Article

Empirical Analysis Foreign Trade Review

54(3) 177–205, 2019
of the Determinants © 2019 Indian Institute of
Foreign Trade
of Trade Balance in Reprints and permissions:
in.sagepub.com/journals-permissions-india
Post-liberalization DOI: 10.1177/0015732519851632
journals.sagepub.com/home/ftr
Ghana

Linda Akoto1
Daniel Sakyi1

Abstract
This study investigates the determinants of trade balance in post-liberalization
Ghana, covering the period 1984–2015. Specifically, we test the validity of the
Marshall-Lerner condition and the J-curve effect, and further assess the effect
of other macroeconomic variables including household consumption expendi-
ture, government consumption expenditure, foreign income, money supply and
domestic prices on trade balance. The bounds testing approach to cointegration
and the error correction model within a symmetric and asymmetric autoregres-
sive distributed lag (ARDL) framework is used for the estimation. Additionally,
to analyse the dynamic interactions of the variables included in the estimated
model, variance decomposition is applied. The results from both symmetric
and asymmetric specifications show the absence of the Marshall-Lerner condi-
tion and the J-curve effect. Further, the study finds that household consump-
tion expenditure, government consumption expenditure and domestic prices
are negative and significant in the long and short run, whereas foreign income
and money supply are positive and significant in the short run. Results from
the variance decomposition show that innovations in household consumption
expenditure highly contribute to the forecast error variance of the trade bal-
ance compared with other explanatory variables. A key finding of the study
suggests that depreciation of the Ghana cedi is not an appropriate step to help
in improving the country’s trade balance position.

1
Department of Economics, Kwame Nkrumah University of Science and Technology, Kumasi, Ghana.

Corresponding author:
Daniel Sakyi, Department of Economics, Kwame Nkrumah University of Science and Technology,
Private Mail Bag, Kumasi, Ghana.
E-mail: dsakyi.cass@knust.edu.gh
178 Foreign Trade Review 54(3)

JEL Codes: C22, F10, F31, F32

Keywords
Trade balance, Marshall-Lerner condition, J-curve effect, cointegration, Ghana

Introduction
International trade has, for years, remained a key driver of economic prosperity
of countries that are increasingly integrated into the global economy. It serves as
an important tool in boosting the economic and social performance of countries,
especially the developing ones (United Nations Conference on Trade and
Development [UNCTAD], 2005). These benefits can be viewed through various
channels, including economic growth and/or development, poverty reduction
and low inflation (Bhagwati & Srinivasan, 2002; Karam & Zaki, 2015; Kim &
Beladi, 2005; Lotfalipour, Montazeri, & Sedighi, 2013; Sakyi, Villaverde and
Maza, 2015). Notwithstanding, it is important to note that whether a country is
reaping the benefits of international trade or not can be viewed through its trade
balance position which when in deficit (surplus) is often viewed as detrimental
(beneficial) for an economy.
While many developed countries have deliberately harnessed the forces of
international trade to improve their trade balance position, many developing
countries though have embarked on various unilateral and multilateral trade pol-
icy reforms have made very little gains (Osoro, 2013; Shawa & Shen, 2013;
UNCTAD, 2005, 2013). Ghana is no exception to this group of countries that have
implemented several unilateral and multilateral trade policy reforms with the aim
of improving its trade balance position. Among these are the Economic Recovery
Programme (ERP); the Structural Adjustment Programme (SAP); the Free Zone
Act, 1995 (Act 504); African Growth and Opportunity Act (AGOA); and the
African, Caribbean and Pacific group of States and the European Union (ACP-EU)
Economic Partnership Agreements. Again, there have been various efforts to
actively improve the country’s trade balance position through the establishment of
the Ghana National Export Promotion Council (GEPC) and the Ghana Export
Trade Information Centre (GETIC).
Prior to the implementation of the reform programmes, an import substitution
industrialization policy aimed at reducing the country’s dependence on imports
was implemented. During this period, state-owned enterprises in various sectors
of the economy were established, ranging from agricultural to manufacturing
industries. The import substitution industrialization policy was, however, not sus-
tainable due to continual political instabilities, low levels of factor productivity
and gross economic mismanagement (Killick, 2010). It is not surprising that for
most periods during this era, economic performance including the position of the
trade balance (see Appendix A) was poor. Although the country recorded a 3-year
trade surplus for the years 1967 to 1969, this was not sustainable as, by 1971, it
experienced one of its largest trade deficits (Killick, 2010). As the Bank of
Akoto and Sakyi 179

Ghana (2008) and Sakyi (2011) observed, the story was not different during the
1970s and the early 1980s. These huge trade deficits coupled with other economic
imbalances experienced in the pre-liberalization era precipitated the launch of the
ERP in 1983 on the basis of liberal policy regime with the aim of addressing,
among others, the trade deficits experienced in the country (Rodrik, 1999).
Though there were early gains in Ghana’s trade balance position after ERP,
and, by 1986, the country recorded a trade surplus of 0.025 per cent of GDP,1 this
was not sustainable, as for most periods thereafter the country recorded deficits in
its trade balance position (see Appendix A). During the early 1990s, the country
experienced trade deficits of 0.42 per cent, 0.51 per cent and 0.58 per cent of GDP
in 1990, 1992 and 1993, respectively. This could be attributed to the twin external
shocks which hit the country in the early 1990s. These shocks resulted in an
increase in crude oil prices and the collapse of two major export commodities of
the country, namely cocoa and gold. Although during some periods after ERP
there were significant improvements in the country’s trade deficits, the trade bal-
ance position in most periods was poor as indicated earlier. For instance, between
2001 and 2003, there were significant improvements in the country’s trade bal-
ance position as deficits improved from 0.35 per cent of GDP in 2001 to 0.25 per
cent of GDP in 2003. This was as a result of the quick recovery in the prices of
cocoa and gold in the international market (Bank of Ghana, 2008). However,
Ghana experienced trade deficits of 0.42 per cent of GDP in 2004 and 0.57 per
cent of GDP in 2008. The story is not so different in recent times as the country
recorded deficits of 0.44 per cent, 0.28 per cent, 0.26 per cent, 0.32 per cent, 0.21
per cent and 0.18 per cent of GDP in 2010, 2011, 2012, 2013, 2014 and 2015,
respectively. It is evident that though the trade liberalization policies embedded in
the ERP were aimed at improving the country’s trade balance position, there have
been very limited results in that regard.
It is important to note that, despite the negative implications of persistent trade
deficits on the economic performance of countries, only a few studies (see
Bhattarai & Armah, 2005; Danquah, 2008; Iyke & Ho, 2017) have been devoted
to the determinants of trade balance in Ghana. However, these studies mainly
focused on the effect of the exchange rate on Ghana’s trade balance controlling
for only domestic and foreign incomes. We further note that virtually no attention
has been given to the impact of other relevant macroeconomic variables on
Ghana’s trade balance. Again, following the recent literature, another limitation
we deduce from Bhattarai and Armah (2005) and Danquah (2008) is their sole
focus on the symmetric impact of the exchange rate on the trade balance. In sup-
port, Bahmani-Oskooee and Bahmani (2015) and Bahmani-Oskooee and
Fariditavana (2015, 2016) argue that the relationship between trade balance and
exchange rate may be asymmetric. However, as per the literature reviewed, only
a few studies have considered both symmetric and asymmetric determinants of
trade balance.
Given that Ghana’s post-liberalization period has mostly been characterized by
trade deficits, the question worth investigating is what the key determinants of
these deficits are? To answer this question, this article provides an empirical test
of the Marshall-Lerner condition and the J-curve effect by investigating the effect
180 Foreign Trade Review 54(3)

of exchange rate movement on trade balance in Ghana. It further assesses the

effect of other macroeconomic variables including household consumption
expenditure, government consumption expenditure, foreign income, money sup-
ply and domestic prices on trade balance. Finally, we analyse the dynamic interac-
tions of the variables considered in the study.
The remainder of the article is structured as follows. The second section pre-
sents a survey on the determinants of trade balance. This is followed by the model,
data and the estimation methods in the third section. The fourth section reports the
empirical results and discussion. Finally, the fifth section is devoted to the impli-
cations of the study results and offers some concluding remarks.

Determinants of Trade Balance: A Review

This section of the article is divided into two parts. The first part provides a review
of the literature on the theoretical links between the trade balance and its determi-
nants, while the second part focuses on the literature related to empirical studies.

Theoretical Review
Theoretically, the exchange rate has an impact on trade balance, which can be
discussed in terms of the Marshall-Lerner condition and the J-curve effect. The
former, as proposed by Alfred Marshall and Abba Lerner, states that in a situation
of currency depreciation (devaluation), the summation of the elasticities of
demand for long-term exports and imports must be more than one to attain a
favourable trade balance position. On one hand, if the summation is exactly one,
there will be no change in the trade balance in the event of depreciation (devalua-
tion). On the other hand, a perverse effect of depreciation (devaluation) on trade
balance occurs if the sum of demand elasticities of exports and imports is less than
one. Also, the J-curve effect posits that a worsening trade balance resulting from
depreciation (devaluation) of a particular currency may only be temporary. The
J-curve effect further postulates that due to the inherent nature of individuals, in
the event of currency depreciation (devaluation), there will be an initial deteriora-
tion of trade balance in the short-run but this gets improved in the long-run. In
other words, depreciation will worsen the trade balance of an economy in the
short run but subsequently improves if both import demand and export supply are
more inelastic in the short-run relative to the long-run.
In addition, the economic theory postulates that the money supply of an econ-
omy can influence its trade balance position. This can be observed from the per-
spective of the monetarist approach to the balance of payments which explains
how changes in demand and supply of money affect the overall balance of pay-
ments. Given that trade balance is the most important component of the balance
of payments of most developing countries, of which Ghana is no exception, this
phenomenon is relevant for the present study. At a given exchange rate, a rise in
Akoto and Sakyi 181

domestic money supply with no change in money demand will worsen the bal-
ance of payments through its effect on the trade balance. This is so because indi-
viduals with large cash balances may increase their demand for both domestic
and foreign (imported) goods. This may first result in a trade deficit which then
trickles down to cause a current account deficit and subsequently, a balance of
payments deficit.
Further, the saving-investment gap is used to explain the relationship between
government expenditure and the trade balance. The saving-investment identity
expresses trade balance as the sum of private savings and government savings.
This implies that private savings and government savings determine imbalances
in a country’s trade balance. Specifically, all other things being equal, an increase
in investment by the private sector will cause deterioration in the trade balance of
an economy. Similarly, an increase in government expenditure will cause deterio-
ration in the trade balance of an economy, ceteris paribus.
Again, the absorption approach to the balance of payments explains the theo-
retical link between domestic expenditures and trade balance. Proponents of this
approach (see, e.g., Alexander, 1952, 1959; Harberger, 1950; Meade, 1951) are of
the view that an economy experiences a favourable balance of payments if its
national income increases more than its domestic expenditures. Hence, an increase
in government and/or household expenditures relative to national income (output)
will worsen a country’s trade balance.

Empirical Review
Several empirical studies have investigated the principal determinants of trade
balance in both developed and developing countries.
Using quarterly data over 1971(Q1) to 1994(Q4) period, Bahmani-Oskooee
(2001) employs the Engle-Granger and Johansen-Juselius cointegration technique
to investigate the effects of nominal and real effective exchange rates of 11 Middle
Eastern countries on their trade performance. The findings show support for the
Marshall-Lerner condition by revealing that real depreciation has a favourable
long-run effect on the trade balance of most non-oil exporting Middle Eastern
countries. The results further indicate that while domestic income obstructs the
trade balance of the respective countries, foreign income is revealed to enhance
trade balance.
Catao and Falcetti (2002) employ the Autoregressive Distributed Lagged
(ARDL) method with a lag structure of the auto-regression selected by the
Schwartz Bayesian Criterion and the vector error correction models to investigate
the main determinants of Argentina’s trade balance. Using time series data span-
ning from 1975 to 2000, the results show that, the major drivers of Argentina’s
trade balance are export sensitivity to commodity prices on the international mar-
ket, domestic absorption and high-income elasticities for imports.
Employing cointegration techniques and the impulse response function (IRF)
on annual data from a period of 1970 to 2000, Narayan (2004) tests the J-curve
effect in New Zealand. The study finds no cointegration between real effective
182 Foreign Trade Review 54(3)

exchange rate, domestic income, foreign income and trade balance. In other
words, the results show no stable long-run relationship amongst the variables. The
results further reveal a causal connection in both directions between foreign
income and trade balance and also find a one-way link between real exchange rate
and trade balance. Again, the study shows a clear existence of the J-curve pattern
for New Zealand’s trade balance based on the IRF results. More succinctly, the
results show that for the first 3 years, depreciation in New Zealand dollars wors-
ens the trade balance but improves it afterward confirming the J-curve effect.
Similarly, Singh (2004) employs the error correction model (ECM) and gen-
eralized autoregressive conditional heteroskedasticity (GARCH) model to ana-
lyze the effects of exchange rate volatility on India’s trade balance. The study
also tests for the presence or otherwise of the J-curve effect using quarterly data
from 1975(Q2) to 1996(Q3). The result shows no support for the existence of the
J-curve in India’s trade balance. The study further reveals that the impact of
exchange rate volatility on India’s trade balance is insignificant. The short-run
results show that domestic income and real exchange rate play a key role in
improving India’s trade balance. The effect of foreign income on India’s trade
balance is also however insignificant.
In a related study, Onafowora and Owoye (2006) use cointegration, vector
error-correction techniques, Granger-causality tests and generalized impulse
response analysis to examine the causal relationship between trade balance and
budget deficit in Nigeria from 1970 to 2001. The study finds a positive long-run
equilibrium relationship between budget deficit and trade deficit. Contrary to
the conventional theory, this study shows a causal link from trade deficit to a
budget deficit and the authors attribute this outcome to the fact that Nigeria is
mainly dependent on the export of petroleum products. The study further shows
a positive correlation between money supply, real exchange rate and trade bal-
ance, and a negative relationship between domestic income, interest rates and
trade deficit in the long-run. The authors further explain that an increase in
domestic income causes a rise in demand for foreign goods and services while
rising interest rates encourage capital inflow which leads to an appreciation of
the domestic currency.
In addition, Saruni (2007) employs the ordinary least square (OLS) regression
analysis for data spanning from 1970 to 2002 in Tanzania and concludes that trade
liberalization, foreign direct investment, government expenditure and private
consumption expenditure are the main contributors to Tanzania’s trade balance
although the coefficient of foreign direct investment is negligible. Specifically,
government expenditure, foreign direct investment and foreign income are found
to exert positive impact on Tanzania’s trade balance while household consump-
tion expenditure, real exchange rate and trade liberalization influence the trade
balance negatively.
Using cointegration techniques—specifically the bound testing approach
within the ARDL framework—Duasa (2007) investigates the main drivers of
Malaysia’s trade balance over the period 1974 to 2003. The author concludes that
money supply and income play a major role in determining the long-run behav-
iour of Malaysia’s trade balance as compared to the exchange rate. Specifically,
Akoto and Sakyi 183

money supply is found to have a negative effect on trade balance while income
has a positive effect. Though real exchange rate has a positive relationship with
trade balance, its impact was statistically insignificant, hence showing no evi-
dence of the Marshall-Lerner condition for Malaysia. Policy-wise, Duasa (2007)
indicates that the difficulties in trade balance with regard to Malaysia’s context
can be corrected through its policies on income or growth and money supply
rather than the exchange rate regime.
Falk (2008) contributes to the subject matter by applying the fixed effects and
linear mixed models within panel data analysis techniques to data on 32 emerging
countries and industrialized economies spanning from 1990 to 2007. The study
shows that trade balance is positively and significantly related to real foreign GDP
per capita of the respective trading partners. It is also revealed that real domestic
GDP per capita has a negative effect on trade balance. Though a depreciation of
the real exchange rate index is found to improve trade balance, countries with
negative trade balance and/or large positive net foreign investment position, how-
ever, have their trade balance less sensitive to movements in the real effective
exchange rate index.
Employing the bounds testing approach to cointegration and error correction
models developed within an autoregressive distributed lag (ARDL) framework,
Waliullah, Khan, Kakar, and Khan (2010) examine the determinants of Pakistan’s
trade balance from a period of 1970 to 2005. The authors find evidence of cointe-
gration between trade balance, foreign income, exchange rate and money supply.
The results further show that the depreciation of the exchange rate is directly
related to the trade balance in the long and short-run which confirms the Marshall-
Lerner condition in the case of Pakistan. Further, the study shows that money
supply and foreign income play a vital role in determining the behaviour of trade
balance. Specifically, money supply has a negative effect on Pakistan’s trade bal-
ance while that of foreign income is positive. Waliullah et al. (2010) further note
that, even though the exchange rate regime can help improve upon Pakistan’s
trade balance, it has a weaker influence as compared to monetary policy and
income. The short-run results are consistent with the long-run results with the
exception of foreign income which is negative and insignificant.
Mohammad (2010) also explores the principal factors of Pakistan’s trade bal-
ance from 1975 to 2008 using the Johansen cointegration approach and error cor-
rection model. The study reveals that foreign income, foreign direct investment,
inflation, domestic household consumption and real effective exchange rate sig-
nificantly affect Pakistan’s trade balance. Specifically, the results show that for-
eign income, foreign direct investment and exchange rate have positive effects on
Pakistan’s trade balance whereas inflation and domestic household consumption
affect trade balance negatively in the long-run. The short-run results for all the
variables are consistent with the long-run findings with the exception of foreign
income and inflation. Whereas the effect of foreign income on trade balance is
negative and significant, that of inflation is positive but insignificant in the
short-run.
Again, Nienga (2010) contributes to the discussion by investigating the deter-
minants of trade balance in Kenya. The author reveals real exchange rate, government
184 Foreign Trade Review 54(3)

consumption expenditure, domestic income and money supply (M3) as the main
significant factors that affect Kenya’s trade balance during the period 1920 to
2010. Using the ordinary least squares technique, the study shows that real
exchange rate and domestic income affect Kenya’s trade balance positively while
the effect of government consumption expenditure and money supply on trade
balance is negative.
Khan and Hossain (2012) employ the unrestricted error correction mecha-
nism (UECM) and the generalized method of moments (GMM) estimator to
examine the determinants of Bangladesh’s trade balance and 50 of its trading
partners from 1980 to 2005. The results show evidence of cointegration imply-
ing the existence of a stable long-run relationship between trade balance and its
determinants, that is, relative GDP, relative per capita GDP and real exchange
rate. The short-run dynamics also show the convergence of trade balance to its
long-run equilibrium.
Sarbapiya (2012) also employs various econometric techniques including
Augmented Dickey-Fuller (ADF), Johansen cointegration, vector error correction
models (VECM) and OLS to observe the long- and short-run trade behaviour for
India over the period of 1973 to 2011. The results reveal a long- and short-run
causality among inflation, exchange rate, FDI, household consumption and for-
eign income. Foreign income and FDI have a significant positive impact on trade
balance while household consumption, inflation and exchange rate exert negative
effects on trade balance in the long-run. The findings from the short-run analyses
are consistent with those of the long-run.
Similarly, Kayhan, Bayat, and Yüzbaşı (2013) use quarterly data for the period
1987Q1 to 2011Q3 and employ the bootstrap process based on the Toda-
Yamamoto causality and frequency domain analysis methods as estimation tech-
niques. The study finds a bi-directional causality between trade deficits and
government expenditure in Turkey. Contrary to the Toda-Yamamoto causality
analysis, the frequency domain causality analysis finds the causality running from
government expenditure to trade deficit to be a short-run or medium-term phe-
nomenon. On the other hand, frequency domain causality analysis further reveals
that there is a causal relationship that runs from trade deficit and government
expenditure in both the long- and short-run. In light of literature and theoretical
frameworks, the findings imply that a reduction in government expenditures will
successfully reduce trade deficits.
In a related study, Sulaiman and Abdul-Rahim (2014) test the J-curve pattern
on Thailand’s trade in forest products with the rest of the world. The study employs
the bounds test within the ARDL framework using data spanning from 1970 to
2010. The study also assesses the short- and long-run dynamics of Thailand’s
trade balance and its determinants. The results show no evidence of the J-curve
pattern in Thailand’s trade in forest products. The long-run results show that
Thailand’s national income improves its trade balance while foreign income,
exchange rate and forest policy cause trade balance to deteriorate. In the short-run,
Thailand’s national income exerts a positive and significant impact on trade bal-
ance. The impact of foreign income, exchange rate and forest policy are however
statistically insignificant in the short-run.
Akoto and Sakyi 185

Turkay (2014) employs the Johansen cointegration and error correction model
on annual time series data spanning from 1980 to 2012 to test the validity of the
Marshall-Lerner condition in Turkey. The study finds support for the Marshall-
Lerner condition in Turkey in the long-run. However, the short-run results show
an insignificant relationship between exchange rate and Turkey’s trade balance.
Further, Tandon (2014) explores the trade balance-real exchange rate multilat-
eral and bilateral relationship from a panel of region’s (Europe, Latin America and
Asia) perspective. Using data covering the period 1980 to 2010, the author
employs the generalized least square (GLS), ARCH and Newey West methods of
estimation to show that real exchange rate is a significant determinant of export,
imports and the overall trade balance in most regions and countries included in the
study. The results further indicate that the response of trade balance to real
exchange rate for countries belonging to the currency union is not symmetric. In
addition, evidence from countries such as Brazil, Mexico, China, Malaysia and
the Philippines supports the theoretically established fact that a devaluation of the
real exchange rate improves trade balance.
Tran and Dinh (2014) employ fixed and random effects models to examine the
effects of FDI inflows on trade imbalances in developing and transition countries in
Asia. The study further extends the trade balance model to assess the impact of the
real exchange rate, foreign income, production capacity of the manufacturing sector
and domestic absorption on trade imbalances. The findings indicate that FDI inflows
cause a decline in trade balance. The study also finds that real depreciation tends to
worsen trade balance, and therefore show no evidence of the Marshall-Lerner con-
dition. The results further show that foreign income, domestic absorption and the
production capacity of the manufacturing sector improve trade balance.
Tutueanu (2015) examines the J-curve effect on international trade of Romanian
forest products. The study applies the bounds testing approach within the ARDL
framework to annual time series data spanning from 1990 to 2014. The results show
the absence of the J-curve pattern for the trade balance of forest products. The study
further shows that Romania’s national income and world income have positive and
significant long-run effects on the trade balance of forest products in Romania.
Using the asymmetric ADRL model and monthly data covering the period June
1999 to April 2012, Aliyu and Tijani (2015) investigate the effect of exchange rate on
Nigeria’s trade balance. The estimated asymmetric error correction models provide
new evidence for slower transmission of exchange rate depreciations into the coun-
try’s trade balance. Again, the findings suggest that, Nigeria’s trade balance responds
to asymmetric changes in exchange rate. Further, the study finds a bi-directional
causal relationship between changes in exchange rates and trade balance.
In a related study on Hungary, Varge (2015) analyzes the impact of the agricul-
tural sector on employment and foreign trade balance. Using time series data
spanning from 2004 to 2014, the study employs descriptive analysis as the ana-
lytical tool. The study reveals the agricultural sector as the second largest con-
tributor to exports and adds that the sector improves Hungary’s trade balance year
after year over the sample period. Again, the study shows that the share of agricul-
ture is higher in exports than in imports of Hungary, hence causing a significant
improvement in Hungary’s trade balance.
186 Foreign Trade Review 54(3)

Using quarterly data from bilateral data of the USA and its six major partners
(Canada, France, Germany, Italy, Japan and the UK) over a period of 1971(Q1) to
2013(Q2), Bahmani-Oskooee and Fariditavana (2016) test the validity of the
J-curve effect using both linear and non-linear specifications. The result shows
evidence of the J-curve effect between the USA and its bilateral trading partners
in the linear specification. However, J-curve effect is evident between only the
USA and China when the non-linear adjustment process is applied.
Bahmani-Oskooee, Halicioglu, and Ghodsi (2016) investigate the symmetric
and asymmetric effect of exchange rate for Britain and eight of its partners using
monthly data spanning the period 1988M1 to 2015M11. The study employs the
non-linear version of the ARDL approach to cointegration and error correction
estimation techniques to examine whether pound appreciations and depreciations
affect trade differently. The findings reveal that the UK–Germany, the UK–Italy
and the UK–Japan data show a long-run asymmetric effect of exchange rate
whereas foreign income is found to exert a positive effect. Domestic income is
revealed to have a negative effect in both cases.
Iyke and Ho (2017) examine the effect of real exchange rate changes on
Ghana’s trade balance using quarterly data spanning the period 1986Q1 to 2016Q3
and employ linear and non-linear specifications for analysis. The study finds evi-
dence of an asymmetric effect of the exchange rate on trade balance. The study
also finds support for the J-curve effect and further reveals that domestic income
and foreign income enhance Ghana’s trade balance.
Ketenci (2017) investigate the effect of the existence of the customs union
between Turkey and the European Union (EU) on Turkey’s trade balance by the
use of an extended gravity model of trade. Estimations were done for two periods,
pre-trade liberalization (1980 to 1995) and post-trade liberalization (1996 to
2012). The results reveal that Turkey’s involvement in the customs union led to
significant improvement in its trade balance with the EU.
Korkmaz and Bilman (2017) adopt the stepwise testing procedure for the
hypothesis of the S-Curve of Turkey’s trade balance using quarterly data for the
period 2003(Q1) to 2011(Q4). The study investigates the S-curve hypothesis at
both aggregate and industry levels. The results from the industry level analysis
show that the S-curve hypothesis which posits that trade balance is uniformly
countercyclical and negatively correlated with the current and future movements
in terms of trade, but positively correlated with past movements was only evident
in the manufactured goods industry. On the contrary, at the aggregate level, there
was no evidence of S-curve behaviour.

Methodology and Data

Model Specification
This article adopts the reduced form of the trade balance model based on the
imperfect substitutes model of international trade originally postulated by
Goldstein and Khan (1985) and Rose and Yellen (1989) and followed by Rose and
Akoto and Sakyi 187

Yellen (1991), Wilson and Tat (2001), Bahmani-Oskooee (2001), Lal and
Lowinger (2002), Singh (2002) and Narayan (2004). The baseline model takes the
functional form:
TB  f (rE , Y , Y *) (1)

where TB represents trade balance, rE represents the exchange rate, Y represents

domestic income and Y* represents foreign income. Following the previous litera-
ture, we modify equation (1) in two ways.
The first modification is motivated by recognizing that different expenditure
components (household and government expenditures) of aggregate demand
or domestic income (Y) have different effects on trade balance. The second
modification is to capture the effect of certain key variables (i.e., money supply
and domestic prices) identified in literature (see, for example, Duasa, 2007;
Kayhan et al., 2013; Mohammad, 2010; Saruni, 2007; Shawa & Shen, 2013;
Waliullah et al., 2010) to have an effect on trade balance. The monetary approach
to balance of payments postulates that the money supply of an economy can
influence its trade balance position. At a given exchange rate, a rise in domestic
money supply with no change in money demand will worsen the balance of pay-
ments through its effect on the trade balance. This is so because individuals with
large cash balances may increase their demand for both domestic and foreign
(imported) goods.
Thus incorporating these two modifications, equation (1) is re-specified as:
TB  f (rE , HCEX , GCEX , Y * , MS , DP ) (2)

where HCEX represents household consumption expenditure, GCEX represents

government consumption, MS represents money supply and DP represents domes-
tic prices. All other variables are as previously defined. The estimable form of
equation (2) which we call Model 1 is expressed in natural logarithm as:
lnTBt = α0 + α1 ln rEt + α2 ln HCEX t + α3 ln GCEX t + α4 ln Yt *
+ α5 ln MSt + α6 ln DPt + µt (3)

ais (i = 1, 2, 3, ..., 6) represents parameters to be estimated, t is the time period

considered for the study, m represent the Gaussian white noise and ln is the natural
logarithm operator. As a priori, a real depreciation indicated by a rise in rE is
expected to make domestic goods more competitive relative to their foreign coun-
terparts. As a result, a rise in the exchange rate will increase exports and reduce
imports all other factors being equal: implying that the coefficient of the exchange
rate is expected to be positive. An economy’s inability to produce enough to offset
local demand makes it immensely reliant on foreign substitutes. We thus expect
increments in household and government expenditures to worsen the trade bal-
ance. That is, a2 and a3 are expected to be negative. In addition, an increase in
world income stimulates demand for goods and services including exported prod-
ucts from Ghana. Hence a4 is anticipated to assume a positive sign. Furthermore,
increases in real money balance lower interest rate and incite the outflow of
188 Foreign Trade Review 54(3)

capital from the domestic economy. Such outflows occasionally cause deteriora-
tion in the trade balance. Thus, a5 is expected to be negative. Finally, high domestic
prices make domestic commodities expensive relative to prices of foreign prod-
ucts. This fuels a high demand for less-expensive imported substitutes, which in
turn worsens the country’s trade balance. Therefore, we expect a6 to be negative.
In addition, the recent literature argues that the effect of exchange rate on trade
balance may be asymmetric rather than symmetric (see Bahmani-Oskooee &
Bahmani, 2015; Bahmani-Oskooee & Fariditavana, 2015). If this is the case then
the relationship between the real exchange rate and trade balance cannot be ade-
quately captured by the specification as represented in equation (3). Following
previous studies (see, for example, Bahmani-Oskooee & Bahmani, 2015;
Bahmani-Oskooee & Fariditavana, 2016; Delatte & Lopez-Villavicencio, 2012;
Verheyen, 2013), we formulate an asymmetric trade balance model by decompos-
ing the movement of exchange rate into its negative (appreciation) and positive
(depreciation) partial sums. Thus,
lnrE = lnrEo + lnrEt+ + lnrEt−

where lnrEt+ and lnrEt– are the partial sum process of positive and negative
changes in LnrE, respectively. More elaborately, they can further be expressed as:
t t

POSt = lnrEt+ = ∑ ∆lnrE+j = ∑ max(∆lnrE j , 0)

j=1 j=1

t t

NEGt = lnrEt− = ∑ ∆lnrE−j = ∑ min(∆lnrE j , 0) (4)

j=1 j=1

To obtain the asymmetric versions of the estimable model, we replace lnrE in

equation (3) with POSt (positives) and NEGt (negatives) and call this Model 2.
Doing so is important as it allows us to determine whether the effect of real
exchange rate on the trade balance is symmetric or asymmetric. If the two partial
sums have the same coefficients in terms of sign and size, then the effect is sym-
metric, otherwise, they are asymmetric. Adding to this, following Bahmani-
Oskooee and Fariditavana (2015), we define a significant and positive normalized
coefficient of POS as evidence of the Marshall-Lerner condition; and a short-run
deterioration combined with a significant and positive normalized coefficient of
POS as evidence of the J-curve.

Data Description and Sources

This article uses annual time series data spanning from 1984 to 2015. The choice
of this time span is premised on the fact that the study seeks to analyse Ghana’s
trade balance in the post-liberalization era. Except for the exchange rate, all vari-
ables included in the model are extracted from the World Bank’s World
Development Indicators (2016). Data on the exchange rate is obtained from the
Akoto and Sakyi 189

International Monetary Fund (IMF)–International Financial Statistics (IFS).

Following Bahmani-Oskooee (1991, 2001), the trade balance is measured as the
ratio of exports and imports. Bahmani-Oskooee (1991) provides two justifications
for this measure; (a) the ratio is not sensitive to units of measurement and (b) the
measure could be interpreted as nominal or real trade balance. In this study, we
use the real effective exchange rate index.2 This index is widely used as a measure
of the real exchange rate (see, for example, Bahmani-Oskooee, 2001; Rose &
Yellen, 1989; Singh, 2004; Tandon, 2014). The IMF computes this index using the
volume quotation system, thus quoting foreign currency in terms of the domestic
currency. However, for this article, we use the price quotation system (i.e., the
inverse of the volume quotation system) where an increase in the exchange rate
implies real depreciation. Again, M2 and consumer price index (CPI) are used as
measures for money supply and domestic prices, respectively. Duasa (2007),
Waliullah et al. (2010), and Shawa and Shen (2013) use the same measure for the
two variables. To obtain a measure of foreign income, we estimate the average
real GDP per capita of Organisation for Economic Co-operation and Development
(OECD) member countries3 due to the close bilateral relationship Ghana has with
these countries.4 Household and government consumption expenditures are meas-
ured as percentages of GDP as used by Saruni (2007), Nienga (2010), and Shawa
and Shen (2013).

Estimation Strategy
In this study, we employ the ARDL framework as proposed by Pesaran, Shin, and
Smith (2001). The choice of this approach rather than the conventional approaches
such as the Johansen (1988) and Johansen and Juselius (1990) is based on several
considerations. The ARDL technique provides consistent results irrespective of
the order of integration of the variables under study; either I(1) or I(0) or a mixture
of both (but not orders greater than one). It provides unbiased estimates of the
long-run model and reliable t-statistics though some of the regressors might be
endogenous. The ARDL estimation technique is very efficient in cases of rela-
tively small samples as in the case of this study. This technique also allows for the
introduction of optimal lags of both the dependent and explanatory variables;
implying that, various variables are allowed to have their optimal speed of adjust-
ment to equilibrium. A generalized representation of the ARDL model is specified
as follows:
k k

∆Q = αO + δ1t + δ2 Qt−1 + δ3 X t−i + ∑βi ∆Qt−i + ∑αi ∆X t−i + εt (5)

t=1 i=1

where Q is the dependent variable; X is a vector of explanatory variables; t repre-

sents the time period and e represents the error term.
The F-test statistic is used in checking the existence of a long-run equilib-
rium relationship among the variables under study. The null hypothesis for no
cointegration among the variables is represented as H0: d1 = d2 = d3 = 0 whereas
190 Foreign Trade Review 54(3)

the alternative hypothesis is represented by H1: d1 ≠ d2 ≠ d3 ≠ 0. The test involves

the use of critical value bounds which depends on the order of integration of the
variables. As a decision rule, if the estimated F-statistic exceeds the upper bound
critical value, the null hypothesis is rejected. We then conclude that the series
are cointegrated and thus there exists a long-run equilibrium relationship among
the variables irrespective of the order of integration and otherwise if the esti-
mated F-statistic is below the lower bound critical value. However, if the test
statistic lies in-between the upper bound critical value and the lower bound
critical value, a conclusive inference cannot be made. In this case, a negative
and statistically significant error correction term is often relied on. If the varia-
bles are found to be cointegrated, it is necessary to estimate the parameters that
characterize the long-run equilibrium relationship between trade balance and its
covariates. The next step in the estimation strategy involves the error-correction
model (ECM) within the ARDL framework which yields estimates of the param-
eters of the short-run and long-run relationships.

Variance Decomposition
To investigate the dynamic interactions between trade balance and its explanatory
variables, we further employ the variance decompositions (VDC) within the vec-
tor autoregressive (VAR) framework. The VDC technique allows us to estimate
the proportion of the variations in trade balance that are due to own innovations
(or shocks) as well as shocks in other variables in the system. In stimulating a
standard VDC, the orthogonalized responses are normally used where the under-
lying shocks of the VAR model are orthogonalized using the Cholesky decompo-
sition. However, Lutkepohl (1991) and Laurens and Cordoso (1998) state that this
approach depends on the ordering of variables in the VAR model when the cor-
relation between the variables is large. We, therefore, adopt an alternative method–
the generalized forecast error variance as proposed by Pesaran and Shin (1998)
and Koop, Pesaran, and Potter (1996). The major advantage of the generalized
forecast error variance over the orthogonalized forecast error variance analysis is
that the order of the variables is distinctively determined in the VAR system such
that the order of the variables in the VAR specification does not matter. This
approach has been used by various authors including, Narayan (2004), Onafowora
and Owoye (2006) and Owoye and Onafowora (2011).

Results and Discussion

Stationarity Test Results

To ensure that the evidence attesting to the presence or otherwise of a unit root in
the individual series are valid, this article employs both the Augmented Dickey-
Fuller (ADF) and Phillips-Perron (PP) tests to investigate the order of integration
of the variables. As reported, the ADF and PP stationarity test results show that
Akoto and Sakyi 191

regardless of the inclusion of trend or not the variables are either integrated of
order zero (I(0)) or order one (I(1)) (see Table 1). This outcome of mixed integra-
tion and absence of I(2) variable further justifies the use of the ARDL bounds test
approach to cointegration framework in this article.

Cointegration Test Results

The symmetric ARDL model selection of (2, 2, 1, 2, 1, 2, 0) for Model 1 and the
asymmetric ARDL model selection of (1, 1, 2, 2, 1, 2, 2, 2) for Model 2, respec-
tively, are based on the Schwartz Bayesian Information Criterion (SBIC).5 The
bounds test results reported show that there exists a stable long-run relationship
among the variables in our models (see Table 2). The computed F-statistic for each
model is greater than the upper bound critical value at 5 per cent level of signifi-
cance. Hence the null hypothesis of no cointegration is rejected and a conclusion
of an equilibrium long-run relationship among the variables is established.

Table 1. Stationarity Tests

ADF Test
Levels First Difference
Variables Trend No Trend Trend No Trend
LnTB −2.877 −3.024 −7.739*** −7.675***
LnrE −4.639*** −4.445*** – –
LnHCEX −2.880 −2.489 −5.737*** −5.843***
LnY* −4.767*** −4.711*** – –
LnGCEX −3.203 −2.778 −5.247** −5.353***
LnDP −1.015 −2.724 −3.406** −3.033**
LnMS −2.396 −1.727 −6.268*** −6.330***
PP Test
Levels First Difference
Variables Trend No Trend Trend No Trend
LnTB −2.771 −2.933 −7.864*** −9.457***
LnrE −6.749*** −4.358*** – –
LnHCEX −2.958 −2.486 −7.132*** −7.042***
LnY* −4.809*** −4.777*** – –
LnGCEX −3.270 −2.856 −9.401*** −8.788***
LnDP 0.306 −3.082** –5.325*** –
LnMS −2.383 −1.721 −6.349*** −6.289***
Source: The authors.
Note: *** and ** denote significance at 1 per cent and 5 per cent, respectively.
192 Foreign Trade Review 54(3)

Table 2. Results of Cointegration Relationship

Test-Statistics
Model 1 Model 2
F-statistics 6.962** 6.288**
Source: The authors.
Notes: ** indicates the rejection of the null hypothesis at 5 per cent significance level. The ARDL
model specification gives the 95 per cent upper and lower bound critical values for Models
1 and 2 as 3.61 and 2.45 and 4.5 and 2.87 respectively.

Long-run and Short-run Relationship

Having established the existence of cointegration, we now report and discuss the
estimates of the parameters that characterize the response of trade balance to vari-
ations in the determining factors in both the long- and short-run. The estimated
long- and short-run coefficients are reported in Tables 3 and 4, respectively. In
each Table, we report results related to Model 1 and Model 2.
The results from Model 1 show that, in the long run, there is a positive but
insignificant relationship between the real exchange rate and trade balance.
Though exchange rate obtains it’s expected sign, its insignificance contradicts
theoretical expectation, which notes that real depreciation in the exchange rate
(signified by a rise in rE) is associated with an improvement in trade balance.
Consequently, the Marshall-Lerner condition under which real depreciation will
lead to an improvement in trade balance in the long run does not hold in Ghana.
Further, the short-run coefficient of the real exchange rate is found to be nega-
tive and significant in Model 1 (see Table 4). This implies that an increase in the
real exchange rate leads to a deterioration of the trade balance in the short run.
The sign of the short-run coefficient of real exchange rate confirms the predic-
tion of the J-curve effect. However, the long-run result suggests that the J-curve
effect is not fully supported. This outcome is not surprising as several authors
have concluded the same. Saruni (2007) and Duasa (2007) confirm the absence
of the Marshall-Lerner condition in the case of Tanzania and Malaysia, respec-
tively. Similarly, Tran and Dinh (2014) also find the absence of the Marshall-
Lerner condition in some selected Asian countries. Singh (2004), Sulaiman and
Abdul-Rahim (2014), Tutueanu (2015) and Bahmani-Oskooee et al. (2016) also
confirm the absence of the J-curve effect in India, Thailand, Romanian and the
UK, respectively.
With regard to Model 2, we find that in both cases the coefficients of POS and
NEG have the same signs and sizes. The outcome is not different when we con-
sider the corresponding short-run results (Model 2). It can thus be concluded that
the relationship between the real exchange rate and Ghana’s trade balance is not
asymmetric. Long-run results from Model 2 shows that depreciation of the Ghana
cedi will have a favourable effect on the trade balance, however, this is insignifi-
cant. It can, therefore, be concluded that in this specification, we also find no
evidence of the Marshall-Lerner condition and the J-curve effect.
Akoto and Sakyi 193

Table 3. Estimated Long-run Coefficients

Dependent Variable LnTB

Regressors Model 1 Model 2
LnrE 0.2759
(0.169)
LnHCEX −3.1210*** −2.2474***
(0.5962) (0.4908)
LnGCEX −0.5364** −0.4853**
(0.222) (0.2243)
LnY* 0.6597** 0.3425
(0.2711) (0.2104)
LnMS 0.0009 0.00005
(0.007) (0.0050)
LnDP −0.0747* −0.0293
(0.0398) (0.1370)
POS 0.0718
(0.5684)
NEG 0.1122
(0.2814)
Constant 11.4111*** 9.8915**
(3.158) (3.4643)
Source: The authors.
Notes: ***, ** and * represent significance at 1 per cent, 5 per cent and 10 per cent, respectively.
Standard errors are in parenthesis.

Table 4. Estimates of Short-run Error Correction Representation

Dependent Variable: LnTB

Regressors Model 1 Model 2
DLNTB–1 −0.2739*
(0.1269)
DLnrE −0.3049**
(0.1291)
DLnrE–1 −0.1431
(0.0857)
DLnHCEX −1.5937*** −1.1113***
(0.2271) (0.2699)
DLnGCEX −0.4330*** −0.3384**
(0.0916) (0.1432)
DLnGCEX–1 −0.4697*** −0.4366***
(0.1044) (0.1061)
(Table 4 Continued)
194 Foreign Trade Review 54(3)

(Table 4 Continued)

Dependent Variable: LnTB

Regressors Model 1 Model 2
DLnY* 0.0030 −0.3261*
(0.1331) (0.1759)
DLnY*–1 0.0876
(0.1028)
DLnMS −0.0045 −0.0040
(0.0028) (0.0030)
DLnMS–1 −0.0059*
(0.0033)
DLnDP −0.0474** −0.4873*
(0.0215) (0.2333)
DLnDP–1 0.5945**
(0.2361)
DPOS 0.0999
(0.4738)
DPOS–1 −0.7571
(0.4498)
DNEG −0.6680**
(0.2330)
DNEG–1 −0.1686
(0.1260)
DECM–1 −0.6354*** −0.8565**
(0.1512) (0.2001)
Diagnostic Tests
Serial Correlation 0.8889 1.1935
[0.4387] [0.3581]
Heteroskedasticity 0.8889 1.1935
[0.4387] [0.3581]
Normality 0.8889 1.1935
[0.4387] [0.3581]
Functional Form 0.8889 1.1935
[0.4387] [0.3581]
Source: The authors.
Notes: ***, ** and * represent 1 per cent, 5 per cent and 10 per cent level of significance
respectively (…) are the standard errors and […] are prob. values.

We now analyze the long- and short-run effects of other macroeconomic vari-
ables including, household consumption expenditure, government consumption
expenditure, money supply, foreign income and domestic prices on the trade bal-
ance in both models.
Akoto and Sakyi 195

Unsurprisingly, household consumption expenditure is found to have a delete-

rious effect on trade balance both in the long- and short-run. These negative
effects are statistically significant at 1 per cent level. This outcome is not surpris-
ing because household consumer goods comprise both domestic and foreign prod-
ucts. Hence, an increase in the consumption expenditure of local residents
increases the demand for both domestic and foreign goods. Further, with a large
component of household consumables in Ghana being imported products (includ-
ing food products, clothing, vehicles and spare parts, second-hand products and
electronics), an increase in household expenditure is expected to increase imports
relative to exports and therefore worsens the trade balance. To add, because of low
levels of domestic production and associated high cost, Ghanaians tend to offset
the excess domestic demand through the importation of foreign consumer goods.
Given the high preference of Ghanaians for foreign goods, an increase in house-
hold consumption expenditure tends to exacerbate the deterioration in the coun-
try’s trade balance position. Similar results are reported by Saruni (2007) and
Shawa and Shen (2013) for Tanzania, Mohammad (2010) for Pakistan and
Sarbapiya (2012) for India.
Similar to the household consumption expenditure, the long- and short-run
results show that government consumption expenditure also exerts a negative and
significant effect on trade balance. Given the fact that government spending is
made up of expenditure on domestic and foreign goods, an increase in govern-
ment consumption expenditure may be diverted to either domestic or foreign
goods. Similarly, expenditures by the government can either be diverted into pro-
ductive ventures such as investing in productive sectors in order to promote
exports or non-productive areas such as the purchase of luxurious cars, military
goods, machines and equipment, mostly not produced domestically. Therefore,
deterioration in trade balance may emerge from increased government spending
on foreign-made durable and non-durable goods that incessantly divert funds out
of the country with its attendant adverse effects on trade balance. Justifiably, a
chunk of government consumption expenditure is diverted to the provision of
inputs for public goods including road construction, health care, education and
agriculture. Given that these inputs are mainly not produced domestically, the
country relies on foreign substitutes. Nienga (2010), Shawa and Shen (2013) and
Kayhan et al. (2013) also note that rising government spending worsens trade bal-
ance stance of Kenya, Tanzania and Turkey, respectively.
As expected, a boost in foreign income is found to significantly improve the
trade balance in the long-run. This implies that an increase in foreign income
increases the quantity of exports goods demanded by the country’s major trading
partners from the rest of the world including exports from Ghana. UNCTAD
(1999) notes that the acceleration of the growth rate of industrialized countries
improves upon the trade balance in most developing countries. However, the
short-run results depend on the model considered. Again, the long-run effect of
foreign income in the asymmetric specification is statistically insignificant.
Studies by Mohammad (2010), Korap, Alacahan, and Irhan (2011), Sarbapiya
(2012) and Shawa and Shen (2013) confirm similar results in Pakistan, Turkey,
India and Tanzania, respectively.
196 Foreign Trade Review 54(3)

Money supply is found to exert a positive but insignificant effect on the trade
balance in the long-run. On the other hand, the short-run results also show that
money supply is negatively related to trade balance. However, this effect is also
insignificant. The short-run results further reveal that the past value of money
supply is negatively related to the trade balance but marginally significant (see
Table 4). Contrary to our results, Duasa (2007), Nienga (2010) and Waliullah et
al. (2010) find support for the monetary approach to the balance of payments for
Malaysia, Kenya and Pakistan, respectively.
Finally, the domestic price also shows up to be another long-run determinant of
trade balance in Ghana. The estimated long-run coefficients of domestic price are
negative and marginally significant in the long- and short-run. As one would
expect, this outcome implies that high domestic prices (due to persistent inflation-
ary pressures) make domestic or exported products relatively expensive to con-
sumers both at home and abroad. Subsequently, this results in increasing domestic
demand for cheaper imported substitutes by Ghanaians and reduction in foreign
demand for Ghana’s exports; the negative substitution and income effects of high
domestic prices reinforce each other. The end result is a persistent deterioration in
the trade balance stance of the country. As a matter of fact, trends in the growth of
Ghana’s domestic prices (inflation) over the years have been very volatile with
most periods recording relatively high levels of growth in domestic prices.
According to Doe (2012), the country has over the years wallowed in high levels
of inflation as it records an end of the year inflation rate of 24.6 per cent and 40.5
per cent for 1986 and 2000, respectively. The situation is not different in recent
times as the inflation rate has been on the rise. Specifically, in 2011, 2012, 2013,
2014 and 2015, the country recorded a year-on-year inflation rate of 8.7 per cent,
9.2 per cent, 11.6 per cent, 15.5 per cent and 17.70 per cent, respectively (The
World Bank, 2015). In addition, this result is not surprising because due to high
cost of production in the economy coupled with upward adjustments of the prices
of petroleum products, the prices of most locally produced goods are relatively
higher as compared to the prices of foreign imported goods. The short-run dynam-
ics also show evidence of a positive relationship between the lagged effect of
domestic price and trade balance in the short run (Model 2). This result is consist-
ent with findings of Mohammad (2010), Sarbapiya (2012), Shawa and Shen
(2013) for Pakistan, India and Tanzania, respectively.
In order to test the statistical adequacy of the models, we conduct diagnostic
and stability tests. In this vein, we test for the presence or otherwise absence of
misspecified functional form, normality, heteroskedasticity and autocorrelation in
the ARDL model and non-linear ARDL specification. As evident by the diagnos-
tic test results reported (see Table 4), the estimated models are free of any diag-
nostic problem. The estimated ECT−1 coefficients are negative and statistically
significant at 1 per cent error level for all specifications (see Table 4). The nega-
tive and significant coefficients of the ECT−1 confirm the cointegration results.
The result shows a high speed of adjustment of convergence to the long-run equi-
librium every year after a short-run shock. Specifically, equilibrium, in the long-run,
will adjust by approximately 64 per cent and 86 per cent in Models 1 and 2,
Akoto and Sakyi 197

respectively, every year after any shock observed in the short run. Lastly, both
specifications of trade balance equations are found to be stable over the sample
period. This is confirmed by the plots of the CUSUM and CUSUMSQ tests as
proposed by Brown, Durbin and Evans (1975) and suggested by Pesaran et al.
(2001) within the ARDL framework. In both specifications, the CUSUM and
CUSUMSQ residuals lines lie within the 5 per cent critical value bounds (see
Appendix B) implying stability in the models. Hence, it can be concluded that the
regression equations are stable throughout the sample period considered.

Variance Decomposition Results

We conduct the generalized VDC for both Models within an unrestricted VAR
model as proposed by Koop et al. (1996) and Pesaran and Shin (1998). From an
estimated VAR model and in order to convey a sense of dynamics, we present the
VDC at 1, 4, 6, 8 and 10 periods (see Table 5).
It is evident from the VDC results that within the 10-year horizon, the forecast
error variance of trade balance is as a result of its own shocks. Specifically, in the
first period, 76.24 per cent of its forecast error variance is attributed to its own
shocks but its contributions gradually fell to 51.95 per cent by the tenth period. In
terms of innovations in the explanatory variables, innovations in household con-
sumption expenditure contribute more to the forecast error variance of trade bal-
ance as compared to the other variables over the horizon. In the same vein,
innovations in the exchange rate, foreign income, domestic prices, money supply
and government consumption expenditure contribute sequentially to the forecast
error variance of trade balance over the specified time horizon. The fluctuations
in household consumption expenditure explain 31.20 per cent of the forecast error
variance in the trade balance after 10 years. Innovations in exchange rate contrib-
ute 23.75 per cent, while shocks in foreign income, domestic prices, money sup-
ply, and government consumption expenditure contribute 16.02 per cent, 10.74
per cent, 10.31 per cent, and 6.09 per cent respectively.

Table 5. Variance Decomposition Results

Variance Decomposition of LnTB

a) % of Forecast Variances Explained by Innovations in
Horizon LnTB LnrE LnHCEX LnGCEX LnY* LnMS LnDP
1 76.235 17.031 36.388 5.537 3.029 12.733 9.668
4 56.094 20.799 30.598 7.068 13.371 11.084 10.595
6 54.269 22.645 30.492 6.868 14.884 10.289 11.035
8 52.493 23.963 30.680 6.727 16.177 9.796 10.826
10 51.950 23.754 31.198 6.092 16.020 10.305 10.737
Source: The authors.
198 Foreign Trade Review 54(3)

The VDC substantiate the significant role played by household consumption

expenditure, exchange rate, foreign income, money supply, domestic prices and
government consumption expenditure in accounting for fluctuations in forecast
error variance of Ghana’s trade balance over the time horizon within the sample
period considered. In terms of explanatory power, innovations in government
consumption expenditure explain very little of the forecast error variance of trade
balance as compared to other variables. The implication is that government con-
sumption expenditure contributes very little to trends in Ghana’s trade balance
over the sample period. Nevertheless, the portion of trade balance variations
accounted for by most of the explanatory variables increases continuously over
the time horizon of which the percentage of forecast error variance in the trade
balance is highly accounted for by innovations in household consumption expend-
iture as it maintains the highest percentage than the other variables. This also
implies that changes in the country’s trade balance are highly attributable to
shocks from household consumption expenditure.

Concluding Remarks
This study analyzes the determinants of trade balance in post-liberalization Ghana.
In order to achieve this, it tests for the presence or absence of the Marshall-Lerner
condition and the J-curve effect by investigating the response of trade balance to
movements in the real exchange rate. It does this in two ways, first by including
the real exchange rate in the model and second by decomposing real exchange
rate into POS (positive) and NEG (negative) in the model. Further, the article
examines the impact of other macro-economic variables on trade balance. Finally,
it investigates how innovations in the explanatory variables and own shocks con-
tribute to the forecast error variance of the trade balance. The bounds testing
approach to cointegration and the error correction model within the ARDL frame-
work is used as the estimation strategy in symmetric and asymmetric specifica-
tions. It also employs variance decomposition to analyze the dynamic interactions
of the variables. We find evidence of both long-run relationship and short-run
dynamics among the variables considered. We also find the absence of the
Marshall-Lerner condition and J-curve effect in Ghana in both estimations (thus
symmetric and asymmetric models). Again, the variance decompositions show
that household consumption expenditure contributes highly to the forecast error
variance of trade balance.
Given this, urgent measures need to be considered in order to help improve
upon the country’s trade balance position to hasten economic growth and devel-
opment. From the results obtained, it can be noted that depreciation of the Ghana
cedi is not an appropriate step to help improve upon the country’s trade balance
position. Unlike in strong economies such as the industrialized ones where mar-
ket forces can operate on their own to produce self-correcting forces in case of
Akoto and Sakyi 199

any currency depreciation hence efficiently controlling the trade balance, for
most developing countries like Ghana, self-correcting measures might not often
work in terms of controlling the trade balance. Therefore, other measures of cur-
rency stabilization could be adopted to improve the trade balance as a stable
exchange rate would enable producers of tradable goods to make long-term
investment plans. Again, policies to divert household consumption expenditures
to domestic goods could also be adopted through the promotion of various made-
in-Ghana goods. Also, government policies aimed at establishing companies
(preferably through public-private partnerships) purposely for the production of
export-oriented goods in order to meet export demand could be implemented.
Finally, the government should embark on policies that would create an enabling
environment for businesses in the country to produce at a relatively lower cost.
The results from the present study notwithstanding, it is important to note that
this study is limited to only macroeconomic factors that affect the trade balance.
Given this, further studies may try to assess other micro-factors that might affect
the trade balance.

Appendix A
.4
.2
Trade Balance (Log)
-.2 -.4
-.6 0

1960 1980 2000 2020

Year

Figure 1A. Trend of Ghana’s Trade Balance

Source: Data from World Bank Development Indicators (WDI) (2016).
Appendix B
Stability Tests

Model 1a. Model 1b.

12 1.6

8
1.2

4
0.8

0.4
-4

0.0
-8

-12 -0.4
03 04 05 06 07 08 09 10 11 12 13 14 15 03 04 05 06 07 08 09 10 11 12 13 14 15

CUSUM 5% Signific anc e CUSUM of Squares 5% Signific anc e

Figure 2B. CUSUM AND CUSUMSQ Tests for Parameter Constancy

Source: The authors.
 Model 2a. Model 2b.
10.0 1.6

7.5
1.2
5.0

2.5
0.8

0.0

0.4
-2.5

-5.0
0.0
-7.5

-10.0 -0.4
2007 2008 2009 2010 2011 2012 2013 2014 2015 2007 2008 2009 2010 2011 2012 2013 2014 2015

CUSUM 5% Significance CUSUM of Squares 5% Significance

Figure 3B. CUSUM AND CUSUMSQ Tests for Parameter Constancy

Source: The authors.
202 Foreign Trade Review 54(3)

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research,
authorship and/or publication of this article.

Funding
The authors received no financial support for the research, authorship and/or publication
of this article.

Notes
1. This outcome is often attributed to the recovery of the cocoa sector.
2. This index as defined by the International Monetary Fund (IMF) is a measure of the
value of a currency against a weighted average of several foreign currencies divided by
a price deflator or index of costs.
3. The countries considered include Australia, Austria, Belgium, Canada, Chile, Czech
Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland,
Israel, Italy, Japan, Korea, Luxembourg, Mexico, Netherlands, New Zealand, Norway,
Poland, Portugal, Slovak Republic, Slovenia, Spain, Sweden, Switzerland, Turkey,
Unites Kingdom and Unites States of America.
4. Bahmani-Oskooee (2001), Narayan (2004), Saruni (2007) and Ju, Wu, and Zeng (2010)
have also used similar measure.
5. The Schwartz Bayesian Information Criterion (SBIC) is used because it provides the
most parsimonious results. We have considered a maximum lag length of 2 for these
models.

References
Alexander, S. S. (1952). Effects of a devaluation on a trade balance (International Monetary
Fund Staff Papers No. 2, pp. 263–278). International Monetary Fund, Washington DC.
———. (1959). Effects of a devaluation: A simplified synthesis of elasticities and
absorption approaches. American Economic Review, 49(1), 22–42.
Aliyu, A. J., & Tijani, S. M. (2015). Asymmetric cointegration between exchange rate and
trade balance in Nigeria. Cogent Economics & Finance, 3(1), 1045213.
Bahmani-Oskooee, M. (1991). Is there a long-run relation between the trade balance and
the real effective exchange rate of LDCs? Economics Letters, 36(4), 403–407.
———. (2001). Nominal and real effective exchange rates of middle eastern countries and
their trade performance. Applied Economics, 33(1), 103–111.
Bahamani-Oskooee, M., & Bahmani, S. (2015). Non-linear ARDL approach and the
demand for money in Iran. Economics Bulletin, 35(1), 1–11.
Bahmani-Oskooee, M., & Fariditavana, H. (2015). Nonlinear ARDL approach, asymmetric
effects and the J-curve. Journal of Economic Studies, 42(3), 519–530.
———. (2016). Nonlinear ARDL approach and the J-curve phenomenon. Open Economies
Review, 27(1), 51–70.
Bahmani-Oskooee, M., Halicioglu, F., & Ghodsi, S. H. (2016). Asymmetric effects of
exchange rate changes on British bilateral trade balances. Retrieved from https://mpra.
ub.uni-muenchen.de/73477/
Bank of Ghana. (2008). Trade balance and policy efficiency: Evidence from Ghana
(Working Paper WP/BOG 2009/02). Bank of Ghana, Accra.
Bhagwati, J., & Srinivasan, T. N. (2002). Trade and poverty in the poor countries. American
Economic Review, 92(2), 180–183.
Akoto and Sakyi 203

Bhattarai, K. R., & Armah, M. K. (2005). The effects of exchange rate on the trade
balance in Ghana: Evidence from cointegration analysis (Research Memorandum 52).
Cottingham: Business School, University of Hull.
Brown, R. L., Durbin, J., & Evans, J. M. (1975). Techniques for testing the constancy of
regression relationship over time. Journal of Royal Statistical Society, 37(1), 149–163.
Catao, L., & Falcetti, E. (2002). Determinants of Argentina’s external trade. Journal of
Applied Economics, 5(1), 19–57.
Danquah, B. A. (2008). The effects of exchange rate on Ghana’s external trade (Master’s
thesis). Department of Economics, Kwame Nkrumah University of Science and
Technology, Kumasi, Ghana.
Delatte, A. L., & López-Villavicencio, A. (2012). Asymmetric exchange rate pass-through:
Evidence from major countries. Journal of Macroeconomics, 34(3), 833–844.
Doe, S. K. (2012). The relationship between inflation, inflation uncertainty and interest rate
in Ghana (Master’s thesis). Department of Economics, Kwame Nkrumah University of
Science and Technology, Kumasi, Ghana.
Duasa, J. (2007). Determinants of Malaysian trade balance: An ARDL bound testing
approach. Journal of Economic Cooperation, 28(3), 21–40.
Falk, M. (2008). Determinants of the trade balance in industrialized countries (FIW
Research Report No. 013). FIW - Research Centre International Economics, Vienna.
Goldstein, M., & Khan, M. S. (1985). Income and price effects in foreign trade. In R. W.
Jones & P. B. Kenen (Eds.), Handbook of international economics (pp. 1041–1105).
North-Holland, Amsterdam.
Harberger, A. C. (1950). Currency depreciation, income and the balance of trade. Journal
of Political Economy, 58(1), 47–60.
Iyke, N. B., & Ho, S. (2017). The real exchange rate, the Ghanaian trade balance, and the
J-curve. Journal of African Business, 18(3), 380–392.
Johansen, S. (1988). Statistical Analysis of Cointegration Vectors. Journal of Economic
Dynamics and Control, 12(2–3), 231–254.
Johansen, S., & Juselius, K. (1990). Maximum Likelihood Estimation and Inference
on Cointegration: With Applications to the Demand for Money. Oxford Bulletin of
Economics and Statistics, 52(2), 169–210.
Ju, J., Wu, Y., & Zeng, L. (2010). The impact of trade liberalization on the trade balance in
developing countries. IMF Staff Working Papers, 57(2), 427–449.
Karam, F., & Zaki, C. (2015). Trade volume and economic growth in the MENA region:
Goods or services? Economic Modelling, 45(C), 22–37.
Kayhan, S., Bayat, T., & Yüzbaşı, B. (2013). Government expenditures and trade deficits
in Turkey: Time domain and frequency Domain analyzes. Economic Modelling, 35(C),
153–158.
Ketenci, N. (2017). The effect of the European Union Customs Union on the balance of
trade in Turkey. Foreign Trade Review, 52(4), 219–232.
Khan, M. Z. S., & Hossain, M. I. (2012). Determinants of trade balance of Bangladesh: A
dynamic panel data analysis. Bangladesh Development Studies, 35(2), 45–65.
Killick, T. (2010). Development economics in action (2nd ed.). Milton Park, Abingdon,
Oxon: Routledge.
Kim, M., & Beladi, H. (2005). Is free trade deflationary? Economics Letters, 89(3), 343–349.
Koop, G., Pesaran, M. H., & Potter, S. M. (1996). Impulse response analysis in nonlinear
multivariate models. Journal of Econometrics, 74(1), 119–147.
Korap, L., Alacahan, D. N., & Irhan B. H. (2011). An empirical model for the Turkish trade
balance: New evidence from ARDL bounds testing analyses. Ekonometri ve İstatistik
Sayı, 14, 38–61.
204 Foreign Trade Review 54(3)

Korkmaz, A., & Bilman, M. E. (2017). The S-curve behaviour of the trade balance: A
stepwise procedure. Foreign Trade Review, 52(1), 1–14.
Lal, A. K., & Lowinger, T. C. (2002). The J-curve: evidence from East Asia. Journal of
Economic Integration, 397–415.
Laurens, B., & Cardoso, J. (1998). Managing capital flows: Lessons from the experience
of Chile (International Monetary Fund Working Paper No. 168). Washington, DC:
International Monetary Fund, Monetary and Exchange Affairs Department.
Lotfalipour, M. R., Montazeri, S., & Sedighi, S. (2013). Trade openness and inflation.
Evidence from MENA region countries. Economic Insights-Trends and Challenges,
2(65), 1–11.
Lutkepohl, H. (1991). Introduction to multiple time series analysis. New York, NY:
Springer-Verlag.
Meade, J. E. (1951). The balance of payments. Oxford: Oxford University Press.
Mohammad, D. S. (2010). Determinants of balance of trade: Case study of Pakistan.
European Journal of Scientific Research, 41(1), 13–20.
Narayan, K. P. (2004). New Zealand’s trade balance: Evidence of the J-curve and Granger
causality. Applied Economics Letters, 11(6), 351–354.
Nienga, E. (2010). The determinants of trade balance in Kenya: An empirical analysis,
1970–2010. European Journal of Scientific Research, 31(3), 14–20.
Onafowora, A. O., & Owoye, O. (2006). An empirical investigation of budget and trade
deficits: The case of Nigeria. The Journal of the Development of Areas, 39(2), 153–174.
Osoro, K. (2013). Kenya’s foreign trade balance: An empirical investigation. European
Scientific Journal, 9(19), 1857–7881.
Owoye, O., & Onafowora, O. A. (2011). The relationship between tax revenues and
government expenditures in European Union and Non-European Union OECD
countries. Public Finance Review, 39(3), 429–461.
Pesaran, H., & Shin, Y. (1998). Generalized impulse response analysis in linear multivariate
models. Economic Letters, 58(1), 17–29.
Pesaran, M. H., Shin, Y., & Smith, R. J. (2001). Bounds testing approaches to the analysis
of level relationships. Journal of Applied Econometrics, 16(1), 289–326.
Rodrik, D. (1999). Where did all the growth go? External shocks, social conflict, and
growth collapses. Journal of Economic Growth, 4(4), 385–412.
Rose, A. K., & Yellen, J. L. (1989). Is there a J-curve? Journal of Monetary Economics,
24(1), 53–68.
———. (1991). The role of exchange rates in a popular model of international trade: Does
the ‘Marshall-Lermer’ condition hold? Journal of International Economics, 30(3–4),
301–316.
Sakyi, D. (2011). Trade openness, foreign aid and economic growth in post-liberalization
Ghana: An application of the ARDL bounds test. Journal of Economics and
International Finance, 3(3), 146–156.
Sakyi, D., Villaverde, J., & Maza, A. (2015). Trade Openness, Income Levels, and
Economic Growth: The Case of Developing Countries, 1970-2009. The Journal of
International Trade and Economic Development, 24(6), 860–882.
Sarbapiya, R. (2012). An analysis of determinants of balance of trade in India. Research
Journal of Finance and Accounting, 3(1), 73–83.
Saruni, M. (2007). Determinants of the trade balance in Tanzania, 1970–2002. In Awarded
theses 2006 (pp. 69–88). Washington, DC: The World Bank.
Shawa, M. J., & Shen, Y. (2013). Analysis of the determinants of trade balance: Case
study of Tanzania. International Journal of Business and Economics Research, 2(6),
134–114.
Akoto and Sakyi 205

Singh, T. (2002). India’s trade balance: the role of income and exchange rates. Journal of
Policy Modeling, 24(5), 437–452.
———. (2004). Testing J-curve hypothesis and analysing the effect of exchange rate
volatility on the balance of trade in India. Empirical Economics, 29(2), 227–2245.
Sulaiman, C., & Abdul-Rhim, A. S. (2014). J-curve effect and Thailand’s trade in forest
products: ARDL bounds testing. World Applied Science Journal, 29(1), 150–155.
Tandon, S. (2014). Trade balance and the real exchange rate: An empirical analysis of
multilateral and bilateral relationship. Foreign Trade Review, 49(2), 117–139.
The World Bank. (2015). World development indicators. Washington, DC: The World
Bank.
Tran, A. T., & Dinh, B. T. T. (2014). FDI inflows and trade imbalances: Evidence from
developing Asia. The European Journal of Comparative Economics, 11(1), 147–169.
Turkay, H. (2014). The validity of Marshall-Lerner condition in Turkey: A cointegration
approach. Theoretical and Applied Economics, 21(10), 21–32.
Tutueanu, G. (2015). Measuring the influence of the J-curve effect on trade in Romanian
forest product. Theoretical and Applied Economics, 22(4), 75–84.
United Nations Conference on Trade and Development (UNCTAD). (1999). Trade and
Development Report. New York and Geneva: United Nations Publication.
———. (2005). Developing countries in international trade: Trade and development
index. New York and Geneva: United Nations Publication.
———. (2013). Trade and development report. New York and Geneva: United Nations
Publication.
Varge, M. (2015). The impact of agricultural sector on employment and foreign trade
balance (Ministers Report). Hungary: Ministry of National Economy.
Verheyen, F. (2013). Interest rate pass-through in the EMU: New evidence using nonlinear
ARDL framework. Economics Bulletin, 33(1), 729–739.
Waliullah, K., Khan M., Kakar, R., & Khan W. (2010). The determinants of Pakistan’s
trade balance: An ARDL cointegration approach. The Lahore Journal of Economics,
15(1), 1–26.
Wilson, P., & Tat, K. C. (2001). Exchange rates and the trade balance: the case of Singapore
1970 to 1996. Journal of Asian Economics, 12(1), 47–63.
World Bank. (2016). World Development Indicators. Washington DC: The World Bank.