ghana, Kenya and Zimbabwe, 71% of firms with more than 100
employees export, but only 35% of those with between 29 and 100 employees do so. For
firms with less than 30 employees, only a negligible proportion enters the export market.
A study by the World Bank (2004) using cross sectional data from Uganda and Tanzania
found that larger firms are more likely to export than smaller firms and typically export more
of their output. Foreign-owned firms were also found to be more likely to export more than
domestically-owned firms. It was also found that firms that produce construction materials,
metals, furniture and wood tend to export less than other firms. This is a cross country study,
however, and the findings have limited policy application.
In recent studies, Granér and Isaksson (2007) show that exporters of Kenyan manufacturing
firms are more efficient than non-exporters, while Niringiye et al. (2010) find no evidence
of self-selection by the relatively more efficient firms into exporting in East African
manufacturing firms. They conclude that factors other than technical efficiency may be
playing a more prominent role as determinants of the export decision.
146
The stylized fact that emerges from the studies reviewed above is that firm size is a major
variable on the existing relationship between productivity and firm export performance. The
evidence on the association of firm age, foreign ownership and propensity to export is mixed.
Most of these studies related to firm level export have so far attempted to identify and test
only a number of operational variables, not taking into consideration technical efficiency.
Excluding possible relevant variables may lead to biased results.
Some studies (Soderling, 2000) employ macro and sectoral level time series and cross-
country data to examine the potential export determinants. In the case of Cameroon, the use
of time series data would run into problems of degrees of freedom and other statistical issues.
This is due to the fact that most of the important reforms that rapidly boosted the export
performance were undertaken between 1993 and 1994 (Soderling, 1999; Njikam and
Cockburn, 2002)11. Since estimations based on pre-reforms information will be less
informative, there is need to employ cross sectional data.
11 In reaction to the slow or negative economic growth of the 1980s, Cameroon embarked upon a trade
liberalization program. In this regard, between 1993 and 1994 the list of firms reserved for public sector was
reduced, quantitative restrictions were dismantled, licensing requirements were drastically scaled back,
reference prices were progressively removed, and the level of tariff rates on most products were reduced.
Within a regional framework i.e. within the CEMAC zone, significant tariff reductions were introduced with
fall not only in the average rate but also in the number of rates. In sum, micro reforms included the removal of
protective trade barriers, privatization, and market deregulation. At the macro level and in January 1994,
Cameroon and the other CFA zone countries realigned the parity of their currency from 50 to 100 CFA francs
to the French franc (50% devaluation). This was a major step towards macro economic adjustment and
competitiveness of the export sector.
147
The key questions in this stream of literature are “Do more efficient firms become
exporters?” and “Do exporters become more efficient firms?” This chapter adds to the
literature by building on previous studies and including firm specific characteristics,
business environment factors and factor intensity variables in an integrative model. This is
based on existing theories on productivity and export performance. The existing studies
suggest that there are important specific firm-level factors, business environment factors and
factor intensity variables that need to be examined to understand the link between technical
efficiency and export performance.
5.3 Methodology, Variable specification and data
The main contribution of this chapter is the test on whether technical efficiency significantly
influences the export performance of individual firms or not. Importantly, the study extends
the range of variables that impact on export performance. Thus, considering that firms within
an industry vary significantly in efficiency and other characteristics, then it is largely
expected that the export activity of an individual firm is influenced by a combination of firm
specific characteristics, business environment factors as well as factor intensity variables. In
traditional trade theory, firm specific characteristics and business environment factors would
just add an extra element to residual variance (Pusnik, 2010). The inclusion of the business
environment factors justifies the fact that no firm operates in a vacuum, but deals with its
external environment as well. These factors might stimulate as well as inhibit export
performance of individual firms.
148
5.3.1 Variable specification and Determinants of firm export performance
The dependent variable in this study (export performance) is defined as the ratio of exports
to total sales of an individual firm – propensity/probability to export. According to Wakelin
(1998) and Niringiye et al., (2010), the propensity to export specification is preferable to
the factor intensity determinants of a firm’s export because the factor intensity variables in
the model specification of export performance is expected to help in predicting whether a
firm exports as well as how much the firm exports. Hence, when export behavior is measured
as a share of foreign sales on total sales (export intensity) it has a positive and significant
effect on productivity growth (Castellani, 2002).
The choice of the dependent variable, the propensity to export, which varies between 0 and
1 by definition, has the advantage that it captures the level of exports of firms. The binary
probit estimate in the study incorporates the key theoretical explanations of firm-level export
performance. This is typical of the literature on firm specific effects on export performance
(Graner and Isaksson, 2007).
5.3.1.1 Firm specific characteristics
Cameroon’s export success is driven by firms which vary widely in size and other structural
characteristics. Therefore, it is interesting to examine significance of these firms’
characteristics on Cameroonian firms’ export performance.
149
Firm size
An influential theory linking firm size to technical efficiency is Jovanovic’s (1982) version of
the passive learning model of firm dynamics. His model predicts that larger firms are more
efficient than smaller ones. A selection process leads to an outcome in which efficient firms
grow and survive, while inefficient firms stagnate or exit the industry. However, a positive
correlation between efficiency and size might also arise if relatively efficient firms have a
superior cost structure, or if larger firms have more competent management, both of which would
allow them to gain market shares.
Firm age
The age of the firm is also a debated factor in the literature. Firm age may capture the extent
of a firm’s learning experience. Older firms are usually considered to be more efficient than
younger ones, because owners, managers and employees have gained experience from past
operations. On the other hand, young firms are usually leaner and more receptive to changing
perspectives. Firm age, indicating a learning-by-doing experience, can also significantly
affect firm export decisions, since old firms are able to participate in competitive foreign
markets due to their cumulative experience, business networks and reputation. Niringiye et
al. (2010), however, pointed out that young firms are more proactive, flexible, and
aggressive compared to old firms. As a result they are more willing to adopt modern
technology, but old firms are stuck with outdated physical capital.
Older firms may have a superior cost structure and may therefore be able to better handle
sunk costs associated with export entry. If this holds true, firm age would be expected to
150
enter positively in the export-decision regression. Furthermore, over time a firm may have
established enough international contacts to decide on engaging in export activities.
Foreign ownership
A number of empirical studies have found a significant and positive association between
foreign investment ownership) and firm export participation (Graner and Isaksson, 2007).
Foreign-owned multinational corporations operating in developing countries are assumed to
be more efficient than domestic-owned firms because of greater experience in management
and superior organizational structure. However, it is quite possible that foreign firms seeking
to acquire domestic ones target relatively efficient firms, that is, domestic firms are efficient
before the ownership structure changes. If so, it may be the case that efficiency explains
foreign ownership and not the other way around.
Location
Location is also another important factor, since the export decision by firms in different
locations may be affected due to transport costs, infrastructure, spillover effects and natural
resources (Niringiye et al., 2010). This variable takes the form of a dummy variable. It takes
the value one if the firm is located in the economic capital city (Douala) and zero otherwise.
Douala is considered as the reference city because it is the main industrial zone in Cameroon
and also because of the seaport facilities for transportation. Firms located in Douala may
have significantly higher export propensity and may enjoy a better international image than
similar plants in other parts of Cameroon.
151
Technical efficiency
Firm efficiency is one of the factors which could affect the propensity export as well as
export decision. This variable will be used to investigate either the evidence of self-selection
hypothesis, where only more efficient firms can participate in the export market.
Cherides et al. (1996) revealed that relatively efficient firms will be exporters, but previous
export participation does not affect the unit costs of firms. Therefore, the efficiency gap
between non-exporters and exporters is because the more efficient firms self-select into the
export market, rather than learn by exporting.
The study measures technical efficiency by using stochastic frontier analysis (SFA). The
measure of technical efficiency is obtained by econometric estimation of Cobb-Douglas
production function (Cobb and Douglas, 1928). Specification of the model assumes that
technical efficiency follows a half-normal distribution as shown in chapter four (Aigner,
Lovell and Schmidt, 1977; Meesuen and van den Broeck, 1977).
5.3.1.2 Factor Intensity variables
Physical capital
The relation between the use of physical capital and export activities is based on the factor
endowments to trade patterns predicted by the Heckscher-Ohlin model of comparative
advantage. For this model to be valid at firm level, manufacturing exports should be
152
concentrated in firms that use the relatively abundant factor intensively, that is, if labor is
abundant, capital intensity is expected to be negatively related to export activities.
Human capital
Human capital is also one of the important determinants of a firm’s export performance.
Human capital is expected to positively correlate with efficiency. A high educational level
within firms facilitates international contacts and export participation. In addition, using the
same reasoning as above for physical capital, it can be argued that intensity with which a
firm avails of human capital influences the decision whether to export at all.
5.3.1.3 Business environment variables
These variables include; firm financial access, tax rates, managers’ education and
experience. These variables are constructed based on the rating of 1 to 5.
5.3.2 Model Specification
According to Wagner (2001), studies that make firms’ exports performance depend on a set
of variables focus their attention on knowing if these firms use a unique decision model to
establish the volume of their export or, on the contrary, if firms first decide to participate or
not in the foreign markets and afterwards, they choose the amount of their production to be
sold abroad. In econometric terms, this consists of estimating a Tobit model with all firms
testing it against the Probit model for the decision to export or not.
153
Wagner (2001, 2007) supports the first option and argues that a firm chooses the export
production that maximizes profits, and this could be zero Therefore, the decision process
should be modeled as unique, implying that a model including all the firms, exporters and
non-exporters, must be estimated. And, since the dependent variable is usually the export
propensity (the percentage of the production directed to international markets) and this is,
obviously, a truncated variable (it takes values from 0 to 100 percent), the best way to
estimate the equation consist on using a Tobit model with the whole sample.
If the assertions made on the previous paragraph are true, then the export activity of firms
follows a double decision model: firms first decide if they export or not, what can be
econometrically approached by a binomial model (Probit); and, as soon as they have decided
to take part in foreign markets, they establish the volume of their exports, which forces to
estimate a truncated model, since the dependent variable only is observed if it is greater than
zero (the export propensity is positive).
5.3.2.1 Efficiency scores Estimation
The present chapter employs a Cobb–Douglas log-linear model to derive firm efficiency
scores as specified in Chapter four. In the model, capital, labor, energy and raw material
consumption are used a key independent variables. These scores are obtained by estimating
a four input Cobb Douglas log-linear model given by the following specification:
154
0 1 2 3 4( ) ( ) ) ( ) 5.9i i i i i iInY In K In L InH In R
The choice of the log-linear specification is based on the inherent advantage of reducing the
heteroskedasticity problems as opposed to the non log specification formulation. The firm
time invariant scores are given by the specification below:
0 5.10i i l i k i m i e iTE y l k h r
The input choice in the model is based on economic theory in regard to the firm profit
maximization with efficient resource allocation. In order to avoid the omitted variable bias,
a four input model is estimated which has been found to be more reliable.
5.3.2.2 Model of export performance
The study defines export performance in a dual manner: as the probability to export and the
intensity of exporting. This distinction is particularly important because if a set of variables
impact on these two types of export behavior differently. To examine the impact of firm
characteristics factor intensity and business environment on the firm’s export performance,
the study estimates the following equation;
0 1 2 3sin 5.11i i i i iExp firm Bu ess factor
155
where iExp is the export activity of firm ,i ifirm are the firm characteristics like age, size,
and ownership; ifactor captures the factor intensity variables in the model while sin iBu ess
captures the impact of prevailing business environment like access to finance and credits,
corruption among others. The business environment factors were covered in Récensement
Général des Enterprises (RGE) and in the RPED survey.
5.3.2.3 Probit model of export performance
The study first builds a binary variable of exporter/non-exporter. When the dependent
variable is binary, estimation can proceed by a probit regression and the sign of estimated
coefficients represent the impact of independent variables on the probability of exporting. In
the Probit model, coefficient estimates indicate impact of explanatory variables on
probability of being an exporter (Wagner, 2007).
The Probit model is preferred to the other binary choice models, since economists are likely
to favor the normality assumption of the Probit model. In addition, the method of maximum
likelihood estimation of the Probit model automatically accounts for the heteroskedasticity
problem (Mok et al, 2010).
5.3.2.4 Tobit estimation procedure
When the dependent variable is defined as export intensity, the OLS is not the suitable
estimation procedure because it may produce biased estimates. According to Wakelin
156
(1998), the OLS can give estimates which are higher than one and lower than zero. Hence,
the OLS cannot be used as it will produce biased results.
Wagner (2001, 2007) suggested two models to deal with problem: a one step model and a
two-step model. In a one step model, one equation is estimated using data for non-exporters
and exporters, whereas in a two-step model, the decision to export is modeled separately
from the question of how much to export.
Therefore, the appropriate procedure is to use the Tobit estimation is the most popular in
empirical studies of firm export behavior (Wagner, 2001; Greene, 2003; Niringiye et al.,
2010). According to Greene (2003), the Tobit model is preferred in the empirical analysis of
export performance since the dependent variable is bounded between zero and one.
The model basically assumes two things. First the probability of a limit observation (a zero)
is given by a Probit model with parameter vector 1. That is:
1( 0) ( ) 5.12t tp y X
Where ty is the dependent variable, tX is the row vector of K explanatory variables, 1 is
a column vector of K parameters, is the standard normal cumulative distribution
function, and 1,2,...,t T indexes observations. Second, it is assumed that the density of
157
,ty conditional on being a non limit (positive) observation, is that of 22( , ),tN X truncated
at zero. Thus,
2
222
1 1 1( 0) exp 5.13
( / ) 22t t t t
t
f y y y XX
Defining the indicator function 1tI if 0,ty 0tI if 0,ty then the log-likelihood
function is given as:
22
1 2 221
11 ( ) 1 2 (2 ) 5.14
2
T
t t t t t t t tt
I In X I In X In X In y X
Where 1 2 .
The Tobit model assumes that any variable that increases the probability of positive export
must also increase the average volume of exports of the exporting firms. The model
incorporates the decision of whether to export and the level of exports relative to sales in
one model. This implies that it imposes the same explanatory factors for the two decisions
(Greene, 2003). The Tobit model is also appropriate for censored data (Wagner, 2001).
The equations to be estimated in this chapter can be defined as:
158
20 1 2 3 4 5 6
7 8 9 10
11 12 1
exp
_ _ 5.15
D EXP size size O w ner Loc TE foreign
Acessfin Taxrates M ngedu M ng
Industrial dum m y Export destination
20 1 2 3 4 5 6
7 8 9 10
11 12 1
exp
_ _ 5.16
PEXP size size O w ner Loc TE foreign
Acessfin Taxrates M ngedu M ng
Industrial dum m y Export destination
where DEXP is a dummy variable valuing 1 if the firm is an exporter and 0 otherwise.
PEXP is defined as share of export in total sales or Pr .Export
opensity to ExportTotal sales
The
models are estimated as follows: the first equation is estimated with a probit specification
with DEXP as the dependent variable and employing the whole sample of firms. The Tobit
model estimates the second equation ( )PEXP using the whole sample of firms. The truncated
regression is used for the second equation as well but only for the sub sample of exporters.
5.3.2.5 The Regression of Exports on Technical efficiency
As the technical efficiency ranges between zero and one, the distribution of efficiency is
truncated above unity. If the ordinary least-squares (OLS) method was applied, then the
parameter estimates would be biased. The usual method for handling this problem is to use
a limited dependent variable model; thus, we employ the Tobit model (Tobin, 1958). The
specification of the equation is as follows:
159
20 1 2 3 4 5( / ) 5.17TE Exports Exports Output capital labor ratio foreign
Where TE =Technical efficiency
In order to describe the potential nonlinear relationship between export ratio and
performance, and to capture the positive and negative export effects, the squared term of
export ratio is included in the model. The inclusion of the quadratic term of export in the
Tobit model is mainly supported by the significant coefficient of export2. To isolate the
relationships between export performance and efficiency, it is essential to introduce into the
model other independent variables that are likely to affect efficiency. Among the various
firm attributes, three representative firm-attribute factors are introduced as control variables
in the model: size, capital/labor ratio, and ownership. In estimation, the study applies a
logarithmic model by transforming the independent and explanatory variables into
logarithms, thus controlling for heteroscedasticity.
160
Table 5.1: Description and Summary Statistics of the Variables
y Total sales
K Capital depreciation
L Total labor cost (wages, salaries and bonuses)
TE Technical efficiency
Export Ratio of export to total sales
Size Number of employees
Capital/labor ratio Ratio of capital to total number of employees
Foreign Dummy variable for foreign invested firms
Variable Mean Standard deviation
y 20.038 2.058
K 15.714 1.617
L 83.00 2.427
TE 0.327 0.211
Export 0.600 0.424
Size 32.854 0.507
Capital/labor ratio 0.1893 0.627
Source: Author’s Calculation
5.3.3 The Data
The study uses data on manufacturing firms in Cameroon from the survey organized by the
World Bank entitled “Regional Program on Enterprise Development (RPED)”. The RPED
was designed to improve the understanding of firm level productivity in Africa and to
develop recommendations to improve enterprise development. RPED surveys have been
conducted since 1991 in several African countries.
Firms in the manufacturing sector are surveyed and information gathered on a variety of
issues including outputs and resource use. However, very few studies have analyzed the data
161
to establish the links between technical efficiency and manufacturing export performance
especially in Cameroon. Among these studies are: Soderling (1999), Njikam (2002), Njikam
and Cocburn (2007) and Njikam et al., (2008).
The RPED data set is complimented with the RGE especially for the business environment
factors. The RGE survey includes a very complete questionnaire about each firm’s structure
and strategic decisions, producing a good insight into the Cameroonian manufacturing firms.
5.4 Empirical Analysis
The empirical analysis begins by looking at some firm-level statistics about exporters. Figure
5.3 shows the sample proportions of exporters in six different industries. The figure shows
that exporting is highly concentrated to the wood sector, even though exporters are spread
out across industries. The high concentration in the wood industry is explained by the fact
that Cameroon lies in the equatorial rain forest including Central Africa Republic, Gabon
and Democratic Republic of Congo, where most of its products are exported to developed
countries. Furthermore, the least export-oriented industry is the textile and garments sector,
which is also the most labor intensive industry.
162
Figure 5.3: Percentage of Exporters, By Industry
Notes: The reported percentages are based on observations over the period 2009.
5.4.1 Choice of specification
In the data set, there are few number of firms which have no exports. The dependent variable,
the propensity to export, which varies between 0 and 1 by definition, therefore frequently
takes a value of zero. As a result OLS regression may not be the most suitable estimation
procedure. The model estimation follows Cragg (1971) specification. The specification
estimates a single censored Tobit model. This uses all the available information from the
explanatory variables, but includes both the decision of whether or not to export and the
level of exports, in one model (see Lin and Schmidt, 1984)12
12 Lin and Schmidt (1984) give more details about the type of specification.
0
0.2
0.4
0.6
0.8
1
Food Textile Garment Wood Furniture Metal Total
Per
cen
t
Source: Author's calculation based on RPED Data, 2009
163
The alternative specification is to separate the decision of whether or not to export from the
decision of how much to export. The first stage uses the whole data set and considers the
decision of whether or not to export using a Probit model.13
The model assumes an underlying *Y which cannot be seen. Instead a variable Y can be
observed which takes a value 1 when Y* is greater than 0, and 0 when it is equal or less than
zero. For the second stage only the subset of firms which export are considered. A truncated
estimation procedure is used as the dependent variable is observed only if it is greater than
zero14.
13 The models are estimated using Newton’s method of estimation for maximum likelihood estimation, taking the OLS estimates as the starting values. 14 The Tobit model is considered as the restricted model, and the probit and truncated are the unrestricted models.
164
5.4.2 The probability of Exporting
Table 5.2: Probit Estimates of the determinants of the decision to Export
Variable 1 2 3 Efficiency 0.162*** 0.162*** 0.170*** (6.79) (6.89) (6.87) Firmsize -0.033** -0.065** -0.031** (-2.12) (-2.23) (-2.11) Sizesq 0.058 0.015 0.011 (0.08) (0.22) (0.16) Foreign 0.178** 0.188** 0.215** (2.00) (2.12) (2.35) Mngexp 0.004 0.004 0.005 (1.06) (1.11) (1.34) Mngedu 0.037** 0.039** (0.043)** (2.19) (2.26) (2.40) Expzone 0.035 0.007 (0.38) (0.08) Induszone 0.124* 0.119 (1.45) (1.36) Acessfin 0.044 0.048* (1.27) (1.40) Taxrates -0.009** -0.014** (-2.34) (-2.49) Dfood -0.214** (-2.45) Dwood -0.013 (-0.15) Dchemicals 0.111 (1.09) Dtextiles -0.131** (-2.21) No. Obs 313 313 313 Wald Chi 85.85 91.42 103.19 Prob>Chi2 0.000*** 0.000*** 0.000*** Pseudo R-sq 0.29 0.30 0.32
Notes: Values in parenthesis are the t-values. ***, **, * show level of significance at 1%, 5%, and
10% respectively.
165
Table 5.3: Probit estimates of Determinants of propensity to export to different regions
Variable
Africa Developed ROW
1 2 1 2 1 2
Efficiency 0.174*** 0.177*** 0.150*** 0.160*** 0.360*** 0.139***
(6.72) (6.51) (4.96) (5.22) (4.25) (4.54)
Firmsize -0.067** 0.003** -0.098** 0.047** 0.817*** 0.306***
(-2.22) (2.01) (2.25) (2.12) (2.81) (2.80)
Sizesq 0.013 0.001 0.002 0.019 0.151 0.054
(0.18) (0.01) (0.02) (0.20) (0.60) (0.57)
Firmage -0.002*** -0.002*** -0.013** -0.014*** -0.023** -0.009**
(-2.80) (-1.69) (-2.93) (-2.93) (2.09) (-2.13)
Foreign 0.178** 0.216** 0.431*** 0.453*** 0.626* 0.273*
(1.96) (2.31) (4.01) (4.24) (1.79) (1.91)
Expzone 0.038 0.002 0.295** 0.289** 0.848** 0.303*
(0.39) (0.01) (2.23) (2.12) (2.05) (1.94)
Induszone 0.073* 0.063* 0.322*** 0.326*** 1.126*** 0.405***
(1.80) (1.68) (2.64) (2.67) (2.72) (3.04)
Acessfin 0.031*** 0.034*** 0.089* 0.092* 0.072* 0.031*
(2.84) (2.92) (1.77) (1.85) (1.52) (1.60)
Taxrates 0.017* 0.015* -0.076* -0.076* -0.078 -0.031
(1.58) (1.48) (-1.83) (-1.82) (-0.79) (-0.82)
Mngexp -0.001 -0.002 0.003 0.003 -0.023* -0.007
(-0.20) (-0.14) (0.55) (0.53) (-1.55) (-1.23)
Mngedu 0.029* 0.031* 0.028* 0.037* 0.112* 0.048*
(1.53) (1.58) (1.66) (1.55) (1.61) (1.90)
Dfood -0.169* -0.157* -0.133**
(-1.69) (-1.46) (-2.11)
Dwood 0.054 0.123*** 0.036
(0.22) (3.02) (0.28)
Dchemicals 0.152* 0.12* 0.166**
(1.47) (1.69) (1.97)
Dtextiles -0.102* -0.017 -0.077*
(-1.82) -0.12 (-1.59)
No. Obs 273 273 163 163 138 138
Wald Chi 93.27 103.14 50.05 58.13 39.23 47.76
Prob>Chi 0.000*** 0.000*** 0.000*** 0.000*** 0.000*** 0.000***
Pseudo R-sq 0.31 0.33 0.29 0.31 0.29 0.30
Notes: Values in parenthesis are the t-values. ***, **, * show level of significance at 1%, 5%, and
10% respectively.
166
Concerning the decision to export to neighboring African countries, Developed countries or
the rest of the world (ROW), firm size, firm age, foreign ownership, accessibility to finance,
manager’s education, being a food processing and being a chemical firms are the only
variables which are significant in all the equations. Efficiency of the firms is highly
significant at a 1% level in all the equations. Therefore, being efficient is very important for
the firms as efficiency boosts their export performance.
5.4.3 The propensity to export
Table 5.4 presents the marginal effects of variables that determine the propensity to export
for Cameroon manufacturing firms. The empirical results are estimated by including and
excluding some variables from the model. Some variables are excluded for the sake of
parsimony, while others are excluded to establish whether there is a significant change in
the results or not.
Marginal effects for continuous variables (capital-labor ratio, firm size, firm age, firm age
squared, efficiency and ownership) are reported in the Table to facilitate interpretation.
According to Wagner (2001) and Niringiye et al. (2010), the direction of causality may go
both ways for some of the variables affecting propensity to export. Therefore, the estimated
marginal effects are interpreted as a nature of association, rather than causation between
export propensity and the independent variables in the model.
Notes: 1. Dependent variable is given as Export sales ratio 2. Reported values are marginal effects and values in parenthesis are the t – values.
167
3. ***, ** and * indicate statistical significance at 1%, 5%, and 10% respectively.
Table 5.4: The Tobit Estimates of Propensity to Export
Variable 1 2 3 4 5 6 7
Efficiency 2.160*** 2.060*** 2.090** 1.980*** 1.970** 1.908**
(2.64) (2.75) (2.47) (2.45) (2.45) (2.45)
Firmsize -0.454** -0.540** -0.790 -0.908 -0.390*** -0.759 -0.691**
(-2.43) (-2.50) (0.68) (-0.70) (-2.60) (-0.60) (-2.57)
Sizesq 0.86*** 1.08* 1.08 0.197*** 0.154 0.157** 0.160
(3.37) (1.44) (0.63) (2.67) (0.55) (2.55) (0.57)
Firmage -0.513** -0.504 -0.642** -0.649 -0.539 -0.521 -0.411***
(-2.20) (-1.22) (-2.15) (-1.15) (-1.04) (-1.02) (2.89)
Foreign 0.962** 0.944 0.100 0.104** 0.107* 0.108 0.126*
(2.24) (1.26) (1.23) (2.25) (1.67) (1.29) (1.45)
Acessfin 1.720 0.187** 0.164 0.164* 0.155 0.159
(1.19) (2.20) (1.11) (1.66) (1.05) (1.08)
Taxrates 0.870*** 0.679*** 0.434** 0.329** 0.490** 0.601**
(2.52) (2.52) (2.36) (2.28) (2.24) (2.51)
Mngexp 0.146 0.143 0.906* 0.792 0.791*
(0.67) (0.63) (1.45) (0.39) (1.41)
Mngedu 0.820*** 0.830*** 0.932* 0.868* 0.108**
(2.64) (2.65) (1.75) (1.66) (1.99)
Expzone 0.363** 0.303 0.303 0.433
(2.00) (0.86) (0.89) (1.22)
Induszone 0.555* 0.548* 0.524* 0.593*
(1.42) (1.40) (1.48) (1.55)
D_food -0.624* -0.662* -0.543*
(-1.74) (-1.80) (-1.60)
D_wood -0.466** -0.469** -0.525
(-2.24) (-2.25) (-1.38)
D_textiles -0.788* -0.809* -0.868*
(-1.66) (-1.67) (-1.77)
Africa -0.190* -0.209**
(-1.63) (-2.69)
Developed 0.190* 0.157***
(1.93) (2.79)
Row -0.139 -0.141
(-0.75) (-0.78)
Constant -4.18*** -4.42*** -4.47*** -4.75*** -4.62*** -4.48*** -1.09
(-3.35) (-3.13) (-2.97) (-2.93) (-2.92) (-2.75) (-1.34)
Pseudo R-sq 0.38 0.48 0.39 0.35 0.30 0.23 0.25
No. Obs 319 319 313 313 313 313 313
F-test F(5, 314) F(7,312) F(9, 304) F(11, 302) F(15,298) F(18, 295) F(17, 296) Prob > F 0.000*** 0.000*** 0.000*** 0.000*** 0.000*** 0.000*** 0.000***
168
The results from Table 5.4 show that firm size is an important determinant of propensity to
export suggesting that sunk costs is an important issue for firms to consider during the start-
up process. According to Bigsten et al., (1997), Niringiye et al. (2010), high fixed and sunk
costs of exporting makes it difficult for small firms to enter the export markets. This result
is in accordance with the existing theory of sunk costs of entering the export market.
169
Table 5.5: Tobit Estimates of propensity to export without firm size
Variable 1 2 3 4 5
Efficiency 0.197** 0.197* 0.189* 0.187* 0.188*
(2.08) (1.84) (1.89) (1.70) (1.66)
Firmage -0.958 -0.131 -0.128 -0.106 -0.873
(-0.94) (-0.98) (-0.98) (-0.86) (-0.74)
Foreign 0.935* 0.986* 0.102* 0.107 0.108
(1.91) (1.69) (1.89) (1.01) (1.03)
Acessfin 0.157** 0.167* 0.142** 0.146* 0.138*
(2.10) (1.89) (2.04) (1.66) (1.59)
Taxrates 0.714 0.773 0.507 0.394 0.569
(0.47) (0.43) (0.30) (0.27) (0.33)
Mngexp 0.683** 0.569** 0.223** 0.122**
(2.41) (2.41) (2.18) (2.10)
Mngedu 0.734** 0.759*** 0.874** 0.806**
(1.96) (2.63) (2.47) (1.98)
Expzone 0.306* 0.246*** 0.245***
(1.89) (2.64) (2.62)
Induszone 0.563* 0.555 0.532
(1.51) (1.22) (1.13)
Dfood -0.67 -0.71
(-1.15) (-1.16)
Dwood -0.482 -0.485
(-1.04) (1.01)
Dtextile -0.867* -0.887*
(-1.65) (-1.68)
Africa 0.209**
(1.99)
Developed 0.17**
(1.21)
Row -0.126
(-0.51)
Constant -0.531* -0.577* -0.590* -0.554* -0.540*
(-1.77) (-1.66) (-1.71) (-1.62) (-1.63)
Pseudo R-sq 0.30 0.37 0.34 0.33 0.35
No Obs 319 319 313 313 313
F-test F(2, 313) F(2, 305) F(3, 302) F(2, 299) F(2, 296)
Prob>F 0.000*** 0.000*** 0.000*** 0.000*** 0.000*** Notes: 1. Dependent variable is given as Export sales ratio 2. Reported values are marginal effects and values in parenthesis are the t – values. 3. ***, ** and * indicate statistical significance at 1%, 5%, and 10% respectively. 4. All robust standard errors in the regression are adjusted for 3 clusters in firm size.
170
Since firm size has consistently been shown as an important determinant of both probability
to export and propensity to export, controlling for firm size investigates the issue of
heterogeneity in the models. Controlling for firm size, the Tobit regression presented in
Table 5.5 shows that firm age is not an important determinant of propensity to export. The
coefficient of firm age is also negative suggesting that older firms in this case are less likely
to export more than younger firms. In the absence of firm size, the most significant variables
affecting the propensity to export are manager’s experience, manager’s education as well as
being efficient. Location at the export zone is also a very important determinant since firms
in the export zone have tax holidays; they are more likely to export than firms paying taxes.
171
Table 5.6: Tobit Model with Interaction effect
Variable 1 2 3 4
Efficiency 0.822** 0.244** 0.679 0.235**
(2.42) (2.33) (0.36) (2.31)
Firmsize -0.376* -0.216** -0.332** -0.133*
(-1.46) (1.97) (2.13) (-1.69)
Firmage -0.588 0.987 -0.777 -0.857
(-0.90) (-1.11) (-1.03) (-1.02)
Size*TE 0.177* 0.153*
(1.47) (1.54)
Size*ownership 0.288** 0.239 0.292**
(2.32) (1.35) (2.33)
Size*food -0.163*
(-1.44)
Wood*size -0.644
(-0.61)
Size*chemicals 0.523
(0.06)
Textile*size -0.294*
(-1.46)
No Obs 319 319 319 319
Pseudo R-sq 0.63 0.65 0.73 0.71
F-test F(4, 315) F(4, 315) F(5, 314 F(8, 311)
Prob>F 0.0316** 0.0342** 0.0579* 0.0519*
Notes: 1. Dependent variable is given as Export sales ratio 2. Reported values are marginal effects and values in parenthesis are the t – values. 3. ***, ** and * indicate statistical significance at 1%, 5%, and 10% respectively.
Table 5.4 shows the cross-level interaction effect between firm size and other firm-specific
determinants. The Positive and significant effect between firm size and technical efficiency
shows that large firms which are more technically efficient are more likely to export. More
so, foreign firms are more likely to export. The negative and significant effects show that
large firms in the textile and food industries are less likely to export.
172
5.4.4 The effect of export orientation on Technical Efficiency
In order to estimate the effect of export on technical efficiency, the SFA was used to calculate
the technical inefficiency for the manufacturing firms in the sample. The average score of
the sample firms is 36.4. This implies average mean average efficiency is 63.4 for
Cameroonian firms. Graner and Isaksson (2007) found that the average (mean) technical
efficiency of Kenyan manufacturing firms was 55% in 1992 – 1994. After computing the
technical efficiency, we then proceeded to test whether export orientation improved
technical efficiency by estimating equation of export on technical efficiency using the Tobit
method. The results are presented in Table 5.7.
Table 5.7: Estimates of the effects of Export orientation and the control variables on TE
Variable Coefficient t - values
Intercept - 3.905 - 9.900***
Export - 0.252 - 2.318**
Export2 0.226 2.901**
Size 0.121 10.592***
Capital/labor ratio 0.111 3.950***
Foreign 0.042 1.841*
Log likelihood function 1772.67
Pseudo R-sq 0.427
Notes: ***, **, * show significant at the 1%, 5% and 10% level, respectively.
173
The result of the pseudo R-sq suggests that about 42.7 percent of the variation in technical
efficiency between the sample firms can be explained by variations in export orientation and
the control variables.
The coefficient of export variable was estimated to be -0.252 and that of the squared term of
export was 0.226, both being significant at the 5% level. The results suggest a U-shaped
relationship between export ratio and technical efficiency. The signs of export and export2
suggest that firms with a high level of export-orientation experience higher technical
efficiency. As such, it is important to estimate the turning point of the U-shaped curve. The
point of inflection is computed by taking the partial deviation of our Tobit model with respect
to export as follows:
252.0exp*226.02)(exp
)(
ort
ort
TE
Setting the above partial deviation to be zero, the inflection point of export is determined to
be 0.557. In other words, technical efficiency begins to decline when the ratio of export sales
to total sales varies from zero to 0.557, and technical efficiency reverts to an upward trend
when the export ratio changes from 0.557 to 1. The inflection point also suggests that for
firms with export ratios less than 0.557, further expansion of domestic market would
improve their technical efficiencies.
174
According to Mok et al. (2010), when firms’ exports take up only a moderate but not
dominant portion of their total sales, the costs of transaction to handle various bureaucratic
procedures in exportation as well as to meet the ever-demanding technical barriers of trade
are considerably high compared with the possible benefits of internationalization through a
moderate degree of export orientation. Hence, beyond a turning point, the marginal returns
of higher level of export orientation may exceed the marginal costs of exportation, and the
net effects of export orientation enter a positive efficiency territory. Therefore, a firm
reaching a relatively high level of exportation may be able to manage information and to
concentrate their resources to engage in the international market while reaping the benefits
of internationalization. This helps explain why the association between export orientation
and technical efficiency is evidenced to be U-shaped in our sample firms.
Besides, export-orientation can have positive impacts on technical efficiency for firms with
a large portion of sales to international market, while firms that try to develop the domestic
and overseas markets simultaneously perform less efficiently. From the prospective of
technical efficiency, it appears that the optimal strategy for Cameroonian manufacturing
firms is to focus on one market, that is, either the domestic or overseas market rather than
splitting the firms’ resources or efforts to target both domestic and international markets.
In order to obtain further indication of the relationship between export and technical
efficiency illustrated by the Tobit model, the empirical study is extended from the base
model by performing the following group-wise analysis. The study attempts to compare the
175
average technical efficiency of various groups of firms categorized in terms of export ratio.
The firms are divided into four groups by export ratio equally and then compare the means
of technical efficiency of the two extreme groups. Mok et al. (2010) suggested a non-
parametric Wilcoxon rank-sum test be performed as there is no reason to assume that the
distribution of technical efficiency is normal15. The null hypothesis to be tested:
.:0 groupsextremetwotheforsamethearemeasuresefficiencytheofondistributiTheH
Table 5.8: Group-wise Technical Efficiency comparisons
Grouping
Variable
Firms with
export ratio
between
0 - 0.250
Firms with export
ratio between
0.751 - 1
Wilcoxon
Statistic
Z -
Statistics Sig.
Export
Mean Rank 110.506 17.195 778 -3.95 0.042***
Number of
firms 79 22
Notes: The null hypothesis of the Wilcoxon test is that the technical efficiency distributions of the two populations are the same.
As shown in Table 5.8 (Wilcoxon), the mean rank of firms with export ratios between zero
and 0.25 is 110.506, while that of the contrasted group is 17.195; this gap is very significant.
It suggests that the two extreme groups do not exhibit the same level of technical efficiency.
15 The Wilcoxon rank-sum test is a nonparametric alternative to the two sample t-test which is based solely on
the order in which the observations from the two samples fall. With the Wilcoxon test, an obtained W is
significant if it is LESS than or EQUAL to the critical value.
176
The results from the Tobit model in Table 5.7 (Tobit) illustrate a significant U-shaped
relationship between export and technical efficiency. By comparing the two extreme groups
from the degree of export ratio, there is an obvious disparity of performance between them.
The results suggest that the technical efficiency gap between firms targeting their major
products to the domestic market and firms focusing on the overseas market is significant. As
long as firms focus on a specific market, whether domestic or overseas, they can obtain their
advantages on performance in terms of technical efficiency.
The above finding does correspond with the common conjecture of a positive relationship
between export and efficiency. The empirical study illustrates that the role of export on
technical efficiency depends on the attributes of a firm’s market orientation. Among the
previous studies, Gomes and Ramaswamy (1999), and Graner and Issaksson (2007)
supported the positive role on efficiency, particularly highlighting that the positive
productivity effects largely occur at the firm level before it enters into the international
market, that is, firms improve their efficiency in order to develop an export market.
However, Bernard and Jensen (1999) rejected the linear positive relationship between export
and efficiency, considering it too simplistic. They performed an empirical study on the
industry level in which the results showed that there was no evidence to suggest significant
productivity gains at the industry level resulting from exports. Mok et al., (2002) suspected
that the negative relationship between exports and efficiency may be partly attributable to
the high transaction costs of exportation that result from the ambiguity, complexity, and
177
inflexibility of government policies in the labor, capital, and product markets. Different from
the literature in which the inverted-U curve finding is largely based on the empirical
evidence from transnational corporations (TNCs) in developed countries (Gomes and
Ramaswamy, 1999), the majority of the manufacturing firms in Cameroon are labor-
intensive.
To ascertain the effects of the three control variables of firm size, capital/labor ratio, and
ownership on technical efficiency, the estimated results show that large-sized firms are
relatively more efficient than their smaller counterparts. The positive and significant
coefficient (significant at 1 percent level) of the variable “size” suggests that large firms take
advantage of the scale economies. A number of empirical studies have examined both linear
and non-linear relationships between firm sizes and their export decision or export
performance (Lundvall and Battese, 1998; Niringiye et al., 2010). Some of these studies found
that firm size, has a linearly significant and positive effect on a firm’s export decision,
implying that there are typically significant sunk costs related to the export decision.
Wakelin (1998) identifies an inverted U-shaped relationship. Therefore, both firm size and
its square are included in the estimated model to test for non-linearity. The coefficient of
capital/labor ratio is positive and significant, which suggests that the capital intensity
measured by the capital-labor ratio has a positive effect on efficiency.
More so, the estimator of the coefficient of the foreign-invested firm dummy variable though
positive is weakly significant at the 10% level, which indicates that the foreign-invested
178
firms in Cameroon are more efficient than their locally invested counterparts. This is
somewhat expected, given that foreign-invested firms are normally expected to have better
knowledge of advanced technology, management skills, and research and development
capability. This finding may be reconciled by the fact that a significant proportion of
foreign-invested firms in Cameroon, though still somehow labor-intensive due concentrate
on product design, innovation, and development. This is especially the case in a number of
textile and garment firms. This result is inconsistent with the finding of Wei et al. (2002)
that wholly foreign-owned firms in China are less productive than firms with other types of
ownership.
5.5 Conclusion
This paper has analyzed the role of firm specific characteristics in influencing export
performance at the firm level, as well as empirically examined the effects of export
orientation on the technical efficiency of manufacturing firms Cameroon. Particular
emphasis has been placed on the technical efficiency characteristic of the firm.
By considering the relationship between export and technical efficiency at the firm level, the
paper has a number of advantages over more aggregate studies. It is only at the firm level
that the influence of firm characteristics can be separated from those of the sector. By
balancing the two groups of firms by their sectors of origin, the analysis abstracts from
differences in sectors and concentrates on the role of firm characteristics (Wakelin, 1998).
179
Examining export behavior at the firm level allows an assessment of the diversity among
firms, and particularly between highly efficient and less efficient firms.
As concerns the effect of export orientation on technical efficiency, instead of a simple linear
relationship between export-orientation strategy and efficiency at the firm level, our results
suggest a U-shaped relationship between export ratio and technical efficiency. This implies
that firms with a high level of export orientation experience higher technical efficiency. The
findings are further supported by the results of a group-wise comparison in the two extreme
groups of firms in terms of the degree of export ratio.
180
CHAPTER SIX
CONCLUSION, POLICY IMPLICATIONS AND LIMITATIONS OF THE STUDY
6.1 Introduction
Improving firm productivity in the industrial sector clearly plays an important role in
promoting economic growth and alleviating poverty in a country. As the literature review
from developing and transitional countries has shown, efficiency and manufacturing export
enhancements in using resources positively impact firm productivity. Given the various
industrial policies implemented over the years to increase firms’ efficiency and export
performance in Cameroon, it was necessary to quantitatively measure the current levels of
efficiencies and manufacturing export performance and their determining factors. Moreover,
as experience from other countries in the developing world shows, quantitative are needed
before developing new policy instruments and adopting new technologies. The topic is very
important for both academics and policy makers as firms face environmental factors and
existing institutional complexities in Cameroon.
In order to provide insights on how to improve efficiency of manufacturing firms in
Cameroon, this study examined the determinants of technical efficiency as well as explored
the impact of technical efficiency on the export performance of manufacturing firms. The
focus was on analyzing the efficiency of manufacturing firms in Cameroon and identifying
the factors likely to increase productivity and export performance of the firms through a
better use of factors engaged in production. The analyses were carried out at the micro level;
181
firstly by examining the factors that affect technical efficiency in manufacturing firms and
secondly by linking technical efficiency to export performance as well as export orientation.
The manufacturing sector was analyzed because it is one of the major contributors to GDP
and a source of foreign earnings to the economy.
The objectives of this Chapter are in two fold. Firstly, to summarize major empirical findings
for policy implications. Secondly, to summarize the contribution of this study to research on
firm efficiency and export performance.
6.2 Summary of findings
The summary of the findings is based on the broad research questions formulated in Chapter
1. The first broad question was answered by examining the determinants of technical
efficiency in manufacturing firms in Cameroon using stochastic frontier analysis. The sources
of technical inefficiency were explained and the levels of technical efficiency/inefficiency in
the various sectors were obtained. The second question was answered by examining the
relationship between technical efficiency and export performance. The determinants of export
performance as well as export orientation were examined using both Probit and Tobit models.
In order to improve firms’ efficiency, and thus stimulate industrial export competitiveness,
the findings of this dissertation suggest that firms should examine their production inputs
182
structures as well as find out opportunities for cost reduction that may improve technical
efficiency of firms and subsequently export performance.
The main finding of this study is that manufacturing firms in Cameroon were technically
inefficient. The most efficient firms were from the food processing sector, followed by the
wood and furniture sector. Firms with 5 to 20 years of operation in Cameroon were found to
be more efficient. More so, medium sized firms were more efficient than small and large
firms. This shows a U-shaped relationship between size and efficiency. Hence, results show
that technical efficiency increases with medium sized scale of operation.
This suggests that medium sized firms should be encouraged to produce more output. This
will not only benefit the firms but also promote Cameroon’s industrial competitiveness at
the international level. The analysis also revealed that foreign ownership; tax rates imposed
by the government, accessibility to financial credit, managerial education as well as
experience were the major variables influencing firms’ technical efficiency in Cameroon.
More so, the findings show that further productivity gains linked to the improvement of
technical efficiency could still be realized in the Manufacturing sector.
As concerns the determinants of manufacturing export performance, the study included factor
intensity variables, firm specific variables and business environment factors in the Tobit and
Probit models. The results show that higher level of efficiency, firm size, foreign ownership,
lower tax rates, producing in the industrial zone, and being in the food processing and textile
183
sectors were the main determinants of propensity to export. The main finding supports the
self-selection hypothesis.
The major determinants to export to Africa, Developed countries and Rest of the World were
firm size, firm age, ownership, and accessibility to finance. On the determinants of the
decision to export or not, highly efficient firms, firm size, foreign ownership, tax rates, being
in the food processing and textile sector were statistically significant variables. To promote
manufacturing export performance, polices should be designed for attracting foreign
investments more especially in the food processing and textile sectors.
This dissertation thus contributes to the efficiency and manufacturing export performance
literature in several ways. It provided resource-use efficiency levels based on the types of
institutional settings that exist in the industrial sector in Cameroon. Specifically, it contributes
to the on-going policy discussion regarding how to improve firm efficiency and
manufacturing export performance. On the academic level, the present study contributes by
dealing with analytical challenges in estimating frontier models. It used parametric models to
estimate technical efficiency and manufacturing export performance, and determined
significant factors associated with inefficiency levels.
184
6.3 Policy Implications
One of the current issues of manufacturing in Cameroon is how to improve firm productivity
and manufacturing export performance under the economic, environmental and institutional
constraints that exist in the country. Because improvements in the efficient use of resources
is related directly to increases in firm efficiency and export performance, our
recommendations can assist policy makers develop effective policy instruments to increase
firm productivity in Cameroon.
With our efficiency analysis, we have provided information on the levels of technical
efficiencies that could be used as an alternative source to evaluate firm performance under
the present institutional setting. Our policy recommendation in this regard is that, there is
still room for technical efficiency improvements with existing firm technologies. In the near
future, however, new technologies must be introduced to sustain higher efficiency levels and
reduce related production costs.
The following policy implications can be drawn from the results of this study. Compared to
all the other sectors, food processing sector has the highest technical efficiency followed by
wood and furniture sector. The government could do well by encouraging especially the food
sector in order to boost food processing activities and to exploit the advantage of the
agricultural sector as an important source of raw materials. Promoting food processing can
185
reduce the problem of post-harvest losses in the agricultural sector and make Cameroon a
competing exporter of food and intermediate products.
Following the evidence of a U-shaped relationship between efficiency and firm size, the
government should design strategies to provide incentives, credit to small and medium sized
firms in order to increase output as well as increasing economic competitiveness and growth
of the firms.
The results consistently show that firm size is an important determinant of propensity to
export suggesting that sunk costs is an important issue for firms to consider during the start-
up process. The findings demonstrate that firm technical efficiency decreases with local
ownership, thus indicating that efficiency in Cameroon manufacturing firms is highly
associated with foreign ownership. This suggests that focus on promoting foreign
participation in industrial production, will help improve technical efficiency. Hence, to
improve manufacturing export performance, polices should be designed towards attracting
foreign investments especially in the food processing and textile sectors.
This study makes contribution to already existing literature in the following ways: analyzing
export performance with main focus on efficiency and firm size for a poor and emerging
country. There has been no previous robust empirical work on answering the age-old, yet
classic question of the effect of efficiency on export performance for Cameroon. Most
studies did focus on the effects of exporting on firm-level productivity.
186
6.4 Limitations of the study and areas of further Research
6.4.1 Limitations
As with any study, there are limitations that should be noted.
1) Cross-sectional design: cross-sectional design used here did not allow analysis of
trends over time. A panel data study might have been able to detect trends in output
variability. A panel design would also allow for more observations and thus would
contend with the criticisms of cross-sectional design that it requires stricter
assumptions than a panel design. Moreover, testing for endogeneity will help
understand many important issues that exist in the current industrial sector.
2) Omitted variables: production function used in this study has only four independent
(input) variables. For an operation as complex as the manufacturing firm, the
specified production function might omit important sub-categories of input in the
production process. Omitted variables result in specification errors that are likely to
confound efficiency estimates. This analysis used a four variable production function
which is consistent with previous literature.
Despite these limitations, we are confident that the results were the best that could be
obtained given the circumstances. The models have permitted not only to estimate the
technical efficiency indexes of manufacturing firms but also to identify the factors that affect
export performance as well as export orientation in Cameroon.
187
6.4.2 Areas for further Research
We must obviously offer the caveat that this empirical study is based on micro firm level
data from manufacturing firms in Cameroon; it would be interesting to set up a survey to
ascertain how much firms’ output affect the economy at the macro level. In the case, it would
be very useful to construct panel dataset, while ideally; time series data is necessary for
dynamic modeling that takes into account technological and policy changes. However, in
the case of Cameroon, surveys are not conducted every year.
More thorough modeling of quality: as discussed, quality in manufacturing firms is an
elusive concept to define and measure. This study included structural, process, and outcome
measures of efficiency. Process measures proved difficult, primarily due to the sample size
and the fact that firms are still at liberty to report quality data or not. A panel data approach
might be one partial remedy, as reporting of quality becomes more widespread (and even
required) in as many firms as possible.
More so, most of the models used are limited in the sense that they do not include market
imperfections due to lack of data. This is another area for further research.
188
REFERENCES
Abuka, C.A. 2005, “An Empirical Analysis of the Impact of Trade on Productivity of South
Africa’s Manufacturing Sector” Unpublished PhD Thesis, University of Pretoria
Aerts J., D. Cogneau, J. Herrera, G. de Monchy, F. Roubaud. 2000. L'Economie
Camerounaise- Un Espoir Evanoui, Paris: Karthala.
Aigner, D.J. and Chu, S.F. 1968. “On Estimating the Industry Production Function”.
American Economic Review, 58(4): 826–39.
Aigner, D., K. Lovell and P. Schmidt. 1977. “Formulation and Estimation of Stochastic
Frontier Production Function Models” Journal of Econometrics 6, 21–37.
Alvarez, R., and R. López, 2005. "Exporting and performance: evidence from Chilean
plants," Canadian Journal of Economics, Canadian Economics
Amara, N., N. Traore, R. Landry, and R. Romain. 1999. “ Technical Efficiency of farmers
attitude toward technological innovation: The case of the Potato Farmers in
Quecbec”, Canadian Journal of Agricultural Economics, 47(1), 31-49
Amin, A. 2002. “An examination of the sources of economic growth in Cameroon” AERC
Research paper 116, Nairobi
Amos, T.T. 2007. “An analysis of productivity and Technical Efficiency of Smallholder
Cocoa farmers in Nigeria” Journal of Social Science 15(2): 127-133
Barnett, W. A., and I. Usui, 2006. "The Theoretical Regularity Properties of the Normalized
Quadratic Consumer Demand Model" Retrieved February 5, 2009, from Munich
Perspnal RePEc Archive: http://mpra.ub.unimuenchen.
de/410/1/MPRA_paper_410.pdf
Battese, G. 1992. “Frontier production function and technical efficiency: A survey of
empirical application in agricultural economics” Agricultural Economics 7:185
Battese G. and T. Coelli. 1988. “Prediction of firm level technical efficiencies: With a
generalized frontier production function and panel data” Journal of Econometrics
38: 387-399
Battese, G. and T. Coelli. 1992. “Frontier production functions, technical efficiency and
panel data: with application to paddy farmers in India”. Journal of productivity
Analysis 3, 153-169
189
Battese, G. and T. Coelli.1993. “A production frontier production function incorporating a
model for technical inefficiency effect” Working paper in Econometrics and
Applied statistics No. 69, Department of Econometrics, University of New
England, Armidale.
Bauer, P.W. 1990. “Recent developments in the econometric estimation of frontiers”
Journal of Econometrics 46: 39-56
Behr, A. and S. Tente. 2008. “Stochastic frontier analysis by means of maximum likelihood
and the methods of moments” Banking and Financial Studies, Discussion Paper,
Series 2, No. 19
Belotti F., S. Daidone, G. Ilardi. 2012. “Stochastic frontier analysis using Stata” The Stata
Journal, Vol. v, Number ii, pp. 1 – 39.
Benjamin N., Devarajan S. (1985). “Oil Revenues and Economic Policies in Cameroon”
World Bank, Washington pp 9.
Benjamin, N. C. Deverjan, and R. Weiner. 1989. “The ‘Dutch Diseases’ in a Developing
Country Oil Reserves in Cameroon,” Journal of Development Economics, Vol. 30
Bernard, A. B., Jonathan Eaton, J. Bradford Jensen, and Samuel S. Kortum, 2003. "Plants
and Productivity in International Trade", American Economic Review, 93(4),
1268-1290.
Bernard, J.and B. Jensen, 1999. “Exceptional exporter performance: cause, effect, or both?”
Journal of International Economics 47 (1999) 1–25
Bigsten. 2000. "Rates of return on physical and human capital in Africa's manufacturing
sector", Economic Development and Cultural Change, 48: 80 1—27
Burki, A. and T. Dek. 1998. “Measuring Production Efficiency a Small Firms in Pakistan”
World Development 26(1), 155-169
Castellani, D. 2002. “Export behavior and productivity Growth: Evidence from Italian
Manufacturing Firms” Weltwirtschaftliches Archiv , Vol. 138 (4)
Caves, R., 1971. “Export-Led Growth and the New Economic History. In J. N. Bhagwati, R. W. Jones, R. A. Mundell and J. Vanek (eds) Trade, Balance of payment and Growth. Papers in International Economics in Honour of Charles P. Kindleberger, Amsterdam
190
Chapelle, K. and P. Plane. 2005. “Technical Efficiency measurement within the
Manufacturing Sector in Cote d’Ivoire: A Stochastic Frontier Approach” The
Journal of Development Studies 41(7), 1303-1324
Charnes, A., W.W.Cooper and E. Rhodes. 1978. “Measuring the efficiency of decision
making units”, European Journal of Operation research, 2, 429-444.
Chen A.Z., W.E. Huffman and S. Rozelle, 2003. “Technical efficiency of Chinese Grain
Production: A Stochastic Production Frontier Approach” Paper prepared for
presentation at the American Agricultural Economics Association Annual Meeting,
Montreal, Canada, 2003.
Cheng, Y. and Lo, D., 2004. “Firm Size, Technical Efficiency and Productivity Growth in
Chinese Industry”, School of Oriental and African Studies University of London,
Working Paper No. 144
Clerides, S. K., S. Lach and J. R. Tybout. 1998. .Is Learning by Exporting Important? Micro-Dynamic Evidence from Colombia, Mexico, and Morocco.. Quarterly Journal of Economics 113(3):903.947.
Coelli, T.J. 1994. A Guide to FRONTIER version 4.1: A Computer Program for Stochastic
Frontier Production and Cost Function Estimation, Mimeo, Department of
Econometrics, University of New England, Amidale
Coelli, T.J. 1995. “Recent Developments in Frontier modeling and Efficiency Measurement”
Australian Journal of Agricultural Economics, 30: 219-245
Coelli, T.J. 1996. “A guide to FRONTIER 4.1: A computer program for Frontier production
function estimation” Centre for Efficiency and Productivity Analysis (CEPA),
working paper 96/07, Department of Economics, University of New England,
Australia
Coelli, T., Prasada Rao, D. S., O'Donnell, C. J., & Battese, G. E. (2005). An Introduction to
Efficiency and Productivity Analysis (2nd ed.). New York: Springer.
Cummins, D., Zi, H., 1998. Comparison of Frontier Efficiency Methods: An Application to
the U.S. Life Insurance Industry. Journal of Productivity Analysis 10 (2), 131–52.
Daraio, A. and L. Simar. 2005. “Introducing environmental variables in nonparametric
frontier models: a probabilistic approach,” Journal of productivity analysis, 24, 93-
121
191
Debreu, E. 1951. “The coefficient of resource utilization” Econometrica 19, 273-292
Doe, L. 1995. “Managing Cameroon’s Banking Sector – In and out of Crisis: The Role of
the Government,” African Development Review, Vol. 7, pp 103 -66
Dunne, P. and A. Hughes. 1994. “Age, size, growth and survival: UK companies in the
1980s.” The Journal of Industrial Economics 42, 115-140.
Dunning, J. H. 1988. Multinationals. Technology and Competitiveness, London, Unwins
Evans, D.S. 1987. “Tests of Alternative Theories of Firm Growth”, Journal of Political
Economy, Vol. 95, No.4, pp. 657-674.
Faruq, H.A and D. T. Yi. 2010. “The Determinants of Technical Efficiency of Manufacturing
Firms in Ghana” Global Economy Journal, Vol. 10, Issue 3, pp 1524-5861
Fare, R. and C.A.K Lovell. 1978. “Measuring technical efficiency of production” Journal of
Economic Theory 19(1), 150-162
Farrell, M.J. 1957. “The measurement of productive efficiency”, Journal of the Royal
Statistical Society, Series A, 120(3): 253-290.
Farrell, M.J. and M. Fieldhouse, 1962, “Estimating efficient production under increasing
returns to scale”, Journal of the Royal Statistical Society, Series A, General,
125, no. 2, 252-267
Fariñas J. C. and A. Martin-Marcos 2007. “Exporting and economic performance: firm-level
evidence of Spanish manufacturing”, The World Economy 30 (4), 618–646.
Førsund, F.R. and L. Hjalmarsson (1987): Analyses of Industrial Structure. A PuttyClay
Approach, Stockholm: Almqvist & Wiksell International
Fraley, C and A. Raftery. 1998. “How many clusters? Which clustering method? Answers
via model-based cluster analysis.” Technical report No. 329, Department of
Statistics, University of Washington
Ghura, D. 1997. “Private Investment and Endogenous Growth: Evidence from Cameroon”
IMF Working Paper, WP/97/165
Giannakas, K., K. C. Tran and V. Tzouvelekas. 1998. “On the Choice of the Functional Form
for Agricultural Efficiency Measurement”, Journal of Agricultural Productivity
192
Giles, J. A., and C.L. Williams (2000). “Export-Led Growth: A survey of the empirical
Literature and some Noncausality Results, part I. Journal of International Trade
and Eonomic Development 9: 261- 337.
Gomes, L. and Ramaswamy, K. 1999. ‘An empirical examination of the form of the
relationship between multinationality and performance’, Journal of International
Business Studies 30: 173-188
Gourlay, A. R., and J. S. Seaton. 2004. “UK Export Behaviour at the Firm Level” Economic
Issues, Vol. 9, Part 2.
Graner, M. and A. Isaksson. 2007. “Firm Efficiency and the Destination of Exports:
Evidence from Kenyan Plant-level Data” United Nation Industrial Development
Organization (UNIDO)
Greene, W. 2001. “Fixed and random effects in nonlinear models” Working paper,
Department of Economics, Stern School of Business, NYU.
Greene, W. 2002. “Alternative panel data estimators for stochastic frontier models” Working
Paper, Department of Economics, Stern School of Business, NYU
Greene, W. 2003.“Distinguishing Between Heterogeneity and Inefficiency: Stochastic
Frontier Analysis of the World Health Organization´s Panel Data on National
Health Care Systems”, Working Paper, Department of Economics, Stern School of
Business, NYU.
Greene, W. 2005. “Reconsidering Heterogeneity in Panel Data Estimators of the Stochastic
Frontier Model” Journal of Econometrics 126, 269–303.
Hecksher, E. F. and B. Ohlin (1991), Heckscher-Ohlin Trade Theory, translated, edited and
introduced by Harry Flam and M. June Flanders, Cambridge, Mass, MIT press
Hirsch, S. 1971. The export performance of six manufacturing industries: A comparative
study of Denmark, Holland and Israel, Praeger Publishers, new York.
Hirsch, S., Adar Zvi .1974. “Firm size and export performance.” World Development, Vol.2,
No. 7, pp 41-4
Ijiri, Y., and H. Simon. 1964. “Business firm growth and size” The American Economic
Review, 54, 77-89.
193
Jondrow, J., C.A.K. Lovell, I. Materov and P. Schmidt. 1982. “On the estimation of technical
inefficiency in the stochastic frontier production function model,” Journal of
Econometrics 19, 223-238
Jovanovic, B. 1982. “Selection and evolution of industry,” Econometrica, 50, 649-670
Kaldor, N. 1970. The case of Regional Policies. Scottish Journal of Political Economy 17
(3): 337-448
Kodde, D.A. and F.C. Palm, 1986, 'Wald Criteria for Jointly Testing Equality and Inequality
Restrictions', Econometrica, Vol.54, No.5, pp.1243-48.
Koopmans, T.C. 1951. “An Analysis of Production as an efficient Combination of
Activities” In T.C. Koopmans (Ed.), Activity Analysis of production and
Allocation, Cowles Commission for Research in Economics, No. 13, Wiley
Kopp, R.J. and V.K. Smith. 1980. “Frontier production function estimates for Steam Electric
Generation: A Comparative analysis. Southern Economic Journal, 47
Kumbhakar, S. C. and Lovell K. C.A. (2000) Stochastic Frontier Analysis. Cambridge
University Press, Cambridge.
Kumbhakar, Subal C. 1991 Estimation of technical inefficiency in panel data models with
firm- and time-specific effects. Economics Letters. 36 (1) 43-48
Kumbhakar, S.C. and Tsionas, E.G. 2006 Estimation of stochastic frontier production
functions with input-oriented technical efficiency. Journal of Econometrics. 133
71 - 96.
Leibenstein, H. 1966. “Allocative efficiency versus X-efficiency” American Economic
Review, Dec: 1252-1258
Little, I. 1987. “Small Manufacturing Enterprises in Developing Countries” The World Bank
Economic Review 1(2), 203-236
Little, I., D. Mazumdar and J. Page. 1987. Small manufacturing Enterprises: A Comparative
Analysis of India and other Economies, New York: Oxford University Press.
Lundvall, K. and E. Battese. 2000. “Firm size, age and efficiency: Evidence from Kenyan
Manufacturing firms” Journal of Development Studies, 36(3), 146-163
Mahadeven, R. 2000. “How Technically Efficient are Singapore’s Manufacturing
Industries” Applied Economics 32, 2007-20014
194
McPherson, M.A. 1996. “Growth of micro and small enterprises in Southern Africa” Journal
of Development Economics 48, 253-277.
Meeusen, W. and J. van den Broeck. 1977. “Efficiency estimation from Cobb-Douglas
productions with composed errors” International Economic Review 18, 435-444
Meesters, L.J. 2009. Efficiency of Financial Institutions: A stochastic frontier analysis
approach, PhD Thesis, University of Groningen, The Netherlands
Melitz, M. J. 2003. .The Impact of Trade on Intra-Industry Reallocations and Aggregate Industry Productivity..Econometrica 71(6):1695.1725.
Mok, V., G. Yeung, Z. Han and Z. Li (2010) “Export Orientation and Technical Efficiency:
Clothing Firms in China” Managerial and Decision Economics, Volume 31(7), pp.
453-463
National Institute of Statistics (2009) Cameroon
Ndulu, B.J. and S.A. O’Connell. 1999. “Governance and Growth in Sub-Saharan African”
Journal of Economic Perspectives, 13(3), pp. 41-66.
Nelson, R. and G. Winter. 1982. An Evolutionary Theory of Economic Change Cambridge,
Mass.: Harvard University Press.
Nikaido, Y., 2004, “Technical Efficiency of Small-Scale Industry: Application of Stochastic
Production Frontier Model” Economic and Political Weekly, Vol. 39
Niringiye A, and E, Luvanda. 2010, 'Determinants of Export Participation in East African
Manufacturing Firms', Current Research Journal of Economic Theory, vol. 2, no.
2, pp55-61.
Niringiye, A., E. Luvanda, and J. Shitundu. 2010. “The relationship between Firm size and
Technical Efficiency in East Africa Manufacturing Firms” Journal of Sustainable
Development in Africa 12(4), 226-236
Njikam, O., E. Bamou and C. N’donga. 2008. The case of the Manufacturing Sector in
Cameroon, An AERC Publication on African imperatives in the New World Trade
order, Vol. 2, Case studies of Manufacturing and Services
Njikam, O and J. Cockburn. 2007. “Trade liberalization and productivity growth: Firm-level
evidence from Cameroon.” PEP research paper, Canada.
195
Njikam, O. 2003. “Trade reforms and efficiency in Cameroon’s Manufacturing Industries”
AERC Research Paper 133
OECD. 2011. Cameroon. Tunis: AfDB/OECD
Oczkowski, E., and K. Sharma. 2005. “Determinants of efficiency in Least Developed
Countries: Further evidence from Napelese manufacturing firms” Journal of
Development Studies, 41(4), 617 – 630.
Olson, J.A., P. Schmidt and D.M. Waldman. 1980. “A montecarlo study of estimators of
stochastic frontier production functions,” Journal of Econometrics 13, 67-82
Orea, L. and S.C. Kumbhakar. 2004. “Efficiency Measurement Using a Latent Class
Stochastic Frontier Model” Empirical Economics, 29, p. 169-183
Papadogonas, T., F. Voulgaris and G. Agiomeirgianakis. 2007. “Determinants of Export
Behavior in the Greek Manufacturing Sector” Operational Research International
Journal Vol. 7 No.1, pp 121-135.
Pack, H. 1984. ‘Productivity and Technical Choice: an Application to the Textile Industry’,
Journal of Development Economics, Vol.16, pp.153–76
Papke, L. E. and J. M. Wooldridge (1996) “Econometric Methods for Fractional Response
Variables with an Application to 401(K) Plan participation Rates. Journal of
Applied Econometrics 11(6): 619-632.
Pusnik, K. 2010. “From Technical and Cost Efficiency to Exporting: Firm level data from
Slovenia” Economic and Business Review, 12(1), 1-28
Reinhard, S, Lovell CAK, Thijssen G. 2002. “Analysis of Environmental Efficiency
Variation” American Journal of Agricultural Economics, 84, 1054 – 1065
Roberts, M. J & Tybout, J. R, 1997. "The Decision to Export in Colombia: An Empirical
Model of Entry with Sunk Costs," American Economic Review
Roudaut, N. and A. Vanhems. 2011. “Explaining firms efficiency in the Ivorian
manufacturing sector: a robust nonparametric approach” Journal of Productivity
Analysis
Richmond, J. 1974. “Estimating the Efficiency of Production”, International Economic
Review 15: 515-521
196
Ritter, C. and L. Simar, L. 1997. “Pitfalls of normal-gamma stochastic frontier models”
Journal of productivity Analysis 8, 168-182
Schenk, R. 2004. “What is economic efficiency” Journal of Economic Education, Cyber
Economics, 34(2): 191.
Schmidt, P. 1986. “Frontier production functions.” Econometric Reviews 4: 289-355
Schmidt, P. and C. Sickles C. 1984. “Production frontiers and panel data” Journal of
Business and Economic Statistics 2: 367-374
Shepard, R. W. 1953. Cost and Production Functions, Princeton, N.J.: Princeton Uni Press
Soderling, L. 1999. “Structural policies for international competiveness in manufacturing:
the case of Cameroon”, Working paper No. 146, OECD Development Centre.
Tambi, E.N. 1984.Agricultural development policy and performance in Cameroon, 1960-
1980', PhD thesis, University of Pittsburgh, USA
Taymaz, E. and G. Saatci, 1997. “Technical change and efficiency in Turkish manufacturing
Industries” Journal of Productivity Analysis 8, 461-475
Tybout, J. R. 2000. “Manufacturing firms in developing countries: How well do they do, and
why?”Journal of Economic Literature, 38, pp. 11-44.
Tybout, J. R. 2003. Plant- and Firm-Level Evidence on "New Trade Theories". In Handbook
of International Trade, ed. E. K. Choi and J. Harrigan. Oxford: Basil Blackwell
chapter 13.
Varian, H. 1992.Microeconomic Analysis, 3th Edition, New York: W.W. Norton &Inc
Wagner, J. 2001. A note on the Firm Size-Export Relationship. Small Business Economics
17(4): 229 – 237
Wagner, J. 2007. “Export and firm characteristics: first evidence from fractional probit panel
estimates,” University of Lunebury working Paper Series in Economics, No 97,
http://hdl.handle.net/10419/28215
Wakelin, K. 1998. “Innovation and export behavior at the firm level” Research Policy, 26(7-
8), 829-841.
Wei Z, Varela O, Hassan MK. 2002. “Ownership and performance in Chinese manufacturing
industry.” Journal of Multinational Financial Management 12: 61-78.
197
World Bank. 2012. The Joint World Bank-IMF Debt Sustainability Report For Low Income
Countries. Washington DC:
World Bank. 2007. The Joint World Bank-IMF Debt Sustainability Report For Low Income
Countries. Washington DC:
Yang, J., and D. Nyberg, (2009). External Debt Sustainability in HIPC Completion Point
Countries: An Update. Washington DC: International Monetary Fund.
Zhu, S., P. Ellinger and C.R. Shumway. 1980. “The choice of Functional Form in Banking
Inefficiency” Applied Economics Letters, 2, 375-379
198
APPENDICES
Appendix 1: Derivation of the Maximum Likelihood functions for half-normal and
exponential model
a) The half-normal model
The component of iu is assumed to be positive representing production inefficiency. Most
often iu is assumed to be half-sided normal: ),0( 2ui Niidu .
The density of u is given as:
2
2
2exp
2
2)(
uu
uuf
With the moments
uuE
2)( and 22
)( uuV
The ML approach for the half-normal model
Assuming independence of the error terms ,uandv the joint density function results as the
product of individual density functions:
2
2
2
2
22exp
2
2)().(),(
vuvu
vuvfufvuf
To obtain the density of the composed error term ,uv the joint density ),( uf is first
obtained. Integration over u results in
)(1)(2
,)( 11
0
duuff
199
Where 222uv and vu
The density distribution of uv is asymmetric and characterized by
uuEuvEuE
2)()()(
The variance of ,uv is given by
222 2)()()( vuvVuVV
The log-likelihood function is given by
n
i
n
iiiInInnInnLIn
1 1
2
2
12
2
1)(1
12),|(
using .loglog iii xy
Having obtained the estimates ,, 222vuvu and
the estimates of the
variance components can be recovered:
2
222
2
2
11
uv and
Estimates of individual inefficiencies
Since it is impossible to obtain estimates for iu and iv for each individual firm, the
inefficiency ratio iTE is obtained as the exponential conditional expectation of u given
the composed error term :
iiuE
i eET |ˆ
The conditional density of u , given , is
200
1
2
2
12
)(exp
2
1
)(
),()|(
u
f
ufuf
Hence, the distribution of u conditional on is ),,( N
where
2
2u
and )1()( 2
22
2222
2
222
uuvu
where ,22 u is the fraction of the variance of the inefficiency to the total variance.
Having obtained the distribution of ,|u the expected value )|( uE can be used as point
estimator for iu (Jondrowet al., 1982):16
)(
)(
1)|(ˆ
2i
iii
z
zzuEu
where
iiz
16Instead of obtaining firm-specific efficiencies from ,)|(exp uE Battese and Coelli (1988) propose the
alternative estimator:
ii
i
iii uu
u
uEET2
exp|)exp(ˆ2
where
Where 22)(log uii xyu
and ,2222 uv noting in general
.|)exp()|(exp iiuEuE Furthermore, both estimators are unbiased, but inconsistent estimators
because 0)ˆ( iuVar for .N
201
a) The Exponential Model
The component iu is assumed to follow the exponential distribution with density given in
alternate parameterization u 1 as
00
0exp1
)(
u
uu
ufuu
The moments are
2uV(u))( anduE u
The Maximum Likelihood approach for the exponential model
Assuming independence of the error terms ,uandv the joint density distribution is the
product of individual density functions
2
2
2exp
2
2)()(),(
vuvu
vuvfufvuf
To obtain the density of the composed error term ,uv the joint density ),( uf is first
obtained and integrating out u from it as
2
2
0 2
1exp
1),()(
u
v
uu
v
vu
duuff
The density distribution of is asymmetric (Behr and Tente, 2008) and characterized by
uuEuvEE )()()(
The Variance of is given by
222 )()()( vuvVuVV
202
Assuming independence across subject ,i the likelihood is the product of individual
densities :)(f
n
i u
i
u
v
v
i
n
u
v
nu
vuyL1
2
222 exp
2
1exp
1),|(log
The log-likelihood is given by
n
i u
ii
u
v
v
ii
u
vuu
xyxy
nnyInL
1
2
22
logloglogloglog
2
1)log(,,|log
Estimates of Individual inefficiencies using exponential model
Given that the conditional distribution )|( uf is distributed as ),~( 2vN and given by
vv
u
f
ufuf
~2
2~exp
)(
),()|(
22
With u
v
2~
According to Kumbhakar and Lovell (2003) the expected value of inefficiency ,u given
estimated residual , in the normal-exponential model can be taken as:
vi
viviiiuE
~
/~~|
The derivatives of the log-likelihood functions for the half-normal and exponential
distributions are shown in appendix 2.
203
Appendix 2: Derivatives of the ML for half-normal and Exponential Models
1) Half-normal Model
The log-likelihood function is expressed as:
n
iii
n
ii xyxyInnInnInyInL1
2
211
112
2
1][1
2),,|(
The derivatives are given by
i
n
i i
ii
n
iii x
Fxxy
nInL
112 1
ii
n
i i
in
iii xyxy
n
In
InL
13
2
1422 12
1
2
1
2
ii
n
i i
i xyInL
14 12
11
Where
1 iii xIny
1 xIny ii
2) Exponential model
The log-likelihood function is given as:
n
i u
vii
v
n
iii
u
n
i u
vii
vu
vuvu
xyInd
d
xyxyInnnInyInL
1
112
222
1
11
2
1)(,,|
n
iui
n
i
u
vii
v
u
vii
v
vi
x
xy
xyxInL
11 1
1
n
iii
u
n
i
u
vii
v
u
u
vii
v
v
u
v
uu
xy
xy
xy
nnInL
12
1 23
2 1
1
1
204
n
i
u
vii
v
u
vii
v
u
ii
vu
v
v xy
xy
xynIn
InL
122
1
1
11
Appendix 3: Doing Business in Cameroon
Doing business
Rank 2010
Rank 2009
change
Doing Business 171 167 -4
Starting a Business 174 174 0
Dealing with Construction Permits 164 154 -10
Employing Workers 126 124 -2
Registering Property 143 142 -1
Getting Credit 135 132 -4
Protecting Investors 119 114 -5
Paying Taxes 170 172 +2
Trading Across Borders 149 147 -2
Enforcing Contracts 174 173 -1
Closing a Business 98 98 0
The statistics show that Cameroon is doing poorly as far as starting a business, employing
workers, getting credits, protecting investors; paying taxes and trading across borders are
concerned. The only variable which is ranked below 100 is closing a business. This may
explain why some of the firms and businesses that were captured in the RPED data until
2002, were no longer in existence during the 2009 survey.