The Journal of The Botswana Economics Association

The Botswana Journal of Economics
The Journal of the Botswana Economics Association
ARTICLES
Fixed Income Market Efficiency: Evidence from Kenya’s 10-Year Local Currency Bond
Pako Thupayagale 1
Capital Account Liberalisation and Foreign Direct Investment in Nigeria:

A Bound-testing Approach
Bankole, Abiodun S. (Ph.D) and Ayinde, Taofeek Olusola 14
Credit demand amongst farmers in Mukono District, Uganda

George Wilson Ssonko and Marris Nakayaga 33
Determinants of Informal Employment: A Case of Tanzania’s Construction Industry

Jehovaness Aikaeli and Beatrice Kalinda Mkenda 51
The Impact of Bank and Stock Market Developments on Economic Growth

in Zimbabwe: 1988 to 2012
Phineas G. Kadenge and Felex Tafirei 74
Fixed Income Market Efficiency: Evidence from Kenya’s 10-Year Local
Currency Bond
Pako Thupayagale1
Abstract
This paper tests for long memory in the yield changes and volatility of Kenya’s benchmark 10-
year government bond, in order to evaluate the informational efficiency of the local currency
market. Using the ARFIMA-FIGARCH model the statistical properties of yield changes and
volatility are simultaneously estimated. Evidence of long memory in both yield changes and
volatility are conclusively demonstrated. This finding suggests a pattern of time-dependence
in the data, which stands against the efficient market hypothesis. In addition, the existence
of long memory in the data is valid for all sample periods, suggesting that the recent bond
markets reforms have not wholly produced the expected efficiency gains.
Keywords: bonds, Long Memory, AFRIMA, ARFIMA-FIGARCH
JEL Classification Code: G1, G 12, G14.
1
Financial Markets Department, Bank of Botswana, e-mail: thupayagalep@bob.bw
1 BOJE: Botswana Journal of Economics

I. INTRODUCTION
Long memory (or long-range dependence) in debt markets has important implications for the
efficiency of the market in pricing fixed income securities. The efficient market hypothesis
(EMH) provides the standard framework to analyse and interpret informational efficiency
in capital market data. While a number of definitions of market efficiency are available,
the random walk version of the EMH proposed by Bachelier (1900), formalised by Osborne
(1959) and refined by Fama (1965, 1970) asserts that for a financial market to be efficient
future prices, or returns, cannot be predicted from currently available information.
If yields in Kenya’s 10-year benchmark bond display long memory, then they exhibit non-linear
behaviour marked by distinct but non-periodic cyclical patterns; and, long-term dependence
between distant observations. This, in turn, suggests that the fluctuations of the 10-year bond
embody a predictable component; and hence, past trends in yield movements can be used to
extrapolate future trends. In this light, long memory provides evidence against the weak-form
version of the efficient market hypothesis.2
There are numerous studies that present evidence for and against long memory in financial
markets. However, most of the research has concentrated on global equity markets (e.g.,
Poon, 2005 and references therein).In terms of the analysis of long memory dynamics in fixed
income markets, the literature has mostly been limited to developed markets. For example,
Bollerslev et al, (2000), McCarthy et al, (2004), Schotman et al, (2008) and McCarthy et al,
(2009) all examined the various aspects of long memory in interest rates and yield spreads in
order to determine its existence, magnitude and investment implications.
Research on long memory dynamics in fixed income markets in Africa is limited; notable
exceptions include Thupayagale (2011), who demonstrates evidence of long memory in the
volatility of South Africa’s local currency 10-year bond using methods based on wavelets, and
Thupayagale (2012),who analyses long memory behaviour among several emerging markets
(including South Africa) and finds that the information content of long memory models does
not generally lead to improved forecast accuracy relative to the standard GARCH process.
The purpose of this paper is to augment this line of analysis concerning the characterisation
of long memory dynamics, by focusing on Kenya’s local currency debt market given recent
the implementation of structural reforms aimed at enhancing market efficiency and secondary
trading activity. This study, therefore, contributes to the broader discussion concerning the
state of capital market development in Africa by examining the efficiency of local bond
markets with a view to evaluate the success of various policy initiatives designed to enhance
the level of debt market operations.
This research extends the existing literature in the following ways. First, by determining if long
memory exists in Kenya’s bond market, since there does not appear to be any previous tests of
long memory in this market. Second, long memory in bond yield changes and volatility are
simultaneously estimated using the ARFIMA-FIGARCH model, which represent a relatively
new innovation in time series analysis. Third, informational efficiency is examined in the
market before and after reforms, in order to evaluate whether these reforms led to efficiency
gains.
2
The weak form of the EMH asserts that the current price incorporates all relevant historical information about bond yields. As
such, changes in bond yields cannot, therefore, be predicted from past trends in yields.
BOJE: Botswana Journal of Economics 2

To summarise key findings from the outset, evidence of long memory in Kenya’s 10-year
bond yield changes and volatility are recorded. Both parameters are statistically significant,
suggesting that they represent an important characterisation of Kenya’s bond market.
Furthermore, this study finds evidence of long memory in both sample periods (i.e., the entire
sample period and the post-reform sample period) indicating that Kenya’s local currency
bond market remains inefficient despite recent implementation of financial market reforms.
However, the magnitude of the long memory parameters in the post reform period are smaller,
suggesting perhaps, that some progress has been made, although further improvements are
still required.
II. BOND MARKET REFORMS IN KENYA
Since the turn of the millennium, local currency emerging market debt has evolved from a
niche market for credit specialists into a mainstream asset class, whereby emerging market
debt is progressively becoming part of strategic holdings of global fixed income managers.
However, it is important to observe that emerging markets are not homogeneous and these
markets differ markedly in terms of the level of development. For instance, African bond
markets (ABMs) are perceived by many investors as the terminus for investors searching for
high yields. In addition, ABMs are, on average, less deep and liquid than their larger emerging
market counterparts, owing to a variety of reasons, including: a narrow investor base; relatively
small volumes of transactions; underdeveloped financial infrastructure; and bottlenecks in the
trading, settlement and clearing infrastructure. Indeed, most ABMs - excluding South Africa -
remain very small by world standards. Small size and associated low levels of liquidity, raise
questions regarding the efficiency of these markets and the process of price determination.
Using the EMH as a criterion, the theme of market efficiency is explored within the context
of Kenya’s bond market.
Promoting capital market development in Kenya has become an important component of the
government’s financial development strategy. In particular, recent reforms recognised the
development of bond markets and the financing of capital formation as key factors bearing
upon the prospects for long-term growth. The importance of the local bond market in Kenya,
not only as a vehicle to fund the country’s budget deficit, but also as a source for investments
that are free of credit risk, that can serve as a benchmark for the development of the domestic
corporate bond market and as an alternative to equity financing and bank lending has motivated
policymakers to initiate measures designed to ameliorate conditions in the bond market.
In order to improve market efficiency in the domestic debt market, the Central Bank of Kenya,
in collaboration with the Capital Markets Authority and the Nairobi Stock Exchange, initiated
a program to enhance the efficiency and liquidity of the Kenyan government securities market
as a key part of its strategy to develop the domestic bond market. Since September 2007, the
Kenyan authorities have embarked on a focused issuance program aimed at building large
and liquid benchmark bonds. This was achieved through larger issuance of new government
bonds and re-openings of existing issues, therefore increasing the free-float available for
secondary market purposes. By proactively helping to re-channel liquidity from off-the-
run issues to benchmark bonds and then conducting more frequent auctions, the authorities
have sought to increase the pool of assets available for secondary trading, thereby achieving
critical sizes that are more easily tradable both by onshore and offshore investors. In addition,
to more effectively conduct fiscal funding and align issuance more closely with budgetary

requirements, the auction calendar and issuance schedule were further streamlined. These
reforms brought operations and policies closer to internationally accepted principles and
practices and have revitalised the government bond market by, among others, supporting
larger and more frequent issuance and improved liquidity in the secondary market. Additional
reforms included the introduction of the automated trading system (ATS) in December 2009.
In fact, all government and corporate bonds are dematerialised and trade in an end-to-end
automated platform, encompassing the placement of orders to matching and finally clearing
and settlement. Furthermore, the Kenyan Government has committed itself to macroeconomic
stability characterised by moderate inflation, trend growth and sustainable government deficits.
Table 1 presents key macroeconomic indicators for Kenya. The growth and inflation outlook,
along with a variety of key fiscal metrics, indicate improving conditions for Kenya’s bond
market to develop.
Table 1: Economic and Financial Indicators for Kenya

2010 2011 2012 2013 2014F
Real GDP growth 5.8 4.4 4.6 4.8 5.9
CPI inflation 4.1 14.0 9.4 5.7 7.1
Current account/GDP -7.9 -9.8 -10.4 -9.2 -10.6
FX reserves (USD, billions) 4.0 3.7 5.7 6.1 7.0
Months of import cover 3.9 4.2 4.3 4.1 4.2
Fiscal balance/GDP 7.2 -5.0 -5.6 -6.8 -8.7
Primary balance/GDP -2.1 -1.9 -1.8 1.4 -1.6
General govt/debt 25.9 27.8 26.1 28.7 28.3
External debt/GDP 22.2 26.4 23.5 23.0 25.4
Policy rate 6.0 18.0 11.0 8.5 8.5
S&P Sovereign Rating B+ B+ B+ B+ B+
Notes: 1. CPI inflation refers to the annual average inflation rate.
2. The government has an inflation target of 5% over the medium term.
3. The ratio of foreign exchange reserves to imports is expressed as months of import cover.
4. The policy rate refers to the year-end central bank rate.
Source: Reuters, Bloomberg, IMF, Central Bank of Kenya, Standard Chartered Research.
Table 2 presents an overview of debt markets in Kenya. In relation to treasury (or government)
bonds, the fixed-rate ‘plain vanilla’ nature of the bond market is highlighted and common
tenors are noted. Of particular interest are secondary market trading conditions where
average daily turnover is estimated at between USD20-25 million. In terms of the taxation
dispensation, a 15% withholding tax applies to listed bonds. The only exception relates to
infrastructure bonds, which are tax exempt. This favourable tax treatment is motivated by the
government’s preference to further develop the level and quality of infrastructural development
by encouraging investments in these bonds.

Table 2: Rates in Kenya
T-bills T-bonds Term auction
deposit
Issuer Treasury
Use of proceeds Liquidity Fiscal financing Liquidity
management/fiscal management
financing
Curve span 91-to 364-day 2Y-30Y 7- to 28-day
Common tenors 2Y, 5Y, 7Y, 10Y,
15Y, 20Y
Coupon Fixed
Coupon frequency Semi-annual
Day count Actual/365
Primary Market
Auction style Multiple-price Multiple price Multiple price
Average issue size KES 2-5 billion KES 18-20 billion variable
Secondary market
Average trade size KES100 million KES100 million NA
Average daily turnover KES1.7-2.2 billion
Quotation convention Yield Yield Yield
Settlement period T+3 T+3 T+0
Bid/offer spread 15bps 50bps NA
Note: 1. 15% withholding tax applies to listed bonds. The only exception relates to infrastructure bonds, which are tax exempt.
2. Secondary market fees comprise a 0.04% brokerage on consideration
3. All trades are required to go through a local broker
Source: Standard Chartered, Reuters, Bloomberg, Central Bank of Kenya and ABSA Capital Research
III. LONG MEMORY IN TIME SERIES
Models analysing long memory dynamics were first introduced by Hurst (1951). His
investigation was motivated by hydrological considerations; in particular, the storage and
distribution of water from Nile River given its non-periodic (flooding) cycles. Mandelbrot
and Wallis (1968) described this feature as the ‘Joseph effect’ hinting to the biblical reference
in which seven years of plenty were to be followed by seven years of famine. Long memory
is associated with a correlation structure over long lags. Specifically, it describes a data series
whereby observations in the remote past are highly correlated with observations in the distant
future. Mandelbrot (1971) pioneered the application of long memory models to financial
markets. His study triggered the examination of long memory dynamics across various
securities and asset markets in order to analyse and interpret financial market data.
To define a long memory model formally, let υτ be the autocovariance function with a time lag
t of a stationary process X t , there exists long memory in X t if its autocovariance function
ρ(τ) decays monotonically and hyperbolically to zero. This asymptotic property can be stated
as:

Xt Xt
ρ (τ ) decays monotonically and hyperbolically to zero. This asymptotic property can be stated
as:
2 d −1
ρτ ≈ τ as τ → ∞ (1)
ρτ ≈ |t| as|t|→∞ .................................................................................................................. (1)
2d-1
Where
d ∈ (0, 0.5) is the long memory parameter. In the case where d> 0.5, the series is
Where d ∈ (0, Meanwhile,
nonstationary. 0.5) is the long
for dmemory
∈ (−0.5parameter.
, 0) , the series In the case where
is described as d> 0.5, the series
antipersistent, which is is a
nonstationary. Meanwhile, for d ∈ (0, 0.5), the series is described
measure of the decline in statistical significance between distant observations. as antipersistent, which is a
measure of the decline in statistical significance between distant observations.
A variety of measures have been used to detect long memory in financial time series. The most
A variety
widely of measures
used have been
in contemporary used to detect
econometric long memory
analysis in financial time series.I(d)
are the fractionally-integrated Thetimemostseries
widely used
models in contemporary
introduced by Grangereconometric
and Joyeuxanalysis (1980) and are the fractionally-integrated
Hosking (1981). Fractionally I(d) integrated
time
series models
processes introduced
are distinct frombyboth
Granger and Joyeux
stationary (1980)processes
and unit-root and Hosking in that(1981).
they are Fractionally
persistent, but
integrated
are processes
also mean are distinct
reverting; from bothprovide
and, therefore stationary and unit-root
a flexible processes
alternative in that they
to standard I(1) are and I(0)
persistent, but are also mean reverting; and, therefore provide a flexible alternative to standard
processes.
I(1) and I(0) processes.
To estimate the long memory parameter, d, in financial time series data, the most familiar model
is
Tothe autoregressive
estimate the long fractionally integrated
memory parameter, d, moving
in financial averagetime (ARFIMA
series data,(p, thed,most
q)) model,
familiarwhich
captures
model is the autoregressive fractionally integrated moving average (ARFIMA (p, d, Baillie
temporal dependencies in the conditional mean process. Subsequent to this, q)) et
al.,
model, (1996)
which developed the fractionally
captures temporal dependencies integrated generalised
in the conditional meanautoregressive
process. Subsequent conditional
heteroskedasticity
to this, Baillie et al., (FIGARCH (p, d, q))model,
(1996) developed which integrated
the fractionally captures long memoryautoregressive
generalised in the conditional
variance
conditional of heteroskedasticity
a time series. Since non-zero values
(FIGARCH of the fractional
(p, d, q))model, differencing
which captures long parameter
memory inimply
dependence
the conditional between
variancedistant observations,
of a time series. Since substantial
non-zeroattention
values of hasthebeen directed
fractional to the analysis
differencing
of fractional
parameter dynamics
imply to testbetween
dependence empirical and observations,
distant theoretical propositions
substantial in financial
attention haseconomics.
been
Against
directed to the analysis of fractional dynamics to test empirical and theoretical propositions in in
this background, a recent innovation has been the joint estimation of long memory
returns
financialand volatilityAgainst
economics. using this
the background,
ARFIMA-FIGARCH model, which
a recent innovation has been is the
the joint
subject of the next
estimation
section.
of long memory in returns and volatility using the ARFIMA-FIGARCH model, which is the
subject of the next section.
IV. EMPIRICAL METHODOLOGY
IV. EMPIRICAL METHODOLOGY
Modelling Returns: ARFIMA Model
Modelling Returns: ARFIMA Model
The general specification for the ARFIMA (m, d, n) class of models can be expressed as:
The general specification for the ARFIMA (m, d, n) class of models can be expressed as:
φ ( L)(1 − L) d y t = θ ( L)ε t (2)
φ (L) (1 - L) d yt = θ (L) εt ..................................................................................................... (2)
Where L is a lag operator, φ (L) and θ (L) are polynomials in the lag operator of orders m and n
Where L is a lag operator, φ (L) and mθ (L) are polynomials inn
the lag operator of orders m and
j j
respectively. Further, φ ( L) = 1 − ∑φ j L and θ ( L) = 1 + ∑θ j L . .AllAll
n respectively. Further, the the
rootsroots of φand
of φ (L) (L) and
j =1 j =1
θ (L) lie outside the unit circle. The residual, εt , follows a white noise process with variance,
θ (L )lie outside the unit circle. The residual, ε t , follows a white noise process with variance, σ2.
σ2. The fixed income yield (change) at time, t, denoted y t and dis to the fractional differencing
parameter. The long memory property arises when fractional differencing parameter, d ∈ (0,
0.5).
7
Modelling Volatility: FIGARCH Model

Modelling Volatility: FIGARCH Model
The FIGARCH process can be derived from the standard GARCH (p, q) model, which can be
expressed as:
ht = ω + α ( L)ε t2 + β ( L)ht .......................................................................................................

(3) (3)
Where ht and ε t2 are conditional and unconditional variances of εt respectively, ω = ε2[1 – β(1) –
q p
α(1)], and φ ( L) = 1 − ∑ φ j L j and β ( L) = 1 + ∑ β j L j . The GARCH (p, q) process in Equation
j =1 j =1
(3) can be rewritten as an ARMA (m, p) process in ε t2 ,
[1 − α ( L) − β ( L)]ε t2 = ω +[1 − β ( L)]ν t .....................................................................................

(4) (4)
where vt ≡ ε t2 − ht2 has a zero mean and is serially uncorrelated, i.e., Et −1 (vt ) = 0. To ensure
covariance stationarity, the roots [1 − α ( L) − β ( L)]and [1 − β ( L)] are constrained to lie outside
the unit circle. When the autoregressive lag polynomial,1 − α ( L) − β ( L) , contains a unit root, the
GARCH (p,q) process is said to be integrated in variance and the process is called an integrated
GARCH (or IGARCH) process (Engle and Bollerslev, 1986) and is given by:
(5)
φ ( L)(1 − L)ε t2 = ω + [1 − β ( L)]vt ............................................................................................. (5)
From this model, the general specification of the FIGARCH model can be obtained by
introducing the fractional differencing operator, (1 − L) d , as such:
φ ( L)(1 − L) d ε t2 = ω + [1 − β ( L)]vt ...........................................................................................

(6) (6)
The FIGARCH (p, d, q) model encompasses other GARCH based models, and is equivalent to
the GARCH process when d = 0 and to the IGARCH when d = 1 (see Baillieet al, 1996). The
( )
FIGARCH 1, d , 1 model will be estimated by maximum likelihood methods, which are both
consistent and asymptotically efficient. In particular, the following log-likelihood is maximised:
η = T {ln Γ[0.5(df + 1)] − ln Γ(0.5df ) − 0.5 ln[π (df − 2)]}

T
{ [
− 0.5∑ ln(ht ) + (1 + df )ln 1 + ε t2 (ht [df − 2]) ]} ......................................................................
(7) (7)
t =1
Where df denotes the degrees of freedom. Following the results of previous findings that
returns are not normally distributed, estimation is based on the student t distribution.
8
V. EMPIRICAL ANALYSIS

This analysis is based on daily yields on Kenya’s benchmark 10-year local currency bond. The
data, which are obtained from Thompson Reuters, span the period from October 1, 2004 to
December 31, 2012.Figure 1 plots the evolution of the 10-year yield over the sample period.
The 10-year yield peaked at 17.2 percent in 2011 from a trough of 5.6 percent in 2010. The
Central Bank of Kenya increased its key lending rate by 1225 basis points to 18 percent during

Bank of Kenya increased its key lending rate by 1225 basis points to 18 percent during 2011 as a
sharp currency depreciation and a severe drought took headline inflation to 19 percent year-on-
year. The yield on the 10-year bond rose in tandem with these developments.
2011 as a sharp currency depreciation and a severe drought took headline inflation to 19
percent year-on-year. The yield on the 10-year bond rose in tandem with these developments.
Figure 1. Daily Yields of Kenya’s 10-YearLocal Currency Bond
Figure 1. Daily Yields of Kenya’s 10-YearLocal Currency Bond
18
16
14
12
10
4
2005 2006 2007 2008 2009 2010 2011 2012
(a) Preliminary Observations

(a) Before
Preliminary
embarkingObservations
on further statistical analysis, the yield on Kenya’s 10-year yield was
Beforechecked
embarking on further and
for stationarity statistical
it was analysis,
determined thethat
yield
the on Kenya’s
yield becomes 10-year yieldafter
stationary wasbeing
checked
for stationarity
differencedandonceit (as
wasreflected
determined that root
in the unit the tests).
yield Hence,
becomes stationary after
all subsequent being
empirical differenced
analysis is
conducted on yield changes (differences). At the same time, LM and normality
once (as reflected in the unit root tests). Hence, all subsequent empirical analysis is conducted on tests highlight
the existence
yield changes of non-normality
(differences). At theand ARCH
same time,effects
LM and in the data. Table
normality tests3 highlight
presents the
thesummary
existence of
statistics on the yield difference in Kenya’s 10-year bond.
non-normality and ARCH effects in the data. Table 3 presents the summary statistics on the yield
difference in Kenya’s 10-year bond.
Table 3. Description of the Data
No. of observations 2151
Mean -0.0003
Standard Deviation 0.1944
Skewness 0.6699
Kurtosis 52.4949
Normality test Chi^(2)=219719
ARCH (5) test F(5, 2141)=82.8436
ARCH (10) test F(10, 2131)=43.9522
Unit root test
Constant9 Constant and Trend
ADF -48.5737** -48.5804**
PP -48.6143** -48.6272**
(b) Results from GARCH Models

The non-white noise characteristics of changes in Kenya’s benchmark 10-year Treasury note
motivate an extension of the model in order to take them into account in the statistical model
to be estimated. To this end, the GARCH (1, 1) model is estimated using the assumption of the

Student t distribution.3This model allows for the modelling of volatility persistence based on
some stylised facts usually observed in high-frequency financial time series data, among them,
the presence of thick tails, time-varying correlations and volatility clustering. Table 4presents
the models and highlights the importance of GARCH effects by showing that the GARCH and
ARCH terms are statistically significant over both sample periods.
Evidence of persistence in variance as measured by the GARCH model is reflected in the

magnitude and significance of the ARCH and GARCH terms (indeed, as this sum approaches
unity the greater the degree of persistence). If the sum of GARCH and ARCH terms are equal
to unity (i.e., α+β=1), then any shock to volatility is permanent and the unconditional variance
is infinite. In this case, the process is called an IGARCH (as shown in equation 5). Therefore,
in order to have an indication of long memory in the 10-year note persistence in variance is
measured. Volatility persistence over the two sample periods differs (i.e., the entire sample
period and the post reform period) are 0.8994 and 0.8337, respectively. In other words, over
both sample periods the level of Kenya volatility is persistent, which indicates evidence of long
memory in the volatility structure. It is also observed that the level of volatility persistence is
lower in the post-reform period relative to the full sample period. This difference may signal
prima facie evidence of the impact of reforms. Diagnostic tests to assess the adequacy of the
model are performed by applying the Ljung-Box Q statistic test to standardised and squared
standardised residuals. These diagnostics suggest that the estimated models are appropriate for
the data considered. Specifically, the Ljung-Box test determines whether a time series consists
of random variables; a large p-value (i.e., p-value› 0.05) suggests evidence of no dependence
in the residuals. This, in turn, indicates that the model is correctly specified. Furthermore, the
various sign bias tests suggest that there are no volatility asymmetries in the data that need be
incorporated into the model.
Table 4. GARCH Estimates

3/6/2004 - 31/12/2012 1/9/2009 – 31/12/2012
Constant -0.4961 [0.4393] -0.0887 [0.7845]
ω 0.0051 [0.0291] 8.3002 [7.9149]
α 0.6968 [0.0508]** 0.6811 [0.0376]**
β 0.2026 [0.0027]** 0.1526 [0.0138]**
df 2.0000 [0.0001]** 2.3001 [0.0292]**
Q(5) 1/ 2.1427 5.7188
Q(5) 2/ 2.6235 6.2580
Sign bias test 1.4017 3.5538
Negative size bias test 1.7790 0.7446
Positive size bias test 3.8367 2.2253
Joint test 8.1031** 2.8839
Note: 1. The Ljung-Box Q test applied to standardised residuals
2. The Ljung-Box Q test applied to squared standardised residuals
The numbers in () and [] refer to lag lengths and standard deviations
(**) and (*) indicates statistical significance at the 1% and 5% levels, respectively.
To select the lag length in the GARCH(p,q) model and other subsequent models used in the study the Schwarz Bayesian
3
information criterion (SBIC) is employed. The SBIC is preferred to other standard information criteria (i.e., the Akaike and
Hannan-Quinn information criteria) given its statistical properties; in particular, the SBIC will asymptotically deliver the correct
model order.

(c) Results from ARFIMA-FIGARCH Models
A disadvantage of the standard GARCH model is its inability to capture long memory in the
data. Therefore, this study now turns to the estimation of the fractional differencing parameter
in Kenya’s bond market using the ARFIMA-FIGARCH model, which is the main interest of
this paper. To model persistence in changes in the benchmark 10-year yield and its volatility
simultaneously maximum likelihood methods are used to estimate the ARFIMA-FIGARCH
model. The ARFIMA part of the equation provides a basis to test for market efficiency by
examining the size of the fractional differencing term, d, in the mean equation. In particular, d
measures the adjustment speed (relative to a stationary ARIMA case where d = 0) and, hence,
permits conclusions based on the EMH as a criterion. On the other hand, the FIGARCH part
of the model captures long memory in the conditional variance of the data; and hence, provides
insights into the behaviour of volatility in the 10-year bond. Table 5 presents estimates of the
ARFIMA-FIGARCH model; in particular, the size, sign and significance of d and d , are
of interest as they capture long memory in changes in the yield and its associated volatility,
respectively. Furthermore, over both the full sample and the post-reform sample the ARFIMA-
FIGARCH and ARFIMA-FIGARCHmodels are employed, respectively.4 The long memory
parameters, d and d are presented in Table 5. In addition, the models highlighted in Table
5 appear to be well specified, since there appears to be no evidence of autocorrelation in the
residuals or volatility asymmetries in the data that need to be accounted for.
Table 5. ARFIMA-FIGARCH Estimates

3/6/2004 - 31/12/2012 1/9/2009 – 31/12/2012
Constant 0.2634 [0.2116] 0.3346 [0.2951]
d 0.3603 [0.0992]** 0.2846 [0.1307]*
ω 1.1132 [0.0731]** 0.0965 [0.0271]**
0.1519 [0.0287]** 0.1468 [0.0624]**
d
0.6645 [0.2558] 0.4474 [0.0085]**
β
0.3170 [0.0122] 0.3082 [0.0643]
Ø
2.6311 [1.7130] 4.4230 [1.3026]*
df
1.7656 0.1192
Q(5) 1/
2.1175 2.2305
Q(5) 2/
0.1325 4.6269
Sign bias test
1.0443 2.1919
Negative size bias test
1.7413 2.4313
Positive size bias test
8.3137 6.1117
Joint test
Note: 1. The Ljung-Box Q test applied to standardised residuals
2. The Ljung-Box Q test applied to squared standardised residuals
The numbers in () and [] refer to lag lengths and standard deviations
(**) and (*) indicates statistical significance at the 1% and 5% levels, respectively.
ARFIMA MODEL
Market efficiency is considered by examining the size of the fractional differencing term, d,
in the mean equation. Over the full sample period, the long memory parameter for Kenya’s
10-year benchmark bond is 0.3603 and is statistically significant. This implies that changes
in the 10-year bond yield are autocorrelated and, hence, future changes in the 10-year yield
can be predicted using past yield data. This result suggests that weak-form efficiency does
not hold; since a predictable component to the data is revealed. For the post-reform period,
Lag lengths are computed on the basis of the Schwarz Bayesian information criterion.
4

the long memory parameter is 0.2846, indicating that the intensity of predictability in the
data generating process is reduced, relative to the full sample period. In addition, the effect
is only significant at the 5 percent level of significance. These results, therefore, indicate the
existence of long memory effects in both sample periods. However, the strength is smaller
and less statistically significant in the post-reform period. Nonetheless, the results highlight
the importance of modelling long-range dependence in bond yield data and points to market
inefficiency in both sample periods. Given that in both periods the fractional differencing term
is statistically different from zero.
FIGARCH MODEL
Evidence of long memory in the volatility of changes in Kenya’s 10-year bond yields is found
in the full sample and post-reform sample period. Findings from this model are shown by
d in Table 5. The fractional differencing term, d = 0.1519, over the full sample period is
significantly different from zero. This indicates a pattern of time dependence in the volatility of
changes in yields that may allow for past information to be used to improve the predictability of
future volatility. Meanwhile, over the post-reform period, d = 0.1468, and is also statistically
significant. These findings indicate the importance of long-range dependence in the volatility
structure of Kenya’s Treasury note. This finding furthermore indicates that future volatility is
a function of its past value and so is predictable from past information. The significant size of
the fractional differencing parameters in this study underscores the importance of modeling
long memory dynamics in Kenya’s 10-year bond.
VI. CONCLUSION
The informational efficiency of Kenya’s local currency bond market has been studied by
examining the time series properties of the benchmark 10-year bond from June 3, 2004 to
December 31, 2012. This study differs from previous research in a number of respects. First,
it focuses on the long memory attributes of Kenya’s local currency market, which appear not
to have been conducted so far. Second, it evaluates market efficiency over the entire sample
and after the implementation of key reforms in Kenya’s capital markets in order to gauge
their success. Third, estimation is conducted using time series techniques that allow for the
simultaneous modeling of persistence in changes in yields and their volatility.
In sum, the results of the ARFIMA-FIGARCH model suggest that changes in Kenya 10-year
bond yields are characterised by stochastic processes which have a predictable component.
This, in turn, implies a departure from the EMH, suggesting that relevant market information
was only partially or gradually reflected in bond yield changes.
Furthermore, this paper presents evidence of long memory in bond yield changes and their
corresponding volatility regardless of the sample period. In relation to the post-reform sample,
the long memory estimates are quantitatively smaller, perhaps suggesting some progress
towards improvement with respect to informational efficiency. However, the fact that the
fractional differencing term is statistically different from zero in both periods indicates that
the market is still informationally inefficient – albeit at lower levels. This paper points to a
number of possible factors behind the absence of improvements in market efficiency, despite
the recent reforms in the bond market.

A prerequisite for bond market liquidity is a framework that balances the demand and supply
side factors. However, the Kenyan market remains characterised by a severe structural shortage
of bonds, with demand from annuity providers far exceeding available supply, hence the lack
of liquidity. Another feature in Kenya pertains to the presence of non-synchronous trading
or non-trading-effects which, in turn, reflect the small market size and further compound
illiquidity and hinder market efficiency, despite the progress made to date in terms of capital
market development. On the other hand, it is also anticipated that the implementation of
primary dealer rules will enhance the price discovery process and connect the buyers and
sellers more easily.
Finally, while this analysis is focused on Kenya’s local currency 10-year bond, further analysis
could usefully be conducted in a number of directions. One extension would be to investigate
the exact causes of the inefficiencies in the Kenyan bond market. Another possibility would
be to expand country coverage and compare our findings with those in other local-currency
markets.
REFERENCES
ABSA Capital, 2014, “African Local Markets Guide 2014,”Emerging Markets Research
January 2014, ABSA Capital.
Adelegan, O.J., and B. Radzewicz-Bak, 2009, “What Determines Bond Market
Development in sub-Saharan Africa,” IMF Working Paper 09/213, (IMF: Washington D.C.:
International Monetary Fund).
Bachielier, L., 1900, Theory of Speculation (Paris: Gauthier-Villars)
Baillie, R.T., 1996, “Long Memory Processes and Fractional Integration in Econometrics,”
Journal of Econometrics, Vol. 73, No. 1, pp.5-59.
Baillie, R.T., T. Bollerslev and H.O. Mikkelsen, 1996, “Fractionally Integrated Generalised
Autoregressive Conditional Heteroskedasticity,” Journal of Econometrics, Vol. 74, No. 1, pp.
3-30.
Bollerslev, T.P., 1986, “Generalised Autoregressive Conditional Heteroskedasticity,”
Journal of Econometrics, Vol. 31, pp. 307-327.
Bollerslev, T., J. Cai and F.M. Song, 2000, “Intraday Periodicity, Long Memory Volatility,
and Macroeconomic Announcement Effects in the US Treasury Bond Market,” Journal of
Empirical Finance, Vol. 7, pp. 37 - 55.
Ding, Z., C.W.J. Granger and R.F. Engle, 1993, “A Long Memory Property of Stock Market
Returns and a New Model,” Journal of Empirical Finance, Vol. 1, pp. 83-106.
Engle, R. F. and T. Bollerslev, 1986, “Modelling the Persistence of Conditional Variances,”
Economic Review, Vol. 5, pp. 1-50.
Fama, E.F., 1965, “The Behaviour of Stock Market Prices,” Journal of Business, Vol. 50,
pp. 34-105.
Fama, E., 1970, “Efficient Capital Markets: A Review of Markets and Empirical Work,”
Journal of Finance, Vol. 25, pp. 383-423.
Granger, C., 1980, “Long Memory Relationships and the Aggregation of Dynamic
Models,” Journal of Econometrics, Vol. 14, pp. 227-238.
Granger, C., and R. Joyeux, 1980, “An Introduction to Long-Memory Time Series Models
and Fractional Differencing,” Journal of Time Series Analysis, Vol. 1, pp. 15-29.
Hosking, J., 1981, “Fractional Differencing,” Biometrika, Vol. 68, pp. 165-176.
Hurst, H.E., 1951, “Long-term Storage Capacity of Reservoirs,” Transactions of the
American Society of Civil Engineers, Vol. 116, pp. 770-799.

Lo, A.,1991, “A Long-term Memory in Stock Market Prices,” Econometrica, Vol. 59, No.
5, pp. 1279-1313.
Mandelbrot, B.B., 1977, “Fractals: Form, Chance and Dimensions,” (New York, Free
Press).
Mandelbrot, B. B., and. J. Wallis, 1968, “Noah, Joseph and Operational Hydrology,” Water
Resources Research, Vol. 4, pp. 909-918.
Mandelbrot, B.B., 1971, “When Can Price Be Arbitraged Efficiently? A limit to the Validity
of the Random Walk and Martingale Models,” Review of Economics and Statistics, Vol. 53,
pp. 225-236.
McCarthy, J., R. DiSario, H. Saraoglu and H. Li, 2004, “Tests of long-range dependence
in interest rates using wavelets,”Quarterly Review of Economics and Finance, 44, 180-189.
McCarthy, J. , C. Pantalone and H. C. Li, 2009, “Investigating long memory in yield
spreads,”The Journal of Fixed Income, Vol. 19, No. 1, 73-81.
Osborne, M.F.M., 1959, “Brownian Motion in the Stock Market,” Operations Research,
Vol. 7, pp. 145-177.
Poon, S.-H., 2005, A Practical Guide to Forecasting Financial Market Volatility, John
Wiley & Sons Ltd, Chichester
Schotman, P.C.,R. Tschernig and J. Budek, 2008, “Long Memory and the Term Structure
of Risk,” Journal of Financial Econometrics, 6, 459–495.
Schwarz, G, 1978, “Estimating the Dimension of a Model,” Annals of Statistics, Vol. 6,
pp. 461-4.
Standard Chartered, 2013, “Local Markets Compendium 2013”, Standard Chartered Bank.
Thupayagale, P., 2011, “Long Memory in the Volatility of an Emerging Fixed-Income
Market: Evidence from South Africa,” South African Journal of Economics, Vol. 79, No. 3,
p. 290-300.
Thupayagale, P., 2012, “Long Memory in the Volatility of Local Currency Bond Markets:
Evidence from Hong Kong, Mexico and South Africa,” in Risk Management – Current Issues
and Challenges (edited by Nerija Banaitiene), InTech, Open Access Publications.

Capital Account Liberalisation and Foreign Direct Investmentin Nigeria:
A Bound-Testing Approach
Bankole, Abiodun S. (Ph.D)1 and Ayinde, Taofeek Olusola2
Abstract
This study focuses on the neoclassical counter-revolution framework to investigate the

relationship between capital account liberalisation and foreign direct investment in Nigeria
for the periods 1980-2011. The technique of analysis employed is the Bound-Testing Approach.
This technique, which was later re-parameterized to investigate the short-run dynamics, is
primarily used to ascertain the long-run equilibrium condition among the variables. The
results obtained largely supports the neoclassical counter-revolution framework which craves
for government involvement in the natural workings of the economy to a minimal level; only
for regulatory purposes. Thus, market-based measures towards FDI should be checked with
legal/political measures; especially for capital control purposes. More so, the rate of inflation
has alternate effect on FDI in Nigeria; implying that the effect of price level on FDI has not
been consistent. On the whole, foreign direct investment in Nigeria is found to be driven by
non-capital account liberalisation. In effect, the liberalisation of capital account transactions
in Nigeria does not matter for FDI; either in the short-run or long-run situation. As such,
government should focus on non-liberalising factors such as qualitative governance, price
stability and institutional development in order to enhance foreign direct investment in Nigeria.
Key Words: Bound-Test, Capital Accounts, Foreign Investments, Neo-Classical.
JEL Classification Code: E22, F 32, F 23.
1
A Reader; lectures at the Department of Economics; University of Ibadan, Oyo State, Nigeria.
2
Department of Economics and Financial Studies, Fountain University, P.M.B. 4491, Isale-Osun, Oshogbo, Osun State, Nigeria
* Corresponding Author: olusolaat@gmail.com

1.0 INTRODUCTION
Essentially, the liberalisation of capital account transactions is predicated on the allocative

efficiency hypothesis (as popularized by the neoclassical propositions pioneered by Solow;
1956) which presupposes that capital should be re-allocated from the capital-poor economies
to capital-rich economies. Consequently, this is expected to cover the saving-investment
gap of these capital-deficit countries and, thus, increases the global capital flow. However,
the neoclassical model predicts a level-effect (temporary) increase and not scale-effect
(permanent) increase in the growth rate of Gross Domestic Product (GDP) per capita for a
capital-poor country that liberalises its capital account transactions. Testing for a permanent
growth effect is not theoretically plausible because capital accumulation, which is subject to
diminishing returns, is the only channel through which liberalization affects growth in the
neoclassical model (Henry, 2007).
Besides, one of the key conclusions of the new literature is that the principal benefit of financial
openness for developing economies may not be access to foreign capital that helps increase
domestic investment by relaxing the constraint imposed by a low level of domestic saving.
Rather, the main benefits may be indirect ones and of collateral effects such as the catalytic
effects of foreign finance on domestic financial market development, enhanced discipline on
macroeconomic policies, and improvements in corporate governance as well as other aspects
of institutional quality (Obadan, 2006; Okogu and Osafo-Kwaako, 2006). It is in this wise that
the IMF (1995) and World Bank (1997) raised a caveat for the liberalisation of capital account
transactions in developing economies. Many studies; employing the neoclassical model,
have considered the investigations between capital account liberalisation and economic
growth (see Rodrik, 1998; Eichengreen, 2001; Edison, Klein, Ricci and Slok, 2004; Prasad,
Rogoff, Wei and Kose, 2003). These empirical efforts are considered theoretically implausible
because marginal returns to capital are considered diminishing and when not augmented with
technological factors or human capital development, it is considered insufficient to generate
economic growth (Romer, 2012). This remains the major void in empirical investigations on
capital account liberalisation and economic growth. As such, this study fills the theoretical
gap in empirical estimations as we seek to employ a modified Harrod-Domar framework – as
opined by Skinnner (2011) as the basis for the neoclassical counter-revolution hypothesis.
Although, there are few studies (Aizenman & Noy 2006; Aizenman and Noy, 2003; Noy
and Vu, 2007; Asiedu & Lien, 2004) that directly investigate the relationship between capital
account liberalisation and foreign direct investment. However, these studies were largely
either of cross-country or involve firm-levels data, and thus, involve panel data structure.
As we seek to cover the major gap of theoretical implausible framework predicated on some
empirical studies (see Hsiao, 2003), the use of annualized data sets within a time-series trend
are more merited than that of the panel data sets (Reisen and Fischer, 1993 and Mathieson
and Rojas-Suarez, 1993; Henry, 2007 and Miller, 2004). Apart from this introductory section,
this study is further divided into five other sections. Section 2 relates the conceptual issues
around capital account liberalisation while section 3 captures the theoretical and empirical
literature. Section 4 borders on the theoretical methodology and model specification with
section 5 conducting the estimations and discussion of findings while section 6, being the last,
concludes and provides policy suggestions.

2.0 LITERATURE REVIEW
Liberalisation of capital account transactions is well rooted in theoretical propositions. It is

generally anchored on four theoretical strands such as the Neoclassical Global Efficiency
Theory; the Dependency School; the Neoclassical Counterrevolution framework and the
Keynesian Hypothesis. Beginning with, the neoclassical global efficiency theory predicts that
capital mobility adds new resources, technology, management and competition to capital deficit
economies in a way that improves efficiency and stimulates change in a positive direction.
This theory presupposes that free flow of external capital should equilibrate and smoothens
a country’s consumption and production paths. However, the Dependence Model which was
developed from Marxism has been reformulated to accommodate structural changes of the
centre-periphery framework.
Nonetheless, the neoclassical counterrevolution framework indicates that market distortions

must be removed and that the price mechanism must be efficient. It advocates a freely
operating economic system with minimalist government intervention in the workings of
the economy before liberalisation of capital account transactions can translates to foreign
inflows of investment. The counter-revolution framework, as a refinement of the classical and
neoclassical propositions, was part of a more general neoliberal reaction that was opposed
to Keynesian, social democracy, state intervention and structuralism, not to mention radical
theories like dependency. The story of this counterrevolution has been told by Toye (1987).
For Toye, the counterrevolution in development economics began when University of
Chicago economist, Harry Johnson (1923-1977) criticized Keynesian economics during the
early 1970s. Johnson (1977) thought that intellectual movements in economics responded to
perceived social needs rather than arising from an autonomous, purely scientific dynamics
(Peet and Hartwick, 2009).
Essentially, the emphasis of the neoclassical counterrevolution in development policy was

on the solution of three main problems claimed to impede development. Firstly, the problem
of an over-extended public sectors; secondly, the problem of an over-emphasis on physical
capital formation, and finally, the proliferation of distorting economic controls (Toye, 1987,
pp. 48-49). The pioneer of this counter-revolution was Bauer (1972, 1984). Bauer (1984,
p.6) attacked all forms of state investment. The first form is when the state intervenes in
order to raise saving and investment and the second is when it intervenes because of lack of
entrepreneurship. He argued that the requirement for entrepreneurship cannot justify state
ownership because, if a society lacks entrepreneurs, there is no source from which the state
sector can acquire them. Two objectives constitute the essence of the neoclassical counter-
revolution. The first is pricism (to get the price right) through laissez-faire policies and the
second is statism (reduce the scope of state intervention to a minimal requirement). The current
literature merges the two, that is, free market with minimal state intervention (Jabbar, 2004).
The neoclassical counter-revolution framework modifies and extends the neoclassical theory
of growth by Solow and Swan (1956).
From an empirical standpoint, available literature on capital account liberalisation (CAL) and
foreign direct investment (FDI) is largely hinged on cross-section, cross-country panel studies.
Beginning with the Montiel and Reinhart (1999) study, most empirical literature employed the
global efficiency framework. Other studies in this line are Alfaro et. al., (2005), Aizenman
and Noy (2006) and Aizenman and Noy (2003) and these studies found that capital controls

have no impact on aggregate capital flow volumes. More so, the study of Noy and Vu (2007)
suggests that FDI inflows are mainly determined by collateral effects such as institutional
impact
factors on
andaggregate capital flow
macroeconomic volumes.
measures More so,capital
as against the study of Noy
account and VuHowever,
policies. (2007) suggests
Desai,
that FDI inflows are mainly determined by collateral effects such as
Foley and James (2003); Boamah, Craigwell, Downes and Mitchell (2005); Kobrin institutional factors and
(2004)
macroeconomic measures as against capital account policies. However, Desai, Foley
and Sarode (2012) conducted independent studies and found positive correlations and strong and James
(2003); Boamah,
significance level Craigwell, Downes and on
of capital liberalisation Mitchell (2005);
investment Kobrin
levels. The(2004)
study ofand Sarode
Asiedu and(2012)
Lien
conducted
(2004) alsoindependent studies andcapital
found that, although, foundcontrols
positivegenerally
correlations and
deter strong
FDI, theysignificance
still remain level of
region-
capital liberalisation on investment levels. The study of Asiedu and Lien (2004) also found that,
specific.
although, capital controls generally deter FDI, they still remain region-specific.
3.0 METHODOLOGY
3.0 Methodology
This study is anchored on the neoclassical counterrevolution framework. This framework
modifies the traditional Harrod Domar model. For this modification, Skinner (2011) posited that
an infusion of foreign capital through liberalisation is equivalent to raising the savings rate which
can be shown to increase GDP. This aligns with the propositions that foreign capital/investment
is mainly sought to fill the savings cum investment gap in the domestic economy. A typical
Harrod-Domar Model has five apriori assumptions viz;
Y = f ( K ) ; which denotes Output as a function of capital stock(1) .............................................. (1)

dY dY Y ................................................................................................................ (2)
=c⇒ = (2)
dK dK K
Equation (2) above implies marginal product of capital is constant; the production function
exhibits constant return to scale. This implies capital’s marginal and average products are equal.
f (0) = 0 ; which suggests Capital is necessary for output(3) ......................................................... (3)
sY = S = I (4).............................................................................................................................. (4)
Equation (4) implies that investment is a multiple of the savings rate and output.
ΔK = I − δK (5) .......................................................................................................................... (5)
Equation (5) denotes the change in capital stock equals investment less depreciation of the
capital stock.
These assumptions thus generate equal growth rates between the two variables. That is;
Y = cK ⇒ d log(Y ) = d log(c) + d log( K ) (6) ................................................................................ (6)
Since the marginal product of capital, c, is a constant, we have;
dY dK . .
d log(Y ) = d log( K ) ⇒ = ⇒ Y = K ..............................................................................
(7) (7)
Y K
Next, with assumptions (4) and (5), we can find capital’s growth rate as;
. 1 Y
K = − δ = s − δ ..............................................................................................................
(8) (8)
K K
.
(9)
⇒ Y = sc − δ .......................................................................................................................... (9)
In summation, the savings rate multiply by the marginal product of capital minus the
depreciation rate equals the output growth rate. Increasing savings rate, increasing the marginal
In summation, the savings rate multiply by the marginal product of capital minus the
product of capital, or decreasing the depreciation rate will increase the growth rate of output;
depreciation
these are the rate equals
means the output
to achieve growth
growth inrate.
the Increasing
Harrod Domar savings rate, The
model. increasing
model the marginal
implies that
economic growth depends on policies to increase investment, by increasing saving, andoutput;
product of capital, or decreasing the depreciation rate will increase the growth rate of using
theseinvestment
that are the means to achievethrough
more efficiently growth technological
in the Harrodadvances. Domar model. The model implies that
economic growth depends on policies to increase investment, by increasing saving, and using
that investment more efficiently through technological advances.
19
In modification, however, the Harrod (1936) and Domar (1945) assumption (precisely, th
In modification,
modification, however,
In however, the the Harrod
Harrod (1936)
(1936) and Domar
and Domar (1945)(1945) assumption
assumption (precisely,
(precisely, the th
assumption
assumption stated
stated in equation
in in
equation 4 above) that the product of the savings rate and output
fourth assumption stated equation44 above) thatthethe
above) that product
product ofsavings
of the the savings
rate andrate and output
output
saving equals investment (i.e. sY = S = I ) can be re-stated as the product of added savi
savingsaving
equals investment sYsY
(i.e.(i.e. = S= =S I=)I can bebere-stated as the
theproduct
product of added savi
equals
foreign equals
capital ratesinvestment
to output equals savings) can
and re-statedcapital
foreign as which of added
then equals
foreignand
savings capital rates
foreign to rates
capital output equals
to output savings
equals andandforeign
savings foreign capital which
capital which then
then equals inve
equals inve
That
That is;
is;
investment. That is;
(( ss + cf
cf )))YY
Y == sY
sY ++ cfY (10)
= III (10)
cfY == (10)
s ++ cf
(This equation
= sY +cfY
modifies
................................................................................................ (10)
the
This equation modifies
This equation modifies the original the original
original
Harrod-Domar
Harrod-Domar
Harrod-Domar model from
model
model
a closed
from
from aa closed
economy closed economy to
to aneconomy
open to aa
economy
economythrough
economy
through
through the
thethe
infusion
infusion
infusion
of foreign
of foreign
of foreign
capital.
capital.capital.
Following
Followingthethe
Following the derivation
derivation
derivation in the
the original
inoriginal
in the original Harrod-Domar
model, wemodel,
Harrod-Domar
Harrod-Domar model, we obtained
wea re-stated
obtained aa re-stated
obtainedmodelre-stated m
m
equations
of equations(8)
(8) and
and (9)
(9)
equations (8) and (9) as; as;
as;
.
. 11 Y
Y δ (11)
K
K= =K− − δδ = = (( ss +
+ cf
cf )) K − − δ (11)
.......................................................................................... (11)
.
K K
.
⇒
⇒Y =Y = (( sc
sc ++ cfc
cfc)) − − δδ (12)
....................................................................................................... (12)
(12)
Equation
Equation(12) (12)
(12) show
show that the the additions of
of the savings rate
rate (s) (s)bymultiplied by thethe marginal
Equation show that that
the additions additions
of the savings theratesavings
(s) multiplied multiplied
the marginalbyproduct marginal
of
of capital
capital (c)
(c) and
and the
the foreign
foreign capital
capital rate
rate (
( cf )
) multiplied
multiplied by the marginal product of cap
of capital (c) and the foreign capital rate ( cf ) multiplied cf by the marginal product of capital of cap
by the marginal product
and
andand
(c) less
less lessthe
thethedepreciation
depreciation
depreciation rate rate equals
equalsthethe
rateequals the output
output
output growthgrowth
growth rate. rate.
GivenGiven
rate. that thethat
Given that the
the diminishing
diminishing diminishing
assumption
returns
assumption assumptionof
of capital,
of capital,
capital, in
in the
in theneoclassical
the hypothesis,
neoclassical hypothesis,
neoclassical hypothesis,presupposespresupposes
presupposes that
that only only
only temporary
that temporary temporary ee
growth
effect
growth is
of growthrealistic and,
is realistic
is realistic thus,
and, thus,
and, thus, validates
validates
validates the need
thetheneed for
needfor the
for the neoclassical
the neoclassical counterrevolution
counterrevolution fram
neoclassicalcounterrevolution fram
The
framework.latter theory
The latteris hinged
theory is on
hingedindividuality
on individuality
The latter theory is hinged on individuality and only minimal government involvem and only
and minimal
only minimal government
government involvem
regulatory
regulatory purposes
involvement for regulatory
purposes can make
make price
canpurposes can make
price appropriate. It
It should
price appropriate.
appropriate. should further
It should
further be
be noted
further be noted
noted that
that the
the mm
that the
Harrod-Domar modified Harrod-Domar
model, stated model,
in stated
equation in
(12)equation
above, (12) above,
which which
translates translates
to an to
open an economy
Harrod-Domar
open economy
model, stated in equation (12) above,
model ofcounterrevolution
the neoclassical counterrevolution
which translates
framework suggests
to there
thatis
an open economy
isexistence
in
of
of the
the neoclassical
neoclassical counterrevolution framework
framework suggests
suggests that
that there
there is in
in existence bb
existence
liberalisation both the liberalisation
effects effects of trade and capital (both market-based measures and
liberalisation
legal/policy effectsIt isof
measures). ofthesetrade
trade and
and capital
liberalisation capital
effects that
(both
(both market-based
market-based
guarantee inflow of foreign
measures
measures and lega
capital.and lega
measures).
measures). It is these
It istogether
these with liberalisation
liberalisation effects
effects that guarantee inflow of foreign capital.
capital. Put
Putting all these equation (12) abovethat being guarantee
divided through inflowbyofYforeign and foreign Put
these together
these together
capital is made subject with equation
with equation
of the formular; (12) above
(12) above being
we havebeing divided
divided
behavioural through
through
equation of theby Y
by form; and foreign
Y and foreign capital capital
subject
subject of of the
the formular;
formular; we we havehave behavioural
behavioural equation equation of of the
the form;
form;
FDI = f (CAOPEN, KAOPEN, TOPEN, CPS_GDP, INFR, GOV _ CONSUM) ...................13
FDI = ff ((CAOPEN
FDI = CAOPEN ,, KAOPEN
KAOPEN ,, TOPEN
TOPEN ,, CPS
CPS _
_ GDP
GDP,, INFR
INFR,, GOV
GOV _ CONSUM )) (13)
_ CONSUM (13)
Where; FDI is Foreign Direct Investment, CAOPEN is the market-based capital account
liberalisation
Where; FDI measure,
is TOPENDirect
is the measure of tradeCAOPEN
liberalisation, CPS_GDP is the credit capital
Where;
to the private is Foreign
FDIsector Foreign
as a ratio Direct
of GDP,
Investment,
Investment,
KAOPEN is CAOPEN
the legal/policy
is the
the ofmarket-based
isindex market-based
capital account capital
liberalisation measure,
liberalisationInmeasure, TOPEN
TOPEN is the measure
is the measure of trade
of trade liberalisation,
liberalisation, CPS_GDP
CPS_GDP is the c
liberalisation. the neoclassical counterrevolution equation above, “Liberalise” captures both is the c
the private
the private sector
sector as a ratio
as a accounts of GDP, KAOPEN
ratio of liberalisation
GDP, KAOPEN is the
is known legal/policy
the legal/policy index of capital
index of capital
capital accounts and current (otherwise as trade liberalisation).
liberalisation.
liberalisation. In
In the
the neoclassical
neoclassical counterrevolution
counterrevolution equation
equation above,
above, “Liberalise”
“Liberalise” captur
captur
capital
Besides, accounts
capital accounts and
since the and current accounts
current accounts
neoclassical liberalisation
liberalisation
counterrevolution (otherwise
(otherwise
hypothesis; known
which known as
as trade
that liberalisatio
trade
presupposes liberalisatio
all
forms of government intervention could be seen as detrimental to the proper workings of the
Besides,
economy, since
since the
Besides, validates neoclassical
thethe counterrevolution
use of the ARDL
neoclassical Bond Tests as hypothesis;
counterrevolution which
which presupposes
the most appropriate
hypothesis; technique for that
presupposes that all
all ff
government
this intervention
study; on the
government basis that the
intervention could
ARDL
could be seen
beBond as
seentest detrimental
as traces the timeto
detrimental the
the proper
todimension workings
of policy
proper on theof
workings of the
the ec
ec
validates
validates the use of the ARDL Bond Tests as the most appropriate technique for this st
dynamics the
of theuse of
economy theasARDL
evident Bond
in the Tests
CAL-FDI as the
nexus,most
the appropriate
above technique
behavioural for
equation this st
the
the basis
can that
that the
the ARDL
be re-specified
basis to further
ARDL Bond test
include
Bond traces
test the the
lagged
traces time
time dimension
the dependent of
of policy
and independent
dimension on
on the
the dynamic
policyvariables. dynamic
It becomes as
economy imperative to trace CAL-FDI
the time dimension of policy due to the lags equation
evident along
economy
policy
as evident
evident
formulations;
in
in the
the CAL-FDI
implementation and
nexus,
nexus, the
the above
reaction/effect
above
(see
behavioural
behavioural
Anyanwu,
equation
Oyefusi,
can
can be
Oaikhenan
be re-sp
re-sp
to
to further
further include
include the
the lagged
lagged dependent
dependent and
and independent
independent variables.
variables. It
It becomes
becomes imperative
imperative
and Dimowo,
the time 1997). Stemming
dimension of policy from
due tothethe
foregoing and in along
lags evident tandempolicy
with the neoclassical impleme
formulations;
the time dimension of policy due to the lags evident along policy formulations; impleme
and
and reaction/effect
reaction/effect (see(see Anyanwu,
Anyanwu, Oyefusi,
Oyefusi, Oaikhenan
Oaikhenan and
and Dimowo,
Dimowo, 1997).
1997). Stemming
Stemming ff
foregoing
foregoing and in
in tandem
andJournal tandem with
with the
the neoclassical
neoclassical counterrevolution
counterrevolution framework,
framework, the app
BOJE: Botswana of Economics 18 the app
20
20
counterrevolution framework, the appropriate ARDL formulation of our behavioural model
specified above is compactly re-specified below thus:
ARDL formulation of our behavioural model specified above is compactly re-specified below
thus:
n n n
ΔFDIt = α0 yt −1 + ∑ αjΔFDIt − j + ∑ µ j ΔCAOPENt − j + ∑ ρjΔZt − j + δ1FDIt −1 + δ2CALt −1 + δ3Zt −1 + εt .........
(14) (14)
j =1 j =1 j =1
n n n
ΔFDIt = α0 yt −1 + ∑ αjΔFDIt − j + ∑ µ j ΔKAOPENt − j + ∑ ρjΔZt − j + δ1FDIt −1 + δ2 KAOPENt −1 + δ3Zt −1 + εt ..(15)
(15)
j =1 j =1 j =1
Where;
Δ is the first differencing operator; yt −1 is the initial value of FDI; Z t −1 is the collection of other
control variables (government size and the per capita GDP) directed towards making price right;
εt is the white noise error term; Expanding the Z t −1 so as to include the specific as well as
explanatory control variables; we have:
n n n
ΔFDI t = α 0 yt −1 + ∑ αjΔFDI t − j + ∑ µ j ΔCAOPENt − j + ∑ ρjΔZt − j + δ1 FDI t −1 + δ 2CAOPENt −1 +
j =1 j =1 j =1
(16)(16)
δ3TOPENt −1 + δ4 INFRt −1 + δ5CPS _ GDPt −1 + δ6OIL _ EXPORTt −1 + δ7GOV _ CONSU t −1 + .............
δ8 INTRt −1 + δ8 ECHRt −1εt
Where; CAOPEN – De-facto (market-based) measure of Capital Account Liberalisation (CAL);

TOPEN – is Trade Openness which serves as the measure of Current Account Liberalisation;
INFR – Inflation rate; it serves as the measure of price stability; CPS_GDP: It measures the level
of financial development as well as the extent of financial deepening in the country. This is the
ratio of credit to the private sector to the Gross Domestic Product (GDP); GOV_CONSU: This is
a measure of government size – A way to capturing government involvement in the workings of
the economy; OIL_EXPORT: This is the measure of natural resource endowment in Nigeria.
n n n
ΔFDI t = α 0 yt −1 + ∑ αjΔFDI t − j + ∑ µ j ΔΚAOPENt − j + ∑ ρjΔZt − j + δ1FDI t −1 + δ 2 KAOPENt −1 +
j =1 j =1 j =1
(17)(17)
δ3TOPENt −1 + δ4 INFRt −1 + δ5CPS _ GDPt −1 + δ6OIL _ EXPORTt −1 + δ7GOV _ CONSU t −1 + .............
δ8 INTRt −1 + δ8 ECHRt −1εt
Where; KAOPEN – De-Jure (legal/policy) measure of Capital Account Liberalisation (CAL) and
Where; KAOPEN – De-Jure (legal/policy) measure of Capital Account Liberalisation (CAL)
other variables are as previously defined.
and other variables are as previously defined.
The measurements of capital account liberalisation have been conducted under two forms in the
The measurements
literature. One form is ofthe
capital account
indirect liberalisation
way; through have been
the dimension conducted
of capital under
control two forms
(indicated here
in the literature. One form is the indirect way; through the dimension
as KAOPEN); otherwise known as de jure (or rule-based) index while the direct form, of capital control
as a
(indicated here
market-based as KAOPEN);
measure, relates tootherwise known as
de facto measure deKose,
(see jure (or rule-based)
Prasad index 2008).
and Terrones; while the
De
directindex
jure form, is as
theaKAOPEN
market-based
indexmeasure,
developedrelates to de and
by Chinn facto
Itomeasure (see Kose,
(2006; updated 2010) Prasad
whichand
is
Terrones;
based 2008).
on four De jure index
(4) indicators such asis multiple
the KAOPEN index
exchange rate,developed by Chinn
current account, and Ito
surrender of (2006;
export
updated 2010)
proceeds which isaverage
and five-year based of on controls
four (4) on
indicators such as multiple
capital transactions exchange
(see Kamar, rate, current
undated; Asiedu
and Lien,
account, 2004).of export
surrender The KAOPEN
proceeds and index usedaverage
five-year in this study onwas
of controls capitalsourced from
transactions
http://graduateinstitute.ch/md4stata/datasets.html.
(see Kamar, undated; Asiedu and Lien, 2004). De Thefacto measure,
KAOPEN as the
index ratioinofthis
used foreign
studydirect
was
investment
sourced from stock to Nominal GDP values, is the direct measure ofDe
http://graduateinstitute.ch/md4stata/datasets.html. capital
factoaccount
measure, liberalisation.
as the ratio
Both FDI Stock
of foreign directand Nominal stock
investment GDP to Values
Nominal- (inGDP
US$)values,
were sourced from an
is the direct online of
measure database:
capital
www.unctad.org/fdistat.
account liberalisation. Both The CAOPEN
FDI Stockisand positively
Nominal related
GDPwhile
Valuesthe- (in
KAOPEN
US$) were is negatively
sourced
from an online database: www.unctad.org/fdistat. The CAOPEN is positively related while
21

the KAOPEN is negatively related to FDI. In accordance with the modified Harrod-Domar
framework (see Skinner, 2011), the TOPEN; which is trade or current account openness, is the
related to FDI.envisaged
precondition In accordance withaccount
for capital the modified Harrod-Domar
liberalisation, framework
hence, positively (see to
related Skinner,
FDI. In2011),
the TOPEN;
tandem withwhich is trade or counter-revolution
the neoclassical current account openness, framework,is the
INFRprecondition envisaged
(as an indicator for capital
for price
level inliberalisation,
account the economy) is the rate
hence, of inflation
positively and it
related toisFDI.
expected to be positively
In tandem related to FDI.
with the neoclassical counter-
revolution framework, INFR (as an indicator for price level in the economy) is thetorate of
CPS_GDP is the indicator for financial deepening and expected to be positively related
FDI while
inflation andGOV_CONSUM,
it is expected as to abe proportion
positively of public
relatedspending,
to FDI. isCPS_GDP
an indicator is for
thethe level for
indicator
of government involvement in the economy. It is expected
financial deepening and expected to be positively related FDI while GOV_CONSUM, to be negatively related to FDI. as a
Accordingof
proportion to public
Dunning (1976), natural
spending, resource endowment
is an indicator for the level remains one of the involvement
of government attractions to in the
foreignersItinistrading
economy. abroad.
expected to be INTR denotes
negatively the rate
related of return
to FDI. on investment
According to Dunningand as(1976),
positednatural
by the simple Keynesian framework, the rate of return determines
resource endowment remains one of the attractions to foreigners in trading abroad. the cost of capital which INTR
suggests to investors on where to domicile their ‘money; EXCHR
denotes the rate of return on investment and as posited by the simple Keynesian framework, the is the cost of foreign
currency
rate in Nigeria’s
of return determinesNaira.
the Itcost
is taken as thewhich
of capital rate of exchange
suggests to of Naira toonDollar;
investors whereDollar
to domicile
being international reference currency. These data are sourced
their ‘money; EXCHR is the cost of foreign currency in Nigeria’s Naira. It is taken from the WDI (2012) andasCBN
the rate of
Statistical Bulletin (2012).
exchange of Naira to Dollar; Dollar being international reference currency. These data are
sourced from the WDI (2012) and CBN Statistical Bulletin (2012).
4.0 ESTIMATIONS AND DISCUSSION OF FINDINGS
4.0 Estimations and Discussion of findings
4.1 TRENDS OF CAPITAL OPENNESS AND CAPITAL RESTRICTIONS IN NIGERIA
4.1 Trends of Capital Openness and Capital Restrictions in Nigeria
The periods prior to 1995 indicates that Nigeria began with a relaxed capital control policy
The periods prior to 1995 indicates that Nigeria began with a relaxed capital control policy
which was relatively improved upon with a coefficient of 0.16097 in 1980 to 0.2431 in 1998 but
which was relatively improved upon with a coefficient of 0.16097 in 1980 to 0.2431 in 1998
since 1989 up until 1995; she employed more stringent policies toward capital account
but since 1989 up until 1995; she employed more stringent policies toward capital account
transactions. However, the market-based agenda for capital openness shows a continuous
transactions. However, the market-based agenda for capital openness shows a continuous
increase
increase trendfrom
trend from1980
1980withwith0.0264
0.0264ratios
ratiostoto0.5373
0.5373inin1995.
1995.InIneffect,
effect,this
thisimplies
impliesthat thatNigeria
embarked
Nigeria embarked upon friendly market-based agenda but unstable capital control policies. the
upon friendly market-based agenda but unstable capital control policies. However,
Post-1995
However, trend indicatestrend
the Post-1995 Nigeria continued
indicates Nigeria on a similar
continued on apattern
similar of market-based
pattern of market-capital
openness agenda Pre-1995 period. It began with a ratio of 0.5698 in 1996
based capital openness agenda Pre-1995 period. It began with a ratio of 0.5698 in 1996 to 0.6266 in 1999to and
thereafter
0.6266 indecline
1999 andcontinuously from a continuously
thereafter decline ratio of 0.5128frominathe year
ratio of 2000
0.5128toin0.3286
the year in 2000
2011. This
sharply contrasts with the legal/institutional policies towards capital control
to 0.3286 in 2011. This sharply contrasts with the legal/institutional policies towards capital (as proxied by
KAOPEN) with strict control in 1996 and continuous relaxation of these
control (as proxied by KAOPEN) with strict control in 1996 and continuous relaxation of policies to a coefficient
ofthese
0.3077 in 2011.
policies Stemmingoffrom
to a coefficient 0.3077theinforegoing, it is observed
2011. Stemming that Nigeria
from the foregoing, it ishas employed a
observed
good and an has
that Nigeria alternate
employedmixaof both
good andlegal/institutional
an alternate mix ofpolicies with market-based
both legal/institutional agenda
policies with in her
capital accountagenda
market-based liberalisation efforts
in her capital for the
account periods under
liberalisation efforts review. This behavior
for the periods under review. is rightly
illustrated in Figure
This behavior 1 below.
is rightly illustrated in Figure 1 below.
Figure
Figure1:1:Capital
CapitalAccount
Account Openness and
and Restrictions
Restrictions(KAOPEN)
(KAOPEN)ininNigeria:
Nigeria:1980-2011
1980-2011
0. 8
0. 6
0. 4
0. 2
0. 0
80 85 90 95 00 05 10
CAO PEN KAO PEN

Source: Author
Source: Author
22

4.2 UNIT-ROOT ESTIMATIONS
The first step in any cointegration technique is to determine the degree of integration of
each variable in the model (Bahmani-Oskooee and Nasir; 2009). Therefore, we conducted
stationarity as well as unit-root tests to ascertain the time-series characteristics of the data.
Also, we provide evidence valid for the justification of Autoregressive Distributed Lag (ARDL)
Bound test as a cointegration technique. As posited by Pesaran et. al., (2001), ARDL Bound
test is more suitable for variables at different order of integration while the Engle-Granger
Cointegration technique is considered suitable for series with same or uniform integration
order.
Table 1: Augmented Dickey Fuller (ADF) Test of Unit-Root

Variables Augmented Dickey Fuller (ADF) Series Ho: I(1)
At Levels At Order 1 Order of Integration
KAOPEN -1.2780 -4.5851 I(1)
CAOPEN -1.5102 -4.4567 I(1)
FDI -4.0480 - I(0)
TOPEN -0.9463 -7.3598 I(1)
INFR -2.9680 -7.3221 I(1)
OIL_EXPORT -5.7148 - I(0)
CPS_GDP -1.7163 -4.6291 I(1)
GOV_CONSU -3.8707 - I(0)
EXCHR -4.5693 - I(0)
INTR -3.0021 - I(0)

Mackinnon Critical Values: 1%: -3.7076 10%: -2.6290 5%: -2.9798
Evidenced from Table 1 above shows that the test of stationarity is mixed among the variables
to be included in our model. While five variables are without unit-root; that is, stationary at
levels – I(0), the remaining five variables are integrated at an order one, I(1), before they
could become stationary. Both measures of Capital Account Liberalisation; CAOPEN and
KAOPEN, have to be differenced at order one before they could be stationary. Also, trade
openness (proxied as TOPEN) which serves as the measure of current account liberalisation;
along side with trade openness (otherwise known as current account liberalisation) which is
proxied as TOPEN, inflation rate (proxied as INFR) and the ratio of credit to the private sector
to the GDP (proxied as CPS_GDP) are also differenced at order one; I(1), before they could
become stationary.

For the other series such as the measure of Foreign Direct Investment (FDI) – proxied as FDI
– which is just the ratio of FDI to GDP coupled with the measures of government size (proxied
as GOV_CONSUM); natural endowment (proxied as OIL_EXPORT), the foreign exchange
rate (proxied as EXCHR) and the rate of return on capital invested (proxied as INTR), they
could not be employed for analyses without resulting in spurious regression and, as such, were
difference at order 1. The fact that those series differenced at an higher order, I(1), would be
used under the same modeling framework with stationary series at levels lend credence to
the use of the Autoregressive Distributed Lag (ARDL) model. As posited by Pesaran et. al.,
(2001), the Engle-Granger Cointegration is found suitable for series of the same differenced
order while the ARDL model can be used to obtain the long-run equilibrium condition of
variables irrespective of the differencing order.
4.3 GRANGER CAUSALITY ESTIMATES
We proceed further to ascertaining which of these variables Granger-causes the other. That is,
from which of the variables do impact analyses flow from. We take to the Granger (1969) due
to its simplicity and also because it is less costly in terms of degrees of freedom (Charemza
and Deadman, 1997). In doing this also, we observe two lag periods as the maximum lag
length adequate for the study; as informed by the optimum lag length selection criteria of
Akaike and Schwarz. The data points employed for this study is annual time-series spanning
1980-2012.
Table 2: Granger Causality Results

Null Hypotheses* Lag Period of 1 Lag Period of 2
F-statistic Probability F-statistics Probability
FDI does not Granger cause
0.0018 0.97 0.4471 0.64
CAOPEN
CAOPEN does not Granger
0.2161 0.64 0.0792 0.92
cause FDI
KAOPEN does not Granger
39.091 2.E-08 1.8620 0.16
cause CAOPEN
CAOPEN does not Granger
7.0801 0.009 0.7798 0.46
cause KAOPEN
KAOPEN does not Granger
0.9342 0.34 0.7292 0.48
cause FDI
FDI does not Granger cause
0.1156 0.73 0.5273 0.59
KAOPEN
Source: E-Views Output. Note: The choice of lag 2 was optimally chosen by Akaike Information Criterion (AIC) and
Schwarz Bayesian Criterion (SBC)
We restrict our discussion of Granger causality test only to our main area of interest; which
are the direction of impact between the legal/policy index of Capital Account Liberalisation
(proxied as KAOPEN) and Foreign Direct Investment, on the one hand, and its counterpart
indicator of market-based measure (proxied as CAOPEN). Under the null hypothesis of no
causality; we do not reject the null hypothesis of no causal link from CAOPEN to FDI and vice
versa; at both lag periods. Similarly, we do not reject absence of causal link from KAOPEN

to FDI at both lag periods. This suggests that Capital Account Liberalisation (CAL) does
not granger causes Foreign Direct Investments (FDI) in Nigeria. However, the test shows
that legal/policy index of Capital Account Liberalisation Granger causes the market-based
measure of Capital Account Liberalisation at the 5 percent level of significance since the
null hypothesis of no Granger causality is rejected with F-statistics ratio of 4.35 with 0.025
probability.
4.4 ARDL ESTIMATES
F-statistics for Testing the Existence of Long-Run Relationship

Table 3: CAOPEN AND FDI
Test Statistics CAOPEN AND FDI KAOPEN AND FDI
Computed F-statistics 1.96* 1.68*
Bound Testing Critical Values at 5% Upper Bound: 4.01 Upper Bound: 4.01
Lower Bound: 2.86 Lower Bound: 2.86
Source: Pesaran et. al., (2001). * denotes accepting the null hypothesis of no cointegration at 5 percent level. The range of
the critical values at 1 percent and 10 percent levels are 5.06; 3.76 and 3.52; 2.45 respectively.
The computed F-statistics ratio is obtained in the OLS estimates detailed in Table 3 above and
then compared with the Bound Testing critical values at the 5 percent level of significance. The
F-statistics ratios obtained are insignificant with coefficients of 1.96 and 1.68 for the market-
based measure and the legal/policy index of capital account liberalisation. These are lesser
than the lower bound critical value of 2.45; therefore, implying that the null hypothesis of no
cointegration should not be rejected even at the 10 percent level of significance. This is so in
that it is also lower than the 5 percent lower bound critical value of 2.86. The implication of
these estimates is that market-based (i.e De-Facto) measure of Capital Account Liberalisation
(proxied as CAOPEN) and Foreign Direct Investment (proxied as FDI) do not have equilibrium
condition that could keep them together in the long-run situation. Similarly, the legal/policy-
based (i.e De-Jure) index of Capital Account Liberalisation (proxied as KAOPEN) does
not have equilibrium condition that could also keep it together in the long run with FDI in
Nigeria. It implies that both capital account liberalisation and foreign direct investment are
independent of each other in Nigeria. This further lend credence to the estimates obtained under
the relationship between policy measures undertaken by the government towards liberalizing
capital account transactions in stimulating FDI; which also showed independent relations.

4.5 RESULTS OF ESTIMATED LONG-RUN AND SHORT-RUN COEFFICIENTS
USING THE ARDL APPROACH
4.5.1 ESTIMATIONS OF LONG-RUN COEFFICIENTS USING ARDL APPROACH
Table 4: ARDL Bound Test (Long Run) Estimates.

Model 1: Market-based Measure Model 2: Legal/Policy Index
ARDL Optimal Ordering: (2, 2, 2, 1, 1, 0, 1, 0, 1, 1) ARDL Optimal Ordering: (2, 2,
2, 1, 1, 0, 1, 0, 1, 1).
Variable Coefficient T-Stat Prob. Coefficient T-Stat. Prob.
C -10.994 -4.153 0.001 -12.029 -3.588 0.004
FDI(-1) -0.368 -2.396 0.032 -0.237 -1.197 0.257
FDI(-2) 0.086 0.540 0.598 0.189 1.079 0.304
CAOPEN(-1) 14.100 3.166 0.007 - - -
CAOPEN(-2) -15.093 -3.334 0.005 - - -
KAOPEN(-1) - - - -7.664 -1.839 0.093
KAOPEN(-2) - - - 13.129 1.966 0.075
GOV_CONSUM 38.181 3.325 0.006 38.943 3.179 0.009
GOV_CONSUM(-1) -20.930 -1.701 0.113 -6.040 -0.354 0.730
INFR 3.749 2.542 0.025 -0.238 -0.052 0.960
INFR(-1) 7.736 5.079 0.0002 -2.521 0.452 0.660
OIL_EXPORT 0.0844 1.966 0.071 -5.64E-05 -0.0009 0.999
INTR -3.201 -2.173 0.049 0.758 0.165 0.872
INTR(-1) -7.649 -4.996 0.0002 2.736 0.489 0.635
TOPEN 0.238 1.907 0.079 0.447 2.404 0.035
CPS_GDP -0.227 -2.844 0.014 -0.077 -0.780 0.452
CPS_GDP(-1) -0.434 4.578 0.0005 0.140 0.780 0.452
EXCHR -0.028 -1.403 0.184 -0.049 -1.921 0.081
EXCHR(-1) -0.004 -0.236 0.817 -0.0006 0.026 0.979
R 2
0.898 0.88
Adj. R2 0.772 0.707
DW Stat. 1.82 2.74
F-statistic ratio 7.122 5.083
Prob.(F-statistics) 0.000473 0.00467
Source: E-Views Output; Note: FDI is the Dependent Variable for both Models. The optimal ARDL ordering is conditioned on
the empirical result that does not violate any decision criteria threshold and relatively produce improved empirical outcomes.
The coefficients of long-run impacts for both the market-based measures and legal/policy
index are as detailed in Table 4 above. Essentially, the estimates indicate that liberalisation on
capital account transactions appear to be time-dependent. The one-period lag effect of market-
based measure is significantly positively related to foreign direct investment in Nigeria with
14.100 coefficient and 0.007 probability value while its two-period lagged effect is significantly

negatively related with -15.093 coefficient and 0.005 probability value. This suggests that
capital account liberalisation is positively related to FDI in the immediate long-run but this
could not be sustainable into the foreseeable future. In contrast, the contemporaneous effect
of legal/policy index show significant negative effect of -7.664 coefficients with probability
value of 0.093 for one-period lag but 13.129 coefficients with 0.075 probability values for
the two-period lag. While the coefficients for the market-based measures are significant at
the 5 percent level, those of the legal/policy index are significant at the 10 percent level. This
supports the findings that legal/policy measures granger causes market-based measures in
Nigeria and that both works in opposite directions. The implication of these estimates is that
even though capital account liberalisation and foreign direct investment in Nigeria do not have
equilibrium conditions that keep them together into the long-run situation, liberalisation of
capital account transactions impact significantly on FDI in the long-run in Nigeria.
The level of price stability (as proxied by INFR) has positively affected foreign direct
investment in Nigeria under a market-based measure of capital account liberalisation
irrespective of the time-dimension (that is, either under the present or immediate past price
level). The current rate of inflation has 3.749 coefficients with 0.025 probability values while
the immediate past inflation rate has 7.736 coefficients with 0.0002 probability values. These
sharply contrast with the coefficient of inflation rate under the legal/policy measures which
are negatively related with -0.0238 and -2.521 coefficients for current and immediate past
periods respectively. These are insignificant; not even at the 10 percent level with 0.960 and
0.660 probability values respectively. More so, government involvement in the economy
(proxied as GOV_CONSUM) and trade (current account) liberalisation (proxied as TOPEN)
are positively related to FDI; especially for the market-based measure. TOPEN has 0.238
coefficient and 0.079 probability value for market-based measure but 0.447 coefficient and
0.035 probability values. The former is significant at the 10 percent level while the latter
at the 5 percent level. The implication is that the liberalisation of trade or current account
transactions; like export and import, impact significantly on foreign direct investment in
Nigeria; in the long-run situation.
On the other hand, the extent of natural resource endowment (proxied as OIL_EXPORT) is
positively related to FDI with 0.0844 coefficient and probability value of 0.071 for the market-
based measure but -5.64E-05 coefficient and 0.999 probability value for the contemporaneous
values of legal/policy index. The level of financial deepening (proxied as CPS_GDP)
in the Nigeria economy does not favour FDI either under market-based measure or legal/
policy index but highly significant for the former. The CPS_GDP has -0.227 coefficient
with 0.014 probability value for the immediate period and -0.434 coefficient with 0.0005
probability value for the past period under the market-based measure while under the legal/
policy measures, the contemporaneous coefficients for immediate and past periods are -0.077
and 0.140 with 0.452 and 0.452 probability values respectively. The implication is that the
extent of financial development in Nigeria has not been favourable to enhancing investment
from foreign countries into the country. The same patterns of behaviors are evident for the
relationship between exchange rates (proxied as EXCHR) and FDI, on the one hand, and,
interest rate (proxied as INTR) and FDI, on the other hand. The interest rate as a measure of the
rate of return on investment in Nigeria has -3.201 and -7.649 coefficients for immediate and
past periods of market-based respectively coupled with 0.049 and 0.0002 probability values.
In effect, the level of inflationary tendencies as well as price instability in the country has
endangered foreign direct investment. Also, the extent of financial development and the rate
of return on investment have not enhanced the attraction foreign investment into the country.

The adjusted R2 obtained for both models 1 and 2 are 0.772 and 0.707 respectively. This show
that the variables included in market-based model (i.e Model 1) accounts for 77.2 percents
movement in FDI while that included in the legal/policy index model (i.e Model 2) explains
for 70.7 percent movement in the FDI. The implication is that the models do not suffer from
any misspecification error. Complementing this is the F-ratio statistics with 7.12 and 5.083
ratios for models 1 and 2 respectively. These are highly significant at the 5 percent levels of
significance with 0.000473 and 0.00467 coefficients respectively; lending credence to the
conclusion that the model has goodness of fit. More so, the Durbin Watson (DW) statistics of
1.82 and 2.74 statistics imply that the models are free from autocorrelation or serial correlation
problem.
From empirical standpoints, the findings in this study supports the results obtained in the
studies of Montiel and Reinhart (1999); Alfaro et al. (2005), and Aizenman and Noy (2003)
where they concluded that imposing capital controls had no impact on volumes of flows but
did shift the composition of flows toward short term – ‘hot money’ – flows. Aizenman and
Noy (2006), found that while capital controls have no impact on FDI gross flows, controls on
the current account do have an indirect impact on FDI inflows through their impact on goods
traded. More so, the findings in this study contradicts that of Boamah, Craigwell, Downes and
Mitchell (2005) study where they provided some evidence that capital flows are significant
in explaining the movements in private investment boom in the region. Additionally,
Somphornsem (undated) found that capital market liberalization makes the process of mergers
and acquisitions easier, and increases alternative sources of capital for domestic companies.
Foreign companies that invest in a liberalized country can acquire funds not only from their
headquarters, but also from an IPO, or issue bonds which can create an incentive for foreign
investors to enter the domestic markets. According to the above reasons, he posited, capital
market liberalization should have a strong positive and significant effect on foreign direct
investment. On the whole, this study produce findings as well as estimates that support the
warning handed to developing economies by the IMF (1995) and World Bank (1997) that they
should sequence their openness on capital account transactions as well as control some of
these capital account items.

4.5.2 ESTIMATIONS OF SHORT-RUN COEFFICIENTS USING ARDL APPROACH
Table 5: ARDL-VECM (Short-Run) Estimates for Model 1: Market-based Model

Model 1: Market-based Measure
ARDL Optimal Ordering: (1, 0, 0, 0, 0, 1, 0, 0, 1)
Variable Coefficient T-Stat Prob.
C 3.667 2.689 0.014
FDI(-1) 1.100 8.382 0.000
D(CAOPEN) -9.428 -2.430 0.025
GOV_CONSUM -12.248 -2.036 0.055
D(INFR) -0.121 -1.696 0.105
OIL_EXPORT -0.052 -3.082 0.006
INTR(-1) -0.363 -5.015 0.0001
D(TOPEN) -0.151 -1.612 0.123
D(CPS_GDP) 0.029 0.674 0.508
EXCHR(-1) 0.009 1.409 0.174
ECM(-1) -0.960 -7.743 0.000
R2 0.902
Adj. R2 0.853
F-statistics ratio 18.35
Prob. F-statistics ratio 0.000
DW Statistics 1.67
Source: E-Views Output; Note: FDI is the Dependent Variable.
Table 6: ARDL-VECM (Short-Run) Estimates for Model 2: Legal/Policy-based Model

Model 2: Legal/Policy-based Measure
ARDL Optimal Ordering: (1, 0, 0, 0, 0, 0, 0, 0, 1)
Variable Coefficient T-Stat Prob.
C -2.498 -2.939 0.009
FDI(-1) 0.503 4.390 0.0004
D(KAOPEN) -10.558 -2.992 0.0078
GOV_CONSUM 11.771 1.925 0.0702
D(INFR) 3.137 1.522 0.145
OIL_EXPORT -0.026 -1.093 0.289
INTR -2.985 -1.456 0.163
D(TOPEN) -0.00031 -0.0023 0.998
D(CPS_GDP) 0.083 1.313 0.206
EXCHR(-1) 0.0098 1.1058 0.283
ECM(-1) -0.994 -6.568 0.000
R2 0.863
Adj. R2 0.788
F-statistics ratio 11.385
Prob. F-statistics ratio 0.000
DW Statistics 1.53
Source: E-Views Output; Note: FDI is the Dependent Variable.

Tables 5 and 6 above detailed the Vector Error Correction Models (VECM) within the
Autoregressive Distributed Lag (ARDL) framework for Models 1 and 2 respectively. These
estimates are optimal as they were based on the optimal values of Akaike Information Criterion
(AIC) selected amidst series of scenario analyses computed. For both Models 1 and 2, the
ECM coefficients are properly signed with -0.960 and -0.994 coefficients respectively. These
are highly significant too with absolute T-statistics values of 7.743 and 6.568 with probability
values of 0.000 respectively. These estimates confirm the long-run equilibrium conditions
evident among the variables of interest and further indicate that all disequilibrium conditions
in FDI are barely recovered on a yearly basis to a tune of 96 percent and 99.4 percent for the
market-based and legal/policy measures respectively.
Admittedly, liberalisation of the economy either of capital or trade are negatively related to
foreign direct investment in Nigeria. Both the market-based and legal/policy measures are
negatively significantly related to FDI in Nigeria with -9.428 and -10.558 coefficients with
probability values of 0.025 and 0.0078 respectively. Trade liberalisation is negatively related
to FDI with -0.151 and -0.00031 coefficients for both the market-based measures and legal/
policy index respectively. These findings contradict the neoclassical counter-revolution
framework but strongly support the dependency theory of capital account liberalisation. Also,
the amount of natural resources endowment (as proxied by OIL_EXPORT) and the return on
investment (proxied as INTR) are negatively related to FDI in Nigeria with -0.052, -0.363 and
-0.026 and -2.985 respectively. Both the natural resource endowment and return on investment
are significant for the market-based measure but insignificant for the legal/policy index of
capital account liberalisation at the 5 percent levels of significance.
On the other hand, the level of financial development as proxied by CPS_GDP follows
theoretical expectation of positive relation with 0.029 and 0.083 coefficients but insignificant
with 0.508 and 0.206 probability values for Models 1 and 2 respectively. More so, the exchange
rate is also directly related with 0.009 and 0.0098 coefficients with probability values of 0.174
and 0.283 for Models 1 and 2 respectively. More so, government involvement (as proxied by
GOV_CONSUM) through the market-based measure is significantly negatively related to the
foreign direct investment (FDI) with -12.248 coefficient and probability value of 0.055 while
its involvement through the legal/policy measure is positively related to FDI with 11.771
coefficient and 0.0702 probability value. While the former is significant at the 5 percent level,
the latter is significant at the 10 percent level. The adjusted R2 obtained for both models 1
and 2 are 0.853 and 0.788 respectively. This show that the variables included in market-based
model (i.e Model 1) accounts for 85.3 percent movement in FDI while that included in the
legal/policy index model (i.e Model 2) explains for 78.8 percent movement in the FDI. The
implication is that the model is properly specified and do not suffer from any specification
error. Complementing this is the F-ratio statistics with 18.35 and 11.385 ratios for models
1 and 2 respectively. These are significant at the 5 percent levels of significance and further
lend credence to the conclusion that the models are well fitted and of best fit. More so, the
Durbin Watson (DW) statistics of 1.67 and 1.53 statistics imply that the models are free from
autocorrelation or serial correlation problem.

4.6 DIAGNOSTIC AND ROBUSTNESS CHECKS
4.6.1 RESIDUAL TESTS
For robustness of results and the reliability of the estimates obtained, it becomes imperative
that we conduct some tests on the estimates obtained. Essentially, four of these tests stand out.
2 2
These are the Ramsey RESET test ( xRESET ); Jarque Bera test ( xNORMAL ); Breusch-Godfrey
2 2
LM test ( xSERIAL ) and the ARCH ( x ARCH ) tests. As detailed in the table 7 below, the estimates
obtained for this study are robust since the diagnostic checks suggest that the model is devoid
of specification bias and that the residual is normally distributed. More so, there is absence of
serial correlation for the overall models 1 and 2 and no evidence of autoregressive conditional
heteroscedasticity.
Table 7: Diagnostic Results

S/N Test Statistics Model 1: Market-based Model 2: Legal/Policy Index
Measure
1 0.0194 0.1727
2
x ARCH
(0.972) (0.842)
2 2.2271 1.2415
2
xSERIAL
(0.109) (0.3216)
3 1.4139 2.7514
2
xRESET
(0.316) (0.115)
4 0.0343 1.4713
2
xNORMAL
(0.7731) (0.512)
Source: E-Views Output. Note: Figures in parentheses are the probabilities of significance.
Specifically,the null hypotheses for these tests are that there is absence of autoregressive conditional heteroscedasticity – for the
ARCH test; the residuals are serially uncorrelated – for the Breusch-Godfrey test; the model does not suffer from functional
specification bias – for the Ramsey RESET test and that the residual is normally distributed – for the Jarcque-Bera test. In line with
these null hypotheses, these estimates suggest that these null hypotheses should all be accepted at the 5 percent of significance since
the probability values for these tests are all above 0.05.
4.6.2 STABILITY TESTS
In order to completely ascertain the reliability level of our estimates, stability tests of CUSUM
and the CUSUM sum of squared were conducted on the error correction estimates obtained.
These tests are considered more apt than the Chow test as it depicts how the estimates depart
or converge to their consistent level. As depicted in figures 2 and 3 below, the estimates lie
within the confidence interval at the 5 percent level of significance; thus, lending credence to
the stability of the model estimated.

Figure
Figure 2:
Figure 2: Stability
2: Stability Test
Stability Test for
Testfor Model
Model1:1:
forModel Market-based
1:Market-based Model
Market-basedModel
Model
15
15
10
10
5
5
0
0
-5
-5
-10
-10
-15
-15
1992 1994 1996 1998 2000 2002 2004 2006 2008 2010
1992 1994 1996 1998 2000 2002 2004 2006 2008 2010
CUS UM 5% S ignificance
CUS UM 5% S ignificance
Figure
Figure 3:
3: Stability
Stability Test
Testfor
forModel
Model2:2:Legal/Policy-based
Legal/Policy-basedModel
Model
Figure 3: Stability Test for Model 2: Legal/Policy-based Model
1.6
1.6
1.4
1.4
1.2
1.2
1.0
1.0
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
0.0
0.0 2008 2009
2008 2009
CUS UM of S quares 5% S ignificance
CUS UM of S quares 5% S ignificance
5.0 CONCLUSION
5.0
5.0 Conclusion and
Conclusion and Recommendation
AND RECOMMENDATION
Recommendation
In
In conclusion,
conclusion, capital
capital account
account liberalisation
liberalisation is
is found
found detrimental
detrimental to
to foreign
foreign direct
direct investment
investment in
in
In
the conclusion,
short-run capital
and bothaccount
could liberalisation
not equilibrium is found detrimental
conditions that to foreign
could keep direct
them investment
together into
thethe
in short-run
short-run andand both could
both could notnot
equilibrium
equilibrium conditions
conditionsthatthatcould
couldkeepkeepthemthem together
togetherinto the
the
long-run situation.
long-run situation. EvenEven though
though the the long-run
long-run impact
impact analyses
analyses suggest
suggest that that capital
capital account
account
into the long-run situation. Even though the long-run impact analyses suggest that capital
liberalisation
liberalisation
account is
is time-dependent,
time-dependent,
liberalisation is time-dependent, the
the result
result obtained
obtained
the result still support
stillstill
obtained supportthe
support thethehypothesis
hypothesis
hypothesis that that
that itit is
is
independent
independent
it is independent of the foreign
of theofforeign direct investment.
directdirect
the foreign investment. The involvement
The involvement
investment. The involvement of government
of government
of governmentin the workings
in the workings
in the of
of
the economy
economy
workings
the indicate
of theindicate
economythatthat itit is
indicate is that
antithetical to foreign
foreign
it is antithetical
antithetical to directdirect
to foreign
direct investment
investment
investment and and
and should
should
should only be
only be
included
only at
at minimal
be included
included minimal level
level for
at minimal regulatory
for level and
and policy
for regulatory
regulatory policyandguidelines. More
More so,
policy guidelines.
guidelines. the
the rate
so, More rateso,of inflation
ofthe rate has
inflation has
of inflation
alternate has
effect alternate
on FDI effect
in Nigeria;on FDI in
suggestingNigeria; suggesting
price has been price has
distortional;been distortional;
perhaps due
alternatedue
perhaps effect
to theoninvolvement
FDI in Nigeria; suggesting
of government price
in the has –been
market distortional;
as proposed by theperhaps due to the
neoclassical
to the
involvement of
involvement of government
government in in the
the market
market –– as as proposed
proposed by by the
the neoclassical
neoclassical counter-revolution
counter-revolution
counter-revolution theorists. On the whole, foreign direct investment in Nigeria is driven by
theorists.
theorists. On
non-capital On the
the whole,
account whole, foreign direct
direct investment
foreign factors
liberalisation investment in
in Nigeria
Nigeria is
but only liberalisation isondriven
currentby
driven non-capital
byaccount
non-capital account
account
(trade)
liberalisation
transactions factors
liberalisationisfactors but only
but for
effective liberalisation
onlyforeign
liberalisation on current
on current in
direct investment account
account
Nigeria. (trade)
(trade) transactions
The transactions is effective
is for
is effective
policy implication for
foreign
that direct investment
investment
non-liberalising
foreign direct insuch
factorsin Nigeria.
Nigeria. The policy
as qualitative
The policy implication
governance,
implication is that
that non-liberalising
aggressive
is non-liberalising
financial development, factors such
factors such
as
as qualitative
stability governance,
of prices
qualitative and sound
governance, aggressive financial
financial development,
fiscal management
aggressive stability
stability inof
should be promoted
development, prices
pricestoand
of order sound
sound fiscal
stimulate
and fiscal
foreign direct investment in Nigeria.
management should be promoted in order to stimulate foreign direct investment in Nigeria.
management should be promoted in order to stimulate foreign direct investment in Nigeria.
REFERENCES
REFERENCES
Adedipe,
Adedipe, B.
B. (2006),
(2006), “Capital
“Capital Account
Account Liberalisation:
Liberalisation: What
What Options
Options for
for Developing
Developing
Economies”,
Economies”, Central Bank of Nigeria Economic and Financial Review, Vol.
Central Bank of Nigeria Economic and Financial Review, Vol.
44, No.
BOJE: 4, pp.
Botswana 63-88.
Journal
44, No. 4, pp. 63-88.
of Economics 30
Aizenman,
Aizenman, J. J. &
& Noy,
Noy, I.,
I., (2006),
(2006), “FDI
“FDI and
and Trade:
Trade: Two
Two Way
Way Linkages?”
Linkages?” Quarterly
Quarterly Review
Review
of Economics and Finance, 46 (3), 317-337.
REFERENCES
Adedipe, B. (2006), “Capital Account Liberalisation: What Options for Developing

Economies”, Central Bank of Nigeria Economic and Financial Review, Vol. 44, No. 4, pp.
63-88.
Aizenman, J. & Noy, I., (2006), “FDI and Trade: Two Way Linkages?” Quarterly Review
of Economics and Finance, 46 (3), 317-337.
Aizenman, J. & Noy, I., (2003) in Aizenman, J. & Noy, I., (2006), “FDI and Trade: Two
Way Linkages?” Quarterly Review of Economics and Finance, 46 (3), 317-337.
Alfaro, L., Kalemi-Ozcan S., & Volosovych, V. (2006), “Why Doesn’t Capital Flow From
Rich to Poor Countries? An Empirical Investigation,” Review of Economics and Statistics.
Anyanwu, J.C., Oyefusi, A., Oaikhenan, H., & Dimowo, F.A., (1997). The Structure of
the Nigerian Economy. Joanee Eductional Publishers Ltd., Onitsha, Anambra State, Nigeria.
Asiedu, E. & Lien, D. (2004), “Capital Controls and Foreign Direct Investment” World
Development, Vol. 32, No. 3, pp. 479-490.
Bahmani-Oskooee, M. & Nasir, A. (2004), “ARDL Approach to Test the Productivity Bias
Hypothesis” Review of Development Economics, Vol. 8, No.3, Pp. 483-488.
Bauer, P., (1984). Reality and Rhetoric: Studies in the Economics of Development.
Weidenfeld and Nicolson, London.
Bauer, P.T. (1972). Dissent on Development. Weidenfeld and Nicolson, London.
Bilel, K; & Mouldi, D. (2011), “The Relationship between Financial Liberalisation, FDI
and Growth: An Empirical Test for MENA Countries” Economics and Financial Review, Vol.
1 (10), pp. 20-26.
Boamah, D; Craigwell, R., Downes, D. & Mitchell, T. (2005), “Does Capital Account
Liberalisation Lead to Investment Booms in Carribean Economies” Carribean Centre for
Monetary Studies, XXXVII Annual Conference.
Cobham, A. (2001), “Capital Account Liberalisation and Poverty” QEH Working Paper
Series – QEHWP570, No. 7, pp. 1-20.
Desai, A.M; Foley, C. F., & James, R. H. (2003). Capital Controls, Liberalisation and
Foreig Direct Investment. International Investment Division, Bureau of Economic Analysis,
U.S Department of Commerce.
Edison, H.J; Klein, M.W., Ricci, L. & Sloek, T. (2004), “Capital Account Liberalisation
and Economic Performance: Survey and Synthesis” IMF Staff Papers, Palgrave Macmillian
Journals, Vol. 51(2).
Eichengreen, B. (2001). Capital Account Liberalisation: What do the Cross-countries
Studies tell Us? University of California, Berkeley.
Greenidge (2008) in Peet, R; & Hartwick, E. (2009). Theories of Development –
Contentions, Arguments, Alternatives.The Guilford Publications Inc; New York, USA.
Henry, P. B. (2007), “Capital Account Liberalization: Theory, Evidence, and Speculation,
Journal of Economic Literature 45(4): 887-935.
Hsiao, C., (2003). Analysis of Panel Data.Second edition, Cambridge University Press.
IMF (2012), “Annual Report on Exchange Arrangement and Exchange Restrictions”,
Washington DC: The International Monetary Fund.
Jabbar, O.A. (2004), “Privatization and Development” Insights from A Holistic Approach,
with Special Reference to the Case of Jordan” Centre for Development Studies, University of
Salford, UK.
Johnson, H. G. (1977). The efficiency and welfare implications of the multinational
Corporation., in C. P. Kindleberger (ed.), The Multinational Enterprises, Theories International
Corporation: A Symposium, Cambridge, MA: MIT Press.

Kamar, B. (undated), “Does Capital Account Liberalisation Spur Economic and Financial
Performance? New Investigation for MENA Countries”
Kobrin, S. (2004) in Henry, P. B. (2007), “Capital Account Liberalization: Theory, Evidence
and Speculation, Journal of Economic Literature 45(4): 887-935.
Kose, M. A; Prasad, E.S. & Terrones, M. E. (2008), “Does Openness to International
Financial Flows Contribute to Productivity Growth?” Discussion Paper No. 3634, IZA,
Germany.
Mathieson, D.J; & Rojas-Suárez, L. (1993), “Liberalization of the Capital Account:
Experiences and Issues” IMF Occasional Paper No. 103.
Miller, H. (2004) in Antonio and Marek (2008), “Determinants of Economic Growth: Will
Data Tell” European Central Bank, Working Paper Series, No. 582.
Montiel, P. & Reinhart, C. (1999), “Do Capital Controls and Macroeconomic Policies
Influence the Volume and Composition of Capital Flows? Evidence from the 1990s” Journal
of International Money and Finance 18, pp. 619-35.
Noy, I; and T.B. Vu (2007), “Capital Account Liberalisation and Foreign Direct Investment”
University of Hawaii, Working Paper, No. 07-8.
Obadan, M.I. (2006), “Capital Account Liberalisation: Experience from the Emerging
Market Economies” Central Bank of Nigeria Economic and Financial Review, Vol. 44, No. 4,
pp. 185-220.
Okogu, B; & Osafo-Kwako, P. (2006), “Sequencing Capital Account Liberalisation”
Central Bank of Nigeria Economic and Financial Review, Vol. 44, No. 4, pp. 157-184.
Omoruyi, S. (2006), “Capital Account Liberalisation: The Way Forward for Nigeria”
Central Bank of Nigeria Economic and Financial Review, Vol. 44, No. 4.
Oyedije, T. A. (2006), “Globalisation and Capital Account Liberalisation” Central Bank of
Nigeria Economic and Financial Review, Vol. 44, No. 4, pp. 31-46.
Peet, R; & Hartwick, E. (2009). Theories of Development – Contentions, Arguments,
Alternatives.The Guilford Publications Inc; New York, USA.
Pesaran M.H., Shin, Y., Smith, R.J. (2001), “Bounds Testing Approaches to the Analysis of
Level Relationships”. Journal of Applied Econometrics 16: 289-326.
Prasad, E.S; Rogoff, K. Wei, S. & Kose, M.A. (2003), “Effects of Financial Globalisation
on Developing Countries: Some Empirical Evidence” IMF Occassional Paper, 220.
Reisen & Fischer (1993), in Peet, R; and E. Hartwick (2009), Theories of Development
-Contentions, Arguments, Alternatives.The Guilford Publications Inc; New York, USA.
Rodrik, D. (1998), “Who Needs Capital Account Convertibility”, Unpublished, Cambridge,
MA: Harvard University.
Sarode, S. (2012), “Effects of FDI on Capital Account and GDP: Empirical Evidence
from India” International Journal of Business and Management, Vol. 7, No. 8
Skinner, G. (2011), “The Neoclassical Counterrevolution and Developing Economies: A
Case Study of Political and Economic Changes in the Phillippines” Social Sciences Journal,
Vol. 7, Issue 1, Article 12.
Solow, R.M. (1956), “A Contribution to the Theory of Economic Growth” The Quarterly
Journal of Economics, Vol. 70, No. 1, pp. 65-94.
Toye, J., (1987). Dilemmas of Development: Reflections on the Counter-Revolution in
Development Theory and Policy. Basil Blackwell, Oxford.

Credit Demand Amongst Farmers in Mukono District, Uganda
George Wilson Ssonko1 and Marris Nakayaga
Abstract
The role of credit in promoting economic activity cannot be underestimated. Nevertheless,

credit extension to the agricultural sector in Uganda is dismal compared to other sectors.
This study uses cross sectional data collected from 127 farmers in Mukono District, Uganda to
shed some light on access to, and the characteristics of demand for credit among the farming
communities. We employ the binary logit model estimation to analyse demand for credit. The
empirical results suggest that the probability of a farmer demanding credit increases with
proximity to credit facility, easier application procedures, customary land tenure system and
membership to farmers’ association. In contrast, the likelihood of credit demand decreases
with increasing farm size. Policy options and recommendations including encouragement of
forming farmers’ associations, leveraging mobile money technologies to reduce distance, and
streamlining application procedures could bolster agricultural credit demand in Uganda.
Keywords: Credit demand; farmers; Mukono District; Uganda – East Africa
JEL Classification Codes: E 51, Q 140
1
Bank of Uganda, Uganda.
* Corresponding Author: gssonko@bou.or.ug

1. INTRODUCTION
Credit is an important component of a firm and household’s survival. It provides a smoother

flow of money in times when there are constrictions of cash flows that would otherwise cause
disruptions in production and consumption. Ray (1998) observes that credit is needed for fixed
capital, working capital, and consumption expenditure. Bhattacharjee and Rajeev, (2010) note
that ‘access to credit on favourable terms and conditions plays an important role in the overall
development of an economy and in poverty eradication’.
On account of the subsistence nature of Uganda’s agriculture coupled with low inputs, credit
extension to farmers would improve farm productivity thereby enhancing returns on investment.
Nonetheless, the inadequate access to credit by farming households due to supply and demand
side bottlenecks remains a major limiting factor to Uganda’s agricultural productivity. As a
result, official credit programmes have been used in an attempt to ensure that credit flows to
the agricultural sector to boost productivity (Atieno, 1997). In India, channeling of credit to
the agricultural sector dates back to the 1870s and culminated into the two tier cooperative
credit structure consisting of short-term and long term arms (Mohan, 2004).
The Government of Uganda (GOU) has had various attempts in the past to remedy the
bottlenecks in credit allocation to the agricultural sector. These attempts took the form of low
interest rate credit schemes such as Rural Farmers Credit Scheme, Entandikwa – “Start-up
Capital” Credit Scheme, Poverty Alleviation Programme, and Bonnabaggawale – “Prosperity
for all” Credit Scheme (Kasirye, 2007; Mpuga, 2004; Matovu and Okumu, 2010). The
performance of these schemes was varied. The timing of introducing the schemes, mechanism
of delivery, and political patronage are cited as some of the causes for the varied performance
(Kasirye, 2007). Despite these credit schemes, the supply of credit to the agricultural sector
does not meet the demand (Owusu-Antwi, 2010) such that a large proportion of farmers in
Uganda are left out (Finscope, 2010).
Studies that have dealt with the agricultural credit problem in Uganda have faced two major
limitations namely, concentration on the supply side of the credit markets and paucity of
reliable micro-level data to estimate credit demand. Furthermore, where data has been
available such as the Uganda National Household Surveys (UNHS) and FINSCOPE surveys,
it has been difficult to disaggregate credit demand data into the various purposes such as
agricultural credit. Consequently, the credit demand studies in Uganda such as Kasirye (2007)
and Mpuga (2008) have focused on the rural credit market in general. Conceptually, a number
of areas have been covered namely, credit demand (Kasirye, 2007), credit source (Mpuga,
2008), credit rationing (Okurut et al., 2005), credit constraints (Zeller, 1994), loan repayment
incentives (Nkurunziza, 2005), and defaults (Udry, 1994). Therefore, the objective of this
study was to predict the odds of a farmer applying for credit, conditional upon information
about personal characteristics of the farmer.
The paper is organized as follows. The methodology used to collect the data used for analysis
is described in the next section. The estimated model is presented in section three. Descriptive
and empirical results are discussed in section four, section five provides policy implications
and recommendations, and section six contains concluding remarks.

2. METHODOLOGY
The study was based on a survey conducted between February and March 2013 in the six parishes
of Katoogo, Bulika, Namawojjolo, Kasenge, Namubiru, and Mpoma in Nama Division, Mukono
District. These parishes were selected because they have been specifically targeted by the agricultural
extension officers for monitoring and evaluation of NAADS (National Agricultural Advisory Services)
programmes. At the parish level, farmers interviewed were picked using simple random sampling with
the parish records maintained by the agricultural extension officers as the sampling frame. Primary
data was collected using questionnaires from 127 farmers.The data collected included socio-economic
and demographic characteristics of households. Data were analysed using statistical techniques such
as descriptive statistics and logistic regression with the aid of the R-statistical programme software.
3. MODELLING CREDIT DEMAND
Demand describes a consumer’s desire, willingness, and ability to pay a price for a specific good or
service (Whelan & Msefer, 1996). Credit demand has various definitions from scholars such as Doan
et al., (2010), Diagne (1999), and Balogun and Yusuf (2011) amongst others. The most commonly used
definition is “the probability that an individual answered yes to the question ‘did you apply for credit
in the last time period?’ The farmers that applied for credit and never received were considered to be
credit constrained and those that got unconstrained.There are various factors affecting credit demand.
3.1 VARIABLES USED IN THE MODEL
Studies that have modeled agricultural credit demand have utilized different variables. Nwaru et al.,
(2011) used education level, household size, farm income, interest amount, gross farm profit of previous
year and farm size while Bokosi (2004) utilised formal education, per capita land owned, number of
animals owned per household, household size and gender. Explanatory variables used in credit demand
are underpinned by the Life Cycle Hypothesis, the Permanent Income Hypothesis, and Keynes’ theory
of consumption. The considerable amount of existing literature on credit demand concur that social,
individual, institutional, and economic factors are key determinants of credit demand (Mpuga, 2008;
Fichera, 2010; Kasirye, 2007; Ferede, 2012).The variables used in the model of this study and their
definitions are presented in Table 1.

Table 1: Definition of explanatory variables for credit demand model in Uganda
Variable Type Description Expected
Sign
AGEFARM Continuous Age of farmer in years ±
FARMAGE Continuous Number of years the farm has been in +
existence
MEMFARASS Categorical Whether an individual belongs to a +
farmers’ association (1 = Yes; 0 = No)
SEXFARM Binary Gender of household head (1 = Male; 2 = ±
Female)
LANDTEN Polychotomous Land tenure system of the farmer (1 = ±
Freehold; 2 = Communal; 3 = Rented)
SIZEFARLANa Categorical The size of farm land (1 = Small; 2 = +
Large)
APPROC Categorical Perception of individual towards ease of +
application procedure (1 = Easy; 2 = Not
easy)
INTRATE Categorical Perception of individual towards -
magnitude of interest rates (1 = Low; 2 =
High)
COLLATERAL Categorical Whether loan provider requires collateral -
or not (1 = No; 2 = Yes)
DISTANCE Categorical Perception of individual towards the -
distance from credit provider / institution
(1 = Near; 2 = Far)
a Size of farm land Less than 1 acre = small; 1 acre or more = Large
3.2 EMPIRICAL MODEL
3.2.1 DETERMINANTS OF CREDIT APPLICATION BY FARMERS
The primary aim of this study was to explain the effects of the explanatory variables on the
response probability. A number of studies such as Berger et al., (2001) have identified the
credit constraints by considering only the supply and not the demand side of credit access.
In order to incorporate the demand side, farmers were asked whether they applied for credit
in the previous 12 months. Based on existing literature, the farmers’ demand for credit is
affected by variables creating a difference in both the capital cost and the returns on capital
(Bigsten et al., 2003). The returns on capital factors are not considered in our model due to
the nature of available data i.e. farmers are not willing to declare their profits. The capital cost
factors included in the model are those that affect the external financing costs like collateral
requirements and variables that affect the transaction costs like distance to credit facility,
as well as interest rates. In addition demographic characteristics such as age, farm size and
membership of farmers’ associations are included. The model can be expressed as shown in
equation 3.1.
Y = a+ biXj + u…………………………………………………………………………… (3.1)

requirements
of farmers’and variables that
associations affect theThe
are included. transaction costs
model can be like distance
expressed as to creditinfacility,
shown equationas3.1.
well
as interest rates. In addition demographic characteristics such as age, farm size and membership
of farmers’ 𝛽𝛽! 𝑋𝑋! + 𝑢𝑢 ……………………………………………………………………………
𝑌𝑌 = 𝛼𝛼 +associations are included. The model can be expressed as shown in equation 3.1. (3.1)
𝑌𝑌 = 𝛼𝛼 + 𝛽𝛽! 𝑋𝑋! + 𝑢𝑢 ……………………………………………………………………………

!! (3.1)
!!
Where;
Where; X=Vector!!
X=Vector
!!
⋮
!! !
!"
Where; X=Vector ! !
Y=0 if the
Y=0 if farmer did not
the farmer did apply for credit;
not⋮ apply Y=1Y=1
for credit; if theiffarmer applied
the farmer for credit;
applied a = the
for credit; 𝛼𝛼=constant;
the constant;
!!"
b= 𝛽𝛽= the coefficient
the coefficient explaining
explaining the independent
the independent variable;
variable; 𝑋𝑋! =Age
X1=Age of farmer;
of the the farmer; 𝑋𝑋! =Farm
X2=Farm age;age;
Y=0 if the farmer didAssociation;
notAssociation;
apply forXcredit; Y=1 if the farmerXapplied for credit; theXconstant;
𝛼𝛼=System;
X ! = Member
=𝑋𝑋Member FarmerFarmer ! = Sex
=𝑋𝑋Sex of Respondent;
of Respondent; ! = Land
=𝑋𝑋Land Tenure TenureSystem; ! = Size
=𝑋𝑋Size
3
𝛽𝛽= the coefficient
of Farmland; explaining the independent
4
variable; 𝑋𝑋 =Age
5
of the farmer; 𝑋𝑋 =Farm 6
age;
of Farmland; ! = Application
X7 =𝑋𝑋Application Procedure;
Procedure; ! = Interest
X8 =𝑋𝑋Interest rates;
! rates; ! = Collateral
X9 =𝑋𝑋Collateral requirement;
requirement;
! X10=𝑋𝑋!" =
𝑋𝑋! =Distance
Member Farmer Association; 𝑋𝑋! = Sex of Respondent; 𝑋𝑋! = Land Tenure System; 𝑋𝑋! = Size
Distance
of Farmland; 𝑋𝑋! = Application Procedure; 𝑋𝑋! = Interest rates; 𝑋𝑋! = Collateral requirement; 𝑋𝑋!" =
Distance
3.2.2 Factors that influence the farmer’s ability to access credit given that they applied for
3.2. FACTORS THAT INFLUENCE THE FARMER’S ABILITY TO ACCESS
the credit
CREDIT
3.2.2 Factors thatGIVEN
influence THAT THEY APPLIED
the farmer’s FOR THE
ability to access creditCREDIT
given that they applied for
the Incredit
order to understand the factors that influence the farmer’s ability to access credit given that
In order to understand
they applied the we
for credit, factors
usedthat influence
a two-step the farmer’s
Heckman / Heckit ability
model.to access
The Heckmancredit given model thatoffers
In order
they appliedto understand
for credit, the
we factors
used athat influence
two-step the
Heckman farmer’s
/ Heckit ability
a means of correcting for non-randomly selected samples. In this case, there are farmers who model.to access
The credit
Heckman given modelthat did
theynot
offers applied
aapply
means forofcredit,
for weso used
correcting
credit, for anon-randomly
using two-step
only Heckman
selected
the proportion /that
Heckit
samples. model.
applied In Thepercent
this
(48.8 Heckman
case, there
of the model
aretotal offers
farmers
sample)
a means
whowould
did notof correcting
apply
result into for non-randomly
forselection
credit, sobias.
using Toonly selected samples.
the proportion
determine whether the In this
thatfarmer case,
appliedwas there
(48.8 are farmers
percent
denied credit who
of (Y=1)
the total did
or not
not apply
denied
sample) for credit,
creditresult
would (Y=0),so using
a binary
into only
selection the
logitbias.proportion
model Towas that applied
estimated.
determine (48.8
Both probit
whether percent
the farmer of
and logit the total
was analyses sample)
denied credit are well-
would
(Y=1) result
or notinto
established selection
approaches
denied creditbias. theTocredit
in(Y=0), determine
demand
a binary whether
logit studies
model the(Mpuga,
farmer
was was
2008;
estimated. denied
Kasirye,
Both credit (Y=1)
2007).
probit and or
The not
choice
logit
denied
of credit (Y=0),
whether to use a abinary
probit logit
or modelmodel
logit was estimated.
is a matter Bothof probit and logitconvenience
computational analyses are (Greene,
well-
analyses are well-established approaches in the credit demand studies (Mpuga, 2008; Kasirye,
established
1997; approaches in the credit demand studies (Mpuga, 2008; Kasirye, 2007). The dependent
choice
2007).
of whether
TheWooldridge,
choice
to use
2009;
of whether
a probit ordummy
Gujarati,
to use a 2004).
logit model
probit
is aand
Logistic
or logitofmodel
matter
regression is a ismatter
computational
used of since
convenience
the
computational
(Greene,after
variable
convenience is a dichotomous
(Greene, 1997; Wooldridge, variable
2009; maximum
Gujarati, likelihood
2004).isLogistic estimation
regression is applied
is used
1997; Wooldridge,
transforming 2009;
the dependent Gujarati, 2004).
into a logit Logistic
variable regression
(Garson, 2008). It used estimatessince thetheodds dependent
of a certain
since
variable the dependent variable is avariable
dichotomous dummy likelihood
variable and maximum likelihood
eventisoccurring.
a dichotomous dummy
The dependent variable is and maximum
a logit, which is the natural estimation
log of the is applied
odds, that after
is,
estimation
transforming is the
applied after transforming
dependent the dependent
into a logit variable (Garson,into 2008). a logit variablethe
It estimates (Garson,
odds of2008).a certain It
estimates the odds
event occurring. The ofdependent
a certain eventvariable occurring.
is a logit,Thewhichdependent variable
is the natural logisofathelogit, which
odds, that isis,the
natural log ! of the odds, that is,
ln = 𝑎𝑎 + 𝑏𝑏𝑏𝑏 .................................................................................................................... (3.2)
(!!!)
! !!!!"
ln 𝑃𝑃 = = 𝑎𝑎!!!" ............................................................................................................
+ 𝑏𝑏𝑏𝑏 ....................................................................................................................
…….. (3.2)
(3.2)
(!!!) !!!
…..............................................................................................................
!!!!" (3.3)
……..
𝑃𝑃 = !!!!!!" ....................................................................................................................... (3.3)
where P is the probability of the event occurring, X are the independent
….............................................................................................................. (3.3)variables, e is the base of
where the Pnatural logarithm and
is the probability of thea and
event b are the parameters
occurring, X are theofindependent
the model. variables,
The empirical form
e is the baseof the
where model P isused
the probability
in the study ofisthe
as event
follows: occurring, X are the independent variables, e is the base of
of
thethe natural
natural logarithm andand a and b arethethe parameters of the model. TheThe empirical form of
PrY = 1 logarithm
/ (1 + e - (a+bX) ) a.............................................................................................................
and b are parameters of the model. empirical form of the (3.4)
the
model model used
usedYinisthe in the
study study
is is
asthe as
follows: follows:
where the logit for dependent variable. The logistic prediction equation for the present
PrY
PrYstudy== 11 // (1
(1 ++ ee -- (a+bX))).........................................................................................................
(a+bX)
.............................................................................................................(3.4)
(3.4)
was
where Y
Y is
is the
the logit
logit for
for the
the dependent
dependent
Y = In (odds (event)) = In (prob (event)/prob (non-event)) variable.
variable. The
The logistic
logistic prediction
prediction equation
equation for
for the
the present
present
study = was In (prob (event) / [1 – prob (event)])
Y = In=(odds bo + b(event))
(odds (event)) ==InIn(prob
(prob(event)/prob
(event)/prob(non-event))
(non-event))
1X1 + b2X2 + ... + bnXn, ............................................................................................ (3.5)
= In
= In (prob
(prob (event)
(event)//[1 [1––probprob(event)])
(event)])
== bboo ++ bb11X X1 ++ bb2X
1
X2++ ...
2 2
... ++ bbnXXn,,........................................................................................
n n
............................................................................................(3.5)
39
(3.5)
where bo is the constant with X1 ... Xn independent variables affecting the probability of credit
demand and b1 ... bn were the coefficients estimated. 39


4.0 RESULTS AND DISCUSSION
4.1 DESCRIPTIVE ANALYSIS
4.1.1 GENDER OF HOUSEHOLD HEAD AND CREDIT DEMAND
A total of 127 farmers were interviewed of which 53 percent and 47 percent were female and
male, respectively. Sixty two farmers (48.8 percent) applied for credit and 30 farmers (23.6
percent) accessed credit as shown in Table 2.
Table 2: Gender of Household Head and credit demand

Gender of HH head Number Applied for Credit Accessed Credit
YES NO YES NO
Male 67 (52.8%) 28 (22.0%) 39 (30.7%) 11 (8.7%) 17 (13.4%)
Female 60 (47.2%) 34 (26.8%) 26 (20.5%) 19 (15.0%) 15 (11.8%)
Total 127 (100%) 62(48.8%) 65 (51.2%) 30 (23.6%) 32 (25.2%)
Source: Primary data; the figures in parentheses are percentages relative to the entire sample
4.1.2 AGE OF FARMERS AND CREDIT DEMAND
Farmers considered their age a very personal matter and as a result, the questionnaire used
class intervals to capture the age groups. Four age groups were categorized namely, below
25 years, 26-35 years, 36-50 years, and above 50 years whose proportion of the total sample
were 2.36 percent, 22.83 percent, 41.73 percent, and 33.07 percent, respectively. As shown
in Figure 1, the age group 36-50 years had the highest proportion of farmers who applied for
Figure
credit as1:well
Graphs
as thoseshowing thedenied
who were relationship
access to between
credit. the age categories with credit
constraint (Left hand panel) and application (Right hand panel)
Figure 1: Graphs showing the relationship between the age categories with credit
constraint (Left hand panel) and application (Right hand panel)
Source: Primary data; on the y axis 1, 2, 3, and 4 represent age cohorts below 25; 26-35; 36-50;
Source: Primary data; on the y axis 1, 2, 3, and 4 represent age cohorts below 25; 26-35; 36-50; & above 50, respectively
&The
above 50, respectively
farmers in the above 50 years category are able to access credit when they apply than the
rest, suggesting that credit constraint reduces with age. However, it can be noted that as the
The farmers in the above 50 years category are able to access credit when they apply than the
farmers grow much older they apply less; this could imply that they have accumulated enough
rest,
fundssuggesting
from theirthat
pastcredit constraint
years of work andreduces withdon’t
therefore age.have
However,
to incuritdebts
can be
fornoted that as the
their businesses.
farmers grow much older they apply less; this could imply that they have accumulated enough
funds from their past years of work and therefore don’t have to incur debts for their businesses.
4.1.3 Farm size and credit demand
4.1.3 FARM SIZE AND CREDIT DEMAND
Farmers with less than 1 acre of land were considered small while those with more than 1
acre of land were categorized as large. The proportion of farmers with small sized farms was
30.7 percent while 69.3 percent had large farms. The farmers with large farm sizes were more
inclined to borrow partly on account of their large number in the sample. Table 3 shows the
reasons advanced by farmers for staying out of the credit market in the previous one year.
Table 3: Farm Size and Reasons why farmers did not borrow
Farm Inadequate Don’t want to High Interest Don’t need a Application
Size Collateral incur debt Rate loan Procedure
too difficult
Small 3 4 0 1 5
Large 3 9 7 3 22
Total 6 13 7 4 27
Source: Primary data
Of the 62 farmers who applied for credit, 27 farmers (43.55 percent) felt the application
procedure was too difficult. In addition, seven farmers (11.29 percent) were of the view that
interest rates were high.
Table 4: Farm Size and credit demand

Farm Size Number Applied for Credit Accessed Credit
YES NO YES NO
Small 39 (30.7%) 23 (18.1%) 16 (13.0%) 10 (7.9%) 13 (10.2%)
Large 88 (69.3%) 39 (30.7%) 49 (38.6%) 20 (15.7%) 19 (15.0%)
Total 127 (100%) 62(48.8%) 65 (51.2%) 30 (23.6%) 32 (25.2%)
Nonetheless, none of the farmers with a small farm thought interest rates were high. The latter observation suggests that borrowers
in microcredit markets have price inelastic demand where they are insensitive to high interest rates (Salazar et al., 2011). Table 4
shows the relationship between farm size and credit application / access.
4.1.4 AGRICULTURAL ACTIVITY AND CREDIT DEMAND
The proportions of farmers involved in agricultural activities are shown in Table 5. Most of
the farmers were involved in crop husbandry. This could mainly be explained by their target
market. Since 98 percent of the farmers produce for the domestic market, production of food
crops would be most appropriate given the availability of the market. Food crop production
has the highest number of farmers who applied / accessed credit.

Table 5: Agricultural Activity and credit demand
Agricultural Number Applied for Credit Accessed Credit
Activity
YES NO YES NO
Cash Crops 16 (12.6%) 7 (5.5%) 9 (7.1%) 4 (3.1%) 3 (2.4%)
Food Crops 94 (74.0%) 43 (33.9%) 51 (40.2%) 19 (15.0%) 24 (18.9%)
Livestock 15 (11.8%) 11 (8.7%) 4 (3.1%) 7 (5.5%) 4 (3.1%)
Forestry 2 (1.6%) 1 (0.8%) 1 (0.8%) 0 (0.0%) 1 (0.8%)
Total 127 (100%) 62(48.8%) 65 (51.2%) 30 (23.6%) 32 (25.2%)
4.1.5 LAND OWNERSHIP AND CREDIT DEMAND
The land tenure systems observed in the sample were free hold, customary, leasehold and
rented with the respective proportion of farmers for each type being 68.5 percent, 17.3 percent,
3.1 percent, and 1.6 percent. As shown in Table 6, 83.5 percent of the farmers possessed
documents that indicated the type of land tenure. Of these only 47 farmers (37.0 percent)
applied for credit and only 13 farmers (10.2 percent) accessed credit.
Table 6: Land ownership and credit demand

Land Number Possess ownership Applied for Credit Accessed Credit
tenure documents
YES NO YES NO YES NO
Freehold 87 (68.5%) 83 (65.4%) 4 (3.1%) 47 (37.0%) 40 (31.5%) 13 (10.2%) 34 (26.8%)
Customary 22 (17.3%) 17 (13.4%) 5 (3.9%) 13 (10.2%) 9 (7.1%) 9 (7.1%) 4 (3.1%)
Leasehold 4 (3.1%) 3 (2.4%) 1 (0.8%) 0 (0.0%) 4 (3.1%) 0 (0.0%) 0 (0.0%)
Rented 14(1.6%) 3 (2.4%) 11(8.7%) 2 (1.6%) 12 (9.4%) 1 (0.8%) 1 (0.8%)
Total 127 106 21(16.5%) 62 (48.8%) 65 (51.2%) 23 (18.1%) 39 (30.7%)
(100%) (83.5%)
4.1.6 FARMER ASSOCIATIONS AND CREDIT DEMAND
Sixty six (66) percent of the farmers reported to belong to farmer associations, while the
other 34 percent did not belong to any farmer association. Out of all the members that belong
to farmer association 56 percent applied for credit while the other 44 of the farmers didn’t
borrow. All the members in the associations that applied for credit were able to access credit.
Farmers’ associations provided various services to the members namely, marketing, input
supply, credit facilities, standards compliance and training opportunities.
4.1.7 DISTANCE TO CREDIT FACILITY AND CREDIT DEMAND
The data shows that most of the farmers live within 0.5 kilometers from the credit facilities.
The details of all the credit sources and their distances are captured in Figure 2.

Figure 2: Histogram and graph showing the relationship between distance to credit facility
(left hand panel) and current source of loan (right hand panel)
Figure 2: Histogram and graph showing the relationship between distance to credit
facility (left hand panel) and current source of loan (right hand panel)
Source: Author’s illustration based on primary data

From Figure 3 (left hand panel) it is evident that distance to the credit facility has impact on
FromtheFigure
farmers’3 (left handtopanel)
decision borrow,it is
thatevident
is, the that distance
nearer to the
the credit creditthefacility
facility, higherhasthe impact
farmer’son the
farmers’
willingness to borrow. There are more people borrowing if the credit facility is lessthe
decision to borrow, that is, the nearer the credit facility, the higher thanfarmer’s
5
willingness
kilometersto compared
borrow. There
to otherare more ranges.
distance people However
borrowing if the credit
closeness to credit facility
sourceisdoes
lessnotthan 5
kilometers comparedborrowers’
greatly influence to other ability
distance ranges.
to get creditHowever
as depictedcloseness
in Figureto3 credit source
(right hand does not
panel).
Figure 3:
The Distance
people in and the
further decision
places are asto borrow
much likely(left
to hand
access panel)
credit as and
the credit
people
greatly influence borrowers’ ability to get credit as depicted in Figure 3 (right hand panel). constraint
close to the andThe
credit
thepeople
distancein to
facilities. creditplaces
further facility
are(right
as muchhandlikelypanel)to access credit as the people close to the credit
facilities.
Figure 3: Distance and the decision to borrow (left hand panel) and credit constraint and
the distance to credit facility (right hand panel)

4.1.8 Interest rates and credit demand
4.1.8 Interest rates and credit demand
4.1.8 INTEREST RATES AND CREDIT DEMAND
The interest rates charged for the loans vary across financial institutions.
The interest rates charged for the loans vary across financial institutions.
Figure 4: Histograms showing the interest charged on the credit (left hand panel) and the
interestFigure
rate categories (right
4: Histograms handthe
showing panel)
interest charged on the credit (left hand panel) and the
interest rate categories (right hand panel)
Source:Source: Author’s illustration based on primary data

Author’s illustration based on primary data
The interest rates range from 0 percent to as high as 50 percent as depicted in the histogram
The interest rates range from 0 percent to as high as 50 percent as depicted in the histogram
(Figure 4). The histogram indicates the classes (1) Didn’t know, (2) zero, (3)1-9, (4)10-19, (5)
(Figure20
4).and
Theabove.
histogram indicates
A substantial the classes
number (1) Didn’t
of farmers know,
were even not (2) zero,
aware (3)1-9,
at how much(4)10-19,
they had(5) 20
and above. A substantial
acquired loans. Thisnumber ofindicates
implicitly farmers that
weretheeven not aware
respondents areatnot
how muchby
bothered they
thehad acquired
interest
loans. This implicitly indicates that the respondents are not bothered by the interest rates at
rates at which they acquire credit.
which they acquire credit.
4.1.9 COLLATERAL AND CREDIT DEMAND
Table 7 shows that the sources of credit were: government aided schemes (3), micro finance
institutions (2), relative and friends (5), private money lenders (4), commercial banks (1)
and farmer association (7). The biggest number45 of respondents (39 percent) borrowed from
government aided schemes. Other credit sources had the following proportions: micro finance
institutions (19 percent); relatives and friends (18 percent); money lenders (13 percent);
commercial banks (10 percent); and 2 percent from farmer associations.
Table 7: Land ownership and Credit Demand

Source of Credit Offered Collateral No Collateral offered
Small Farm Large Farm Small Farm Large Farm
size size size size
Commercial Bank 0 6 0 0
Microfinance Institution 3 7 1 1
Government aided schemes 2 5 8 8
Private money lenders 0 1 3 4
Relatives and Friends 0 0 4 7
Farmer Groups 0 0 1 0

The use of collateral is mostly reflected in the formal and semi-formal sector, whereas almost
all loans within the informal sector were obtained without collateral. Although we also see
a high number of uncollateralized loans in the semi-formal sector, it is mainly because the
government credit schemes that offer credit to farmers are subsidized and some of them
do not require any form of security to acquire funds for agricultural purposes. A sum of 24
farmers offered collateral for the loans they acquired while 36 of the members acquired credit
without offering any form of security. The assets used as collateral by those who applied for
loans included land, personal belongings and other items; with the respective percentages
i.e. 41 percent, 45 percent and 14 percent. Providing guarantors is another form of security
required by most lending institutions. In the sample we find that 62 percent of the farmers who
borrowed had guarantors while 38 percent did not. The relationships between the respondents
and the guarantors included trade/business partners, family members and friends.
Table 8 shows the applications received by the lending institutions, successful applicants as
well as those that failed.
Table 8: Land ownership and Credit Demand

Source of Credit Applied Accessed Denied
Commercial Bank 6 4 2
Microfinance Institution 12 4 8
Government aided schemes 24 13 11
Private money lenders 8 5 3
Relatives and Friends 11 3 8
Farmer Groups 1 1 0
62 30 32
The farmers acquired loans for the following reasons: improve productivity (44 percent),
add to capacity (26 percent), produce new output and consumption (11percent), replace old
equipment and produce new output (4 percent), and repayment of existing loan (2 percent).
Cognisant, of the fact that education level is a major determinant of agricultural credit demand
such as Nwaru et al., (2011), pretesting the questionnaire indicated that most farmers were
not very forthcoming regarding their education. The hesitation by farmers to respond to this
question prompted the researchers to eliminate the variable.
4.2 THE CREDIT DEMAND MODEL OF FARMERS IN MUKONO DISTRICT
4.2.1 FACTORS THAT INFLUENCE WHETHER A FARMER APPLIED FOR

CREDIT
As shown in Table 9, the factors that were significant were: distance to credit facilities;
application procedures; farm size; land tenure system; and being a member of farmer
associations. In contrast, age, farming experience, gender, land tenure system leasehold/rented
land tenure system, interest rates and collateral requirement were not statistically significant at
the p = 0.01, p = 0.05 and p = 0.1 significance levels.

Table 9: Binomial logistic regression showing credit application factors amongst farmers
in Mukono District
Coefficients: Estimate Std. Error z value ME Pr(>|z|) Signif.
(Intercept) -2.877 1.460 -1.971 -0.718 0.049 *
Age -0.030 0.019 -1.545 -0.007 0.122
Farmage 0.016 0.026 0.628 0.004 0.530
MeFarAss 1.053 0.573 1.837 0.254 0.066 .
SexResO 0.599 0.503 1.191 0.150 0.234
as.factor(Landten)2 1.780 0.751 2.370 0.395 0.018 *
as.factor(Landten)3 0.019 0.748 0.026 0.005 0.980
as.factor(Sizefarlan)2 -1.040 0.524 -1.984 -0.254 0.047 *
AppProc 1.239 0.502 2.469 0.299 0.014 *
InterestRate 0.235 0.327 0.720 0.059 0.472
Collateral 0.557 0.516 1.079 0.138 0.281
Distance 2.084 0.545 3.826 0.520 0.000 ***
Signif. Codes: ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1
No. Obs 127
Log likelihood -56.617 (df=12)
Likelihood ratio statistic 58.666 (df=11)
3.95E-
pchisq(58.6661,12,lower.tail=F)[1] 08
McFadden pseudo R squared 0.341
Bptest BP = 9.007, df = 11, p-value = 0.621
Source: R statistical software output
The fitted probabilities of credit demand increases per unit increase in these explanatory
variables namely, member farmer association, customary land tenure system , easier application
procedure, low interest rate and being nearer to credit facilities. Being nearer to credit facility
is significant at all levels. Both easier application procedure and customary land tenure system
are significant at 5 percent level. Membership in farmer association is only significant at 10
percent level.

4.2.2 FACTORS THAT INFLUENCE WHETHER A FARMER ACCESSED CREDIT
FOLLOWING APPLICATION
Table 10: Showing two step estimation results for credit access factors amongst farmers
in Mukono District
Probit selection equation: Estimate Std. Error t value Pr(>|t|) Signifi.
(Intercept) -1.652 0.837 -1.973 0.051 .
Age -0.019 0.011 -1.671 0.098 .
Farmage 0.010 0.015 0.643 0.521
MeFarAss 0.638 0.327 1.950 0.054 .
SexResO 0.341 0.288 1.186 0.239
as.factor(Landten)2 0.905 0.416 2.174 0.032 *
as.factor(Landten)3 0.006 0.443 0.013 0.990
as.factor(Sizefarlan)2 -0.624 0.301 -2.072 0.041 *
AppProc 0.708 0.287 2.465 0.015 *
InterestRate 0.157 0.192 0.819 0.415
Collateral 0.408 0.307 1.327 0.188
Distance 1.194 0.303 3.944 0.000 ***
Outcome equation: Estimate Std. Error t value Pr(>|t|) Signifi.
(Intercept) 0.783 0.465 1.685 0.095 .
Age 0.015 0.006 2.484 0.015 *
Farmage -0.007 0.007 -1.117 0.267
MeFarAss -0.339 0.174 -1.950 0.054 .
SexResO -0.154 0.131 -1.175 0.243
as.factor(Landten)2 -0.232 0.172 -1.348 0.181
as.factor(Landten)3 -0.030 0.198 -0.153 0.879
as.factor(Sizefarlan)2 -0.082 0.144 -0.567 0.572
AppProc -0.195 0.148 -1.316 0.191
InterestRate -0.035 0.087 -0.405 0.686
Collateral 0.208 0.156 1.334 0.185
Signif. codes: ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1
Multiple R-Squared:0.2799, Adjusted R-Squared:0.1114
121 observations (62 censored and 59 observed)
Source: R statistical software output
Considering the outcome equation, age of the farmers and being a member of farmer associations
are the only significant variables. Low Interest rates are not significant as indicated in Table
10. The fitted probability of being credit constrained increases per unit increase in age. Age is
significant at 5 percent level. While fitted probabilities of being credit constrained decreases
per unit increase in obtaining membership in farmer associations and low interest rates. Being
a member of farmer association is only significant at 10 percent level whereas low interest rate
is insignificant at all levels.

Membership to farmer association was significant and positively influenced the probability of
a farmer demanding for credit. The positive effect of membership to a farmer association could
be attributed to the services provided by the association in terms of financial management
training and the social capital such membership confers. The aspect of training provides
information and mentorship about savings and credit as observed in the work of Huppi and
Feder (1990) who studied cooperatives and group lending in developing countries. In addition,
membership to a farmer’s association provides social capital which is critical in small scale
credit access as found out by Darie (2012) who studied farmers in Masaka, Uganda.
The regression results provide evidence that customary land system is an important factor
in influencing farmer’s demand for credit. This is because people who have lived with the
customary land tenure for a long time, clearly understand how the system operates. This result
is in line with Deininger (2003) who suggests that farmers who have secure and long-term
land rights are likely to invest a lot of their effort to make long term investments leading
to economic growth. The significance of the relationship between application for credit and
customary land tenure suggests that lenders are willing to accept such land as collateral. The
acceptance of such land as collateral is probably motivated by the fact that it is communally
owned. The community that owns the land is likely to serve the role of peer monitoring to
the member who staked the land as collateral. From the peer monitoring hypothesis by Hoff
and Stiglitz (1990), this should reduce monitoring costs on the lender’s part and minimize
default either strategic or unintended. The capacity to access credit using customary land
could explain the observation by Mugambwa (2007) that it is possible to achieve economic
development under this land tenure system.
The easing of application procedures has a significant impact on credit demand. The study
found that easing the application procedures by one unit increases the marginal contribution
to the probability to apply for credit by 29.9 percent compared to having complex procedures.
The findings of this study corroborate the findings of Namasaka (2007) whose findings
strongly suggest that regulatory application frameworks and credit access are strongly related.
The probability of applying for loans decreases per unit increase in large farm size. The
negative relationship could be attributed to the fact that large farms have sufficient resources
for their needs. In addition, the persistent credit rationing when loans are being approved
could be discouraging to the large farms. The discouragement could arise if the final loan
amount approved is below the value required. The findings of the study are in agreement with
Kumar and Francisco (2007) whose findings show that size strongly affects credit demand.
Kumar and Francisco (2007) postulate that size effects on credit demand are more evident for
longer term maturities.
Distance from financial institution had a positive effect on the probability of applying for
credit. This could be on account of lower transaction costs for going to the financial institution
for the borrower as well as the lender to monitor the borrower. In addition, this deals with the
challenge of information asymmetry (Hoff and Stiglitz, 1990) common in developing country
credit markets.
There is a positive relationship between interest rates and loan application. Lowering interest
rates by one unit would increase the marginal contribution to the probability to apply for credit
by 5.9 percent. This finding is in agreement with Keynes (1936) who suggests that investment
is partly induced by interest rates and investment demand. In this study however, the observed

relationship was not significant. In other words, interest rates are not a limiting factor in credit
constraint. It is not unusual to find a farmer who accessed credit without necessarily knowing
the interest rate. This is in line with the findings of Karlan and Zinman (2008) who postulated
that loan amounts are more critical in the case of developing countries than the rate of interest.
5. POLICY IMPLICATIONS AND RECOMMENDATIONS
This study showed that socio-economic and demographic characteristics of households are
important determinants of credit demand. Some policy issues could be drawn from these
results when programmes for improving credit access by small scale entrepreneurs and / or
farmers are designed.
The positive correlation of membership to farmer association with credit demand is crucial
regarding the need to encourage farmers to join such organizations. Beyond providing social
capital for use when applying for credit, they can be used for technological transfer awareness,
quality control for produce, and mutual insurance (Ray, 1998). Indeed, they provide more
organic groupings that can be utilized by governments interested in influencing rural finance
rather than adhoc set-ups that normally arise whenever a government rural finance package is
announced.
Land is the most important asset to a peasant farmer. Nevertheless, the mixed mode tenure
system in Uganda implies that the farmer rarely has full control of the ownership except
in leasehold and freehold tenure systems. Consequently, credit providers may be wary of
extending credit to individuals who own only user rights under the mailoland tenure system.
However, capacity to access customary land titles and pledge it as security points to the
possibility that the use of the title as collateral has been sanctioned by the group. Since the
group would not want to lose their land, they serve as peer monitors to the one borrows based
on the customary land title. The implication is that in its effort to modernize agriculture,
government can still encourage use of customary land as a viable alternative to individualized
mailoland, leasehold and freehold tenure systems.
The direct relationship between the ease of application procedures and credit demand is
indicative of the need for financial services providers to reduce the bureaucratic nature of the
procedure of accessing loans. The co-existence of formal and informal providers in developing
countries’ credit markets is partly explained by the bureaucratic procedures prevalent in formal
financial services providers (Daniel, 2010). As a way of increasing access to formal financial
services and products, streamlining these procedures with a view of leaving only the necessary
minimum would go a long way in promoting credit demand.
The inverse relationship between credit demand and farm size suggests that large farmers are
unlikely to go through with the loan application if credit rationing is likely. The endowments
of such large land owners imply that they are more sensitive to credit rationing and interest
rate changes. Hence, financial services providers ought to design products that take such
customers into perspective given that they possess the necessary collateral to guarantee their
loans.
Distance to financial institution from farmer’s dwelling having a positive effect on credit
demand suggests that credit providers ought to establish service points closer to the clientele.

However, establishing a service point is a function of many factors including potential clientele
and infrastructural capacities of the region. Given that the latter factor is a major hindrance in
most developing countries, leveraging technologies like mobile money could be a potential
solution to deal with this challenge.
Insensitivity to interest rate changes by farmers is largely a result of the fact that credit is
scarce in Uganda. On account of the low numbers of banked people and poor saving culture,
the deposits available to financial services providers to lend clients are low. This coupled
with a relatively high fiscal deficit of the government implies that farmers who manage to get
access to credit are relatively insensitive to interest rate changes. The implication is that there
is unfulfilled appetite for credit which both private and public stakeholders ought to address.
6. CONCLUSIONS
The need to modernize agriculture through mechanization and agro-processing form the
backbone to Uganda’s development strategy (Poverty Eradication Action Plan (PEAP).
Nonetheless, mechanization and agro-processing cannot take off without sufficient credit to
fund these activities. Hence there is a need to boost credit access through encouraging credit
demand. Factors identified as promoting credit demand include proximity to credit facility,
application procedures, farm size, land tenure system, and membership to farmers’ association.
Mechanisms of leveraging these factors to enhance credit demand have been highlighted in
the policy options.
ACKNOWLEDGEMENT
The authors gratefully acknowledge the financial support from World Bank Joint Japan
Group Scholarship Programme (WB/JJGSP) who supported the corresponding author’s
study programme at the University of Reading and the DANIDA Fellowship Programme
for supporting the co-author’s study at the University of Copenhagen. The data used in the
analysis was collected for the latter’s M.Sc. dissertation. They also thank the anonymous
reviewers for their useful comments and valuable suggestions that improved the paper.
REFERENCES
Atieno, R. 1997. Determinants of Credit Demand by smallholder farmers in Kenya: an

empirical analysis. Der Tropenlandwirt, 98, 63-71.
Balogun, O and Yusuf, S. 2011. Determinants of Demand for Microcredit among the Rural
Households in South-Western States, Nigeria. Journal of Agricultural and Social Sciences,
Vol. 7(2), 41-48.
Berger, A.N., Klapper, L.F. and Udell, G.F. 2001.The ability of banks to lend to
informationally opaque small businesses. Journal of Banking and Finance, Vol. 25, 1-47.
Retrieved from http://www.federalreserve.gov/pubs/feds/2001/200134/200134pap.pdf on
July 08, 2013.
Bigsten , A. Collier, P. Dercon, S. Fafchamps, M. Gauthier, B. Gunning, J.W. Oduro, A.
Oostendorp, R. Patillo, C. Sorderbom, M. Teal, F. and Zeufack, A., 2003. Credit Constraints in
manufacturing Enterprises in Africa.Journal of African Economies, Vol. 12, 104-125
Bhattacharjee, M., Rajeev, M. and Vani, B.P. 2009.Asymmetry in Information and Varying
Rates of Interest. A study of the informal credit market in West Bengal. The Journal of Applied
Economic Research, Vol. 3(4), 339-364.

Bokosi, F.K. 2004. Determinants and characteristics of Household Demand for Small holder
Credit in Malawi. Retrieved from http://128.118.178.162/eps/get/papers/0408/0408001.pdf
on June 10, 2013
Daniel, B. 2010.Informal Credit Markets – An Overview.Retrieved from http://shodhganga.
inflibnet.ac.in/bitstream/10603/79/18/08-chapter2.pdf on December 16, 2013.
Darie, Y. 2012. Liquidity as a challenge in rural economy: does social capital facilitate
access to credit in Masaka District, Uganda? Masters Dissertation, Norwegian University of
Life Sciences.
Deininger, K. 2003. Land policies for growth and poverty reduction.Retrieved from http://
www-wds.worldbank.org/external/default/WDSContentServer/WDSP/IB/2003/08/08/00009
4946_0307250400474/Rendered/PDF/multi0page.pdf on March 06, 2013.
Diagne, A. 1999. Determinants of Household Access to and participation in Formal and
Informal Credit Markets in Malawi. International Food Policy Research Institute (IFPRI),
Food Consumption and Nutrition Division (FCND) Discussion Paper No. 67. Retrieved from
http://www.ifpri.org/sites/default/files/pubs/divs/fcnd/dp/papers/dp67.pdf on June 08, 2013.
Doan, T., Gibson, J. and Holmes, M. 2010. What determines credit participation and credit
constraints of the poor in peri-urban areas, Vietnam? Retrieved from http://mpra.ub.uni-
muenchen.de/27509/1/MPRA_paper_27509.pdf on June 08, 2013
Ferede, K.H. 2012. Determinants of Rural Households Demand for and Access to Credit
in Microfinance Institutions: The Case of Atamata Woreda – Ethiopia. Master’s Dissertation,
Wageningen University Research Center.
Fichera, E. 2010.An analysis of households’ credit markets in Ethiopia and Malawi.PhD
thesis, University of Nottingham.
FinScope, 2010. FinScope Uganda 2009: Results of a National Survey on Demand, Usage
and Access to Financial Services in Uganda. Kampala, Uganda: DFID & Steadman/Synovate
Garson, G.D. 2008. Logistic Regression: Statnotes from North Carolina State University.
Retrieved from http://www2.chass.ncsu.edu/garson/PA765/logistic.htm on January 08, 2013.
Greene, W.H. 1997. Econometric Analysis. 5th ed. Upper Saddle River, NJ: Pearson
Education.
Gujarati, D. N. 2004. Basic Econometrics. 4th ed. Boston, USA: McGraw Hill.
Hoff, K. and Stiglitz, J.E. 1990. Introduction: Imperfect Information and Rural Credit
Markets: Puzzles and Policy Perspectives. The World Bank Economic Review, Vol. 4(3), 235-
250.
Huppi, M. and Feder, G. 1990. The Role of Groups and Credit Cooperatives in Rural
Lending. World Bank Research Observer, Vol. 5(2), 187-204.
Karlan, D. S., and Zinman, J. 2008. Credit Elasticities in Less Developed Economies:
Implications for Microfinance.” American Economic Review, Vol. 98 (3), 1040-1068.
Kasirye, I. 2007. Rural Credit Markets in Uganda: Evidence from the 2005/6 National
Household Survey. Retrieved from http://www.afdb.org/fileadmin/uploads/afdb/Documents/
Knowledge/25120349-EN-AFECONF-PAPERECV-CLS1-2-12-IBRAHIM-KASIRYE-15P.
PDF on April 05, 2013.
Keynes, J.M. 1936. The General Theory of Employment, Interest, and Money. Cambridge,
UK: Macmillan Cambridge University Press.
Kumar, A. and Francisco. 2007. Enterprise Size, Financing Patterns and Credit Constraints
in Brazil: Analysis of Data from the Investment Climate Assessment Survey. World Bank
working paper.Retrieved from http://papers.ssrn.com/sol3/papers.cfm?abstract_id=960152 on
July 09, 2013.
Matovu, J. and Okumu, L. 2010. Credit accessibility to the Rural Poor in Uganda. Retrieved
from http://www.eprc.or.ug/pdf_files/researchseries/series1.pdf on April 08, 2013.

Mohan, R. 2004. Agricultural Credit in India: Status, Issues and Future Agenda. Reserve
Bank of India Bulletin, November, 993-1007.
Mpuga, P. 2008. Constraints in access to and demand for rural credit: evidence
from Uganda. Retrieved from http://www.afdb.org/fileadmin/uploads/afdb/Documents/
Knowledge/30753249-FR-122-MPUGA-ACCESS-AND-DEMAND-FOR-RURAL-
CREDIT-ADB-CONF.PDF on June 10, 2013.
Mpuga, P. 2004. Demand for credit in rural Uganda: who cares for the peasants? Retrieved
from http://www.csae.ox.ac.uk/conferences/2004-GPRaHDiA/papers/2f-Mpuga-CSAE2004.
pdf on March 04, 2013.
Mugambwa, J. 2007. A comparative analysis of land tenure law reform in Uganda and
Papua New Guinea. Journal of South Pacific Law, 11(1): 39.
Namasaka, B.N. 2007.Access to micro credit financing by rural women entrepreneurs in
Bungoma County.Msc. Dissertation.
Nkurunziza, J.D. 2005. Reputation and Credit without collateral in Africa’s formal banking.
University of Oxford, Centre for the Study of African Economies Working Paper WPS/2005-
02.Retrieved from http://www.economics.ox.ac.uk/index.php/centre-for-the-study-of-african-
economies-series/reputation-and-credit-without-collateral-in-africa-s-formal-banking on
April 27, 2013.
Nwaru, J.C., Essien, U.A. and Onuoha, R.E. 2011. Determinants of Informal Credit
Demand and Supply among Food Crop Farmers in Akwa Ibom State, Nigeria. Journal of Rural
and Community Development, Vol. 6(1), 129-139.
Okurut, N., Schoombee, A. and Van der Berg, S. 2005. Credit demand and credit rationing
in the Informal Financial Sector in Uganda. South African Journal of Economics, Vol. 73(3),
1-29.
Owusu-Antwi, G. and Antwi, J. 2010.The Analysis of the rural credit market in Ghana.
International Business & Economics Research Journal, Vol. 9(8), 45-56.
Ray, D. 1998. Development Economics. New Jersey, USA: Princeton University Press.
Salazar, G., Bogan, V. and Turvey, C. 2011.The Elasticity of Demand for Microcredit.
Retrieved from http://ssrn.com/abstract=1646048on June 04, 2013.
Udry, C. 1994. Risk and insurance in a rural credit market: an empirical investigation in
Northern Nigeria. The Review of Economic Studies, Vol. 61(3), 495-526.
Whelan, J. and Msefer, K. 1996. Economic Supply and Demand.MIT System Dynamics
in Education Project Working Paper D-4388.Retrieved from http://www.core.org.cn/NR/
rdonlyres/Sloan-School-of-Management/15-988Fall-1998-Spring-1999/82AE042B-6ADA-
41F9-B0C5-688DB0669C92/0/economics.pdf on June 05, 2013.
Wooldridge, J.M. 2009. Introductory Econometrics: A Modern Approach. London,UK:
South-Western Cengage Learning.
Zeller, M. 1994. Determinants of Credit rationing: A study of informal lenders and formal
credit groups in Madagascar. World Development, 22(12), 1895-1907. Retrieved from http://
www.ifpri.org/sites/default/files/pubs/divs/fcnd/dp/papers/dp02.pdf on June 10, 2013.

Determinants of Informal Employment: A Case of Tanzania’s
Construction Industry
Jehovaness Aikaeli and Beatrice Kalinda Mkenda1
Abstract
This paper analyses the determinants of informal employment in Tanzania’s construction

industry. A Logit regression model is employed in estimating factors that influence the choice
of type of employment (formal versus informal) for micro and small entrepreneurs (MSEs).
The results reveal that higher earnings in the informal compared to the formal settings – given
the professional status of the micro and small practitioners – is among the major reasons for
workers in this industry to choose informal rather than formal employment. The other factors
that contribute to choosing informal employment include; lack of capital, which deters micro
and small entrepreneurs from starting large formal firms, and low education. For firms, the
possibility of paying the workers low salaries, and being female are factors that increase the
possibility of informal employment. Policies suggested that can enhance creation of decent
employment are; improving financial services through risk mitigation, credit information
dissemination and outreach to MSEs; enhancing and rationalizing earnings in the economy;
and improving the quantity and quality of education as an enabling instrument.
Key words: informal employment, logit, construction, workers.
JEL Classification Codes: O 170, J 46, C 21.
1
The authors would like to thank the four anonymous reviewers for providing very useful comments and suggestions, which
helped to improve the paper. The authors are also deeply indebted to the International Development Research Centre (IDRC),
for providing funding for the research, and all the participants to the research workshop that was held in Entebbe (January 22nd
to 23rd, 2014), for their useful suggestions.
* Corresponding Author: bkmkenda@udsm.ac.tz

1. Introduction
Informal employment is a type of employment that is not bound by formal contractual

arrangements. An alternative definition uses job security as a measure of formality, defining
participants in informal employment as those who do not have employment or work security,
and thus go without social security (Lewis, 1955). Both definitions imply lack of choice
for involvement in employment in the informal economy especially in the developing
world2. However, participation in informal employment may also be driven by a wish to
avoid regulations or taxation. It is important to note from the outset that although informal
employment is largely an informal economy3phenomenon, it is also manifested in some formal
organizations of developing countries, including Tanzania.
The rationale of informal economy as a mainstay of informal employment has been viewed in
two major perspectives regarding its position in the national policies of developing countries;
the first relates to those who believe that the informal economy is an important driver of
income and growth, and hence it deserves due support. The second relates to those who
associate the informal economy with economic disorganization, and thus it does not deserve
support, but instead, requires reorganization to formalize it. The growth of informal economy
activities has been due to the fact that many informal workers find their activities the major
means for survival, and for some formal workers, informal economic activities are necessary
for supplementing their earnings.
The informal economy is estimated to account for 42% of Gross Domestic Product (GDP) in
sub-Saharan Africa (SSA), and specifically 34% of the national economy in Tanzania (Becker,
2004; and Economic and Social Research Foundation (ESRF), 2011). According to the
International Labour Organization (ILO) (2013), the informal economy comprises half to three-
quarters of all non-agricultural employment in developing countries. Some of the characteristic
features of informal employment are lack of protection in the event of non-payment of wages,
compulsory overtime or extra shifts, lay-offs without notice or compensation, unsafe working
conditions and the absence of social benefits such as pensions, sick pay, leave and health
insurance. Women, migrants and other vulnerable groups of workers who are excluded from
other opportunities have little choice but to take informal low-quality jobs. In view of this, the
informal economy has remained a useful concept to activists, policymakers and researchers
since a large share of employment and income is outside the regulated formal economy (Chen,
2007).
2
Parallel with growth of the informal sector in many countries, an increase in various forms of informal employment has happened.
The concept of informal employment is relevant to developing, transition, and developed countries. From a broader context of
informal employment, the question of informal sector is of limited relevance since informal employment can be attributed to
globalization of economic processes and so can include employment of not only in the informal but also some of the formal and
regulated sector’s employment. Various forms of sub-contracting of workers render employment an informal status, especially
where people would judge employment as alternative, atypical, non-standard, irregular and precarious or unsecure. A conceptual
framework for defining informal employment was proposed in the ILO’ report on ‘Decent Work and the Informal Economy’
(Hussmans, 2001; ILO, 2002). The conceptual framework retains informal employment as a terminology distinct from informal
sector employment, albeit the two being closely related. Informal sector employment comes from disaggregation of jobs in terms
of type of production units while informal employment disaggregates jobs in terms of underlying characteristics of particular
employment regardless of the production unit where these jobs are created.
3
The term informal economy is used interchangeably with informal sector. The context of informal sector was originally adopted
by Lewis in 1950s but it has increasingly been replaced by informal economy in the recent literature.

The informal economy exhibits a high degree of vibrancy in job creation in Tanzania.
However, it is faced with a number of constraints and suffers from low labour productivity
and use of low technology or rudimentary tools. The reason for this is that informal economy
organizations in the country, like elsewhere in SSA, are basically micro and small enterprises
(MSEs). Nevertheless, the Integrated Labour Force Survey (ILFS) of 2006 shows that the
informal economy in Tanzania is expanding rather than contracting (National Bureau of
Statistics (NBS), 2007). For example, the proportion of all households in Tanzania mainland
with informal sector activities increased from 35% in 2001 to 40% in 2006. Among the
lingering questions so far are; why is the formalization process slow? What spurs informal
employment? Using a case study of construction industry in Tanzania, this paper investigates
the determinants of informal employment. The construction industry is chosen because it is
one of the fastest emerging and growing sectors with a significant number of informal workers.
Tanzania’s construction sector is currently experiencing high growth, primarily driven by

recent developments in road construction, housing and mining industries. The growth rate
of the construction sector increased from 0.8% in 2000 to 9% in 2011. Its contribution to
GDP rose from 5% to 8% during the same period. While employment in road works and
mining construction activities is largely formal, it is informal in the housing sector, save
for large corporate construction projects. Residential construction activities are regularly
done informally, and substantial sub-contracting of informal contractors occurs in formally
contracted projects. This study uses primary data which was collected in six most vibrant urban
centres in Tanzania, namely Dar es Salaam, Mwanza, Arusha, Tanga, Mbeya and Dodoma.
The rest of the paper is organized as follows; after an introductory section, Section 2 gives an
overview of informal employment status in Tanzania. Section 3 contains a literature review,
and Section 4 gives an overview of empirical work on informal employment. The methodology
is described in Section 5. Section 6 describes the data used for estimation and discusses the
results, as well as provides analyses and technical interpretations. The conclusion and some
policy implications are discussed in Section 7.
2. OVERVIEW OF INFORMAL EMPLOYMENT SITUATION IN TANZANIA
Labour participation rate in this paper’s view conforms to that of the ILFS of 20064, which
takes the age of 15 years and above as the active age in the labour market in Tanzania (NBS,
2007). The employment context adopted in this overview is based on the national definition of
employment, which excludes all persons who were temporarily absent from work during the
reference period. Moreover, people who were working but whose work was not reliable with
regard to its availability and adequacy in terms of hours were considered unemployed (NBS,
2007). This definition is different from the standard international definition of employment,
which refers to people who worked for at least one hour in the reference period.
2.1 SIZE OF INFORMAL EMPLOYMENT IN TANZANIA
Official statistics from the ILFS show that about 40% of all non-agriculture sector workers
in Tanzania are engaged in informal sector employment, which is offered largely by MSEs.
The group of workers in this category is informally or casually employed, and in addition,
4
The ILFS of 2006 is the latest available and most reliable source of labour and employment statistics in the Tanzania. It is indeed
far behind the current developments.

group of workers in this category is informally or casually employed, a
number of employees engaged in the formal sector have an “informal” sta
in tandem
a good number with the stylized
of employees engagedargument
in the formalthat inhave
sector developing countries,
an “informal” status. This about h
observation is in tandem with the stylized argument that in developing countries, about half to
the employees are informally engaged (Figure 1).
three-quarters of the employees are informally engaged (Figure 1).
Figure
Figure 1: 1: Employment
Employment by sectorby sector
(percent) (percent) in 2006
in 2006
Household act. Parastatal Government
3.2% 0.4% 2.6%
Other private
8.6%
Informal
10.1%
Agriculture
75.1%
Source: (NBS, 2007)

Source: (NBS, 2007).
The ILFS explored reasons why a large proportion of workers were in informal employment,
The ILFS
and the leadingexplored
two causes reasons why
were: (i) lack a large
of formal proportion
employment; and (ii)ofa need
workers were in info
for additional
income to supplement insufficient earnings from formal employment. From this study’s
the leading
survey, two causes
there is evidence were:
that a wide range(i)
of lack ofinformal
activities employment;
the construction and (ii) a nee
industry of Tanzania
to supplement
is informally insufficient
done. The earnings
reasons why firms may optfrom formal
to employ employment.
informally From this
rather than formally
evidence that causes
provide additional a wide rangeemployment,
of informal of activities whichinare:the construction
(i) lack industry
of sufficient capital to of
establish large formal firms; (ii) hassles of tax laws; and (iii) low knowledge about procedures
done. The reasons
and requirements why offirms
for formalization may
micro and smallopt to employ informally rather
enterprises.
additional causes of informal employment, which are: (i) lack of suffici
large formal firms;
2.2 TANZANIA’S POLICY (ii)
VIEWhassles of tax laws;
OF INFORMAL ECONOMYand (iii) low knowledge
requirements for formalization
The current Employment of micro
Policy of Tanzania, anddrafted
which was smallin enterprises.
2008 as a revision of
the previous National Employment Policy of 1997, states its main objective as increasing
employment opportunities. This would lead to poverty reduction by creating an enabling

environment for all stakeholders
to participate fully in human capital development and decent
6
The ILFS of 2006
employment promotion. The national employment is the latest available and policy
most acknowledges,
reliable source of labour
among and employm
other things,
Tanzania.
that Tanzania It is hasindeed to: (i) put far behind in place the measures current
thatdevelopments.
will ensure that the pattern of economic
growth is made more employment-intensive and pro-poor; (ii) increase formal jobs because
formal paid employment opportunities are increasing at a far lower rate than the rate at which
the actual demand for those jobs is growing; and (iii) transform the 59 informal sector so that it
provides decent employment and increase labour productivity in the economy.

Other important policies in which creation and formalization of employment have been
mainstreamed include: national youth development policy, 2007; sustainable industrial
development policy, (1996 – 2020); national population policy (1992, reviewed in 2006);
policy on women in development in Tanzania, 1992; and construction industry policy, 2003.
In all these policies, there are some statements emphasizing the transformation of the informal

sector into formal to enhance decent employment, inter alia. Overall, government policy is clear
in that it does not support informal sector employment but strives to achieve formalization for
some reasons comprising, among others, tapping the tax potential of the informal economy.
Nevertheless, a question that most analysts and researchers worry about is on strengths and
weaknesses of existing policies and the missing link, given that the transformation process is
slow, and at the same time, creation of jobs is increasing faster in the informal economy than
in the formal economy.
2.3 GENERAL ROLE OF INFORMAL SECTOR EMPLOYMENT
In the literature, the informal economy, which creates informal employment in developing
countries, may be good or bad for economic performance. Resolving the dilemma around the
relevance of the informal sector would generate a solution to another dilemma as to whether
expanding or contracting the informal sector will provide a positive gain to the economy as
a whole. While private benefits of informal employment may not be reduced by the informal
settings, the fiscal dividend of growth of informal employment is compromised by tax
avoidance or evasion. Although there can be a substantial difference in significance of the
informal economy and its employment between developed and developing countries, some
studies of the informal sector in developing countries underline its merits and demerits from
economic, social and political dimensions (Tanzi, 1982; Harding and Jenkins 1989; Portes
et. al., 1989; Feige, 1989; Renooy, 1990). Although these advantages and disadvantages cut
across countries, they are different in their magnitude from one country to the other.
A key message to researchers in this subject is that when evaluating the informal economy,
a critical look at its economic, social and political contributions is necessary if one wants to
establish its net impact to the whole economy5 (Portes, et. al., 1989; Harding and Jenkins,
1989).
3. CONTEXT AND THEORETICAL PERSPECTIVES
Informal employment in developing countries is in both formal and informal sectors, though
it is fundamentally a phenomenon of the informal sector. Most studies adopt Harding and
Jenkins’ (1989) criteria to define the informal sector and hence informal employment. The
institutional patterns that shape the informal sector comprise political, economic and social
scopes. Whether in developing or developed countries, the formal-informal employment
dividing factors are similar; some of them being social in nature (Breman, 1980). In Tanzania,
and in the construction sector in particular, most residential buildings in the booming cities
are constructed informally and without government regulations. This explains why a number
of areas in the cities have turned into slums as they have been developed without prior
surveying and planning. There have been several cases of buildings collapsing due to sub-
standard quality of work emanating from lack of enforcement of regulations. The illegality of
activities carries a reasonable weight in characterizing Tanzania’s construction industry and
its employment. This is because construction takes place in some reserved public places, and
the demolishing of such buildings has usually been evident especially when roads and other
5
This study does not seek to resolve the dichotomy between formal and informal sectors, and as to whether the informal sector is
a useful concept in terms of its economic benefits. Rather, it takes a positive view that already exists of employing a reasonable
amount of the workforce. The interest is not to resolve this dichotomy – probably prematurely – but in the dynamics that bolster
employment in the informal sector, including a focus on informal employment that takes place in the formal sector. The approach
adopted is disaggregation of jobs by their characteristics rather than sectoral distinctions.

infrastructure developments are done. Of course, such buildings would have been situated in
the wrong places for a number of years.
From an economic point of view, informal activities are a sum of all economic engagements,
excluding those outside the regulated and legally recognized sector. The sub-criteria used in
economic context of informal employment are several, but this study only identifies the most
critical criteria that are closely related to informal employment for the sake of specificity and
focus. From the ILO’s point of view, which is underpinned by Harding and Jenkins (1989)
and Renooy (1990), the economic sub-criteria of interest include: (i) labour market or status
of labour; (ii) professional status; and (iii) national statistics and tax evasion.
On status of labour, if the market is characterized by undeclared wages or salary entitlements,

absence of social benefits, inappropriate working conditions, and other organizational
difficulties due to being outside formal settings, its employment is regarded as informal. In the
booming construction sector of Tanzania, a large number of active workers who do manual
work do so without safety tools, are under-paid, and have no social security or hardship
allowances. Regarding professional status, wage employees and non-wage employees are
distinguished, with a view that wage employees are in the formal category, while non-wage
are in the informal category. While in developed countries it can sound absurd for a formal
entity to hire some workers informally, in the construction industry of Tanzania for example,
there are cases of formally licensed firms that operate informally, and thus employ some of the
wage employees informally. The other aspect of professional status is about self-employment
which is ideally considered informal (Hart, 1973). In Tanzania, those engaged in constructing
residential buildings are largely self-employed and some of them use unremunerated family
workers or servants entitled to very low wages. The national statistics and tax evasion context
regards the informal sector as all economic activities hidden from statistical systems, either
to avoid reporting altogether or to under-report information (Feige, 1989). In Tanzania,
employment in the construction industry is not entirely transparent to records and thus can be
regarded as either hiding or under-reporting to evade tax.
The conceptual framework for describing informal employment suggests that jobs rather than
persons should be used to assess informality. The reason is that a person can hold two or
more jobs and among those, one or more can be informal. According to the ILO (2003), the
conceptual framework of informal employment disaggregates total employment according to
two dimensions: type of production unit; and type of a job as given in Table 1.
Table 1: Conceptual framework of informal employment

Production Jobs by status in employment
units by type
Own-account Employers Employers Employees Members of
workers Contributing producers’
family cooperatives
Informal Formal Informal Formal Informal Informal Formal Informal Formal
Formal
sector 1 2
enterprises
Informal
sector 3 4 5 6 7 8
enterprises (a)
Households (b) 9 10
Source: ILO (2003).

Notes on rows denoted (a) and (b):
(a) As defined by the Fifteenth International Conference of Labour Statisticians (excluding
households employing paid domestic workers).
(b) Households producing goods exclusively for their own final use and households employing
paid domestic workers.
The dark grey cells in Table 1 refer to jobs that, by definition, do not exist in the type of
production unit in question while the light grey cells refer to formal jobs. Un-shaded cells
represent the various types of informal jobs. Definitions by cells of the framework are as
follows:
Cells 1 and 5: Contributing family workers: no contract of employment and no legal or
social protection arising from the job, in formal sector enterprises (Cell 1) or informal
sector enterprises (Cell5).
Cells 2, 6 and 10: Employees holding informal jobs, whether employed by formal sector
enterprises (Cell 2), informal sector enterprises (Cell 6) or as paid domestic workers by
households (Cell 10).
Cells 3 and 4: Own-account workers (Cell 3) and employers (Cell 4) employed in their
own informal sector enterprises.
Cell 7: Employees holding formal jobs in the informal sector enterprises.
Cell 8: Members of informal producers’ cooperatives. The informal nature of their jobs
follows directly from the characteristics of the producers’ cooperative.
Cell 9: Own-account workers engaged in the production of goods exclusively for own final
use by their household (e.g. subsistence farming).
In developing countries, the debate on the informal sector has been mainly conceptual, while
in industrialised countries it has been methodological, focusing principally on measurement
techniques. The disagreement in developing countries is over what the informal sector
comprises. Therefore, its employment distinction is difficult to make, although there is
agreement over what to call it. In contrast, for advanced countries there is general agreement
over what it is but no agreement on what to call it. Thus, it is referred to as employment.
This is why several terms evolved to describe production which escapes taxation and/or GDP
estimates, like informal, parallel, black, shadow, underground, unrecorded, irregular, hidden,
subterranean economy (Bernabè, 2002).
The informal sector debate which dominated much of the 1970s and 1980s took a duality
approach that focused on the informal-formal sector relationship. This approach distinguishes
between two urban economies: (i) the poor where workers are informally employed; and (ii)
the rich where workers are formally employed. The second strand was the critic of the first
view, which dominated most of the late 1980s and 1990s especially in Latin America. It looked
at both the poor-informally employed and the rich-formally employed as two sides of the same
phenomenon that reinforce each other (de Soto, 1989; Moser, 1994; Weeks, 1975; Mazumdar,
1976; and Roberts, 1990). If we consider characteristics of informal employment in Tanzania
and other developing countries, aspects of labour belong to both of these categories (Allen,
1998; Birkbeck, 1979; Bromley and Gerry, 1979; Moser, 1994; Portes, 1978; Portes, et. al.,
1989).

4. EMPIRICAL PERSPECTIVES
According to Yamada (1996), the central question is whether individuals choose to work in the
informal sector to earn competitive incomes and obtain returns to their entrepreneurial abilities
or they opt to work in the sector as the only alternative at their disposal. Using data from Lima
and Peru for 1985, 1986 and 1990, Yamada tested the basic hypothesis that people self-select
themselves to informal jobs in the urban areas by choice. This understanding contrasts sharply
with the alternative popular view that informal sector employment is an involuntary and
transitory option that provides meagre incomes. The results from the study generally support
the hypothesis of voluntary self-selection and higher earnings in informal self-employment.
There are other studies in developing countries that show that informal employment may be a
desirable job choice (Maloney, 1999; Marcoullier et al., 1997; Pradhan and van Soest, 1997).
These views are insightful for this study as there might be a good number of people opting to
work informally who could also be employable formally but chose to employ themselves in
the informal sector to either earn more, escape the tax net, or for any other reasons.
Kay (2011) uses the South African Statistics definition to explain informal employment as
“… economic activity that occurs outside the purview of state regulation and
… originating from a business or firm that is not registered with the state”
(p.1).
Kay attempts to establish the relationship between formal and informal employment in South
Africa. This study does not make strong conclusions on the determinants of informal sector and
its employment in South Africa but identifies heterogeneous sub-sectors within the informal
sector of the country. This reflects the case of Tanzania in terms of heterogeneity, but the study
seems to have left a vacuum on the causes of informality. Kay’s approach is narrow in that it
is limited to only registration and regulation aspects of informality.
Stoevska (2012) analyses a case of Jordan and highlights factors spurring informal employment
in the country including, loss of jobs and decreased earnings (where former wage workers,
who are unemployed and underemployed, seek work in the informal economy especially after
the job crises). There is also an indication that growth led to increasing vulnerability, job
insecurity and inequality. This is a kind of growth that has been referred to as “immiserizing”,
which comes with offsetting negative effects especially to the poor. Stoevska indicates that
in Jordan, informal employment evolves as a survival strategy following lack of formal jobs,
obstacles to employment in the formal sector, and a need to supplement family income, among
others.
Different studies have found some key determinants of informal employment. Rodman’s
(2007) study on employment and shared growth in North Africa and Middle East finds that
lack of formal employment and government controls, including tax hassles, increase informal
employment. Owing to a shortage of well-paying decent jobs, low incomes contribute to
employees being pushed to accepting informal employment. A World Bank (1999) private
sector assessment for Morocco also finds fiscal restraints (including taxes) as one of the factors
that enhance informality of jobs. Other determinants of informal employment especially in the
Middle East and North Africa are discussed in Diego and Tanabe (2012), and they include

education (taken a proxy for human capital), and gender. The findings are that education is
negatively related to informal employment, and women’s pay is lower than that of men. This
motivates employers to recruit more women than men for low salaried jobs both in the formal
and informal sectors.
While there has been an argument that some workers move to informal employment to earn
more than they could get from the formal sector, this is not always the case. The experience of
Serbia shows that some employees engage in informal employment only as a last resort due
to earnings that are relatively lower compared to what they could otherwise earn in formal
employment (Krstić and Sanfey, 2010).
Even though most studies done in Tanzania on informal employment do not focus on the
construction sector as a case study, one study by Milinga and Lema (2000) analyse informal
contractors focussing on their characteristics and reasons for informality. The study uses the
National Informal Sector Survey (NISS) of 1991, the Dar-es-Salaam Informal Sector Survey
(DISS) of 1995, and a Study on Tanzania Informal Contractors (STIC) 1999/2000. They
examine why contractors would not register with the Contractors Registration Board (CRB)
and instead employ and operate informally. The reasons the study finds include, among others,
existence of possibilities to operate without registration, high cost of registration dynamics,
and difficult requirements.
Mlinga and Lema (2000) explain that there is strong collaboration between formal and
informal constructors in Tanzania (i.e., reinforcement theory holds) and recommend that this
has to be nurtured to enhance transfer of technology to the informal sector so as to increase
employment opportunities. They judiciously note that,
“Any meaningful development programmes for construction industry

should also aim at developing this important but usually ignored sector”
(p.9).
Concerning cost of the formalization process in Tanzania, currently, the registration fees
have been adjusting downwards over time from the cross-cutting amount of Tsh. 2.5 million
(equivalent to US$ 1,700) reported by Mlinga and Lema (2000) during the 1990s and early
2000s. The adjustments are according to the relative size of the enterprises. Nonetheless,
there are annual contributions to the CRB (excluding income tax) which can in a way hinder
registration for some informal construction firms. Figure 2 shows the CRB requirements, and
in this framework, this study looks at civil and building sub-sectors ceilings that projects are
supposed to observe.

registration for some informal construction firms. Figure 2 shows the CRB requirements, a
this framework, this study looks at civil and building sub-sectors ceilings that projec
supposed to observe.
Figure 2: Framework for contractor classification (Civil & Building – Value in
Figure
Million2:Tsh.)
Framework for contractor classification (Civil & Building – Value in Million Ts
This group could be considered to be Established

Established
operating illegally as they have all the Large
Large
necessary r esources and capability to Establishe
Established Contractor
contractor
register a construction company d Medium
Medium
CLASS I,
Established II & III
small CLASS IV Civil =
Emerging & V 3000 (III)
Small
Civil = 5000 (II)
contractor Unlimited
Establishe
Established CLASS VI 750 ( IV)
(I)
d Informal Civil = 300 1500 (V) Building =
Emerging informal CLASS VII Building = Building = 2200 (III)
informal Civil = 150 600*** (IV) 3000 (II)
200**
contractor Building = 1200 (V) Unlimited
Labourer/
120* (I)
(I)
Employee
INFORMAL CONTRACTORS
FORMAL
FORMAL CONTRACTORS
CONTRACTORS

** Class Seven
Class Seven nottoallowed
not allowed build storey to build storey building.
building.
** Class Six are restricted to 3 storeys structures.
*****Class Six
Class Five are are restricted
restricted to 4 storeysto 3 storeys structures.
structures.
*** Class
Source: FiveRegistration
Contractors are restricted to 4 storeys structures.
Board (2013).
Source: Contractors Registration Board (2013).
Figure 2 shows that emerging and small informal contractors take small projects. If they
register2 ashows
Figure bid forthat
higher values while
emerging in class
and small IV, theircontractors
informal values should notsmall
take exceed Tsh. 120If they re
projects.
million (equivalent to US$ 75,472) and Tsh. 150 million (equivalent to US$ 90,340) for
a bid for higher values while in class IV, their values should not exceed Tsh. 120 m
building and civil construction categories, respectively. But as Mlinga and Lema (2000) put it,
some of the informal contractors undertake a number of projects worth far more values than
the required ceilings. This simply means that even the 65 interface between formal and informal
sectors is not clear-cut.
5. MODELLING WORKERS CHOICE BETWEEN FORMAL AND INFORMAL

EMPLOYMENT
The study adopts a conventional approach to model employment participation decision based
on earnings and other individual characteristics, since an individual engages in either formal
or informal employment depending on his/her characteristics. We model determinants of
informal employment using a multivariate logit regression model, a variant of the probabilistic
statistical model (Balakrishnan, 1991; Hosmer and Lemeshow, 2000; Agresti, 2002; Green,
2003). It is a model that can be applied to predict the probabilities of employees to take
either of the two types of employment given a set of independent variables (i.e., employees’
characteristics).
The model specification is based on Green (2003). The logic behind this model is to construct a
linear predictor function that can make a score from a set of weights that are linearly combined
with the explanatory variables of a given observation using a dot product, as in equation (1);

The model
linear predictorspecification
function is based
that can on Green
make (2003).
a score fromThe logic
a set behind this
of weights thatmodel is to construct
are linearly combineda
linearthe
with predictor
explanatory function that can
variables of amake
givenaobservation
score from using a set of weights
a dot thatasare
product, in linearly
equationcombined
(1);
with the explanatory variables of a given observation using a dot product, as in equation (1);
1) 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑿𝑿! , 𝑘𝑘 = ( 𝜷𝜷! ∙ 𝑿𝑿! ).
1) 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑿𝑿! , 𝑘𝑘 = ( 𝜷𝜷! ∙ 𝑿𝑿! ).
Xi denotes
denotes aavector vectorofofexplanatory
explanatoryvariables
variables(or(orfeatures)
features) describing
describing observation
observation i (an employee
i (an employee
Xi this
in denotes case), a vector
βkk isis aaof explanatory
vector
vector of variables
of weights
weights (or (or features)
(or regression
regression coefficients)
coefficients) observation i to
describingcorresponding
corresponding (an employeek
tooutcome
outcome
in this case),
k (typeofofemployment
(type β is
employment
k a vector
inin thisof weights
this case), (or regression
case), and score
score (X coefficients)
(Xi,i, k)k)isisthe
thescore corresponding
scoreassociated
associated with to
with assigningk
outcome
assigning
(type of employment in this case), and86 score (X , k) is the score associated with assigning
individual i to category category of employmentkk. .
of employment i
individual i to category of employment k8.
The
The explanatory
explanatory variables variables and and employment
employmentoutcomes
outcomesrepresent
representobserved
observedproperties
propertiesofofthethe data
The
data explanatory
points points(employees), variables
(employees), which and
whichimplyemployment
theythey
imply outcomes
originate
originatefrom represent
fromthethe observedofproperties
observations
observations of theinin
ofN Nindividuals
individuals data
the
points
labour (employees),
market. While which
the imply
choice they originate
options are from
two, the the observations
strength of the of N
model individuals
is that it in the
allows
the labour market. While the choice options are two, the strength of the model is that it allows
labour
plugging market. While the choice options are two, the strength of the model is that it allows
plugging in in any
any alternative
alternativenumbernumberofofexplanatory
explanatoryfeaturesfeatures(in(indifferent
differentforms)
forms) to to
ascertain
ascertain their
plugging
influence inonany
the alternative
two number ofoutcomes.
employment explanatory features (in logistic
Multivariate differentregression
forms) to ascertain
uses a their
linear
their influence
influence on on the
the two two employmentoutcomes.
employment outcomes.Multivariate
Multivariate logistic
logistic regression
regression uses
uses a linear
a linear
predictor
predictor function
predictor function f(k,
function
i) to
to calculate the
i) to calculate
f(k, i)
f(k,
the probability that
calculate the probability
that employee ii has
probability that employee
has employment outcome
employee i has employment
outcome k,
employment outcome k,
of the following form;
k, of
of thethe following
following form;
form;
2) 𝑓𝑓 𝑘𝑘, 𝑖𝑖 = 𝛽𝛽!,! + 𝛽𝛽!,! ∙ 𝑥𝑥!,! + 𝛽𝛽!,! ∙ 𝑥𝑥!,! + ⋯ + 𝛽𝛽!,! ∙ 𝑥𝑥!,! .
2) 𝑓𝑓 𝑘𝑘, 𝑖𝑖 = 𝛽𝛽!,! + 𝛽𝛽!,! ∙ 𝑥𝑥!,! + 𝛽𝛽!,! ∙ 𝑥𝑥!,! + ⋯ + 𝛽𝛽!,! ∙ 𝑥𝑥!,! .
In equation (2), x ,i represents mthth explanatory variable for employee i, while βm,k defines
In equation (2), 𝑥𝑥m !,! represents mth explanatory variable for employee i, while 𝛽𝛽!,! defines a
a regression coefficient associated with the mth explanatory
variable forvariable andi,the kth employment
In
regression equationcoefficient (2), 𝑥𝑥!,! represents associatedm with explanatory
the mth explanatory employee
variable and while
the k 𝛽𝛽!,!
th defines a
employment
outcome.
regression Usually,
coefficient the regression
associated
outcome. Usually, the regression coefficients and explanatory variables are grouped into vectors coefficients
with the m th and explanatory variables are grouped
explanatory variable and the k th into
employment
vectors
outcome. of size
Usually, M+1,
of size M+1, to have the predictor function written compactly as, the to have
regression the predictor
coefficients function
and written
explanatory compactly
variables as,are grouped into vectors
of size M+1, to have the predictor function written compactly as,
! ∙ 𝑿𝑿! .

3) 𝑓𝑓 𝑘𝑘, 𝑖𝑖 = 𝜷𝜷
3) 𝑓𝑓 𝑘𝑘, 𝑖𝑖 theory,

8
In discrete choice 𝑿𝑿! . observations
= 𝜷𝜷 !
∙where represent people and outcomes represent choices like this, the score is
In
8
considered this specification,
In discrete choice
the gaintheory, associated βk is
where with the
observations set i of
person regression
represent
choosing peoplecoefficients
employment andk.outcomes associated
represent
The predicted with
choices
outcome or likeemployment
choicethis,
ofthe score
work is is
In
outcome
considered this specification,
the
k,the and gain Xi is score.
associated 𝜷𝜷
a𝜷𝜷row ! is
with the set iof regression coefficients
k. associated
vector of the set of explanatory variables associated with employee
person choosing employment The predicted outcome with
or employment
choice of work is outcome
In this
the one with specification, highest ! is the set of regression coefficients associated with employment outcome
k,
the
k, There
i. and
one
and Xare
X
with i is the a
two row
highest
rowmain
vector
score.
ways of
of that
the set
the of of
logit explanatory
model canvariables variables
be specified, associated
and it is with
as follows:employee
employee i. There are
i. There are
i is a vector the set explanatory associated with
two
two (i) main
mainSet ways
ways of independent that
that the logit the logit linear model
model binary can
can be be specified,
regressions.
specified, This and it
and formulationis as
it is as follows: follows:
takes simple log-linear
66
(i) Set
(i) Set
form ofofof independent
predictors.
independent This
linearlinearis binary
the
binaryapproachregressions.
adopted
regressions.
66 This This
in this formulation
study
formulation since ittakes
takes issimple
simplesimpleand log-linear
log-linear

form
suffices
form of of the predictors.
predictors. need of identifying This Thisis isthethe approach
employment
approach adopted
selection
adopted in this
infeatures.
this studystudysincesinceit isitsimple
is simple
and and
sufficesthe
suffices
(ii) Latent-variable theneed needmodel.
ofofidentifying identifying employment
employment
It is usually selection
selection
used because isfeatures.
itfeatures.
easy to compare logistic
(ii) Latent-variable
(ii) Latent-variable
regression with model. themodel. related It Itisprobit
isusually
usually used
used
model, because
because
and also to isiteasy
it extendis easy to compare
to compare
simple logistic logistic
logistic regression
regression
regression
with
with
model the
thetorelated related
more complex probit
probit model, model,
models. and
andSince also
also this to extend
to extend simple
simple
is not the logistic
logistic
objective regression
in regression
this study, itmodel model
to more
is ruled to more
complex
complex
out. models. Since this is not the objective
models. Since this is not the objective in this study, it is ruled out. in this study, it is ruled out.
5.1 As
5.1 AsASaa generalized
generalized
A linearmodel
linear
GENERALIZED model
LINEAR MODEL
The rationale
The rationale for
rationalefor taking
fortaking
taking the
the thenaturalloglog
natural
natural logof
of the
theof probabilities
the is toistransform
probabilities
probabilities to
is transform the variable
to transform to meet
the tovariable
the variable meet to meet
the
the continuous
the continuous criterion,
continuouscriterion,
criterion, and
andand since
since
since it
it also also
it alsohas
has the the practical
haspractical
the practicaleffect
effect ofeffectof converting
of converting
converting the probability
the probability
the probability
(which
(which is
(which is bounded
isbounded
boundedtototobe
bebebetween
between
between 00and
0and
and 1)to
1) toato
1) avariable
variable
a variablethatthat
that canrange
can range
can fromfrom
range
from −∞
to. to +∞.
−∞
Thus, Thus,Thus,
to +∞.
!
4) Logit 𝐸𝐸[𝑌𝑌
4) Logit 𝐸𝐸[𝑌𝑌! ! 𝑋𝑋𝑋𝑋!,!!,!, …
, …𝑋𝑋𝑋𝑋
!,! ==logit ln ln!!!! !! = 𝛽𝛽=!,!𝛽𝛽+ 𝛽𝛽+!,! 𝛽𝛽∙ 𝑥𝑥!,!∙ 𝑥𝑥+ 𝛽𝛽+
logit𝑝𝑝!𝑝𝑝!= = ∙ 𝑥𝑥!,!∙ +
!,! 𝛽𝛽
!,! !
!!! !,! !,! !,! !,! 𝑥𝑥!,! +
!
… + 𝛽𝛽 ∙ 𝑥𝑥
… + 𝛽𝛽!,! ∙ 𝑥𝑥!,! . .
!,! !,!
Equation (4) can be written more compactly as,

Equation (4) can be written more compactly as,
!!
5) Logit 𝐸𝐸[𝑌𝑌! 𝑿𝑿𝒊𝒊 = logit 𝑝𝑝! = ln ! = 𝜷𝜷! ∙ 𝑿𝑿! .
5) Logit 𝐸𝐸[𝑌𝑌! 𝑿𝑿𝒊𝒊 = logit 𝑝𝑝! = ln ! !
!!! = 𝜷𝜷! ∙ 𝑿𝑿! .
!!!!
The formulation in equation (5) expresses logit regression as a generalized linear model. It
The
predicts
6 formulation
In discrete choice theory,in
variables withequation
where various (5)
observationstypesexpresses
ofpeople
represent andlogit
probability regression
outcomesdistributions as by
represent choices a this,
like generalized
fitting a linear
the score linear model. It
predictor
is considered
predicts
function variables
of the with
the gain associated above withform
person various
in an
i choosing types k.ofThe
arbitrary
employment probability
transformation distributions
predicted outcome of of workby
the expected
or choice fitting
is thevalue a the
one withofthe linear predictor
variable.
highest
score.
function
Both of the aboveand
the probabilities form the in an arbitrary
regression transformation
coefficients of the expected
are unobserved, and they are value of the variable.
determined by
Both
the the probabilities and the regression coefficients are BOJE:
61 model. unobserved, and they
Botswana Journal are determined by
of Economics
the model.
5.2 As a log linear model
The formulation in equation (5) expresses logit regression as a generalized linear model. It
predicts variables with various types of probability distributions by fitting a linear predictor
function of the above form in an arbitrary transformation of the expected value of the variable.
Both the probabilities and the regression coefficients are unobserved, and they are determined
by the model.
5.2 AS A LOG LINEAR MODEL
Instead of writing the logit of the probabilities pi as a linear predictor, a more practical
formulation is adopted. It is one of the standard specifications borrowed from the multinomial
logit, which takes a combination of the generalized linear model and “the two-way latent
variable” specification.
There is a simple way to arrive at the log linear multivariate logit model for estimation of
employment selection between formal and informal jobs. This involves coming up with
possible employment decisions or outcomes, running K-1 independent binary logit regressions
for K (K = 2 in this logit case), in which one outcome is chosen as a pivot and then the
other employment outcome is separately regressed against the pivot employment outcome. If
employment outcome 2 (the last one) is chosen as the pivot for example, then:
!" (! !!)
!" (! !!!)
6) ln !" (!! !!) = 𝜷𝜷!! ∙ 𝑿𝑿!!.
!" (!!!!!)
By exponentiating both sides and then solving for the probabilities, we get;
7) 𝑃𝑃𝑃𝑃(𝑌𝑌!! = 1) = 𝑃𝑃𝑃𝑃 (𝑌𝑌!! = 2)𝑒𝑒 𝜷𝜷𝜷𝜷!!∙𝑿𝑿

∙𝑿𝑿!!
.
The sum of the two probabilities of choice of formal or informal categories of employment has to
be 1. Solving for the probability of second choice (informal employment) gives,
! !
8) 𝑃𝑃𝑃𝑃(𝑌𝑌!! = 2) = !!!𝜷𝜷𝜷𝜷!!∙𝑿𝑿∙𝑿𝑿!! .
!!!
By the same token, the probability of formal employment selections can be specified as:
! !
9) 𝑃𝑃𝑃𝑃(𝑌𝑌!! = 1) = !!!𝜷𝜷𝜷𝜷!!∙𝑿𝑿∙𝑿𝑿!!.
!!!
5.3 Estimating
5.3 ESTIMATINGthe parameters
THE PARAMETERS
The unknown
The unknown parameters
parameters inin each
each vector are ββ
vector are kk k which
whicharearejointly
jointlyestimated
estimatedbybymaximum
maximuma
posteriori (MAP) estimation (an extension of maximum likelihood through
a posteriori (MAP) estimation (an extension of maximum likelihood through regularizing regularizing the
weights to avoid extreme solutions). The solution is typically found using an iterative
the weights to avoid extreme solutions). The solution is typically found using an iterative procedure
such as iteratively
procedure such as reweighted least squares
iteratively reweighted (IRLS).
least squares The(IRLS).
logarithm
Theoflogarithm
the probability of seeing a
of the probability
given employment choice is modelled using the linear predictor as
of seeing a given employment choice is modelled using the linear predictor as well well as an additional
as an
normalization factor, call it Z. An additional term, –lnZ, is needed to enter the separate
additional normalization factor, call it Z. An additional term, –lnZ, is needed to enter the
probability estimations to ensure the whole set of probabilities forms a probability distribution
separate
such probability
that they all sum estimations
to one, as thetotheory
ensurerequires,
the whole set of probabilities forms a probability
distribution such that they all sum to one, as the theory requires,
!!!
!!!
10) 𝑃𝑃𝑃𝑃(𝑌𝑌!! = 𝑘𝑘) = 1 .

!!!
!!!
The normalization term is additive rather than the usual multiplicative factor. The logarithm of
theBOJE:
probabilities is taken
Botswana Journal to give;
of Economics 62
10) 𝑃𝑃𝑃𝑃(𝑌𝑌! = 𝑘𝑘) = 1 .
!!!
The normalization term is additive rather than the usual multiplicative factor. The logarithm of
The normalization
the probabilities term to
is taken is give;
additive rather than the usual multiplicative factor. The logarithm
of the probabilities is taken to give;
,

,
Exponentiating both sides turns the additive term into a multiplicative factor, and in this process,
itExponentiating
shows why the term
both is written
sides turns theinadditive
the form of –lnZ
term into arather than simply
multiplicative +lnZ.
factor, and Exponentially,
in this process,
it shows why the term is written in the form of –lnZ rather than
expressions in the system of equations in equation (9) are transformed into, simply +lnZ. Exponentially,
expressions in the system of equations in equation (9) are transformed into,
68
value of Z is computed by applying the constraint that requires the sum of all probabilities to
The
be equal to 1:
!!! !
13) 1 = !!! 𝑃𝑃𝑃𝑃(𝑌𝑌! = 𝑘𝑘) = 𝑒𝑒 𝜷𝜷! ∙𝑿𝑿! + 𝑒𝑒 𝜷𝜷! ∙𝑿𝑿! .
!
Solving for Z gives,

!!! !! ∙!!
14) 𝑍𝑍 = !!! 𝑒𝑒 = 𝑒𝑒 𝜷𝜷! ∙𝑿𝑿! + 𝑒𝑒 𝜷𝜷! ∙𝑿𝑿! .
Substituting equation (14) in (12), the probabilities equations resulting from these manipulations
are:
𝑒𝑒 𝜷𝜷1 ∙𝑿𝑿𝑖𝑖
15) 𝑃𝑃𝑃𝑃(𝑌𝑌𝑖𝑖 = 1) =
𝑒𝑒 𝜷𝜷1 ∙𝑿𝑿𝑖𝑖 +𝑒𝑒 𝜷𝜷2 ∙𝑿𝑿𝑖𝑖
𝑒𝑒 𝜷𝜷2 ∙𝑿𝑿𝑖𝑖
𝑃𝑃𝑃𝑃(𝑌𝑌𝑖𝑖 = 2) =
𝑒𝑒 𝜷𝜷1 ∙𝑿𝑿𝑖𝑖 + 𝑒𝑒 𝜷𝜷2 ∙𝑿𝑿𝑖𝑖
The general form is specified as,
𝑒𝑒 𝜷𝜷! ∙𝑿𝑿!
16) 𝑃𝑃𝑃𝑃 𝑌𝑌! = 𝑟𝑟 = .
𝑒𝑒 𝜷𝜷! ∙𝑿𝑿! + 𝑒𝑒 𝜷𝜷! ∙𝑿𝑿!
Further extension to the “softmax” function is done, which is specified as,
! !!
17) 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑘𝑘, 𝑥𝑥! , … , 𝑥𝑥! = ! ! !! .
!!!
By exponentiating
By exponentiatingthe thevalues
valuesx1x, 1...,x
, ...,x , there is exaggeration of the differences between them
n, nthere is exaggeration of the differences between them and
and isthis
this theisreason
the reason this function
this function is named is named softmax.
softmax. Therefore,
Therefore, softmaxsoftmax (k,xnx)1,will
(k, x1, ..., ..., xreturn
n
) willa
returnclose
value a valueto 0close to 0 whenever
whenever xk is significantly
xk is significantly less than less
the than the maximum
maximum of all theofvalues,
all theandvalues,will
and will
return returnclose
a value a value
to 1 close
whento 1 when
applied to applied to the maximum
the maximum value,
value, unless it is unless it isclose
extremely extremelyto the
close to the value.
next-largest next-largest value. The
The softmax functionsoftmax function is
is employed to employed
construct ato weighted
construct average
a weighted that
behaves
average as a smooth
that behavesfunction, which approximates
as a smooth function, which the non-smooth
approximates function max(x1, ..., xfunction
the non-smooth n ) given
as,
max(x1, ..., xn) given as,
!
18) 𝑓𝑓 𝑥𝑥! , … , 𝑥𝑥! = !!! 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑖𝑖, 𝑥𝑥! , … , 𝑥𝑥! 𝑥𝑥! ≈ max (𝑥𝑥! , … , 𝑥𝑥! ).
The
63 probability equation for one to choose either type of employment can be written
BOJE: Botswana Journal ofas,
Economics
19) 𝑃𝑃𝑃𝑃 𝑌𝑌! = 𝑟𝑟 = 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑟𝑟, 𝜷𝜷𝟏𝟏 . 𝑿𝑿! , 𝜷𝜷𝟐𝟐 . 𝑿𝑿! , .

next-largest value. The softmax function is employed to construct a weighted average that
behaves as a smooth function, which approximates the non-smooth function max(x1, ..., xn) given
as,
!
18) 𝑓𝑓 𝑥𝑥! , … , 𝑥𝑥! = !!! 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑖𝑖, 𝑥𝑥! , … , 𝑥𝑥! 𝑥𝑥! ≈ max (𝑥𝑥! , … , 𝑥𝑥! ).
The probability equation for one to choose either type of employment can be written as,
19) 𝑃𝑃𝑃𝑃 𝑌𝑌! = 𝑟𝑟 = 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑟𝑟, 𝜷𝜷𝟏𝟏 . 𝑿𝑿! , 𝜷𝜷𝟐𝟐 . 𝑿𝑿! , .

All
All vectors 𝜷𝜷𝒌𝒌 coefficients
vectorsofof coefficients are uniquely
are uniquely identifiable
identifiable since thesince theprobabilities
sum of sum of probabilities
must be must
69
equal to 1, making one of the respective probabilities completely determined once the other
equal to 1, making one of the respective probabilities completely determined once the other is
known. Equation(19)
known. Equation (19) shows
shows the types
the types of empirical
of empirical factors factors or that
or features features that will the
will maximize maximize
employee’s softmax function as modelled. It is common parlance that
employee’s softmax function as modelled. It is common parlance that the assessment of suchthe assessment of su
features
features isisdone
done based
based on significance
on the the significance
of the of the estimated
estimated parameters,
parameters, the. Fromthe 𝜷𝜷
equation From equati
𝒌𝒌 ′s.(19)
(19)
we can wespecify
can specify the empirical
the empirical model asmodel as
20) 𝑌𝑌 = 𝛽𝛽! + 𝛽𝛽! 𝑥𝑥! + 𝛽𝛽! 𝑥𝑥! + 𝛽𝛽! 𝑥𝑥! + 𝛽𝛽! 𝑥𝑥! + 𝛽𝛽! 𝑥𝑥! + 𝜀𝜀.
This study
This studyemploys
employsequation (20) (20)
equation to estimate the determinants
to estimate of employment
the determinants choice in choice
of employment
Tanzania’s construction
Tanzania’s industry.
construction The The
industry. definitions of the respective
definitions variables variables
of the respective which enterwhich
this enter t
equation for both micro and small entrepreneurs and the firms are given in sub-section 5.5.
equation for both micro and small entrepreneurs and the firms are given in sub-section 5.5.
5.4 Definition
5.4 DEFINITION OF VARIABLES
of Variables
Two data
Two datasets
setscontain information
contain informationon employment selection.
on employment These are
selection. (i) a set
These areof(i)micro
a setand
of micro a
small entrepreneurs, and (ii) a set of constructions firms. Variables included are based
small entrepreneurs, and (ii) a set of constructions firms. Variables included are defined defined based
on the theoretical and empirical underpinnings, and the way they are hypothesised in both
the theoretical and empirical underpinnings, and the way they are hypothesised in both contex
contexts. This means that two respective models explaining informal employment (specified
This means that two respective models explaining informal employment (specified by equati
by equation (20) are estimated.
(20) are estimated.
5.5 MICRO AND SMALL ENTREPRENEURS’ INFORMAL EMPLOYMENT
5.5 Micro and smallMODEL
ESTIMATION entrepreneurs’ informal employment estimation model
In estimatingthis
In estimating thismodel,
model, a type
a type of employment
of employment (Y) is(Y) is a dependent
a dependent variable,
variable, which depends
which depends
on theiXi’s (independent variables). Y is drawn from the survey information since the workersthe work
the X ’s (independent variables). Y is drawn from the survey information since
indicated whetherthey
indicated whether theyareare in formal
in formal (denoted,
(denoted, 1) or informal
1) or informal (denoted,(denoted, 2) employment.
2) employment. The T
probability
probability thatthatY Y
== r ranges
r ranges from from 0 to
0 to 1. While  is an�
1. While is an
error error
term, the term, thepredetermined
included included predetermin
variables
variables are aredefined
definedand and hypothesized
hypothesized as follows:
as follows:
(i) Education (x ): is constructed
(i) Education (x1):1 is constructed by different by different
levels levels of education,
of education, with a continuum
with a continuum starting starti
from
from no noeducation
education (denoted
(denoted by 0)byto0) to college/university
college/university (denoted(denoted
by 5). It isbyhypothesized
5). It is hypothesiz
that thehigher
that the higherthethelevellevel of education
of education the lower the lower the probability
the probability of an to
of an employee employee
choose to choo
informal
informal employment
employment (a (a negative relationship between
negative relationship between education
educationand and employment
employment choice)
(ii) Income
choice). (x2): is reported directly in the data set. The overriding hypothesis is that
(ii) informal ): is reported
Income (xsector
2 attracts directly
micro andin the
smalldataentrepreneurs
set. The overridingowinghypothesis is that
to relatively the earning
better
informal
for this sector
categoryattracts micro and small
of businesses entrepreneurs
– than that of the owing to relatively
formal sector.better earningsrelationsh
A positive
– for this income
between categoryand of businesses
employment – than that of
choice is the
thusformal sector. A positive
hypothesized relationship
in this case.
between income andconstraint
(iii) Capital/financing employment choice
(x3): this isis thus hypothesized
a binary variable,inwhere
this case.
capital access is perceiv
(iii) as a constraint (denoted by 1).3 It increases the probability of one capital
Capital/financing constraint (x ): this is a binary variable, where access is employ
to be informally
perceived as a constraint (denoted by 1). It increases the probability of
Where it is not a constraint (denoted by 0), it decreases the probability of being informa one to be informally
employed. Where
employed. A positiveit is not
signaisconstraint (denoted by 0), it decreases the probability of
thus hypothesized.
being informally employed. A positive sign is thus hypothesized.
(iv) Tax hassles (x4): it is also a binary variable. If this variable is perceived as a reason
micro and small entrepreneurs to engage in construction industry informally, it is given
BOJE:value of 1;
Botswana otherwise
Journal of Economics it is given a value of 0 if it is not mentioned as a factor. 64 A dir
relationship is expected between tax hassles and the probability of informal employm
choice.
(iv) Tax hassles (x4): it is also a binary variable. If this variable is perceived as a reason for
micro and small entrepreneurs to engage in construction industry informally, it is given
a value of 1; otherwise it is given a value of 0 if it is not mentioned as a factor. A direct
relationship is expected between tax hassles and the probability of informal employment
choice.
(v) Gender (x5): this variable was recorded in the interviews, with 1 representing male and
2 female. In this case gender (x5) is not included as one of the explanatory variables
because activities of informal construction undertaken by micro and small entrepreneurs
are unambiguously male-dominated.
5.6 FIRMS’ INFORMAL EMPLOYMENT ESTIMATION MODEL
The second estimation model takes advantage of employment choice option that was
gathered in the firms’ questionnaire. Three hypothesized explanatory variables under this
case are education, income and gender. The education variable (denoted as x2) is defined and
hypothesized exactly as in the former case. There is, however, a change in the relationship
between income and the two job categories in the construction companies. Gender is also
included here as one of the variables of analytical interest since there is no overwhelming
gender bias for official formal or informal employment as it was presupposed for the selection
of manual construction employment of the micro and small freelancers. In firms there are
some activities that can be attractive to either gender.
(i) Income (x2): it is reported in the data set. The context of income of informal
employment taken here is that of last resort. Thus, the informally employed workers
in the formal sector may be increasing with the possibilities of paying low wages
that can reduce costs to the firms. In this respect, the lower the income the higher
is the probability of one to be informally employed in the firm. Thus, an inverse
relationship with the employment variable is predicted.
(ii) Gender (x5): in the firms’ case, gender enters in the estimation. In the construction
firms’ context, contrary to micro and small construction entrepreneurs, females may
be more informally employed to take charge of a number of activities that are not too
masculine in nature. Therefore, the sign of the parameter on gender is positive.
6. ESTIMATION OF DETERMINANTS OF INFORMAL EMPLOYMENT
6.1 DATA DESCRIPTION
The data used to analyse the determinants of informal employment were obtained from a
survey7 which was done in 2013 in six urban areas of Tanzania, namely, Dar es Salaam, Tanga,
Arusha, Mwanza, Dodoma and Mbeya. The focus was on urban areas given the interest in
informal non-agricultural activities, and particularly, the construction industry. It is reasonable
to argue that non-agricultural informal employment competes with formal employment more
than it does with agricultural activities in any developing economy like Tanzania. Furthermore,
the construction industry can ideally be better modelled in urban than in rural areas given that
the scale and intensity of these activities is more in cities and towns than in rural areas.
The reason we opted for a survey is that the available labour force statistics are of 2000/02 and 2006, which may not adequately
7
reflect the current status. Further, some key information we wanted like inputs and outputs of construction firms are unavailable
in the accessible data sets.

The sample population comprised of construction industry stakeholders in the top six urbanized
regions in Tanzania. Three main categories of stakeholders were interviewed, namely,
firms (both formal and informal); micro and small entrepreneurs; and policy makers and
practitioners. It is worth mentioning that for the firms, owners and workers were interviewed,
while for policymakers, all those who were in positions that can influence policies were
interviewed.8 The number of sample elements chosen from each region was based on the NBS
population census weights. The weights were based on urban population in each town and
were applied to decide on the number of respondents to be interviewed. For Dar es Salaam,
sampling was done differently, with the weights based on population in each district as a way
to distribute the number of envisaged respondents according to the population density of the
municipals. The distribution of interviewees is given in Table 2. The construction activities
that micro and small entrepreneurs do are as follows; masonry, ceiling board fixing, carpentry,
electrical installation, painting, architecture, brick making, iron welding for door and window
grills, plumbing, and floors finishing. For the 73 construction firms interviewed, 328 of their
workers were also independently interviewed. This distribution makes an aggregate set of
1,874 interviewees.
Table 2: Regional distribution of interviewed respondents

Micro & Small Entrep. Firms Policymakers
Number Percent Number Percent Number Percent
Dar es salaam 785 54.3 26 35.6 16 57.1
Tanga 105 7.3 7 9.6 2 7.1
Arusha 132 9.1 12 16.4 2 7.1
Mwanza 211 14.6 13 17.8 4 14.3
Mbeya 143 9.9 9 12.3 4 14.3
Dodoma 69 4.8 6 8.2 – –
Total 1445 100 73 100.0 28 100
Following the theoretical perspective, variables that are most influential to a worker’s decision
to opt for either of the choices, and are the ones hypothesized, entered into the estimation
equation (equation 20). This equation includes employment choice as the dependent variable,
and its regressors are; education, income prospects, capital or financing constraints and
perceptions of tax hassles for the micro and small entrepreneurs’ case. For the firms’ case,
equation (20) includes employment choice (the dependent variable) and the adopted regressors
are education, income and gender.
6.2 SOME DESCRIPTIVE STATISTICS
Employment and earnings or income are usually intertwined. The context of income that is
used for micro and small entrepreneurs in this study is that of income as an incentive for choice
of employment. Figure 3 shows that for micro and small entrepreneurs, income is normally
distributed. This means that the median and mode of this variable are on average the same as
Three separate questionnaires were used, namely: (i) firms – this captured employment statistics and views of owners or
8
management of the firms on one side, and the same for their workers on the other; (ii) micro and small enterprises – this
captured data on entrepreneurs and employees. This category of respondents is important to this study in a special way because
in general, employees are hired informally, and owners are also employees, and (iii) policy makers – this questionnaire captured
data from those in government ministries, agencies and departments. Further, the same questionnaire gathered information
from the members of research and academic institutions and non-governmental organization (both local and multinationals).
The understanding was to use this questionnaire to collect information from all who are in influential positions regarding policy
making.

distributed.
distributed. ThisThis means
means thatthat the
the median
median and and mode
mode of of this
this variable
variable areare on
on average
average the
the
mean, and in this view there may be some kind of stability
mean, and in this view there may be some kind of stability of income distribution of income distribution
participating
the mean, and inindividuals.
participating this view there Micro
individuals. Micro and
may be some small
and kind
small entrepreneurs
of stability of incomeseem
entrepreneurs to
to have
distribution
seem among
have been
been establ
establ
manner
manner that distributes their earnings across with some reasonable equity among the
the that
participating distributes
individuals. their
Micro earnings
and small across with
entrepreneurs some
seem toreasonable
have been equity
establishedamong the
in thethis
and manner
can that
be distributes
one of their
the earningsvoluntary
reasons across with formalization
some reasonable may equitybe among the Among th
slow.
and this
participants
can be one of
and this can be55%
the reasons
one ofof thethem
voluntary
reasonshave
voluntary
formalization
formalization may
may be
be slow.
slow.
Among
Among th
small
small entrepreneurs,
entrepreneurs, 55% of primary school level of education, and
the micro and small entrepreneurs, 55%them have
of them have primary school
primary school level level of education,
of education, and a and
number
number of
of tertiary
tertiary education
education qualifications.
qualifications. In
In contrast
contrast to
to MSEs,
MSEs,
negligible number of tertiary education qualifications. In contrast to MSEs, the firms’ income
the
the firms’
firms’ income
income d
d
non-normal
non-normal (see Figure
(see Figure
distribution is non-normal 4).
4). The
(see Figure The major
4). Themajor reason
reason
major reason behind
behind
behind this could
thisbe could
this could be firms
be firms hav
firms having hav
characteristics,
characteristics, and
andand
diverse characteristics, hence
hencethe
hence the income
the income
income gap gap
gap is
is significant
significant
is significant across
across different
across different different categories
categories
categories of
individual
individual workers.
workers.
firms and individual workers.
Figure
Figure 3:3:
Figure Income
Income
3: distribution
distribution
Income among among
distribution micro andmicro
among and
and small
small entrepreneurs
small entrepreneurs
micro entrepreneurs
88
7.5
7.5
variable
incomevariable
77
Logofofincome
6.5
6.5
Log
66
5.5
5.5
0 50 100 150 200 250

0 50 100 150 200 250
Frequency
Frequency
Figure
Figure 4:4:
Figure 4: Income
Income
Income distribution
among among
distribution
distribution the firms’the
among the firms’
firms’ workers
workers workers
77
66
income
Logofofincome
55
Log
44
33
0 10 20 30 40
0 10 20 30 40
Frequen cy
Frequen cy
6.3 REGRESSION RESULTS

6.3
6.3 Regression
Regression results
results
6.3.1 MICRO AND SMALL ENTREPRENEURS
6.3.1
6.3.1 Micro
Micro and
and small
small entrepreneurs
entrepreneurs
Table 3 gives the estimation results for the micro and small entrepreneurs. Data cleaning,
Table 3
3 gives
which involved
Table the
the estimation
givesdropping some of the
estimation results for
for the
respondents
results micro
thewith and
missing
micro small
smallforentrepreneurs.
andvalues some variables, Data
entrepreneurs. Data clea
clea
involved
involved dropping
reduced the dropping some
estimated some of
sample of the
fromthe respondents
respondents with
1,445 to 1,230. The with missing
samplemissing values
was still values for
large for some
enough some variables,
to give variables,
reliable estimates for thefrom
estimated logit regression.
estimated sample
sample from 1,445 1,445 toto 1,230.
1,230. The
The sample
sample was was still
still large
large enough
enough to to g
g
estimates
estimates forfor the
the logit
logit regression.
regression.
73
73

Table 3: Regression results for micro and small entrepreneurs
Number of obs = 1230
Wald chi2(4) = 135. 04
Prob > chi2 = 0
Log likelihood = -30.622
Employment Coefficient Std. Err. z P>z [95% Conf. Interval]
Education 0.099 0.548 0.18 0.857 -0.976 - 0.174
Income 0.638 0.235 2.72 0.007*** 0.177 - 1.098
Capital constraint 1.665 0.954 1.74 0.081* -0.206 - 3.534
Taxes 0.572 1.159 0.49 0.622 -1.701 - 2.845
Note: ***1%; ** 5%; and * 10%
The goodness of fit of this model is generally fine. The results confirm that while education and
tax hassles are insignificant factors for employment choice, income and capital constraints are
positive and significant at 1% and 10% levels, respectively. Starting with income, these results
show that the higher income earned from the informal sector than the formal sector, the higher
the probability of choosing informal employment. This validates the hypothesis. The finding
implies that micro and small entrepreneurs choose informal employment owing to better
return prospects than they can actually earn from formal employment. Capital constraints are
also significant and in line with the hypothesis, which indicates that one of the features that
lead to high probability of choice of informal employment is lack of capital.
6.3.2 MARGINAL EFFECTS – MICRO AND SMALL ENTREPRENEURS
Marginal effects were then estimated to see the impact of change in the significant predictor
variables on the probability of choice of informal employment. The results are given in Table
4.
Table 4: Marginal effects for micro and small entrepreneurs

Number of observations = 1230
Expression: Pr(employment), predict()
dy/dx w.r.t. : x1education x2income x3capital constraint x4taxes at
Education = 0.436 (mean)
Income = 6.582 (mean)
Capital constraints = 0.759 (mean)
Taxes = 0.206 (mean)
Dy/dx Std. Err. z P>z 95% Conf. Interval
Education -0.001 0.001 -0.97 0.332 -0.003 – 0.001
Income 0.002 0.001 1.80 0.072* -0.000 – 0.004
Capital constraint 0.005 0.003 1.81 0.070* -0.000 – 0.009
Taxes 0.001 0.003 0.42 0.673 -0.005 – 0.008
Note: *** 1%; ** 5%; and * 10%

The results show that the marginal effects of income and capital constraints are significant
at 10% level. An increase in income by a unit can lead to a rise in probability of one to be
informally employed by 0.002, while an increase in capital constraint by a unit raises the
probability of informal employment by 0.005.
6.3.3 INFORMAL EMPLOYMENT IN FORMAL CONSTRUCTION FIRMS
Regression equation (20) was fitted to the dataset of workers of the construction firms. Data
cleaning downsized the sample from 385 workers to 247. This is a reasonable sample size for
estimation of a logit model. Table 5 presents the estimation results.
Table 5: Regression results for construction firms

Number of obs = 247
Wald chi2(3) = 63.23
Prob > chi2 = 0
Log likelihood = -119.755
Employment Coefficient Std. Err. z P>z [95% Conf. Interval]
Education -0.593 0.224 -2.64 0.008*** -1.033 – (-0.154)
Income -0.957 0.303 -3.16 0.002*** -1.551 – (- 0.363)
Gender 7.213 1.392 5.18 0.000*** 4.485 – 9.942
Note: ***1%; ** 5%; and * 10%

All explanatory variables are significant at 1% level, and the goodness of fit of this model is
good. All signs are in tandem with the prior hypotheses. Regarding education, the higher the
level of education, the less likely one will be employed informally in a firm. This implies that
formal construction companies mostly hire informal workers with low or inadequate education,
training and skills. For income, the results give a clear message that the lower the accepted
payment or income for those who are informally employed in construction firms, the more
likely the employers will hire them informally. In light of these results, informal employment
is inter alia, propagated by a trade-off entailed in labour cost reduction versus engagement of
high quality workers. Employing people formally means abiding by the labour laws to pay
specified wages or salaries, and providing the necessary packages of social security.
On the aspect of gender, the results indicate that in construction firms, women are more
informally employed than formally. This means that as more women seek work in these firms,
the more the possibility of the firms to engage them informally for low wages and salaries.
From intuitive deduction, since low education and informal employment choice are positively
related, and because there is still imbalance between men and women trained in technical
skills in Tanzania, it is most likely that gender would be directly related with informality in
employment.
6.3.4 MARGINAL EFFECTS – FIRMS’ INFORMAL WORKERS
Marginal effects were estimated to ascertain the impact of change in the respective significant
predictor variables on the probability of the firms’ choice of employing informally. The results
are reported in Table 6.

Table 6: Marginal effects – firms’ informal workers
Number of obs = 247
Expression: Pr(yempoy), predict()
dy/dx w.r.t.: x1education x2income x3gender at
Education = 2.502 (mean)
Income = 4.734 (mean)
Gender = 1.040 (mean)
Dy/dx Std. Err. z P>z 95% Conf. Interval
Education - 0.089 0.345 -2.59 0.010*** -0.157 – 0.002
Income - 0.144 0.041 -3.51 0.000*** -0.224 – 0.063
Gender 1.084 0.167 6.49 0.000*** -0.757 – 1.411
Note: *** 1%; ** 5%; and * 10%
The marginal effects for the three predictors are all significant at 1%. Raising education by one
level will reduce probability of one being informally employed in the firms by 0.09. If income
increases by a unit, the probability of being informally employed is reduced by 0.14, while
being a woman increases the probability of being informally employed by 0.08.
7. CONCLUSION AND POLICY IMPLICATIONS
This study aimed to empirically ascertain the factors that determine informal employment
in Tanzania’s construction industry. The study employed a logit regression model to unearth
features that significantly influence the choice of informal employment for micro and small
entrepreneurs. The results revealed that the higher earnings in informal jobs compared to
those in formal ones, given the professional status of the micro and small entrepreneurs, is
one of the major incentives to choose informal rather than formal employment. Another factor
that exacerbates choice of informal employment is lack of capital, which deters micro and
small entrepreneurs from starting large formal firms but instead resort to unregistered petty
undertakings.
From the firms’ data set, the results show that low education (i.e. inadequate skills and
knowledge) is one of the key reasons workers are hired informally by formal firms. The second
reason is the possibility of the formal firms to hire these employees at low wages. Lastly, on
the gender issue, the findings show that more females are employed informally than men, most
likely in jobs related to office handling and clerical matters.
In line with the findings of this study, there are important policy issues for different
employment-creating stakeholders to note. First and foremost, improving financial services
through risk mitigation, credit information dissemination and outreach to MSEs are critical to
capital access and business operations finance. Improving these services are an incentive to
start up formal construction firms. Secondly, policies to enhance and rationalize earnings in
the economy could provide an incentive to micro and small entrepreneurs to formalize. Lastly,
regarding formal employability of workers, improving quantity and quality of education is
important as an enabling instrument. In pursuit of educational achievements, gender imbalance
has to be addressed as a way of increasing professionalism of women and to emancipate them
from being preys of informal employers. While the government plays its role in addressing
these implied solutions, firms in the construction industry have to play their part in terms of
providing on-the-job-training and educational sponsorships for their staff.
REFERENCES
AGRESTI, A., (2002), Categorical Data Analysis, New York, Wiley-Interscience.

ALLEN, T., (1998), “From Informal Sectors to Real Economies, Changing Conceptions of
Africa’s Hidden Livelihoods”, Contemporary Politics, vol. 4 (4): 357-373.
BALAKRISHNAN, N., (1991), Handbook of the Logistic Distribution, Marcel Dekker,
Inc.
BECKER, K.F., (2004), “Informal Economy: Facts Finding Study”, Swedish International
Development Cooperation Agency (SIDA), Stockholm.
BERNABÈ, S., (2002), “Informal Employment in Countries in Transition: A conceptual
framework”, Centre for Analysis of Social Exclusion, London School of Economics, Centre
for the Study of African Economies (CSAE), Paper 56.
BIRKBECK, C., (1979), “Garbage, Industry, and the ‘Vultures’ of Cali, Colombia,” in
Casual Work and Poverty in Third World Cities, Ray Bromley and Chris Gerry (eds.), New
York, Wiley.
BREMAN, J.C., (1980), The Informal Sector in Research: Theory and Practice, Volume 3
of CASP (Series), CASP Erasmus University, Rotterdam.
BROMLEY, R. AND C. GERRY, (1979), “Who are the Casual Poor?”, in Casual Work and
Poverty in Third World Cities, Ray Bromley and Chris Gerry (eds.), New York, Wiley.
CHEN, A. M., (2007), “Rethinking the Informal Economy: Linkage with the Formal
Economy and the Formal Regulatory Environment”, DESA Working Papers No. 46.
CONTRACTORS REGISTRATION BOARD, (2013), “Class Limit for any Single
Contract (In Million TShs.)”, http://www.crbtz.org/contractorsclass.asp; accessed: September
2013.
DE SOTO, H., (1989), The Other Path: The Invisible Revolution in the Third World,
Harper and Row, New York.
DIEGO F. AND A. TANABE, (2012), “Micro-Determinants of Informal Employment in
the Middle East and North Africa Region”, Social Protection and Labour Discussion Paper,
World Bank.
ECONOMIC AND SOCIAL RESEARCH FOUNDATION (ESRF), (2011), Informal
Sector Taxation, ESRF Publication, Dar es Salaam.
FEIGE, E.L., (1989) (ed.), The Underground Economies: Tax Evasion and Information
Distortion, Cambridge University Press, Cambridge and New York.
GREENE, W. H., (2003), Econometric Analysis, Prentice Hall.
HARDING, P. AND R. JENKINS, (1989), The Myth of the Hidden Economy: Towards a
New Understanding of Informal Economic Activity, Open University Press, Milton Keynes,
Philadelphia.
HART, K., (1973), “Informal Income Opportunities and Urban Employment in Ghana”,
Journal of Modern African Studies, vol. 11 (1): 61-89.
HOSMER, W. AND S. LEMESHOW, (2000), Applied Logistic Regression, Wiley.
HUSSMANNS, R., (2001), “Informal sector and informal employment: elements of a
conceptual framework”, Paper presented at the Fifth Meeting of the Expert Group on Informal
Sector Statistics (Delhi Group), New Delhi, 19-21, September.
ILO, (2002), “Decent Work and the Informal Economy”, Report of the Director-General;
International Labour Conference, 90th Session; Report VI; Geneva.
ILO, (2003), “Guidelines concerning a statistical definition of informal employment”,
adopted by the Seventeenth International Conference of Labour Statisticians (November-
December 2003); in: Seventeenth International Conference of Labour Statisticians, Report of
the Conference, Geneva.

ILO, (2013), “Informal Economy”, http://www.ilo.org/global/topics/employment-
promotion/informal-economy/lang--en/index.htm; accessed: November, 2013.
JACK K., (1983), “A Theory of Qualitative Methodology”, in R. M. Emerson (ed.),
Contemporary Field Research, Waveland Press, Prospect Heights, Illinois.
KAY, D. D., (2011), “The Relationship between Formal and Informal Employment in
South Africa”, unpublished Master’s Thesis, Urbana, Illinois.
KRSTIĆ, G. AND P. SANFEY, (2010), “Earnings Inequality and the Informal Economy:
Evidence from Serbia”, European Bank Working Paper 114.
LEWIS, A. W., (1955), The Theory of Economic Growth, Allen and Unwin London.
MALONEY, W.F., (1999), “Does informality imply segmentation in urban labor markets?
Evidence from sectoral transitions in Mexico”, World Bank Economic Review, vol. 13 (2):
275-302.
MARCOULLIER, D., V. RUIZ DE CASILLA, AND C. WOODRUFF, (1997), “Formal
measures of the informal-sector wage gap in Mexico, El Salvador, and Peru”, Economic
Development and Cultural Change, vol. 45 (2): 367-92.
MAZUMDAR, D., (1976), “The Urban Informal Sector”, World Development, 4 (8): 655-
79.
MILES, M. B., AND A.M. HUBERMAN, (1994), Qualitative Data Analysis, Newbury
Park.
MILNE, J. C., (1994), Guidelines for Emerging Contractor Development, Development
Bank of South Africa, Pretoria.
MLINGA, R.S. AND N. M. LEMA, (2000), “Informal Contractors in Tanzania-their
Characteristics and Reasons for Informality”, http://www.irbnet.de/daten/iconda/CIB8923.
pdf; accessed: December 2013.
MOSER, C., (1994), “The Informal Sector Debate, Part 1: 1970-1983,” in C. Rakowski
(ed.), Contrapunto: The Informal Sector Debate in Latin America, State University of New
York Press, Albany.
NATIONAL BUREAU OF STATISTICS (NBS), (2007), Analytical Report for Integrated
Labour Force Survey (ILFS) 2006, Dar es Salaam.
PORTES, A., (1978), “The Informal Sector and the World Economy: Notes on the Structure
of Subsidized Labour”, IDS Bulletin, vol. 9 (4): 35-40.
PORTES, A., M. CASTELLS, AND L.A. BENTON, (1989) (eds.), The Informal
Economy: Studies in Advanced and Less Developed Countries, Johns Hopkins University
Press, Baltimore.
PRADHAN, M. AND A. VAN SOEST, (1997), “Household Labor Supply in Urban Areas
of Bolivia”, Review of Economics and Statistics, 79 (2): 300-310.
RENOOY, P.H., (1990), The Informal Economy: Meaning, Measurement and Social
Significance, Netherlands Geographical Studies, Amsterdam.
ROBERTS, B., (1990), “The Informal Sector in Comparative Perspective,” in Perspectives
on the Informal Economy, M.E. Smith (ed.), University Press of America, London.
RODMAN, S., (2007), “God Jobs, Bad Jobs, and Economic Performance: A View from
the Middle East and North Africa Region, Employment and Shared Growth”, in Pace and
Serneels (ed), World Bank, Washington D.C.
SOLOW, R.N., (1957), “Technical Change and the Aggregate Production Function”, The
Review of Economics and Statistics, vol. 39 (3): 312–320.
STOEVSKA, V., (2012), “Employment and Gender Differences in the Informal Economy”,
Department of Statistics Presentation, International Labour Organization, Geneva
TANZI, V., (1982) (ed.), The Underground Economy in the United States and Abroad,
Lexington Books, Massachusetts.

UNITED REPUBLIC OF TANZANIA (URT), (2007), Household Budget Survey, Dar es
Salaam.
WEEKS, J., (1975), “Policies for Expanding Employment in the Informal Sector of
Developing Countries”, International Labour Review, vol. 111 (1): 1-13.
WORLD BANK,(1999), “Kingdom of Morocco: Private Sector Assessment Update”,
Report No. 19975-MOR, Washington D.C.
YAMADA, G., (1996), “Urban Informal Employment and Self-Employment in Developing
Countries: Theory and Evidence”, Economic Development and Cultural Change, vol. 44 (2):
289-314.

The Impact of Bank and Stock Market Developments on Economic Growth
in Zimbabwe: 1988 To 2012
Phineas G. Kadenge and Felex Tafirei1
Abstract
The paper examines the short and long run impact of bank and stock market developments on
growth in Zimbabwe using annual data from 1988-2012, inclusive. The study uses a financially-
augmented production growth function and applies the Auto Regressive Distributed Lag
(ARDL) approach and the error correction mechanism to simultaneously examine the short
and long run relationships. Bank and stock market developments indices are used to measure
developments in the banking sector and the stock market. The results indicate the existence of a
steady long run relationship between growth, bank and stock market developments. Banks are
found to have a greater impact on economic growth than stock markets. Evidence of supply-
leading hypothesis is found. The results suggest that economic growth is better promoted
through a financially-based economic system than a stock market based one.
Key Words: Zimbabwe, financial development, bank developments, economic growth, stock
market developments
JEL Classification Codes: E44, F43, O11, O16, O43
1
Department of Economics, University of Zimbabwe and Bindura University of Science Education, Zimbabwe respectively
* Corresponding Author: P.kadenge@gmail.com

1. INTRODUCTION
Finance-growth proponents like Greenwood and Jovanovic (1990) and King and Levine
(1993a, 1993b) argue that well developed financial and equity markets enhance economic
growth and development through their effects on capital accumulation, technical progress
and productivity. In developing countries where financial markets are repressed, economic
growth is retarded due to deep rooted inefficiencies in credit and equity markets (Shaw, 1973).
Liberalisation of financial and equity markets is thus expected to lead to financial and equity
market developments that will eventually accelerate growth than if the financial and equity
markets were to remain repressed. However, according to Lucas (1988) proponents of finance-
driven growth seem to overstress the role that finance plays in promoting economic growth at
the expense of other more important growth enhancing factors.
Empirical evidence from Zimbabwe as well as from developed and other developing countries
is not conclusive as to the role that banks and stock markets play in promoting economic growth.
Besides, prior studies2 on finance and growth in Zimbabwe do not simultaneously examine
bank and stock market developments and growth in a unified framework yet both banks and
stock markets are channels through which physical capital can be accumulated. By using fewer
financial variables and studying the effects of bank and stock market developments separately
these prior studies might have failed to capture not only some of the developments that have
occurred in the financial market but also the effects that financial market developments might
have had on economic growth in Zimbabwe. This paper attempts to fill this gap by examining
simultaneously the contribution of bank and stock market developments to both the short and
long run economic growth in Zimbabwe.
An Autoregressive Distributed Lag Model (ARDL) was used to estimate the relationship
between growth and bank and stock market developments in Zimbabwe. Annual time series
data for the sample period 1988-2012 was used in the analysis. The main source of data was
the World Bank complemented by the Reserve Bank of Zimbabwe (RBZ) and the Zimbabwe
Statistical Agency (ZimStat).
The rest of the paper is organised as follows: Section 2 gives an overview of Zimbabwe’s
macroeconomic performance and the bank and stock market developments over the period
under study. Section 3 contains an analysis of related theoretical and empirical literature on
finance and growth. Section 4 presents the methodology used in the study. In section 5, the
results are analysed and presented and the final section concludes the study and gives some
policy recommendations.
2. AN OVERVIEW OF ZIMBABWE’S MACROECONOMIC PERFORMANCE,

BANK AND STOCK MARKET DEVELOPMENTS
2.1 ZIMBABWE’S MACROECONOMIC PERFORMANCE (1988-2012)
Zimbabwe’s macroeconomic performance, as measured by real gross domestic product

(GDP) during the period 1988-2012, was not impressive. The average real growth rate over
this period was nearly 0.32% annually. Domestically the performance of the economy was
Ndlovu (2013), Zivengwa et al. (2011) and Jecheche (2010)
2

my was also affected by external factors that included the global oil crisis of the 1980
l recession of the 1990s and the 2008-2012 international financial crisis. The major
argely weak and persistently
negatively negative
affected by droughts, growth
the poor policyfor the ten year
environment, period
unstable political1999-2008
climate, (Re
the hyperinflationary situation and instability in the financial sector. The performance of
of Zimbabwe, 2012).
the economy was Figure 1.1 bybelow
also affected externalshows realincluded
factors that GDP the growth
global rate forofZimbabwe
oil crisis the
2012. 1980s, the global recession of the 1990s and the 2008-2012 international financial crisis. The
major result was largely weak and persistently negative growth for the ten year period 1999-
2008 (Reserve Bank of Zimbabwe, 2012). Figure 1.1 below shows real GDP growth rate for
Zimbabwe from 1988-2012.
Figure 1.1: 1.1:

Figure Zimbabwe: Annual
Zimbabwe: Real
Annual RealGDP
GDPGrowth Rate,1988-2012
Growth Rate, 1988-2012
15%
Period 1 Period 3
(Busts & Booms) Period 2
(Recovery)
10% (The Lost Decade)
5%
0%
-5%
-10%
-15% Annual Real GDP Growth
-20%
88 90 92 94 96 98 00 02 04 06 08 10 12
Source: Authors’ computations based on World Bank Data, 2013
Source:
As shown inAuthors’ computations
Figure1.1 above based onweak
Zimbabwe experienced World Bankgrowth
but positive Data,rates
2013
averaging
3.9% per annum between 1988 and 1998 (period 1 in Figure 1.1). Growth during this period
was largely influenced by drought in 1992, low investment and low industrial output due to
own in Figure1.1 above Zimbabwe experienced weak but positive growth rates aver
foreign currency shortages (Reserve Bank of Zimbabwe, 2010).
per annum between
Between 1999 1988
and 2008andthe1998 (period
Zimbabwean 1 in Figure
economy 1.1). Growth
was characterised during
by a sustained this perio
decline
y influenced
in realby drought
GDP averagingin-6.5%
1992,per low
annuminvestment
reaching a lowand low industrial
of -17.67% output
in 2008 at the due to fo
height of
the hyperinflationary period. This period, now referred to as the ‘lost decade’ (period 2 in
ncy shortages (Reserve Bank of Zimbabwe, 2010).
Figure 1.1) was characterised by four major droughts (Martens, 2012), the banking crisis of
2003-2004 and hyperinflation that reached a peak of 231 million percent by end of June 2008
een 1999 (ZimStat,
and 2008 theHowever,
2012). Zimbabweanfollowing economy
dollarizationwas characterised
in 2009, growth averagedby7.5%
a sustained
in the 4 decl
GDP averaging -6.5%
year period per toannum
from 2009 reaching
2012 (World a low of -17.67% in 2008 at the height o
Bank, 2013).
inflationary period. This period, now referred to as the ‘lost decade’ (period 2 in F
2.2 BANK DEVELOPMENTS IN ZIMBABWE (1988-2012)
was characterised by four major droughts (Martens, 2012), the banking crisis of 2003
yperinflation that reached
At independence, a peak ofinherited
the government 231 million
a largely percent byand
oligopolistic endstagnant
of June 2008 (Zim
banking
sector which was repressed as elsewhere in Africa. However, the financial reforms undertaken
. However, following
in 1991 dollarization
led to significant changes ininthe2009, growth
architecture of theaveraged 7.5%
financial sector. in athe
From 4 year p
sector
2009 to 2012 (World
dominated Bank,
by only 2013).
four foreign banks in 1980 and no indigenous banks, 17 out of a total of 24
banking institutions (71%) were locally owned by December 2012.
Cumulatively, indigenous banks held about US$3 billion (70%) in total deposits while foreign
owned institutions held US$1.3 billion (30%) of total deposits as at end of 2012. Deregulation
of the financial market also led to an increase in total advances to the private sector and a
significant shift in loan allocations among sectors even though they remained largely short
term in nature (Reserve Bank of Zimbabwe, 2013).
BOJE: Botswana Journal of Economics

82 76
In the mid-1990s up to 2008 the financial sector was in a crisis largely caused by the
hyperinflationary situation, the international liquidity crunch and low domestic saving, volatile
deposits and unsound banking practices. The instability in the financial sector culminated in
the banking crisis of 2003-2004 during which thirteen indigenous banks were liquidated to
protect the depositors (Reserve Bank of Zimbabwe, 2012). It is important to note that over the
same period real GDP growth rate averaged -6.5% annually (World Bank, 2013).
However, following dollarization of the economy in early 2009, the hyperinflationary situation
ended and the financial sector stabilised. From 2009 to 2012 the financial services sector was
the fastest growing sector in the economy with an average growth rate of 13% per annum. At
the same time annual GDP growth rate averaged 7.5% between 2009 and 2012 largely due to
renewed confidence in the financial sector, political stability and improved fiscal discipline
(Reserve Bank of Zimbabwe, 2013).
2.3 STOCK MARKET DEVELOPMENTS IN ZIMBABWE (1988-2012)
The Zimbabwe Stock Exchange (ZSE) which was formed in 1896 was opened to foreign
players in May 1993 in line with the Economic Structural Adjustment Programme (ESAP).
This increased trading activity on the exchange, number of counters, market capitalisation
and integration with the world financial markets. As stock market activity increases economic
growth is expected to rise as the stock market helps attract investment necessary for growth
to occur.
Millions of shares worth billions of dollars are traded annually on the ZSE. In 1997, the value
of shares traded on the stock exchange more than doubled to around US$26.5 billion making
the ZSE one of the star emerging market performers. The ZSE’s mean capitalisation for the 25
year period was 78.1% of GDP (World Bank, 2013).
During 1998, the ZSE turnover and value of shares sold declined by 60% and 88%, respectively.
However, market capitalisation continued to increase. The high market capitalisation that
the ZSE continued to experience up to 2008 might have reflected speculative tendencies as
it was not matched by trading. The poor economic performance, rapid deindustrialisation
and inconsistent policy issues especially with regards to indigenisation led to even more
speculative activities on the ZSE. The result was the suspension of trading on the ZSE in late
2008. Trading only resumed in early 2009 following dollarization of the economy. However,
trading has remained very thin.
In light of the bank and stock market developments described above and the general poor
macroeconomic performance of the economy since 1988, one cannot help but ask what effect
these developments have had on economic growth in Zimbabwe. Is there any relationship
between bank and stock market developments and growth or the pattern that seems to be there
is some mere coincidence? Assuming that there is some relationship, does causality run from
growth to financial development or growth follows financial development? Finally which part
of the financial sector promoted growth more than the other, banks or stock markets?

3. LITERATURE REVIEW
Various theories have been put forward as to how governments may increase
economic growth but none have generated greater interest and attracted controversy
in recent years than the role that financial development can play in promoting the
growth and development of a country. The empirical evidence gathered, both in
developed and developing countries, has generated more questions than answers. This
section reviews the theoretical as well as the empirical literature on finance and growth.
In the traditional neoclassical Solow-Swan (1956) growth model the output of an economy
grows in response to larger inputs of physical capital and labour. Non-economic variables such
as human capital and finance are excluded in this model. The model assumes that as capital
stock increases, growth of the economy slows down since capital is subject to diminishing
returns. This implies that, because of diminishing returns, at some point the amount of new
capital produced would just be enough to make up for capital consumption. When this point
is reached the economy will cease to grow or the economy will converge to a constant steady
state rate of growth. In the long run, the economy will only grow if there is technological
progress. The model, however, treats technology as exogenous.
On the other hand, Romer (1986) treats technology as an endogenous growth variable.
According to this theory there are a number of endogenous variables that drive technological
progress and hence, spur economic growth. This is the basis of finance-growth theory.
Financial variables enter the growth equations through their effects on capital accumulation,
technological progress and productivity growth. The financial variables are thus endogenous
to growth (King and Levine, 1993a, 1993b). Growth in this case can be modeled as where Y
is output, A is total factor productivity (TFP), K is capital, L is labour, H is human capital and
FMD is financial market development. The variable A is a function of FMD.
According to Patrick (1966) causality between financial development and economic growth
can either be supply-leading or demand-following. The supply-leading hypothesis postulates
that financial development has a positive impact on growth and causality runs from financial
development to growth while the demand-following hypothesis postulates that causality runs
from growth to financial development.
McKinnon’s Outside-Money Model and Shaw’s Inside-Money Model support Patrick’s

supply-leading argument. McKinnon (1973) argues that financial markets in developing
countries are underdeveloped and external financing is very limited. Entrepreneurs, therefore,
have to rely on self-financing. This requires that potential investors must first accumulate huge
savings to allow them to undertake lumpy investment expenditures. On the other hand, Shaw
(1973) in his Inside-Money Model argues that investors need not necessarily save before they
can invest but can borrow from financial institutions (debt intermediation view). Financial
institutions offer high interest rates to attract savings. As saving increases, the supply of credit
rises allowing financial institutions to promote investment and hence growth through lending
and borrowing.
The demand-following hypothesis postulates that causality in the finance growth-nexus runs
from growth to finance. According to Patrick (1966), as the economy grows and becomes
sophisticated, the demand for financial services to support economic activities also rises. This
view is also supported by Gurley and Shaw (1967), and Beck and Levine (2004), among
others.

King and Levine (1993b) carried out a cross-section analysis of 80 countries for the period
1960-1989. Their results indicate that economic growth was strongly predicted by the level of
financial development. This study is supported by Levine (1997) who concludes that where
countries have large and efficient financial systems, growth occurs at a faster rate than in
countries where financial systems are inefficient. Allen and Ndikumana (2000) also found
financial development and economic growth to be positively correlated in the case of Southern
Africa.
Levine (1991) argues that liquid stock markets reduce liquidity risks making investors more
willing to commit their funds to high return capital investments. Stock markets also allow
investors to diversify their portfolios and thus reduce idiosyncratic productivity risks thus
raising investment as well. This position is supported by Saint-Paul (1992) but Obstfeld (1994)
and Bhidhe (1993) warn about the possible negative effects that stock markets might have of
depressing saving rates and slowing down economic growth.
Rioja and Valev (2011) studied stock markets, banks and the sources of economic growth using
a large cross country panel of low and high income countries over the period 1980-2009. They
found banks to have a sizable positive effect on capital accumulation in low income countries.
Stock markets were found not to have contributed anything towards capital accumulation or
productivity growth in these countries. In high income countries, however, stock markets were
found to play a significant role in affecting both productivity and capital growth with banks
affecting capital accumulation only.
Choe and Moosa (1999) found financial intermediaries more important than capital markets
in promoting growth in Korea for the period 1970-1992. Similarly, Demetriades and Luintel
(2001) using time series data to study bank and stock market developments in five developed
countries, concluded that while developments in both the banking sector and stock markets
foster economic growth, the influence of bank developments was much more pronounced
than stock market developments. On the other hand Guryay et al. (2007) found out that for
Northern Cyprus, financial development does not cause economic growth and causality runs
from economic growth to the development of financial markets.
Kargbo and Adamu (2009) studied financial development and economic growth in Sierra Leone
from 1970-2008. A financial sector development index was used to measure developments in
the sector. The researchers found that real GDP, financial development, investment and real
deposit rate are uniquely co-integrated. Khan et al. (2005) used an ARDL approach to study
the relationship between financial development and economic growth in Pakistan from 1971-
2004. The study found that the relationship between financial depth, real deposit rate and
economic growth was only significant in the long run and not in the short run.
Hondroyiannis et al. (2005) assessed empirically the relationship between bank system
development and the stock market and economic growth in Greece for the period 1986-
1999. They found a bi-directional causality between finance and growth in the long run.
The contribution of stock market finance on growth was found to be substantially smaller
compared to that of bank financing.
Hamdi et al. (2013) examined the causality between financial development, investment and
economic growth in Tunisia for the period 1961-2010. They used a multivariate framework
based on Vector Error Correction Model and cointegration techniques. Their short run

estimates revealed that finance does not lead to economic growth while the long run results
showed the opposite conclusion. They also found out that investment was the main engine for
economic growth both in the short and long run.
Zivengwa et al. (2011) studied stock market development in Zimbabwe for the period 1980-
2008 and found a uni-directional relationship running from stock market development to
economic growth. Jecheche (2010) found similar results when he looked at the banking sector
while Ndlovu (2013) found evidence of demand-following in Zimbabwe. These results are
conflicting indicating the need for further research in the area.
The reviewed theoretical literature is robust in explaining the relationship between financial
development and economic growth. However, a distinct consideration given to low income
countries on one hand and high income countries on the other seems to suggest that the
theoretical literature is overgeneralised in relation to the effect of financial development on
growth across countries. As reviewed in the empirical literature, there is some evidence of
financial development positively affecting economic growth (King and Levine, 1993b, Allen
and Ndikumana, 2000). Contrary to these findings other studies like Guryay et al. (2007)
found that financial development does not cause economic growth. The empirical literature
suggests that stock markets play an insignificant role in promoting capital accumulation and
productivity in low income countries. Further to this, banks have been found to contribute
more to economic growth than stock markets. Heterogeneity in the empirical evidence may
be a result of different time periods, estimation methods, measures of financial and stock
market developments and growth considered as well as the type of study (cross sectional, time
series or panel time series). This study attempts to provide more evidence of the bank and
stock market developments on economic growth relationship in Zimbabwe by considering
a different time period, estimating method and using bank and stock market development
indices simultaneously.
4. METHODOLOGY AND DATA ISSUES
Several econometric methodologies have been used by researchers to test the relationship
between financial development and economic growth. The Ordinary Least Squares (OLS)
method has been the most widely used method although it has been found to have some
limitations when studying long run relationships between economic variables. Other
methodologies like the Johansen technique require that the economic variables under study be
of the same order of integration. The ARDL technique, on the other hand, can be conducted
even if the series are integrated of order zero (I (0)), integrated of order one (I (1)) or mutually
integrated (Pesaran and Pesaran, 1997). However, the ARDL technique collapses if any variable
is integrated of an order higher than 1.The approach has also been found to yield better results
in small samples than other cointegration techniques (Narayan, 2004). The ARDL approach
also eliminates the problem of serial correlation and endogeneity if appropriate lags are used
(Jalil and Ma, 2008). The ARDL method is also able to estimate simultaneously the long run
and short run parameters of the model. Simultaneously estimating the long run and short run
relationships also corrects for the problem of variable omission and autocorrelation (Khan et
al. 2005).
The use of the neoclassical production function to model growth relationships is common in
economics having been used by Kohpaiboon (2003), Herzer et al. (2006) and Sultan (2012)

also corrects for the problem of variable omission and autocorrelation (Khan et al. 2005).
The use of the neoclassical production function to model growth relationships is common in
economics having been used by Kohpaiboon (2003), Herzer et al. (2006) and Sultan (2012) and
and many
many otherother researchers.
researchers. Following Following
SultanSultan(2012) (2012) the neoclassical
the neoclassical aggregate aggregate
production production
function
function
can be expressed can be expressed as: as:
GDP! = f A ! , K ! , L! ........................................................................................................
(1) (1)
𝑤𝑤ℎ𝑒𝑒𝑒𝑒𝑒𝑒 GDPt is real GDP in time period t, A, K and L are as defined before.
Financial development is expected to contribute to output growth through its effects on TFP.
Thus in this case TFP is a function of developments in the banking sector captured by the bank
sector development index (BDI) and also a function of developments in the stock market sector
captured by the stock market development index (SMDI) respectively. This can be expressed as:
𝐴𝐴! = f BDI! , SMDI! ...........................................................................................................(2) (2)
Substituting equation (2) into equation (1) we have:
GDP! = f(BDI! , SMDI! , K ! , L! ) .................................................................................................
(3) (3)
In its intensity form equation 3 can be written as:
gdp! = f(bdi! , smdi! , k ! ) .............................................................................................
(4) (4)
!"#! !"#! !"#! !!
where gdp! = ! , bdi! = ! , smdi! = ! and k ! = ! respectively.
! ! ! !
Assuming a Cobb-Douglas production function, equation 4 can be explicitly expressed as

β β β
(5)
gdpt = bdit 1 smdit 2 kt 3 ..................................................................................................................... (5)
which can be log transformed into an explicit estimable function of the form
(6)
ln gdpt = β 0 + β1 ln bdit + β 2 ln smdit + β3 ln kt + ε t ....................................................................... (6)
where β! is a constant and β! , β! and β! are the elasticity coefficients of real GDP per capita with
respect to bank and stock market developments and the stock of capital respectively and 𝜀𝜀! is
the error term.
The choice of the ARDL model was based on the assumption that there was a unique relationship
between bank, stock market developments and economic growth. Following Pesaran et al. (2001)
the conditional ARDL-error correction model is thus specified as:
p q r
Δ ln( gdp) t = δ 0 + ∑ α1Δ ln( gdp) t −i + ∑ α 2 Δ(bdi87 ) t −i + ∑ α 3 Δ( smdi) t −i .......................................
(7) (7)
i =1 i =0 i =0
where the other
p
variables are as defined
+ ∑ α 4 Δ(ln k ) t −i + ϕ1 ln gdpt −1 + ϕ 2bdit −1 + ϕ 3 smdit −1 + ϕ 4 ln kt −1 + ε t
i =0
before; ∆ is the
difference operator, δ! is
the drift and 𝜑𝜑! , 𝜑𝜑! , 𝜑𝜑! 𝑎𝑎𝑎𝑎𝑎𝑎 𝜑𝜑! are the long run multipliers respectively. The coefficients 𝛼𝛼!
measure the short run effects. The long run coefficients are expected to have a positive sign
each.The coefficients show how strongly the regressors influence economic growth respectively.
High coefficients signal an important influence that the regressors have on the regressand.
The ARDL estimation involves testing the null hypothesis of no cointegration relationship
against the alternative hypothesis of the existence of a cointegration relationship. The null and
alternate hypotheses were therefore stated as:
𝐻𝐻! : 𝜑𝜑! = 𝜑𝜑! = 𝜑𝜑! = 𝜑𝜑! = 0 ..........................................................................................................
(8) (8)
𝐻𝐻! : 𝜑𝜑! ≠ 𝜑𝜑! ≠ 𝜑𝜑! ≠ 𝜑𝜑! ≠ 0 ..........................................................................................................
(9) (9)
The F test is used to establish whether a cointegration relationship among variables exists or not.
After calculating the F test, it is compared with the critical values tabulated by Narayan (2004)
for small samples. The lower critical bound assumes that all variables are I (0) and upper bound
assumes that all the variables are I (1). If F-test statistic calculated is greater than the upper
bound critical value then we reject the null hypothesis of whether variables are I (0) or I (1). If
F-test statistic calculated is less than the lower bound critical value, at the specific level of
significance, the null hypothesis cannot be rejected. If F-test falls between the upper and lower
81
bounds the result is inconclusive and the decision is made based on the
BOJE: significance
Botswana Journal ofof the error
Economics
correction term.
Finally the short run relationships are captured through error correction. The error correction
After calculating the F test, it is compared with the critical values tabulated by Narayan (2004)
for small samples. The lower critical bound assumes that all variables are I (0) and upper bound
assumes that all the variables are I (1). If F-test statistic calculated is greater than the upper
bound critical value then we reject the null hypothesis of whether variables are I (0) or I (1). If
F-test statistic calculated is less than the lower bound critical value, at the specific level of
significance, the null hypothesis cannot be rejected. If F-test falls between the upper and lower
bounds the result is inconclusive and the decision is made based on the significance of the error
correction term.
Finally the short run relationships are captured through error correction. The error correction
representation of the series was specified as follows:
p q r
Δ ln( gdp) t = δ 0 + ∑ α1Δ ln( gdp) t −i + ∑ α 2 Δ(bdi) t −i + ∑ α 3 Δ( smdi) t −i
i =1 i =0 i =0
p
+ ∑ α 4 Δ(ln k ) t −i + ξecmt −1 + ε t ....................................................................................................... (10)
i =0 (10)
where ξ is the speed of adjustment parameter and ecm is the error correction term. The
coefficient of the lagged error correction term (ξ) is expected to be negative and statistically
significant to further confirm the existence of a cointegration relationship.
Annual time series data over the sample period 1988-2012 was used to estimate the relationship.
Economic growth is the dependent variable proxied by annual real GDP per capita. The choice of
this proxy is consistent with other empirical studies. Domestic credit provided by the banking
sector
sector was
sector wasrefersused
used toconstruct
to to
net construct thebank
credit advances
the bank to development
all economic
development index
sectors
index (BDI).
(BDI). Domestic
on Domestic
a gross basis credit
credit provided
(excluding
provided credit
by by the
the
banking
to the sector
central refers to
government). net credit
Following advances
banking sector refers to net credit advances to 88 Naceur to all
and economic
Ghazouani sectors
(2007) on
anda gross
Levine
all economic sectors on a gross basis (excludingbasis
and (excluding
Zervos
credit
(1996)
credit
toanthe
to the central
annual
centralBDI government).
was constructed
government). Following Naceurand
by subtracting
Following Naceur and
the Ghazouani
domestic
Ghazouani credit(2007)
(2007) andand
sample mean Levine
Levine and and
from
Zervos
Zervos
the annual (1996)
(1996) an
an of
value annual
annual BDI was
BDI credit
domestic constructed
was constructed
and dividing by
by by subtracting
subtracting
the sample the
themean. domestic
domestic
The creditcredit
index sample sample
mean
takes integer mean
from
from the
valuesthewith annual
annual value
value
positive and of domestic
ofhigher
domesticvalues credit
credit and dividing
and dividing
indicating positive by the
bydevelopments sample
the sample mean. mean.
in theThe The index
takestakes
indexsector.
banking
integer
integer
The study values
values with positive
with apositive
expects positiveand and higher
higher values
relationship values indicating
indicating
between BDI and positive
positive
growth. developments in
developments in the bankingthe banking
sector. The study expects a positive relationship between
sector. The study expects a positive relationship between BDI and growth. BDI and growth.
Market capitalisation
Market capitalisation andand stock
stocktraded
tradedvaluevalueasasratios
ratiosofofGDP GDPwere wereusedusedtotocome
comeupupwith witha stock
a
Market
stock capitalisation
market and stock
development traded
index. A value asindex
simple ratiosofofstock
GDP market
were used to come up(SMDI)
development with a stock
was
market development
market development index. index. AA simplesimpleindex indexofofstock stockmarketmarketdevelopment
development (SMDI)(SMDI) was
constructed by
constructed by subtracting
subtractingthe the mean
themean
meanof ofofeach series
eachseries
seriesfrom
from each
each series value
series and dividing
value bywasthe
constructed by subtracting each from each series value andand dividing
dividing by theby the
mean series.
mean series. The final
finalindex
The final indexwaswasobtained
obtainedby averaging thethemeans-removed values of the
of two
mean series. The index was obtained bybyaveraging
averaging means-removed
the means-removed values
values of the the
two two
stock market
stock market series.
market series. The
series. The index
Theindex takes
indextakes integer
takesinteger values
integervalues with
valueswithwith large
largeand positive values indicating
stock large andand positive
positive values
values indicating
indicating
better stock
better stock market development.
marketdevelopment.
development.A AApositive
positiverelationship
relationship between
between SMDISMDI and growth waswasthus
better stock market positive relationship between SMDI andand growth
growth was thusthus
expected.
expected. The The functional
Thefunctional computation
functionalcomputation
computationofofBDI of BDI
BDI and
and SMDI
SMDI is as
is as follows:
follows:
expected. and SMDI is as follows:
BDI t
(11)(11)
BDI == ((xxt −−xx))//xx ......................................................................................................................... (11)
1 ⎡y − y z − z⎤
SMDI == 1 ⎡⎢yt t − y ++zt t− z ⎤ ⎥ .........................................................................................................
SMDI (12)(12)
(12)
22⎢⎣⎣ yy zz ⎥⎦ ⎦
where xxtisis the
where theannual
annualvalue
valueofofdomestic
domesticcredit, x is
credit,x is thethe mean
mean valuevalue of the
of the domestic
domestic credit
credit for the
for the
t
sample period,
sample period, yytisis the
t
theannual
annualratio
ratioofofmarket
marketcapitalisation
capitalisation to to
GDP,GDP, y isy the
is the
meanmean
ratioratio
of of
market capitalisation
market capitalisationvaluevaluetotoGDP
GDPfor forthethesample
sample period,
period, is the
zt iszt the annual
annual ratio
ratio of stock
of stock traded
traded
value to
value to GDP
GDP and and zzisisthe
themean
meanratio
ratio ofofstock
stocktraded
traded value
value to toGDPGDP for for
the the sample
sample period.
period.
Given equations
Given equations (11)(11)and
and(12),
(12),log
logtransformation
transformation was
was done
done afterafter adding
adding a constant
a constant of ρof and and
ρ
λ to BDI and SMDI respectively such that the log transformed variables were restricted to be
λ to BDI and SMDI respectively such that the log transformed variables were restricted to be
greater or equal to zero.
greater or equal to zero.
Gross capital formation (as a % of GDP) was used to measure the increase in capital stock and is
Gross capital formation (as a % of GDP) was used to measure the increase in capital stock and is
one of the major determinants of economic growth according to Keynesian growth theory. It was
one of the major determinants of economic growth according to Keynesian growth theory. It was
used as a control variable. The study expects a positive relationship between investment and
used as Botswana
BOJE: a controlJournal
variable. The study expects a positive relationship between investment and
of Economics 82
economic growth.
economic growth.
Data was mainly obtained from the World Bank and complemented by data from the Reserve
Gross capital formation (as a % of GDP) was used to measure the increase in capital stock and
is one of the major determinants of economic growth according to Keynesian growth theory.
It was used as a control variable. The study expects a positive relationship between investment
and economic growth.
Bank of Zimbabwe and ZimStat. The lack of monthly or quarterly data led to the use of
annual data from 1988 as this was the year in which the World Bank started measuring stock
exchange data for Zimbabwe.
5. PRESENTATION AND ANALYSIS OF RESULTS
Table 1 below shows the descriptive statistics of the time series data used. There is a wide
variation in means of the variables implying that any attempt to carry out regression estimates
in levels will produce biased results. The low values for bank development and stock market
development indices show that there has not been any meaningful development in these
sectors over the sample period. The Jacque-Bera statistic shows that the variables are normally
distributed.
Table 1: Descriptive Statistics

gdp k smdi bdi
Mean 570.8801 15.00015 8.95E-11 -3.28E-08
Median 620.7075 17.37694 -0.530711 -0.060314
Maximum 718.4184 25.61923 3.034266 2.230333
Minimum 344.7421 1.525176 -0.924099 -0.951317
Std. Dev. 127.8618 7.578565 1.101837 0.632196
Skewness -0.452538 -0.410819 1.523996 1.586284
Kurtosis 1.589236 1.919245 4.560658 7.431023
Jarque-Bera 2.926477 1.919918 12.21449 30.93661

Probability 0.231485 0.382908 0.002227 0.000000
Sum 14272.00 375.0038 2.24E-09 -8.20E-07

Sum Sq. Dev. 392367.6 1378.431 29.13706 9.592113
Observations 25 25 25 25

Source: Authors’ computations using based on based on World Bank Data (2013)

Table 2 below shows the estimated correlation matrix.
Table 2: Correlation Matrix
Variable gdp k smdi bdi
gdp 1.0000
k 0.2934 1.0000
smdi -0.4180 -0.4059 1.0000
bdi 0.4125 -0.2651 0.4248 1.0000
Source: Authors’ computations using Eviews 7
The results in Table 2 above show that capital stock and bank development index are positively
related to real GDP per capita whilst stock market development index and real GDP per capita
are negatively related. There is also a negative correlation between bank and stock market
development indices and capital stock.
Table 3 below reports the ADF and the PP unit root test results.
Table 3 ADF and PP Unit Root Test Results
Variable Augmented Dickey Fuller (ADF) Test Phillips–Peron (PP) Test
Log Level First Order of Log Level First Order of

Difference Integration Difference Integration
lngdp -1.0798 -3.4088*** I(1) -0.9654 -3.4088*** I(1)
lnk -0.3508 -4.7754*** I(1) -0.3509 -4.7775*** I(1)
lnbdi -3.0068*** I(0) -2.9763*** I(0)
lnsmdi -3.0047*** I(0) -2.9015*** I(0)
Source: Authors’ computations using Eviews 7
Notes: The ADF and PP tests are based on McKinnon critical values and the lag length on
SBC criterion. *** shows rejection of the null hypothesis at 1%.
As shown in Table 3 above both the ADF and the PP unit root test results show that bank and
stock market development indices were stationary in levels and therefore I (0). On the other
hand, real GDP per capita and capital stock became stationary when first differenced implying
that they were I(1). Since the order of integration was mixed the ARDL model became the
most appropriate estimating technique.
Having made sure that no series was integrated of an order higher than one the ARDL model
for the Zimbabwe growth function was estimated and the results are reported in Table 4.

Table 4: Autoregressive Distributed Lag Estimates
ARDL (1,0,2,2) selected based on Schwarz Bayesian Criterion
Dependent variable is lngdp, 23 observations used for estimation from 1990 to 2012
Regressor Coefficient Standard Error T-Ratio Prob

lngdp(-1) 0.6331 0.1076 5.8842 0.000
lnk 0.0614 0.0209 2.9331 0.011
lnsmdi 0.0022 0.0221 0.1001 0.922
lnsmdi(-1) -0.0346 0.0170 -2.0408 0.061
lnsmdi(-2) -0.0563 0.0171 -3.2865 0.005
lnbdi 0.0246 0.0349 0.7066 0.491
lnbdi(-1) 0.0179 0.0280 0.6402 0.532
lnbdi(-2) 0.0697 0.0325 2.1454 0.050
c 2.1466 0.6707 3.2004 0.006
R-Squared 0.9767 R-Bar-Squared 0.9633
S.E. of Regression 0.0477 F-Stat. F(8,14) 73.2646[0.000]
Mean of Dependent Variable 6.3072 S.D. of Dependent Variable 0.2492
Residual Sum of Squares 0.0319 Equation Log-likelihood 43.0511
Akaike Info. Criterion 34.0511 Schwarz Bayesian Criterion 28.9414
DW-statistic 1.6972 Durbin’s h-statistic 0.8477 [0.397]
Source: Authors’ computations using Microfit 5.0
As reported in Table 4, with an adjusted R2 of 0.96 the model is of good fit as 96% of the
variations in real GDP per capita are explained by the independent variables. The DW-
statistic is also around 2 and the F-statistic is significant showing that some coefficients for
explanatory variables in the model are different from zero. The diagnostic tests are reported
in Table 5 below. The results show that the model passed all the tests meaning that there was
no problem of serial correlation, the model was correctly specified, the disturbances were
normally distributed and that there was no problem of heteroscedasticity.
Table 5: Diagnostic Tests

Test Statistics LM Version F Version
A:Serial Correlation CHSQ(1) = 0.6637 [0.415] F(1,13) = 0.3863 [0.545]
B:Functional Form CHSQ(1) = 1.0569 [0.304] F(1,13) = 0.6262 [0.443]
C:Normality CHSQ(2) = 0.5943 [0.743] Not applicable
D:Heteroscedasticity CHSQ(1) = 2.2183 [0.136] F(1,21) = 2.2416 [0.149]

Notes: A: Lagrange multiplier test of residual serial correlation
B: Ramsey’s RESET test using the square of the fitted values
C: Based on a test of skewness and kurtosis of residuals
D: Based on the regression of squared residuals on squared fitted values

In Table 6 are the bounds test results.
Table 6: Bounds Test Results
Critical Value Bounds Dependent variable Estimated Outcome
F-statistic
I(0) I(1) Flngdp( lngdp| lnk, lnbdi, lnsmdi) 9.8671*** Cointegration
1% 4.614 5.966 Flnsmdi ( lnsmdi| lngdp, lnk, lnbdi) No lagged-error No Cointegration

correction term1
5% 3.272 4.306 Flnbdi( lnbdi| lngdp, lnk, lnsmdi) No lagged-error No Cointegration
correction term1
10% 2.676 3.586 Flnk( lnk| lngdp, lnbdi,lnsmdi) 10.1898*** Cointegration
Notes: critical values are for the model with intercept but no trend with k=3 regressors. Critical
values3 were extracted from Narayan (2004) Appendix A1, A2 and A3.
As shown in the table above when real GDP per capita is treated as the regressand (Flngdp(lngdp|
lnk, bdi, smdi), the F calculated statistic of 9.8671 was greater than the upper critical bound
value of 5.966 at 1% level of significance. The conclusion is therefore that there was a
long run relationship between real GDP per capita, capital stock and bank and stock market
development indices. The results also show that capital is cointegrated with economic growth,
bank and stock market development. This study therefore found no unique cointegration
relationship between real GDP per capita and capital, bank and stock market development.
In Table 7 are the results of the estimated long run coefficients.

Table 7: Estimated Long Run Coefficients using the ARDL Approach
ARDL(1, 0, 2,2) selected based on Schwarz Bayesian Criterion
Dependent variable is lngdp, 23 observations used for estimation from 1990 to 2012
Regressors Coefficient Standard Error T-Ratio Prob

lnk 0.1673 0.0655 2.5551 0.023**
lnsmdi -0.2419 0.0491 -4.9237 0.000***
lnbdi 0.3056 0.0769 3.9812 0.001***
c 5.8509 0.1781 32.8540 0.000***
Source: Authors’ computations using Microfit 5.0. Notes: ** and *** indicates the rejection of the null hypothesis at 5% and 1%
level of significance respectively
The estimated long run results in Table 7 above show a positive and a statistically significant
relationship between capital stock to GDP ratio and real GDP per capita at 5% level of
significance. This means that real GDP per capita in Zimbabwe increases by about 0.17%
following a 1% increase in capital stock to GDP ratio. This result was consistent with the
theoretical model and is supported by Kargbo and Adam (2009) and Jecheche (2010).
Critical values are for 30 observations since no critical values are reported for 25 observations.
3

The results also show a positive and highly significant relationship between bank development
index and real GDP per capita with a 1% increase in the bank development index leading to a
0.31 % increase in real GDP per capita. The coefficient of the bank development index is the
highest showing that the banking sector is the major driver of growth in Zimbabwe. This result
was consistent with the theoretical model and is supported by Rioja and Valev (2011) who
found that the financial sector had a substantial role to play in promoting economic growth in
low income countries between 1980-2009.
An inverse relationship between stock market development index and real GDP per capita
was found with a 1% increase in the stock market development index leading to a decline
in real GDP per capita of about 0.24% per annum. This result was not consistent with the
theoretical model. However, the result is supported by Rioja and Valev (2011) who found
stock markets not to have contributed anything towards capital accumulation or productivity
growth in low income countries for the period 1980-2009. Obstefeld (1994) and Bhide (1993)
also warn about the possible negative effects that developments in the stock market might
have on economic growth.
This weak association between the stock market and economic growth may be reflective of
the strict exchange control regulations imposed to prevent possible capital outflows following
the introduction of the Fast Track Land Reform Programme. The Indigenisation and Economic
Empowerment Policy also had widespread restrictions with respect to ownership of capital
resources (Indigenisation and Economic Empowerment Act, 2007). The hyperinflationary
situation also led to negative real return on capital investments. These developments led to
very high market capitalisation to GDP ratio not supported by trading. Thus investment on
ZSE became speculative in nature hence suspension of trading in late 2008.
The short run dynamics were captured through error correction. The error correction results
are reported in Table 8.

Table 8: Error Correction Representation for the Selected ARDL Model
ARDL(1,0,2,2) selected based on Schwarz Bayesian Criterion
Dependent variable is dlngdp, 23 observations used for estimation from 1990 to 2012
Regressors Coefficient Standard T-Ratio Prob
Error
dlnk 0.0614 0.0209 2.9331 0.010***
dlnsmdi 0.0022 0.0221 0.10007 0.922
dlnsmdi1 0.0563 0.0171 3.2865 0.005***
dlnbdi 0.0246 0.0349 0.70663 0.490
dlnbdi1 -0.0697 0.0325 -2.1454 0.048**
ecm(-1) -0.3669 0.1076 -3.4098 0.004***
ecm = lngdp -.16733*lnk + .24187*lnsmdi -.30598*lnbdi -5.8509*c
R-Squared 0.7589 R-Bar-Squared 0.6211
S.E. of Regression 0.0477F-Stat. F(6,16) 7.3428[0.001]
Mean of Dependent Variable -0.0180 S.D. of Dependent Variable 0.0775
Residual Sum of Squares 0.0319 Equation Log-likelihood 43.0511
Akaike Info. Criterion 34.0511 Schwarz Bayesian Criterion 28.9414
DW-statistic 1.6972
Notes: *, ** and *** show significance at 10%, 5% and 1% level of significance respectively.
In [ ] are p-values.
The error correction term (-0.37) was negative and statistically significant at 1% level. This
also confirmed the existence of a long run relationship between real GDP per capita and
its regressors. In this case approximately 37% of the disequilibria caused by shocks in the
previous year got corrected in the current year. This was a moderate speed of adjustment. The
high R2 and the F statistic which were significant at 1% show that the short run model was of
good fit.
Results in Table 8 also show that capital’s contribution to real GDP per capita was still positive
as it was in the long run. However, its contribution to short run real GDP per capita was now
less but significant at 1% level up from 5% level in the long run. Bank development index
which was positive and significant at 1% in the long run has its two year lag negative and
significant in the short run. This means that bank developments contributed more to growth in
the long run than in the short run. This result is supported by Hamdi et al. (2013) who found
that for the period 1961-2010 finance did not lead to growth in Tunisia in the short run but
in the long run. Khan et al. (2005) also found the relationship between finance and growth
significant in the long run and not the short run in the case of Pakistan from 1971-2004.
The short run negative contribution of banks to growth may have been largely due to the
hyperinflation which wiped out domestic savings leading to widespread disintermediation as
people moved to the parallel market for financial services. However stock market development

gnificant at 1%. In short, banks contributed more to growth in the long run than stock markets
hile stock markets contributed more in the short run than banks. These results are supported by
ondroyiannis et al. (2005).
index which was negative and significant at 1% in the long run had its second lag positive
e ECM andmodel was tested for stability using the CUSUM and CUSUMSQ techniques. Since
significant at 1%. In short, banks contributed more to growth in the long run than stock
significant
e coefficients at 1%.
ofwhile
the In short,model
estimated banks fell
contributed
within more to growth in the
at long run than stock markets
markets stock markets contributed more the
in thecritical bounds
short run than banks.5%These
level of significance
results are
while
s meant thatstock markets
the model
supported wascontributed al.more
stableetover
by Hondroyiannis theinsample
(2005). the short run than
period. Thisbanks. Theseinresults
is shown figureare 1.2.supported by
Hondroyiannis et al. (2005).
Figure
The ECM1.2: Stability
model wasTest;testedCUSUM and
for stability CUSUMSQ
using the CUSUMPlots and CUSUMSQ techniques.
The ECM
Since themodel was
coefficients tested
of the for stability
estimated
Plot of Cumulative Sum of Recursive Residuals
using
model fell the CUSUM
within the and CUSUMSQ
critical bounds at 5%techniques.
level of Since
20
thesignificance
coefficientsthisof meant
the estimated modelwas
that the model fellstable
within thethe
Plot of over critical
Cumulative ofbounds
sample
Sum ofat
period.
Squares 5% islevel
This
Recursive shownof significance
Residuals in
thisfigure
meant that
1.2. the model was stable over the sample period. This is shown
1.5 in figure 1.2.
10
Figure
Figure 1.2: Stability
1.2: Stability Test; CUSUM
Test; CUSUM and CUSUMSQ
and CUSUMSQ
1.0
Plots Plots
0
Plot of Cumulative Sum of Recursive Residuals 0.5
Plot of Cumulative Sum of Squares of Recursive Residuals
20
-10
0.0 1.5
10
-20 1.0
-0.5
1990 1996 2002 2008 2012 1990 1996 2002 2008 2012
The straight lines represent critical bounds at 5% significance level The straight lines represent critical bounds at 5% significance level
0
0.5
-10
urce: Authors’ computations using Microfit 5.0 0.0
-20
-0.5
confirm the goodness of fit of the model Figure 1.3 shows a residual graph for the actual and
1990 1996 2002 2008 2012
The straight lines represent critical bounds at 5% significance level
1990 1996 2002 2008 2012
The straight lines represent critical bounds at 5% significance level
ted observations.
Source:
gure 1.3: Plot
To Authors’
of Actual
confirm computations
and Fitted
the goodness ofusing
of fitValues Microfit
of lngdp
the model 5.01.3 shows a residual graph for the actual
Figure
and fitted observations.
To confirm the goodness of fit of the model Figure 1.3 shows a residual graph for the actual and
fitted observations.
Figure of Actual and Fitted Values of lngdp
1.3: Plot
Figure 1.3: Plot of Actual and Fitted Values of lngdp
6.4

6.2
6.0 6.4
5.8 6.2
5.6 6.0
1990 1996 2002 2008 2012
5.8 lngdp Fitted

5.6
1990 1996 2002 2008 2012
The residuals graph shows that the fitted observations are very close to the actual observed
values and they move together. lngdp Fitted
urce: Authors’ computations using Microfit 5.0
89 96 BOJE: Botswana Journal of Economics

Weak erogeneity was tested using Granger causality test and the results are reported in Table 9.
Table 9: Granger Causality Test Results

Short run (Weak Exogeneity) Test
Dependent Variable
d(lnk) d(lngdp) d(lnsmdi) d(lnbdi)
d(lnk) - 0.8884 0.35828 0.7040
(0.4296) (0.7040) (0.7026)
d(lngdp) 0.4000 - 1.23393 1.50206
(0.6765) (0.3159) (0.2508)
d(lnsmdi) 5.81615 4.3736 - 0.3485
(0.0119)*** (0.7107)
(0.0294)**
d(lnbdi) 2.88039 0.13890 0.3768 -
(0.0837)* (0.8713) (0.6916)
The results show that capital, BDI and SMDI are weakly exogenous. In the long run, as shown
in Table 10, causality is unidirectional and runs from capital, and bank and stock market
developments to economic growth lending support to the supply leading hypothesis also
found by Zivengwa et al. (2011). The result is in contrast to the study by Guryay et al. (2007)
who found that in Northern Cyprus causality runs from economic growth to the development
of financial markets.
Table 10: Pairwise Granger Causality Tests

Null Hypothesis: Obs F-Statistic Prob.
lnk does not Granger Cause lngdp 23 4.53604 0.0254
lngdp does not Granger Cause lnk 0.55900 0.5814
lndc does not Granger Cause lngdp 23 5.89526 0.0107
lngdp does not Granger Cause lndc 1.67093 0.2160
lnsmdi does not Granger Cause lngdp 23 2.69544 0.0946
lngdp does not Granger Cause lnsmdi 0.81868 0.4568
Source: Authors’ computations using Eviews 7.0
SUMMARY OF RESULTS
The results show that bank developments contributed more to economic growth than stock
markets in the long run. However, in the short run banks are found to have a negative impact
on growth while stock markets have a positive impact. These results are supported by Choe
and Moosa (1999), Demetriades and Luintel (2000) and Hondroyiannis et al. (2005) who
found out that the influence of financial development on economic growth was much more
pronounced than that of stock market developments. Since the lagged error term is negative
and significant it implies that there exists a long run relationship between real GDP per capita
and capital, bank and stock market development.

In Zimbabwe, strict controls on capital movements, speculative investment due to the
hyperinflationary situation, low incomes and lack of knowledge by the population on how
stock markets operate may have resulted in low investments through the stock markets.
This leaves banks as the only organised financial markets through which financial resources
from the resource poor communities are raised as banks run deposit accounts for as low as
$1. Moreover, banks have a wider infrastructure across the whole country and thus have a
very wide base from which they can mobilise funds for investment purposes. However, the
contribution by banks to growth was heavily limited by widespread disintermediation that
occurred during the hyperinflationary period which substantially reduced domestic savings
and also the collapse of indigenous banks as people lost confidence in the financial system. In
the long run as the negative effects are corrected, banks were found to contribute positively to
economic growth.
In Zimbabwe, low incomes and lack of knowledge by the population on how stock markets
operate may result in low investments through the stock markets. This leaves banks as the
only organised financial markets through which financial resources from the resource poor
communities are raised as banks run deposit accounts for as low as $1. Moreover, banks have
a wider infrastructure across the whole country and thus have a very wide base from which
they can mobilise funds for investment purposes.
6. CONCLUSIONS AND POLICY RECOMMENDATIONS
There is a steady long run relationship between bank and stock market developments and
growth in Zimbabwe. The positive developments in the banking sector significantly influence
economic growth in the long run despite a negative and statistically significant second lag
impact in the short run. The impact of stock market development on growth has been found
to be negative and statistically significant in the long run whilst it is significantly positive in
the short run. The study further concludes that there is a unidirectional impact running from
capital, bank and stock market developments to economic growth. Capital stock was also
found to have a positive and significant contribution to both long run and short run economic
growth with weak and relatively strong linkages to bank and stock market developments.
The major policy implications are that there is need for both short and long run economic
policies that will promote the well-functioning of the financial system and equity markets so
that more financial resources are mobilised to boost economic growth in the country.
The positive impact of bank developments on economic growth is evidence of supply-leading
hypothesis and an indication of the important role that financial institutions can play in the
economy. There is therefore, need to create modern financial institutions and improve access
to credit, financial intermediation and inclusion.
Zimbabwe has a large unbanked informal economy as well as rural sector. More than US$3
billion dollars was said to be circulating outside the formal financial system (Reserve Bank
of Zimbabwe, 2013).Thus it is critical for government to formulate policies that will help
the financial system harness these huge resources and channel them towards productive use.
Government must restore confidence in the financial sector that was lost during the decade
that followed the introduction of the Fast Track Land Reform Programme; a number of banks

failed and those that remained experienced difficulties in meeting their depositors’ cash
requirements. The government must capitalise the central bank so that it plays its role as
the lender of last resort and strengthen corporate governance measures to help prevent bank
failures and reduce the incidence of non-performing loans.
The negative contribution of the stock market to economic growth may have been due
to deindustrialization precipitated by the Fast Track Land Reform Programme, the
hyperinflationary environment and the Indigenisation and Economic Empowerment Policy.
This led firms to operate with reduced capacity in the face of weakening domestic demand and
increased competition from imports.
The negative contribution of the stock market to economic growth may have been due to
deindustrialisation with firms operating with reduced capacity in the face of weakening
domestic demand and increased competition from imports. General lack of investor confidence
and speculation may also impact negatively on economic growth. If the stock market is an
important channel through which long term capital can be raised as proposed in the literature,
the government must relax the listing rules on the local bourse and promote fair, efficient and
effective trading. The regulatory environment should also be enabling but stringent enough
to curb speculation. Sound macroeconomic policies will help attract investors to the local
bourse and a stable environment will see industrial capacity going up. Dual listing must also
be promoted to increase integration with the world financial market. Currently there is only
one stock exchange. There is thus need to open up an exchange for the small investor and other
secondary exchanges like the market for derivatives.
The impact of capital investment was found to be positive and significant both in the short and
long run. This calls for policies that will further increase productive investment and attract
foreign investors in the country as this would greatly increase economic growth.
More importantly banks were found to positively contribute more to economic growth in the
long run than stock markets. The main policy implication is that growth is better promoted
through a financially-based system rather than through stock exchange promotion. This
however does not mean that government must turn a blind eye on the stock exchange as the
negative effects must be contained. The negative contribution is only an indication of the
need to put in place measures that will force companies to seek expansion through organic
growth as this will promote economic development. Widespread mergers and takeovers in
industry do not promote growth. No doubt there is need to reduce share turnover, improve
corporate control and governance and the regulatory environment for a more organised trading
environment on the stock exchange.
Given a choice, the results seem to suggest that a bank-based system for Zimbabwe is the first
channel of choice for the promotion of economic growth. However, with the rapid expansion
of stock exchanges in developing countries, one would expect them to play a more leading
role in promoting economic growth as they grow and mature. Finally a stable enabling
macroeconomic environment is critical for finance-led growth in Zimbabwe.

REFERENCES
Allen, D. S. and L. Ndikumana, 2000. Financial Intermediation and Economic Growth in

Southern Africa. Journal of African Economies,Vol. 9 (2):132-160.
Beck, T. and R. Levine, 2004. Stock Markets, Banks, and Growth: Panel Evidence. Journal
of Banking and Finance,Vol. 28(3): 423-442.
Bhide, A. 1993.The Hidden Costs of Stock Market Liquidity. Journal of Financial
Economics, Vol. 34: 31-51.
Choe, C. and I. A. Moosa, 1999.The Financial System and Economic Growth: The Korean
Experience.World Development, Vol. 27(6): 1069-1082.
Demetriades, P. O. and K. B. Luintel, 2001. Financial Restraints in the South Korean
miracle. Journal of Development Economics, Elsevier, Vol. 64(2): 459-479.
Government of Zimbabwe. Indigenisation and Economic Empowerment Act: 14/2007.
Greenwood, J. and B. Jovanovic, 1990. Financial Development, Growth and the
Distribution of Income.Journal of Political Economy, Vol. 98: 1076-1107.
Gurley, J. and E. Shaw, 1967. Financial Structure and Economic Development. Economic
Development and Cultural Change, Vol. 15 (2): 333-346.
Guryay, E., O.V. Safakli and B. Tuzel, 2007. Financial Development and Economic
Growth: Evidence from Northern Cyprus. International Research Journal of Finance and
Economics.Issue 8: ISSN 1450-2887.
Hamdi, H., A. Hakimi and R. Sbia, 2013.Multivariate Granger Causality Between
Financial Development, Investment and Economic Growth: Evidence from Tunisa.Journal of
Quantitative Economics, Vol. 11(1 & 2).
Herzer, D., F. D. Nowak-Lehmann and B. Siliverstovs. 2006. Export-led Growth in
Chile: Assessing the role of Export Composition in Productivity Growth, The Developing
Economies, Vol. 44(3).
Hondroyiannis, G., S. Lolos and E. Papapetrou, 2005. Financial Markets and Economic
Growth in Greece, 1986-1999. Journal of International Financial Markets, Institutions and
Money, Vol.15 (2):173-188.
Jalil, A., and Y. Ma, 2008. Financial Development and Economic Growth: Time Series
Evidence from Pakistan and China. Journal of Economic Cooperation, Vol. 29(2): 29-67.
Jecheche, P. 2010. An Empirical Investigation of the Financial Development - Economic
Growth Nexus: The Case of Zimbabwe (1990 – 2008). Journal of Comprehensive Research.
Kargbo, S. M. and P. Adamu. 2009. Financial Development and Economic Growth in
Sierra Leone. West African Journal of Monetary and Economic Integration, Vol 9(2):30–67.
Khan, M. A., A. Qayyum and A. S. Saeed. 2005. Financial Development and Economic
Growth: the Case of Pakistan. The Pakistan Development Review 44, Vol.4(2): 819-837.
King, R.G. and R. Levine. 1993a. Finance and Growth: Schumpeter Might Be Right.
Quarterly Journal of Economics, Vol. 108(3): 717-737.
King, R. G. and R. Levine. 1993b. Finance, Entrepreneurship, and Growth: Theory and
Evidence. Journal of Monetary Economics, Vol.32: 513-42.
Kohpaiboon, A. 2003. Foreign Trade Regime and FDI-Growth Nexus: A case study of
Thailand. The Journal of Development Studies, Vol.40 (2):55–69.
Levine, R.1997. Financial Development and Economic Growth: Views and Agenda.
Journal of Economic Literature, Vol. (XXXV): 688-726.
Levine, R. 1991. Stock markets, Growth and Tax Policy. Journal of Finance, Vol.46
(4):1445-65
Levine, R. and S. Zervos 1996. Stock Market Development and Long-run Growth. Policy
Research Working Paper 1582, World Bank Economic Review, Vol.10: 323-339.

Lucas, R. E. 1988. On the Mechanics of Economic Development. Journal of Monetary
Economics, Vol. 22(1): 3-42.
Martens, J. 2012. How Zimbabwe Lost its Food Sovereignty and Security. Sustainable
Development, Rosa Luxemberg Stiftung, Southern Africa.www.rosalux.co.za. Date accessed
20/03/2014.
McKinnon, R. I. 1973. Money and Capital in Economic Development. Washington, DC:
Brookings Institution.
Naceur, S. B. and S. Ghazouani. 2007. Stock Markets, Banks and Economic Growth:
Empirical Evidence from the MENA Region. Research in International Business and
Finance,Vol. 21:297-315.
Narayan, P. K. 2004. Reformulating Critical Values for Bounds F-Statistics Approach to
Cointegration: an Application to the Tourism Demand Model for Fiji. Discussion Papers,
Department of Economics: 1441-5429, No. 02/04, Monash University, Australia.arrow.
monash.edu.au/hdl/1959.1/2167. Date accessed 20 April 2014.
Ndlovu, G. 2013. Financial Sector Development and Economic Growth: Evidence from
Zimbabwe. International Journal of Economics and Financial Issues Vol. 3( 2): 435-446.
Obstfeld, M. 1994. Risk-Taking, Global Diversification, and Growth. American Economic
Review, December 1994, Vol.84 (5): 1310-1329.
Patrick, H. 1966. Financial Development and Economic Growth. Economic Development
and Cultural Change, Vol. (XIV): 451-65.
Pesaran, M. H. and B. Pesaran.1997. Working with Microfit 4.0: Interactive Econometric
Analysis, Oxford: Oxford University Press.
Pesaran, H., Y, Shin and R. Smith. 2001. Bounds Testing Approaches to The Analysis of
Level Relationships. Journal of Applied Econometrics, Vol. 16(3): 289-326.
Reserve Bank of Zimbabwe Monetary Policies, 2010, 2012 and 2013.www.rbz.co.zw/
pdfs. Date accessed 25 January 2014.
Rioja, F. and N. Valev. 2011. Stock Markets, Banks and the Sources of Economic Growth
in Low and High Income Countries. Journal of Economics and finance: 1-9.
Romer, P. M. 1986. Increasing Returns and Long run Growth. Journal of Political Economy,
Vol. 94: 1002-1037.
Saint-Paul, G. 1992. Technological Choice, Financial Markets and Economic Development.
European Economic Review, Vol. 364: 763-81.
Schumpeter, J. A. 1911. The Theory of Economic Development. Harvard University Press.
Shaw E. S. 1973. Financial Deepening in Economic Development. New York, Oxford
University Press, London and New York.
Solow, R. M. 1956. A Contribution to the Theory of Economic Growth. Quarterly Journal
of Economics Vol 70 (1): 65-94.
Sultan, R. 2012. An Econometric Study of Economic Growth, Energy and Exports
in Mauritius: Implications for Trade and Climate Policy. International Journal of Energy
Economics and Policy, Vol.2 (4): 225-237.
Swan, T.W. 1956. Economic Growth and Capital Accumulation. Economic Record, Vol
32(23): 34-61.
World Bank. 2013. World Development Indicators Database, The World Bank, Washington,
DC.data.worldbank.org/data-catalog/world-development-indicators. Date accessed 27
January 2014.
ZimStat. 2012. Compendium of Statistics. www.zimstat.co.zw/dmdocuments/
Compendium. Date accessed 25 January 2014.
Zivengwa, T., J. Mashika, F. Bokosi, and T. Makova. 2011. Stock Market Development
and Economic Growth in Zimbabwe. International Journal of Economics and Finance Vol.
3(5): 140-150.
The Botswana Journal of Economics (BOJE)
Aims and Objectives

BOJE is a refereed journal. The main objective of the journal is to promote the discussion
of economic issues at the theoretical, applied and policy levels. The journal welcomes
contributions that discuss these issues in new and imaginative ways, offering constructive
ideas and analysis and highlighting lessons that may be learned from the experiences national
and regional economies as well as the global economy at large.
Notes For Contributors

Articles submitted to the BOJE should be original contributions. The articles should not be
under consideration for any other publication. Authors should indicate if there is any version
or part of the article that has been or will be published elsewhere at the time of submission.
Articles should be submitted by email to the Managing Editor: Siphambe@Mopipi.ub.bw

with copies to: Okururtf@Mopipi.ub.bw. All pages should be numbered consecutively.
Articles should be submitted to the Editor in a form suitable for sending anonymously to
the reviewers. Articles should be accompanied by a record of the author’s name, address,
affiliation and contact telephone, fax number and e-mail. Authors should select up to six words
that describe their article. Two of these should describe geographical location; namely region
(e.g. Africa) and country (e.g. Botswana). A list of non-geographic key words from the editor
are printed at the back of this journal. Authors are free to choose key words for their article
that are not in this list.
Because of heavy pressure on space, articles should not exceed 25-30 double-space pages;
including summary, tables, figures, notes and references.
Each article should include an indented and italicised summary of not more than 150 words,
which describe the main arguments and conclusions of the paper.
Presentation Format and Style

1. BOJE’s main focus is economic development, which affects both men and women.
Because of this, BOJE encourages gender-neutral language.
2. All references cited in the article should be included in the Reference Section. Authors
should follow the following format of reference presentation:
Todaro, M. P. 1987. Economic Development in the Third World, 3rd edition, Longman,
New York and London.
Atta, J. K., K. Jefferis and I. Mannathoko, 1996. Small Country Experiences with
Exchange Rates and Inflation: the Case of Botswana. Journal of African Economies, Vol.
5(2):293–336.
3. Broad divisions and sections of the paper should be clearly marked. Quotations should
also be cleared marked. Quotations that exceed 40 words should be indented in the text.

4. Mathematical workings that include details that are not necessary to be included in the
main text, should be provided in a separate sheet of paper. These will be helpful for
referees but will not be published.
5. Tables and figures should be clearly marked and provided on a separate sheet of paper.
The position for tables and figures should also be clearly marked in the text.
6. Authors are expected to correct proofs expeditiously to keep alterations to a minimum.
7. Notes should be numbered consecutively and placed at the end of the article. Authors
acknowledgement should be given in a separate sheet. These will be presented on the first
page of the article.
Comments on Published Articles

Anyone wishing to make comments on a published article should first send a copy to the
authors, inviting them to give their views on the issues raised. The commentators should send
their comment to BOJE together with the author’s response.
Accepted Articles
Following acceptance for publication, articles should be submitted by emails in Microsoft
Word. All figures and tables should be provided within the text.
Authors are entitled a copy of the issue in which their articles appear. Further copies may be
provided at a discount to the author.
Copyright in articles published in BOJE rests with the publisher.
Call for Papers

The Botswana Journal of Economics (BOJE) hereby invites submissions from Economics or
professionals in related disciplines, working on theoretical, applied, policy and other areas
of Economics for the next issue of the Journal. Submissions are to be made to the Editor,
according to the guidelines to contributors that accompany this issue. All contributions are
encouraged to follow the guidelines in order to ease delay in the editorial process.

Call for Papers
The Botswana Journal of Economics (BOJE) hereby invites submissions from Economics or
professionals in related disciplines, working on theoretical, applied, policy and other areas
of Economics for the next issue of the Journal. Submissions are to be made to the Editor,
according to the guidelines to contributors that accompany this issue. All contributions are
encouraged to follow the guidelines in order to ease delay in the editorial process.

Managing Editor
Happy Siphambe
Co-Editor
S. N. Okurut, L. Setlhare, T. Feger and M. Bakwena
Editorial Board
Dr. N.H. Fidzani – Botswana Institute for Development Policy Analysis
Professor Folayan Ojo – Department of Economics, University of Swaziland
Dr. Roy Love - Department of Economics, Sheffield Hallam University, United Kingdom
Professor M.S. Mukras - Department of Economics, Arap moi University, Kenya
Professor J. Salkin – Advisor, Bank of Botswana
Professor J. Ddumba-Ssentamu – Institute of Economics, Makerere University, Uganda
Professor C. Perrings – York University, United Kingdom
Professor R. Hassan – University of Pretoria, South Africa
Dr P. Lawrence – Department of Economics, Keele University, United Kingdom
Professor V. Murinde – University of Birmingham, United Kingdom
The Botswana Journal of Economics (BOJE) is a publication of the Botswana Economics Association
(BEA). The Association was founded in 2002 with the objectives of creating an impartial and autonomous
environment through which debate, free interchange of information, and free contact between different
sectors of the Botswana society can take place. This is with the aim of enriching economic policy formulation
and implementation in the country. At different fora, which is one of the key activities of BEA, members and
concerned professionals get the opportunity to express opinions on any matter provided that they correspond
with the objectives of the Association. Membership is open to professional economists who are trained and/or
practicing in the field of economics and related professions as well as any persons who subscribe to the aims
and objectives BEA.
The Botswana Journal of Economics is a professional journal established for the dissemination of
contemporary economic issues – theoretical, methodological, and policy relevant – in the context of both
the immediate environment and the wider international community. It is open to all researchers in the field of
economics. Non-members as well as members of the Association are welcome to submit unpublished research
articles for editorial consideration. Prospective contributors will find the submission procedure on the inside
cover of each issue of the journal. Editorial correspondence should be sent to the Managing Editor, The
Botswana Journal of Economics, Department of Economics, Private Bag UB 00705, Gaborone, Botswana.
Email: siphambe@mopipi.ub.bw.
The Botswana Economics Association (BEA) and the Botswana Journal of Economics (BOJE) gratefully
acknowledge the support of Friedrich Ebert Foundation, Botswana Office towards the publication of this issue
of the Journal.
Subscriptions
The Journal is published two times in a year: March and October. Copies shall be sent to all paid up members
of the Association. Membership fee for the Association is P100 per year. A single copy of the journal may be
obtained for P60. Subscription rates are as follows:
Individuals Institutions
1 – year subscription P110 P150
2 – year subscription P200 P230
Copyright
All rights reserved. Apart from fair dealing for the purposes of research or private study, or criticism or review
as permitted under the copyright laws of Botswana, no part of this publication may be reproduced, stored
or transmitted in any form, electronic, mechanical or by any other means without, either the prior written
permission of the Publisher, or in accordance with the terms of photocopying licenses issued by organizations
authorized by the Publisher to administer reprographic reproduction rights.
A commercial resale of published brochures, books and general printed matters by the Friedrich-Ebert-
Stiftung are prohibited unless the Friedrich-Ebert-Stiftung gives its explicit and written approval beforehand.

The Journal of The Botswana Economics Association

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Journal of The Botswana Economics Association

Uploaded by

Copyright:

Available Formats

The Botswana Journal of Economics

The Journal of the Botswana Economics Association

Capital Account Liberalisation and Foreign Direct Investment in Nigeria:

Credit demand amongst farmers in Mukono District, Uganda

Determinants of Informal Employment: A Case of Tanzania’s Construction Industry

The Impact of Bank and Stock Market Developments on Economic Growth

Keywords: bonds, Long Memory, AFRIMA, ARFIMA-FIGARCH

JEL Classification Code: G1, G 12, G14.

1 BOJE: Botswana Journal of Economics

BOJE: Botswana Journal of Economics 2

II. BOND MARKET REFORMS IN KENYA

3 BOJE: Botswana Journal of Economics

Table 1: Economic and Financial Indicators for Kenya

BOJE: Botswana Journal of Economics 4

III. LONG MEMORY IN TIME SERIES

5 BOJE: Botswana Journal of Economics

BOJE: Botswana Journal of Economics 6

ht = ω + α ( L)ε t2 + β ( L)ht .......................................................................................................

(3) can be rewritten as an ARMA (m, p) process in ε t2 ,

[1 − α ( L) − β ( L)]ε t2 = ω +[1 − β ( L)]ν t .....................................................................................

φ ( L)(1 − L) d ε t2 = ω + [1 − β ( L)]vt ...........................................................................................

η = T {ln Γ[0.5(df + 1)] − ln Γ(0.5df ) − 0.5 ln[π (df − 2)]}

7 BOJE: Botswana Journal of Economics

(a) Preliminary Observations

(b) Results from GARCH Models

BOJE: Botswana Journal of Economics 8

Evidence of persistence in variance as measured by the GARCH model is reflected in the

Table 4. GARCH Estimates

9 BOJE: Botswana Journal of Economics

Table 5. ARFIMA-FIGARCH Estimates

BOJE: Botswana Journal of Economics 10

11 BOJE: Botswana Journal of Economics

BOJE: Botswana Journal of Economics 12

13 BOJE: Botswana Journal of Economics

Bankole, Abiodun S. (Ph.D)1 and Ayinde, Taofeek Olusola2

This study focuses on the neoclassical counter-revolution framework to investigate the

Key Words: Bound-Test, Capital Accounts, Foreign Investments, Neo-Classical.

JEL Classification Code: E22, F 32, F 23.

BOJE: Botswana Journal of Economics 14

Essentially, the liberalisation of capital account transactions is predicated on the allocative

15 BOJE: Botswana Journal of Economics

Liberalisation of capital account transactions is well rooted in theoretical propositions. It is

Nonetheless, the neoclassical counterrevolution framework indicates that market distortions

Essentially, the emphasis of the neoclassical counterrevolution in development policy was

BOJE: Botswana Journal of Economics 16

Y = f ( K ) ; which denotes Output as a function of capital stock(1) .............................................. (1)

Where; CAOPEN – De-facto (market-based) measure of Capital Account Liberalisation (CAL);

CAO PEN KAO PEN

Table 1: Augmented Dickey Fuller (ADF) Test of Unit-Root

At Levels At Order 1 Order of Integration

KAOPEN -1.2780 -4.5851 I(1)

CAOPEN -1.5102 -4.4567 I(1)

FDI -4.0480 - I(0)

TOPEN -0.9463 -7.3598 I(1)

INFR -2.9680 -7.3221 I(1)

OIL_EXPORT -5.7148 - I(0)

CPS_GDP -1.7163 -4.6291 I(1)

GOV_CONSU -3.8707 - I(0)

EXCHR -4.5693 - I(0)

INTR -3.0021 - I(0)

21 BOJE: Botswana Journal of Economics

4.3 GRANGER CAUSALITY ESTIMATES

Table 2: Granger Causality Results