Bank of Uganda

Working Paper Series

Working Paper No. 01/2016

Private and Public Investments and Economic Growth in

Uganda: A Multiplier Model Analysis

Thomas Bwire¶1, Wilson Asiimweظ, Grace, A. Tinyinondi¶2, Priscilla Kisakyeظ, and Denise,

M. Ayume¶1

Ministry of Finance, Planning and Economic Developmentظ ¶1Research Department, Bank of Uganda ¶2Statistics Department, Bank of Uganda

July 2016

Working papers describe on-going research by the author(s) and are published to elicit comments and to further debate. The views expressed in the working paper series are those of the author(s) and do not in any way represent the official position of the Bank of Uganda. This paper should not therefore be reported as representing the views of the Bank of Uganda or its management

Bank of Uganda WP No. 01/2016

\

Private and Public Investments and Economic Growth in Uganda

A Multiplier Model Analysis

Prepared by

Thomas Bwire¶1, Wilson Asiimweظ, Grace, A. Tinyinondi¶2, Priscilla Kisakyeظ, and Denise, M. Ayume¶1

Ministry of Finance, Planning and Economic Developmentظ ¶1Research Department, Bank of Uganda ¶2Statistics Department, Bank of Uganda

July 2016

Abstract

This paper applies a multiplier model on Uganda’s 2009/10 Social Accounting Matrix

(SAM) to investigate the impact on growth of private and public investments. A positive

impact of general investment on growth is found, but with varying degree of sectoral

intensity. Infrastructural investments depict the highest income multiplier effect on

growth than the general industry and services sectors, while excess capacity exist in the

agricultural sector and the labor market and returns to capital from each unit of

investment more than double the returns to labor.

Capital is increasingly driving growth, but at the expense of the agriculture sector and

labor market. It appears necessary that capital utilization needs to be complemented with

strategies to utilize the excess capacity in agriculture and the labor market, if growth is to

be all inclusive. One way to address the labor market challenge is to use policy to steer

the productive system towards adopting capital-led labor intensive technologies

simultaneously with increasing government infrastructural investment.

JEL Classification: C02, C22, C67, E22, H30, J20

KEYWORDS: Public investment; Private sector investment, GDP, Cointegration, Multiplier model,

Social Accounting Matrix and Economic growth

To cite this article,

Bwire, T., Asiimw, W., Tinyinondi, A. G., Kisakye, P. and Ayume, M.D. (2016), “Private and Public

Investments and Economic Growth in Uganda: A Multiplier Model Analysis”, Bank of Uganda

Working Paper No. 01/2016.

Correspondence Address: Research Department, Bank of Uganda, P.O. Box 7120, Kampala, Tel. 256 414 230791, Fax. 256 414 230791. Authors E-Mail addresses: tbwire@bou.or.ug; Wilson.Asiimwe@finance.go.ug, GAinomugisha@bou.or.ug; priscilla.kisakye@finance.go.ug and damugyenyi@bou.or.ug in the order above. The

views as expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the Bank of Uganda

and the Ministry of Finance, Planning and Economic Development of Uganda and its affiliated organizations.

1 | P a g e

1.0 Introduction

In the past decade, Uganda’s economic growth has averaged 6.7 percent per annum, though this trend

deteriorated in recent years to an annual average of 4.3 percent in the past four years. Amidst

retarding growth, the annual growth in private sector credit (PSC) continues to be robust and

improved from an annual average of 7 percent in the past decade to 13.1 percent between FY 2011/12

and 2014/15. Credit offered by commercial banks to the personal and household sector has declined,

while increasing credit is offered to the agriculture, trade and real estate sectors. Government has

recently intensified investment in infrastructural developments in a bid to boost economic growth.

The share of development expenditure in Uganda’s budget has steadily increased from an average of

4.3 percent of GDP during FY 2002/03 to 2007/08 to 7.6 percent of GDP in FY 2008/09 to 2014/15,

with the bulk of it directed to the energy and transport sectors.

Government has also, together with the central bank, implemented reforms in the financial sector

aimed to foster rapid expansion of the banking system and improve the business and macroeconomic

environment for private sector investment. Despite these developments, real GDP growth has

remained far below the historical trend levels of 7 percent. Export growth also declined during FY

2012/13-2014/15 to an average of about 13 percent of GDP from about 15 percent in the three years

to 2012/13, despite Uganda’s membership to many regional economic blocks and export promotion

zones, weighing down further on economic growth.

It might be the case that the weakening response of growth to private and government investments is

indicative of weak inter-linkages within the income-multiplicative process, i.e. the general

equilibrium interactions of the economic blocks and its capacity to multiplicatively generate incomes

from a unit of investment has slowed down. The paper therefore investigates the amount of national

income generated from each unit of investment in each sector. The paper uses a multiplier model

analysis, which is complemented with suitable time series techniques to investigate the impact of

public and private investments on economic growth and to establish a comparative contribution of

public infrastructural and private investments to economic growth. It also assesses the existing

strategies to revive economic growth to its long-run potential level.

The rest of the paper is organized as follows. Section 2 provides a review of the relevant literature

while Section 3 explains the methodology. Empirical results are given in section 4 while section 5

provides a conclusion.

2 | P a g e

2.0 Literature Review

There is growing interest in analysis of the relative effects of public and private investment on

growth. The recent literature on long-run growth and convergence in real per capita incomes across

countries emphasizes the role of aggregate investment (Khan and Kumar, 1997; Warner 2014, among

others). In the case that public and private investments have different impacts on growth, it bears

significant implications for the determination of the steady-state growth path (Khan and Kumar,

1997). While studying a sample of 95 developing countries for the period 1970 to 1990, Khan and

Kumar use a suite of cross sectional and time series pooled data estimation techniques and establish a

substantial difference in the impact of private and public sector investment on growth. Private

investment was found to bear a much larger impact than public investment especially during the

1980s. In related studies on sub-Saharan Africa, Tokunbo and Lloyd (2010) find a positive

relationship between foreign private investments; domestic investment and net-export growth and

economic growth in Nigeria. In Uganda, Adam et al. (2016) show that public investments in

infrastructure is key for sustainable private sector growth.

Warner (2014) studies whether public investment could drive economic growth on a sample of 124

lower and middle income countries during the period 1960 to 2011. He shows that the public

investment boom variable had a consistent and negative association with private investment rates,

while financial depth was associated with higher private investment rates than with overall growth.

Inflation and export growth showed little association with private investment. However, these

findings run counter to the assumption in models that use Cobb-Douglas specifications that public

capital will stimulate private capital formation, and also undermines a key selling point of public

investment booms – that they will stimulate private investment and thus have a magnified impact on

the economy.

Notwithstanding the positive impact on growth of investment in public goods, elsewhere growth has

been found to be a reducing function of public investment in infrastructure. In South Africa, Fedderke

and Garlick (2008) show that relative to Sub-Saharan standards, transport infrastructure quality is

high, but the quality and accessibility of the infrastructure varies noticeably between the urban and

rural areas. This suggests that the impact on growth of infrastructure investment could be negligible,

though not strictly negative.

3 | P a g e

Using the vector autoregressive (VAR) approach, Ghani and Din (2006) investigate the relationship

between public investment and economic growth in Pakistan over the period 1973 to 2004. They

show that public investment crowds out private investment, which is surprising because public

investment should be complementary to private investment. They also uncover both ways causality

between private investment, public consumption and economic growth and one way causality running

from public investment to growth. The results also showed that growth was largely driven by private

investment, with negative but insignificant impact from public investment.

Llop Llop (2005) uses the Social Accounting Matrix (SAM) model to decompose the endogenous,

exogenous and leakages and assess the factors that contribute to the changes in the SAM multipliers

for the Catalan economy. They used two SAMs, for 1990 and 1994. The SAM multiplier model was

extended by decomposing the multipliers using both the additive and the multiplicative

decomposition approaches. In this paper, the endogenous accounts are categorized as input-output

relations, private consumption, factors of production, factorial income distribution and the interactive

term. The paper established that changes in the multipliers are mainly driven by the structural

coefficients.

Pyatt and Round (1979) used the SAM to develop a dis-integrated multiplier model. The SAM is first

used to deduce the structure of the economy at the base date. The multipliers are decomposed into the

three matrices including; one that captures the transfer effects, second the circular effects and lastly

that captures the cross-effects of the multiplier process, where injections into the economic system

has impacts on the rest of other sectors or economic agents. The analytical work in the paper shows

that decomposition of the structure of the economy can be directly derived from the accounting

balances. The paper captures the fixed price aspects which represent a single step beyond the

structure of the accounting balances.

Breisinger et al. (2009) discuss both the theoretical and practical application of the SAM-based

analysis with a number of exercises for scholars. The paper discusses the economic linkages and

multiplier effects in the SAM framework. The paper shows that for every shock, there are both the

direct and indirect effects. The indirect effects are in the form of consumption and production

linkages. The production linkages are categorized into backward and forward linkages. All these

effects are categorized into multiplier effects of a unit of exogenous investment. In this paper, the

Multiplier model is further decomposed into constrained multipliers and unconstrained multipliers.

Unconstrained multipliers assume prices to be fixed and that any changes in demand will lead to

4 | P a g e

changes in physical output rather than prices. However, this would require an additional assumption

that the economy‘s factors resources are unlimited or unconstrained so that any changes in demand

can be matched by an increase in supply. In the constrained multipliers, the supply responses are

unconstrained by fixing the level of output in certain areas. This approach generates the SAM based

multiplier formula of the constrained multiplier matrix.

Punt et al. (2003) use the SAM to design the SAM-based multiplier model. The authors postulate that

the SAM alone does not constitute an economic model though it can be used to present economic

theory. The paper shows that the SAM approach to modelling is a useful way of showing how the

model statement (price and demand system, behavioral equations and macroeconomic

balances/closures) of a macroeconomic or general equilibrium model can be developed. The paper

adds that the SAM approach of modelling clearly shows that the various properties of a SAM,

particularly the fact that the account row and column total are balanced, are very important when a

SAM is used as the database for a general equilibrium model.

This paper applies a multiplier model on the 2009/10 SAM for Uganda to investigate the impact on

growth due to public and private investments and analyzes the amount of national income generated

from each unit of investment in each sector.

3.0 The Multiplier Model

3.1 The Model Set up

A multiplier model analysis is applied on Uganda’s 2009/10 Social Accounting Matrix (SAM) to

assess the contribution of private and public infrastructural investments to economic growth by sector

and economic agent. To facilitate easy reference, the paper categorizes structural coefficients building

the SAM into policy instrument variables (exogenous), policy target variables (endogenous) and

leakage variables (earnings to government and the rest of the world, ROW- hereafter). This

disintegration of the income creation (multiplicative) process or the structural multipliers will capture

the contributions of sectoral private sector investments to economic growth. The multiplicative

process will quantify the amount of income (direct and indirect) created from each unit of investment

per sector. The direct and indirect effects of private sector investment on growth, is further

categorized into backward and forward linkages of each unit of growth. This approach will also

provide detailed information of the indirect effects on growth, that is, income created by a given

5 | P a g e

sector due to private investment in other sectors. The private investment is also categorized by its

origin that is domestically sourced investments and foreign or externally sourced investments. We

shall use the same method to capture the contribution of other sources of investments (like

government) and shocks to economic growth. The application of this multiplier model on the Social

Accounting Matrix (SAM) will reveal the detailed contributions of each sector’s private investment to

economic growth.

An exogenous expenditure shock, e.g. an increase in government demand for construction services is

associated with both direct and indirect effects (Breisinger et al., 2010). The effect on the

construction sector itself is the direct effect, while that on other sectors, other than construction is the

indirect effect. The indirect linkages encompass production and consumption linkages. The

production linkages are a series of agglomeration of backward and forward linkages. A combination

of direct and indirect effects is what constitutes the multiplier effects. This SAM multiplier analysis

model used in this paper follows the approaches proposed by Llop Llop (2005), Round (2006), Pyatt

and Round (1979), Punt et al., (2003), Bellù (2012), Stone R. (1978) and Breisinger et al., (2010).

3.2 Structural analysis

Using the Row-Column framework of the SAM, we denote payments made by columns as j and

receipts received (rows) as i. As in Asiimwe and Kisakye (2015), the SAM matrix is then denoted Zij

with i rows and j columns. However, for economic and statistical consistency, payments and receipts

are equal per sector (matrix element) and the row-column totals are equal. For instance the row and

column equlity of ROW) account is expressed as;

n

jjROW

n

iROWi zz

1,

1,

(1)

Embedding this in the SAM, we derive the technical coefficients, while differentiating the policy

objectives from policy instruments and leakages. The technical coefficients are derived as:

j

ij

ijz

z

(2)

6 | P a g e

Where ijZ is the individual SAM element and jz is the column totals. We then categorized the

corresponding matrix of technical coefficients into endogenous accounts (policy objectives), ij ;

exogenous accounts (policy instruments or shocks), i and leakages, jl in matrix ijZ . The resultant

ijZ matrix is:

n

n

n

nn

n

n

i

n

m

m

m

mm

m

mmmm

nm

nm

m

mm

m

mm

ij

z

z

z

ll

z

zi

z

z

z

z

z

z

z

z

z

z

z

ll

z

ll

z

ll

z

z

z

z

z

z

z

z

z

z

z

z

z

z

z

z

z

z

Z

1

...

1.....11

.....

.....

............

.....

.....

1

2

11

2

2

1

1

2

22

1

11

2

22

1

11

22

2

2222

1

2121

11

2

1212

1

1111

The total receipts of each economic account say row account in the SAM is a summation of the

endogenous and exogenous expenditure denoted as:

n

j

ijiji zz1

(3)

Where jz is the row total of the SAM, i is a vector of exogenous variables and the rest is the

multiplicity of technical coefficients and the column totals. The right hand side of this equation

should be able to reproduce the original SAM with equality with the sum of the columns and rows.

Satisfying this condition, the technical coefficients can now be used for deriving economic

multipliers using the multiplier model.

3.3 The Multiplier Model Analysis

The multiplier model uses the above technical coefficients ( ij ) to generate economic multipliers

which then define the amount of growth (or income) derived from each unit of investment per sector.

Given that the row and column totals are equal, (3) becomes:

7 | P a g e

)...3,2,1(,

1

mji

n

j

ijiji zz

(4)

Eqn. (4) represents a series of simultaneous equations that define the SAM. The endogeneity across

these equations is captured by their respective technical coefficients denoted by ij . The nominal

flow of income from one economic agent to another, denoted by jz must equate to the receipts

received by each of the given economic agent iz . Dropping subscripts from (4) for ease, we derived

the following simultaneous equation:

zz (5)

Where z represents income and expenditures flows, represents the coefficient matrix and is the

shock component of the SAM. Collecting like terms we derive the following multiplier model.

1)(Iz (6)

The matrix of multipliers ( ) is used to derive the impact of sectoral investment on growth. It’s these

multipliers that are used to derive the backward and forward linkages resulting from each unit of

investment by sector. The sectoral investment that creates more income is considered to contribute

more to the economic system and thus to growth (Roberts, 1998). Thus the sector with the highest

multipliers is the one that can rejuvenate growth prospects in the medium term. The multiplier effect

can be decomposed into direct and indirect effects. Following, Alarcó et al., (2011) the indirect

effects are categorized as backward and forward linkages through the respective column and row

summations. If we define the elements of the multiplier matrix ( ) as ji , , the sector receiving the

investment shock as jsis, , and excluding the multiplier element of the sector receiving investment, the

remaining sectors becomeji~

,~ . The multiplier elements can be expressed as:

jijsisji ~,

~,, (7)

In the equation the direct effect of investment is captured by jsis, whereas the indirect effect is

captured byji~

,~ . It is then possible to decompose the indirect effect into backward and forward

linkages and it takes the form:

8 | P a g e

isn

ijijBW

~~

,~ (8)

and

jsn

jjiiFW

~~

,~ (9)

Where jBW and jFW are the backward and forward linkages, respectively. Thus sectors with strong

backward and forward linkages are those that stimulate growth faster than any other (Breisinger et

al., 2010; Jorge et al., 2011).

3.4 The Data

The paper uses the 2009/10 SAM alongside its corresponding Supply and Use Table (SUT) for the

same year to capture the contribution to growth emanating from a unit of general investment in each

of the sectors and disintegrate the multiplier effect of general investment on growth. The theoretical

structure of the SAM is shown in Appendix 2. The Ministry of Finance, Planning and Economic

Development (MoFPED) in collaboration with the Uganda Bureau of Statistics (UBOS) finalized the

construction of the 2009/10 SAM that replaces the 2002 SAM. It is the corresponding Supply and

Use Table (2009/10) that was used as a basis for rebasing GDP to the FY 2009/10. The economic

sectors in the 2009/10 SAM were increased to 161 compared to 61 sectors in the 2002 SAM, but

these are merged in this study to focus the analysis to addressing the set out objectives. The Supply

and Use Table (SUT) and the SAM comprise a summary table of payments and receipts between

economic agents in a specified period of time usually a year (Stuttard and Frogner, 2003). In this

paper, the 2009/10 Uganda’s SAM was mapped to thirty two (32) accounts as shown in Appendix 3.

As a robust check, a Vector Error Correction Model (VECM) based on quarterly time series data for

the period 1997Q1-2015Q3 is estimated. The VECM consists of five variables: an indicator of output

growth ( tGDP ), development expenditure (Devtexp), exports (Xt), imports (Mt) and private sector

credit (PSCt). GDP is measured in billions of Uganda Shillings in constant 2009/10 prices. Devtexp is

used to proxy public sector investment while PSC, a measure of financial institutions contribution to

the private sector is used to proxy private sector investment. In Uganda, commercial banks hold about

80 percent of the total assets of the financial system and as a result, most private sector firms seeking

finance for investment have to rely on loan finance, for which the most important source is the

banking system.

9 | P a g e

As a share of GDP, growth in PSC has been improving steadily from an annual average of 7 percent

in the past decade to 13.1 percent in the most recent four years. PSC may be an inadequate measure

of private fixed capital formation, but the available aggregate data on private sector investment dates

only to 2009 and on annual basis. We therefore acknowledge the use of PSC as an appropriate price

to pay, despite the fact that it is a significant under estimate of private sector investment. Quarterly

GDP data is from UBOS, Devtexp data is from MoFPED, while data on exports, imports and PSC is

from BoU. All data are in logarithms and are given in levels and in first differences in Figure 1. The

series are also adjusted for seasonality effects, but only where it has been detected.

Figure 1: Series in level and first difference

-800

-400

0

400

800

98 00 02 04 06 08 10 12 14

DEVTEXP_DIFF

4.5

5.0

5.5

6.0

6.5

7.0

7.5

98 00 02 04 06 08 10 12 14

LDEVTEXP

-200

-100

0

100

200

300

400

98 00 02 04 06 08 10 12 14

EXPORTS_DIFF

4

5

6

7

8

98 00 02 04 06 08 10 12 14

LEXPORTS

-400

-200

0

200

400

600

800

98 00 02 04 06 08 10 12 14

IMPORTS_DIFF

5

6

7

8

9

98 00 02 04 06 08 10 12 14

LIMPORTS

-200

0

200

400

600

800

98 00 02 04 06 08 10 12 14

PSC_DIFF

5

6

7

8

9

10

98 00 02 04 06 08 10 12 14

LPSC

-2,000

0

2,000

4,000

6,000

98 00 02 04 06 08 10 12 14

QGDP_DIFF

7.2

7.6

8.0

8.4

8.8

9.2

9.6

98 00 02 04 06 08 10 12 14

LQGDP

Source: Uganda Bureau of Statistics, UBOS, Ministry of Finance Planning and Economic Development,

MoFPED and Bank of Uganda, BoU

The statistical description of the variables used in the VECM is presented in Table 1. The small

coefficient of variation (at most a sixth of the total sample) suggests small dispersion of the data

points from the mean for all the series. The Jarque-Bera (J-B) statistics (Mukherjee, White and

Wuyts, 1998) suggests normal distribution is rejected for all the variables. Nonetheless, as these are

to be analysed in a system framework of the VAR, estimates of a VAR model are robust to deviations

10 | P a g e

from normality, providing system residuals are not auto correlated (Juselius, 2006), as is indeed the

case as we show later.

Table 1: Descriptive Statistics

LQGDP LEXPORTS LDEVTEXP LIMPORTS LPSC

Mean 8.502 6.279 5.895 6.966 7.584 Std. Dev. 0.799 0.956 0.713 0.923 1.096 Coefficient of variation (%) 9.397 15.226 12.093 13.249 14.457 Skewness -0.135 -0.037 0.488 -0.062 0.131 Kurtosis 1.528 1.446 2.012 1.586 1.599 Jarque-Bera 6.813 7.364 5.863 6.132 6.177

Probability 0.033 0.025 0.053 0.047 0.046

Observations 73 73 73 73 73

Notes: LQGDP is the log of quarterly GDP; LEXPORTS is the log of exports; LDEVTEXP is the log of development expenditure;

LIMPORTS is the log of imports and LPSC is the log of Private Sector Credit.

4.0 Empirical Results

The results are generated from the 2009/10 SAM using a multiplier model. The SAM was

compressed to 28 sectors so as to focus the analysis on addressing the contribution of private and

public investment on growth. In this analysis, we presume that private firms receive income from

sales and subsidies from government. The expenditure of producing firms is composed of

intermediate consumption, wages and salaries, payment for use of capital, taxes on activity and

depreciation of capital. Thus the impact of investment on GDP growth mainly follows the input-

output structure embedded in the 2009/10 SAM.

4.1 General Contribution of sector investment to growth

As noted in sections 2.0 and 3.1 of this paper, the impact of each unit of investment on growth is

categorized into direct and indirect effect, for instance an investment in a given sector, increases both

income in the direct beneficiary sector as well as incomes of other sectors interlinked to it through

backward and forward mechanisms as is shown in Figure 2.

11 | P a g e

Figure 2: Sectoral direct and indirect impacts

Source: Author’s computation

It is evident in Figure 1 that investment in all sectors generates income over and above the initial

value of investment, since all the respective multipliers are above unity (total impact). The industrial

sector possesses the highest direct multipliers (1.59) and also indirect (0.9) impacts on growth. The

service sector is second to the industrial sector in steering growth directly (1.51) and indirectly (0.74),

while agricultural sector has the least direct multiplier (1.05) and indirect (0.5) impact. This implies a

unit of investment in the industrial sector generates more income and yields more economic growth

than does a unit of investment in the service and agricultural sectors. The direct multiplier and

indirect impact due to industrial sector on growth that we detect from Figure 1 is however too general

an impact and as is, is difficult to tell the contribution of private and public investment. So in what

follows, we decompose growth impact of industrial investment by source that is ‘private’ and

‘public’.

4.2 Public and private investment contribution to economic growth

The key drivers of economic growth are those economic agents with the highest multipliers. Sectors

with the highest backward linkages possess capacity to propel stronger economy wide growth, mainly

because they are inter-linked to other sectors in their production process through their demand of raw

materials. Conversely, sectors with the higher forward linkages generate more market potentials from

each unit of investment. Thus, investment in sectors with the highest backward and forward linkages

would boost growth back to its potential level. The breakdown of multipliers into public and private

investments is presented in Table 2, whereof the columns are sectors that receive the investment

whereas the rows are endogenous sectors that benefit from the income multiplicative process.

12 | P a g e

Table 2: Comparative impact of private and public investment 2009/10 SUT/SAM Multipliers Public Goods &Infrastructure Investments

Agriculture Industry Services Govt_InvestPublicGds Prv’t_InvestPublicGds

Agriculture 1.05 0.21 0.15 0.08 0.09

Industry 0.20 1.59 0.59 1.45 1.60

Services 0.31 0.70 1.51 1.21 1.01

Total sectoral income 1.56 *2.50 2.25 *2.73 2.70

Memo:

Backward linkages 0.51 *0.90 0.74

Forward linkages 0.35 0.80 **1.01

* Highest level of contribution

Source: Author’s computations.

As can be seen, general investment in industrial sector drives growth faster (2.5), followed by the

services sector (2.25) and then the agricultural sector (1.56). The industrial sector has strong

backward linkages (0.9) compared to services (0.74) and agriculture (0.51). However, the industrial

backward linkages are stronger with the service sector (0.7) compared to its linkages with the

agricultural sector (0.21). This shows that economic growth is mainly driven by the industry and

service sectors given their strong input-output inter-linkages.

Investment in the agricultural sector currently yields less comparative impact on economic growth,

implying, in part, the sector possess high levels of excess capacity. It is probably this dismal

economic viability that makes it less attractive for capital investment, attracting only 0.9 percent of

the Foreign Direct Investment (FDI) in 2013 (Bank of Uganda, 2015).

In addition, government investment in public goods and infrastructure generate more economic

growth (2.73) compared to investment in any other sector, the highest in the category being industry

(2.5). This indicates that the economy is ‘infrastructure hungry’.

4.3 Accelerating Growth

As revealed in the previous section, a significant portion of the current economic growth is

shouldered mainly around government infrastructural investments. This is ‘good news’ for an

infrastructure ‘hungry’ economy like Uganda but could be ‘bad news’ for other sectors, as any

deterioration in a sector’s interlinkage with the rest of the economy is not felt; though it does not add

up to an internal Dutch disease effect. For instance, deterioration of the agriculture sector interlinkage

with other sectors could be compensated by a coincidental increase in infrastructure investments, thus

delimiting potential growth from agriculture. In such circumstances, growth may not reduce in the

short run but hurt the economy through a weakened agricultural sector in the long-run. This generates

13 | P a g e

perpetual excess capacity in the agricultural sector as it’s for the case depicted in the 2009/10 Social

Accounting Matrix (SAM). We now turn to the detailed approach of revamping agricultural sector.

Khan and Kumar (1997) note increasing acknowledgement that public investment in infrastructure

projects may not lead to an increase in growth in some cases. For example, in Latin America in the

1970s and Africa and Asia in the 1980s, which were characterized by several large projects

undertaken of somewhat questionable quality including electric power plants, roads and railways,

built at exorbitant costs and ended up either incomplete or under-utilized. In addition, these projects,

if financed through increased taxes, could drive up costs of inputs and lead to an adverse effect on

growth and investment in an economy.

4.3.1 Revamping agricultural contribution to growth

The rejuvenation of agriculture’s contribution to growth is key and investment in the sector is a

necessary but not sufficient condition to maximize growth returns. The sufficient condition is to

improve the backward and forward linkages of the agricultural sector with the rest of the economy

(industry and service sector). Thus, investment in agricultural sector should be complemented with

investments in the sub-sectors of industry and services sector that have strong backward and forward

linkages with the agriculture sector. In the services sector, Hotels and restaurants have strong

backward linkages with agriculture, whereas in the industrial sector the ‘agro-processing’ possess

strong linkages with the agricultural sector as is shown in Table 3.

Table 3: Investment Drivers of agriculture growth

Agriculture sector Agro-Processing

1) Productive sectors Agriculture Cash crops

Other

Agric

Industry

Food_

Processing

Agro_

processing

Agriculture 1.05 1.05 1.05 0.21 **0.49 0.45

Industry 0.20 0.31 0.18 1.59 1.37 1.50

Services 0.31 0.49 0.27 0.70 0.52 0.49

Total sectoral impact 1.56 **1.84 1.51 **2.50 2.38 2.43

2) Factor payments

Labor 0.16 **0.22 0.15 **0.27 0.21 0.20

Capital 0.88 0.83 0.89 0.86 0.86 0.86

Total factorial impact 1.04 **1.05 1.04 **1.13 1.07 1.06

**highest level of contribution

The sub-sectors of agriculture show that the linkage between agricultural sector and the rest of the

economy is mainly driven by the cash-crops production. The total impact of investment in cash crop

production is at 1.84 units compared to 1.51 units of the rest of agriculture. The cash crops agro-

14 | P a g e

processing dominate agro-processing industry thus the respective strong backward and forward

linkages.

Also as shown in the table, investment in non-agro-processing industries generates more economic

growth than investments in agro-processing. However, investment in agro-processing and food

processing within the agro-processing generate stronger backward linkages with agriculture (0.45 and

0.49, respectively) than with general industry (0.21). Since 66 percent of the working population is

employed in agriculture (UBOS, 2010), it’s eminent to diagnose the contribution of these sectoral

investments to employment and income. As shown in Table 3, labor receives more income from

investments in cash-crop production (0.22) compared to that received from investments in the rest of

agricultural sector (0.15). Thus, agricultural sub-sector linked to agro-processing provides stronger

employment benefits compared to the rest of agriculture.

4.3.2 Revamping labor market contribution to economic growth

Private and public investments have consequential impacts on both economic growth and factor

employment. Sectors that generate more factor income from a unit of investment are considered to

have great potential in rejuvenating national income towards its long run potential. Table 4 shows

that sectoral investment generates more income to capital than it does to labor. This information may

not be conclusive to assert whether Uganda is a capital intensive country or the capital cost in

production is extremely higher than the unit labor cost. What we can conclude is that a unit of

investment yields more labor income to the nation if and only if the investment is in the service sector

(0.29). However, this could reflect the notion that the service sector in Uganda is labor intensive, thus

any unit of investment is translated to a larger wage bill to the labor employed.

Next to services is investment in the industry sector, which generates 0.27 units of marginal labor

income and the agriculture sector is the least paying with marginal labor income of 0.16 units. This

probably explains the rampant rural-urban migration and the youth disorientation in agriculture

opting for activities in urban areas which puts pressure on social services provided.

Table 4: Factor multiplicative income

Agricultural sector Industrial sector Services Sector

Factor payments Agric Cash crops

Industry Agro-

Processing

Services

Public

Admin Transport

Labor 0.16 *0.22 *0.27 0.20 0.29 *0.67 *0.40 Capital 0.88 0.83 0.86 0.86 0.86 0.82 0.75

Total impact 1.04 1.05 1.13 1.06 1.15 1.49 1.15

Source: Author’s computations.

15 | P a g e

Table 4 also shows that returns to capital are higher with investment in the agricultural sector (0.88)

but the same for the industrial and service sectors (0.86). In general, returns to capital are more than

twice the wage income resulting from each unit of investment across all sectors. Thus, capital

endowments are crucial in boosting economic growth. This is ‘good news’ for economic growth but

could be harboring ‘bad news’ for the inclusive growth agenda because capital is driving growth at

the expense of agriculture and labor efforts. In addition, the capacity of the economic system to

multiply income depends on the ratio of the return re-invested in the economy to that repatriated

abroad. The portion of these returns that is repatriated abroad acts as a leakage that discounts the

income multiplicative system which incapacitates the ability of the economic system to propel

economic growth further.

This evidence is consistent with the findings contained in the Uganda National Household Survey for

2009/10 (UBOS, 2010). The survey shows that 66 percent of the working population is employed in

agriculture, a sector which receives the least wage return (0.16) from each unit of investment as

expressed in Table 3. The service sector which has the highest wage returns (0.29) from each unit of

investment employs only 28 percent of the total working population (UBOS, 2010). This shows the

magnitude of excess capacity existing in the use of labor resources tied up in the agricultural sector

thus resulting in unemployment of labor as their exist rigidities for it to join the service and industrial

sectors. It is further shown that the literacy rate of persons above 10 years is 73 percent at national

level and 69 percent for the rural areas. This shows that a large share of the excess capacity in labor

market possess skills that can be utilized to boost economic growth.

4.4 Robustness checks

4.4.1 Vector Error Correction Model (VECM) Analysis

The generality of the result for the relative impact of private and public sector investment on growth

is qualified using a vector error correction model which is estimated within the Johansen’s (1988) and

Granger Non causality (Granger 1969) frameworks. This is applied on a 5-dimensional vector

autoregressive model: tx ( tGDP , tDevtexp , tX , tM , tPSC ) where (for year t), GDP is gross

domestic product, Devtexp is development expenditure, X is exports, M is imports and PSC is private

sector credit, covering the period 1997Q1-2015Q3. Details of measurement of these variables, their

time series characterization, including choices of the lag length are discussed in the online Appendix.

16 | P a g e

The presence of one equilibrium (stationary) relation, even when corrected for small sample bias

among the variables at the conventional 5 percent level of significance, could not be rejected, and the

roots of the companion matrix and graphs of the potential cointegrating relations (available on

request) suggested that r = 1seems reasonably well supported by the data. Unless otherwise stated,

the only existing cointegrating relation, as shown in Table 5, is normalized for the quarterly change

of GDP in order to interpret the estimated coefficients. Consistent with economic theory and

estimates of the multiplier model, results of the long-run analysis reveal that DEVTEXP, exports and

PSC have a positive significant long-run correlation with GDP, and in addition, the correlation of

GDP with imports is negative. The estimated coefficients imply, ceteris paribus, over the long-run a 1

percentage point increases in DEVTEXP, EXPORTS and PSC increases GDP by about 1.7, 5.1 and

4.2 percentage points, respectively, while 1 percentage point reduction in IMPORTS stimulates GDP

growth by 7.8 percentage points.

Table 5: Normalized co-integrated equation

LGDP LIMPORTS LPSC LDEVTEXP LEXPORTS

β

1.000

-7.858 (-4.385)

4.258 (8.946)

1.700 (2.678)

5.065 (3.342)

α 0.008 (0.628) 0.039 (4.705) -0.205 (-4.630) 0.001 (0.031) 0.019 (1.716)

In parentheses are p-values

Source: Author’s Computations. Estimates from E-Views

The error correction term is statistically significant with a correct sign in the PSC equation, adjusting

by about 20 percent of the previous quarter’s deviation from equilibrium. This suggests, for the

system variables considered here that once the system is pushed out of the long-run steady state,

equilibrium is reinstated through adjustments in the PSC. It is not surprising therefore that wherever

the government is considering possible means of funding its development expenditure, it is always

cautious not to hurt PSC growth.

Turning to the question of direction of causality, two elements are of interest to us: i) Have measures

of public and private investment collectively influenced GDP growth in Uganda so that causation

runs from investment to Growth?, or ii) it is the robust and sustained GDP growth over the last two

decades that Uganda has achieved that has been an impetus to its observed investment trends. If it

happens to be the case that we fail to reject both (i) and (ii) simultaneously, it may well be case that a

nexus exists in Uganda; where public and private investment are key in accelerating the desired rate

of growth and at the same time, growth is a stimulus to public and private investment. Results of both

pairwise and block exogeneity are shown in Table 6.

17 | P a g e

Table 6: VAR Granger Causality/Block Exogeneity Wald Tests

Macroeconomic indicators

Development Expenditure

Private Sector Activity

GDP

Development Expenditure --- 4.920 (0.178) 12.467 (0.006)

Private Sector Activity 15.889 (0.001) --- 7.889(0.048)

GDP

4.875 (0.181)

4.087 (0.252)

---

ALL 13.778 (0.032) 10.444 (0.107) 15.395 (0.017)

Notes: “---”denotes dependent variable, and p-values in parentheses

Source: Author’s Computations. Estimates from E-Views

Collectively, the null hypothesis that the system variables considered here do not influence growth in

the short-term is rejected at the conventional 5 percent level of significance; so that in general, this set

of variables, over the short-term are growth enhancing. Individually, public development expenditure

has a more pronounced growth enhancing effect in the short-term than does private sector investment.

The result on government development expenditure though consistent with the multiplier model

results, is somewhat surprising because development expenditure is expected to be associated with a

lagged/delayed impact on output growth. In this case however, it reflects in part the multiplier effect

it has on aggregate demand which, over the short-term, may distort or stimulate economy-wide

output. The level of national output does not seem to drive, in the short run, development expenditure

and private sector activity.

5.0 Conclusion and Policy Implications

The paper investigated the impact on growth due to private and public investments in Uganda using a

suite of multiplier model based on the 2009/10 Social Accounting Matrix (SAM) and the vector error

correction (VECM) and Granger non-causality frameworks for the period Q3-1997 to Q3-2015.

Based on the multiplier model, government investment in public goods and infrastructure generate

more economic growth than does private investments in public goods like education, health and

others, implying that government investment is shouldering substantial growth out turns compared to

private investment. This general result is robust to VECM and Granger non-causality evidences

where a steady-state long-run relationship between economic growth, public and private sector

investment, and exports and imports of goods and services is found, and a strong unidirectional

causality running from both public and private investments to growth is uncovered, with public

investment having a more pronounced growth enhancing effect.

Disaggregating the sectoral effect, the paper uncovers excess capacity in both the agricultural sector

and the use of labor and that the current economic growth is shouldered by industry and service

sectors. We find that the industry sector is mainly driven by government infrastructure investments

18 | P a g e

and that it possesses the highest growth multiplicative income effect. This suggests that to maximize

economic growth, investments should concentrate in industrial sector. However, the two key growth

driving sectors in general and the industry sector in particular, have weak backward linkages with

agriculture sector, except for ‘agro-processing’ and ‘hotels and restaurants’.

The excess capacity in the agricultural sector implies that investment in industry is a necessary but

not sufficient condition. The sufficient condition would require utilizing the excess capacity in

agriculture to boost growth. This requires scaling up investments in the agricultural production as

well as improving on the inter-linkages between agriculture and industrial sector through promotion

of agro-processing. The impact on growth is maximized when investments are directed to agricultural

sub-sectors that provide raw materials to the agro-processing industries. In addition, policies to

expedite the inter-linkages between agricultural sector and the rest of the industrial sector need to

focus on investing more in agro-processing with special concentration in food-agro-processing. The

results also carry the implication that the capacity of the industry sector to revive agricultural growth

strongly relies in the agro-processing sub-sector that demand raw materials from the agricultural sub-

sector of interest.

The study also finds that factor income response to each unit of investment is skewed more towards

capital returns than wage returns, and the returns to labor from investment is comparatively weaker in

agriculture sector compared to the service sector. Capital driving growth at the expense of agriculture

and labor efforts is good news, but this may hurt inclusive growth agenda particularly because of the

economic importance of agriculture. Therefore, in addition to promoting agro-processing,

infrastructural development and agro-steered agriculture, the economy needs to also devise a policy to

utilize the excess capacity in labor market. Since less leakages exist in the labor market, a holistic

approach could be followed where a capital-driven labor-intensive technologies are adopted in the

development process as investments in agro-processing is scaled up. For instance in infrastructure

development, as capital is employed, it could be combined with more labor to boost both productivity

and employment of labor as well as increase the capacity of the income multiplicative system to

generate or steer more economic growth.

Overall, accelerating growth will require three simultaneous approaches. First, promoting

industrialization and in particular agro-processing and ensuring a non-declining government

infrastructure investments at least in the short- to the medium-term. This is because government

investments have strongest multiplier effect for each unit of investment and is strongly linked to

19 | P a g e

industrial and service sectors. Agro-processing has strong linkages with the agriculture sector which

employs more than 66 percent of Uganda’s workforce. The second approach is to strengthen the inter-

linkages within the economic system. Currently the inter-linkages in the economy are low especially

between agricultural sector and the rest of the economy. Still a large share of the economy’s

intermediate inputs are imported and input-output relationships in the economy are still strong

between industry and service but weak with agricultural sector with exception of the agro-processing

industry. The third policy approach is to endeavor to rejuvenate the productivity in agriculture by

promoting agro-based demand-driven agriculture output. Revamping agriculture will require a

holistic approach including strengthening its linkages to other sectors as well as focusing on import

substitution of imported raw agricultural materials.

Finally, the paper also finds that a substantial portion of investment in Uganda is sourced from the

rest of the world, thus increasing a potential risk for leakages from the economic system and this is

constraining the multiplicative process. The impact on economic growth, of the leakage-injection of

foreign capital on the multiplicative income should be judged based on the net effect of these

investments against the cost of repatriation. A suitable candidate for extension of this analysis could

be to establish the net-effect of foreign capital and repatriation on economic growth.

References

Adam C., D. Bevan, and T. Ohlenburg (2016). Public Investment and Growth: Implications of

recurrent costs in Uganda – Project Memo, International Growth Centre-IGC-Uganda.

Alarcón, J., C. Ernst, B. Khondker, and P. D. Sharma (2011). "Dynamic Social Accounting Matrix

(DySAM): Concept, Methodology and Simulation Outcomes; The case of Indonesia and

Mozambique", International Labor Organization Employment Working Paper, No. 88.

Asiimwe, W. and P. Kisakye (2015). “Structural Transformation of the Ugandan Economy: A

Multiplier Model Analysis of the 2002 and 2009/10 Social Accounting Matrices”, Ministry of

Finance, Planning and Economic Development Working Paper, No. 2015/04

Bellù, L. G. (2012). "Social Accounting Matrix (SAM) for analysing agricultural and rural

development policies: Conceptual aspects and examples", FAO

Bank of Uganda (2015). Private Sector Investment Survey 2014 - Report. Kampala: Bank of Uganda.

Breisinger, C., T. Marcelle, and J. Thurlow (2009). Social Accounting Matrices and Multiplier

Analysis: An Introduction with Exercises, International Food Policy Research Institute

(IFPRI).

Bwire, T., O. Morrissey, and T. Lloyd (2016). “A time series analysis of the fiscal effects of aid in

Uganda”, Journal of Development Studies (forthcoming)

_______. (2013a), “A time series analysis of the impact of foreign aid on central government’s fiscal

budget in Uganda”, WIDER Working Paper, No. 2013/101

Bwire, T., F. L. Anguyo, and J. Opolot (2013b). “Exchange Rate Pass-Through to Domestic

Prices in Uganda: Evidence from a Structural Vector Auto-Regression (SVAR)",

Journal of Statistical and Econometric Methods 2(4), 41-68.

20 | P a g e

Cruz, K. T. G., J. L. C. Limkin, N.B. Del Castillo, K. K. T. Chua, R. J. E. Van Den Brink, J. E.

Rogier, R. M. N. Galang (2016). “Philippine Economic Update moving full speed ahead:

Accelerating reforms to create more and better Jobs”, World Bank Working Paper, No. 2016/01.

Cheung, Y-W. and K.S. Lai (1993a). “A Fractional Cointegration Analysis of Purchasing Power

Parity”, Journal of Business and Economic Statistics 11, 103-112.

______. (1993b). “Finite-Sample Sizes of Johansen’s Likelihood Ratio Tests for Cointegration’’,

Oxford Bulletin of Economics and Statistics 55, 313-328.

Dickey, D.A. and W. A. Fuller (1979). "Distribution of Estimators for autoregressive time series and

unit roots", Journal of American Statistical Association 74(2), 427-431.

Fedderke J. and R. Garlick (2008). "Infrastructure Development and Economic Growth in

South Africa: A review of the accumulated evidence", School of Economics -University of

Cape Town and Economic Research Southern Africa Policy Paper, No. 12

Ghani, D. and M. Din (2006). "The Impact of Public Investment on Economic Growth in Pakistan",

The Pakistan Development Review 45(1), 87-98

Granger, C.W.J (1969). "Investigating Causal Relationships by Econometric Models and Cross

Spectral Methods", Econometrica 37, 424-438.

Harris, R. and R. Sollis (2005). Applied Time Series Modelling and Forecasting, John Wiley and Sons

Ltd. England

Johansen, S. (1988). "Statistical Analysis of Cointegration Vectors", Journal of Economic Dynamics

and Control 12 (2-3), 231-254.

Juselius, K. (2006). The Cointegrated VAR Model: Methodology and Applications, Advanced Textsin

Econometrics, Oxford University Press, Oxford. Khan, M.S. and M. S. Kumar (1997). "Public and Private Investment and the growth process in Developing

Countries", Oxford Bulletin of Economics and Statistics 59/1.

Llop Llop, M. (2005). "Comparing Multipliers in the Social Accounting Matrix Framework",

Department d'Economia, Universitat Rovira i Virgili. Mukherjee, C., H. White and M. Wuyts (1998). Econometrics and Data Analysis for Developing Countries,

New York: Routledge

Punt, C., K. Pauw, E. Mohube, B. Gilimani, L. Rantho, R. Leaver, S. McDonald, L. Chant, and C.

Valente (2003). "Social Accounting Matrices and Economic Modelling", P. Project,

Elsenburg.

Pyatt, G. and J. I. Round (1979). "Accounting and Fixed Price Multipliers in a Social Accounting

Matrix Framework", The Economic Journal 89 (356), 850-873.

Reimers, H. E. (1992). "Comparison of tests for multivariate cointegration", Statistical Papers 33,

335-359.

Roberts, D. (1998). "Rural-Urban Interdependencies: Analysis Using an Inter-Regional SAM Model",

European Review of Agricultural Economics 25 (4), 506-527.

Round, J. (2006). "Social Accounting Matrices and SAM-based Multiplier Analysis", Department of

Economics, University of Warwick, UK.

Stone, R. (1978). "The disaggregation of the household sector in the national accounts", in

proceedings of the World bank conference on social accounting methods in development

planning.

Stuttard, N. and M. Frogner (2003). "Developing a pilot Social Accounting Matrix for the United

Kingdom", London, Office for National Statistics, UK.

Tokunbo S. O. and A. A. Lloyd (2010). "Foreign Direct Investmnet and Economic Growth in

Nigeria", University of Santiago de Compostela Journal 10 (2)

Uganda Bureau of Statistics (2010). Uganda National Household Survey - Report. Kampala, Uganda.

Warner, A. (2014). “Public Investment as an Engine of Growth”, International Monetary Fund

Working Paper, No.2014/148

21 | P a g e

Appendix 1: Mapping of economic sectors

SUB SET SUB-SET ELEMENT

SUB SET SUB-SET ELEMENT

Ag

ricu

ltu

re

1 cashcrop c-cotton, c-flowers, c-

coffee, c-tea, c-

othcashcrops

Ind

ust

ry

17 TransptEqvip c-transportequipment

18 OtherManuf c-othmanufact

2 Grainseeds c-maize, c-rice, c-

othcereal, c-oilseeds, c-

potatoes, c-cassava, c-

beans, c-othlegumes, c-

bananas

19 govInvst c-wtsngov, c-oinfagr, c-

oinfroadtp, c-oinfwatertp, c-

oinfrailtp, c-oinfenerg, c-

oinfoth, c-healthgov, c-

edupgov, c-edusgov, c-

edutgov 3 other_agric c-othagr

28 Pvt_Invest c-wtsnprv, c-edupprv, c-

edusprv, c-edutprv, c-

healthprv

4 Animal_Husbandry c-cattlebuffaloes, c-

dairyfarming, c-othanimal

21 Construction c-construct

5 Forestry c-forestry

Ser

vic

e

20 OthrServ c-othsvc

6 Fish c-fishing 22 Trade c-trade

Ind

ust

ry

7 Petroleum c-crudepetroleumgas, c-

oilref

23 TrsprtServ c-transport

8 Mining c-othmining 24 HotelRest c-hotelsrest

9 foodProcesg c-processedmeat, c-

processedfish, c-dairy, c-

grainmillstarch, c-bakery,

c-manufactsugar, c-

manufactcoffee, c-

manufacttea, c-othfood

25 Telecom c-telecomict

26 FinServices c-financialsvc

27 gov_cur c-publicadminist

29 labour f-labn, f-labs, f-labt

30 capital f-capprv, f-land, f-timber, f-

fish, f-ext-

crudepetroleumgas, f-ext-

othminerals, f-ext-quarrying

10 Oils&fats c-edibleoilsfats

11 Animal_feed c-animalfeeds

12 Text_tobacco c-bebtob, c-textiles HH_Rural h-rur

13 Chem_RubPlast c-chemplastrubber HH-Urban h-urb

14 NonMetal c-nonmetprod

15 Metal c-metalprod

16 Energy c-electequipment, c-

electgasaircond

22 | P a g e

Online Appendix

Unit root and specification of the Data Generating Process (dgp)

Results of the Augmented Dickey Fuller (ADF) unit root test (Dickey and Fuller, 1979) are provided

in Table 7 and indicate that all variables are non-stationary, but become stationary upon first

differencing at the conventional 5 percent level of significance.

Table 7: The Augmented Dickey-Fuller (ADF) Unit root test ADF Test

Level First difference

Macrovariables Ho: 0 Lag-length Inference Ho: 0 Inference

LDEVTEXP -1.670(-3.475) 2 Accept -17.856 (-2.904) Reject

LX

-2.654 (-3.473) 0 Accept -8.066 (-2.903) Reject

LM -2.257 (-3.473) 0 Accept -2.600 (-2.904) Reject

LPSC -2.145 (-3.473) 0 Accept -3.396 (-2.903) Reject

LGDP -1.819 (-3.474) 0 Accept -8.907 (-2.903) Reject

Notes: L = Log; Akaike Information criterion [AIC], Schwarz Bayesian criterion [SC] and Hannan-Quinn criterion [HQ] were used

(maximum set at 6 lags). An unrestricted intercept and restricted linear trend were included in the ADF equation when conducting unit

root test of all the series in the levels. Numbers in parentheses are the 5 percent critical values, unless otherwise stated. All unit-root

non-stationary variables are stationary in first differences.

Source: Author’s Computations. Estimates from E-Views

On the basis of unit root testing and the fact that all variables are integrated of the same order, i.e. I(1)

implies they could be cointegrated. The 5-dimentional vector autoregressive model is estimated with

an unrestricted constant. Including an unrestricted constant allows for linear trends in both

cointegrating space and in the variables in levels and produces a non-zero mean in the cointegrating

relation. Furthermore, it avoids creation of quadratic trends in levels, which would arise if both the

constant and the trend are unrestricted (Bwire et al., 2016; Juselius, 2006). The lag-length is

determined as the minimum number of lags that meets the crucial assumption of time independence

of the residuals, based on a Lagrange Multiplier (LM) test, starting with 5 lags. Akaike, Schwarz and

Hannan-Quinn information criteria all favour one lag, but LM suggests a lag of 3, and it is this that

we adopt for our forward model estimation.

An assessment of VAR (3) system residual misspecification test reveals that although the residuals

meet the crucial assumption of no auto-correlation and residual heteroskedasticity test is satisfactory

[ 2 (450) = 444.2[0.569]], the assumption of multivariate normality of the residuals is not met

[ 2 (10) = 98.2 [0.000]]. Rejection of joint non-normality of residuals notwithstanding, estimates of

the VAR model are robust to deviations from normality provided residuals are not auto-correlated

(Bwire et al., 2013b).

Johansen’s Cointegration Test

Having determined the appropriate specification of the d.g.p, we then evaluated the existence of long-

run equilibrium relation (s) using the Johansen (1988) trace statistic test for cointegration. The

determination of the cointegrating rank, r , relies on a top-to-bottom sequential procedure; this is

asymptotically more correct than the bottom-to-top Max-Eigen statistic alternative [Juselius, 2006:

131-134]. The trace-test has been shown to have finite sample bias when the number of observations

is as small as in the present study with the implication that it often indicates too many cointegrating

relations so that the test is oversized (Juselius, 2006; Cheung and Lai, 1993b; Reimers 1992). The

trace-test in the current study is adjusted for finite sample bias using the correcting procedure

23 | P a g e

suggested by Reimers (1992) as shown in Harris and Sollis (2005:122-124) for 71 included

observations after adjustment. Based on the results in Table 8, the presence of one equilibrium

(stationary) relation, even when corrected for small sample bias among the variables at the

conventional 5 percent level of significance, cannot be rejected. Even more, roots of the companion

matrix and graphs of the potential cointegrating relations (available on request) suggest that

1r seems reasonably well supported by the data.

Table 8: Johansen’s Cointegration test and Long-run Analysis

Hypothesized No. of CE(s) Eigen value Trace Statistic Trace Statistic

# 0.05 Critical Value

None * 0.4588 86.995 70.603 69.819

At most 1 0.2847 44.633 36.221 47.856

At most 2 0.1729 21.512 17.458 29.797

At most 3 0.1131 8.4155 6.828 15.495

At most 4 0.0019 0.1308 0.107 3.841

Notes: i) Trace test indicates 1 cointegrating eqn(s) at the 0.05 level; ii) * denotes rejection of the hypothesis at the 0.05

level; iii) # is the small sample corrected test statistic;

Source: Author’s Computations. Estimates from E-Views