Petro Sauti Magai

Open Access article distributed in terms of the

Creative Commons Attribution License
[CC BY 4.0] ([Link]

AN ECONOMETRIC ANALYSIS OF TRADE AND ECONOMIC GROWTH IN

TANZANIA: EVIDENCE FROM TIME SERIES DATA
1
Petro Sauti Magai

ABSTRACT
This article investigates the relationship between trade and economic growth in Tanzania for the period from
1970 to 2016. The article utilises the Autoregressive Distributed Lag Model known as the ARDL bounds testing
to co-integration. In this article, it utilises a general-to-specific technique using the Ordinary Least Square
(OLS) method on estimates, to come up with significant variables. Foreign direct investment, population growth
and exchange rates were added to the model as explanatory variables. The empirical evidence confirms the
existence of a long-run relationship between selected variables, implying that in the long-run, all variables can
move together. The empirical results of the analysis reveal that exports, imports, foreign direct investment and
exchange rates have a robust and significant influence on economic growth in Tanzania. However, population
growth seems to have less insignificance compared to the other variables. As far as policy is concerned, the
government should revisit trade policy measures to control imports and minimise trade deficit. This will in turn
lead to momentous economic growth.

Keywords: Economic growth, trade, time series, econometric analysis, Tanzania.

INTRODUCTION
International competitiveness between countries has traditionally been assessed based on exports and market
shares. Thus, an increasing part of international trade involves the importation of intermediates to be integrated
into the export of final and further processed intermediate goods (Wastyn & Sleuwaegen, 2013). Therefore, a
country’s exports not only reflect the embodied technology and relative endowments which characterise its
domestic production activities, but also the technology and factor endowments of the partner countries from
which a partner country imports intermediate goods (Moussiegt et al., 2012; Wastyn & Sleuwaegen, 2013).
Several economic theories have tried to identify various channels which could facilitate growth effects. Apart
from trade being regarded as an engine for growth, it is also believed to promote the efficient allocation of
resources and allow a country to realise the economies of scale (Busse & Königer, 2012). With this, the role of
trade on economic growth has received considerable attention and several studies have been conducted to
determine the causal relationship between trade and economic grow (Makki & Somwaru, 2004). In Tanzania,
however, the sector has not received as much attention and it is difficult to find studies which quantify the
subject sufficiently. It is against this background that this article sought to analyse the relationship between trade
and economic growth in Tanzania for the period from 1970 to 2016. The rest of the article is organised in five
sections. The introduction is provided in Section 1.0, while Section 2.0 gives a brief picture of trade
performance in Tanzania. Section 3.0 presents a survey of literature together with theoretical models, while,
methodology, analysis and empirical findings are discussed in Section 4.0. The final part, Section 5.0, consists
of the conclusion and policy recommendations.

TRADE PERFORMANCE IN TANZANIA

Tanzania noted an increase in total trade of US$ 20.6 billion in 2015 from US$ 19.7 billion in 2014, an
approximate of 4.57% increase. The increase in total trade was facilitated by the increase of imports from US$
12.8 billion in 2014 to US$ 14.7 billion in 2015, which is an increase of almost 15.32%. This further led to an
increase in trade deficit by almost 51% from US$ 5.8 billion in 2014 to US$ 8.9 billion in 2015. India, China,
EU and Kenya were noted as the main trading partners (EAC, 2015). Tanzania also observed an increase in
imports by 16.6% from 12.8 billion in 2014 to US$ 14.7 billion in 2015. By 2015, Tanzania’s main import

1
Petro Sauti Magai (PhD), Lecturer in International Trade, University of Dar es Salaam Business School.
email: sauti@[Link] or sautimagai@[Link]
Business Management Review 21 (1), pp.74-84 ISSN 0856-2253 (eISSN 2546-213X) ©Jan-June 2018 UDBS. All rights of reproduction in
any form are reserved

74
 Petro Sauti Magai

countries were China, India and EU which recorded import values of US$ 2.9 billion, US$ 2.7 billion and US$
1.8 billion, respectively. These imports included mainly petroleum products, motor vehicles, wheat bran,
pharmaceuticals, chemical products, electrical equipment and machinery.

On the other hand, Tanzania recorded a 15.27% decrease in exports from US$ 6,909.6 million in 2014 to US$
5,854.25 million in 2015. It is difficult to quantify the causes of the huge percentage decrease in exports but
since there was a presidential election in 2015, the perceived instability during elections may have scared away
some traders. Nevertheless, the major exports were gold, cashew nuts, precious metals, tobacco, coffee, sesame
oil and yellow tuna. The exports amounted to US$ 1.1 billion. Of these, US$ 833 million’s worth of exports
were destined for India and SADC, respectively. Exports to COMESA countries and Japan were 5.85% and
3.94% of total exports, respectively. In 2015, the volume of re-exports increased from US$ 1.2 billion in 2014 to
US$ 2.0 billion. The share of re-exports to total exports increased by 17.3%, from 2014 to 2015. The re-
exported products among other things included light vessels, fire-floats, motor vehicles, electrical equipment,
spare parts, mineral fuels, fertiliser and machinery parts (EAC, 2015). Figure 1 shows the value of exports and
imports in Tanzania between 1970 and 2016. Likewise, in the beginning of 2017, Tanzania registered a surplus
balance of payment of US$ 636.7 million, though the country recorded a drop of its exports and imports. This
was a significant recovery from a deficit of US$ 183.9 million in 2016. The surplus was a result of current
account, which narrowed by half to a deficit of $1.6 billion due to a fall in imports. The annual import bill
decreased to US$ 7.8 billion in 2017 from the US$ 9.3 billion recorded in 2016 (BoT, 2017).

15000

10000

5000

-5000

-10000

Import Export Trade deficit

Figure 1: Tanzania's Exports and Imports (US$ at current price in millions)

LITERATURE SURVEY
Theoretical Models: International trade theories
Among the pioneers of trade theories are Ricardo and Heckscher-Ohlin (H-O), both of whom concentrated on
the determinants of global production. The Ricardo Model investigated the association between technology and
production location of multinational firms. The model envisaged that production location is acquired by the
divergence in labour productivity that may be grounded by the gap of production techniques between countries.
Thus, each country has to produce goods with relatively higher yields and to be able to import other goods.
Advanced technology increases productivity; that is, production is concentrated in regions with higher
technology. The H-O model argues that production location is beaconed on the endowment factors rather than
technical differences. Each country generates goods using available factors and sometimes exchanges goods
using its available resources via international trade. The model also examines the effects of the endowment
factor on production, location and decision, arguing that production is concentrated in regions with abundant
resources. However, the model has several challenges where technology cannot be acquired freely by any
business and also the factor endowment does not result in the factor price gap, as the factor price is equalised
through international trade. Reinert (2008) concluded that the Ricardian failure was a result of inappropriate
assumptions which always produced misleading answers.

Rybczynski Theory
This theory was developed by Rybczynski (1955) and investigates the effects of an increase in the quantity of a
factor of production against production, consumption and terms of trade within the context of the H-O Model.

75
 Petro Sauti Magai

The theory argues that an increase in the factor endowment causes an absolute expansion in the production of
goods and an absolute reduction in the production of the commodity using relatively little of the same factor.
The theory also argues that an increase in factor endowment is necessarily beneficial because a country can
export more, thus import more and consume more. However, Daniel (2000) calls this an export-biased trade
strategy, stating that this could worsen terms of trade by offsetting the positive impact of the increase in factor
endowment. Nevertheless, while emphasising the importance of increase in factor endowment on growth,
Colombatto (1900) argues that the export-biased trade strategy can play an important role in the growth process
of developing countries. Three points were tabled out to support his argument. First, growth of developing
countries depends considerably on industrialisation, though in most cases their development is low. Second, the
export promotion policies are not overly emphasised as the macro-economic factors are not conducive. Third,
exports make growth easier and lead to more savings, higher technological advancement and easier access to
foreign loans.

Endogenous growth theories

The Endogenous Growth Theory emphasises the importance of economic growth within the economic field.
Some of the main contributors to this theory are Romer (1986), Lucas (1988), and Mankiw et al. (1992). The
model proposes ways by which growth in less developed countries could be accelerated by making maximum
and efficient use of available resources. The theory aims at explaining both the degree of differences in
economic growth rate across countries and, a greater deal of the growth observed and technological
advancement as a form of capital accumulation. Generally, the theory states that the output per worker (growth
per unit of labour) increases with the output per capita (growth per unit of capital) with increasing rate. Capital
is assumed to include both human and physical capital. This theory capitalises on the demise of the Solow
Model. Solow (1956) had failed to explain how to determine the GDP growth rate (Brzezinski & Dzielinski,
2009). Nevertheless, endogenous theories can be used to determine the growth rate. Therefore, this article
employs the Endogenous Growth Model.

Selected Studies
The effects of trade have been analysed as a major factor for economic growth by many authors (Frankel &
Romer, 1999; Rodriguez & Rodrik, 2000). However, some of them noted the positive relationship between trade
and economic growth while others noted the opposite. Nevertheless, several studies by Were (2015) and
Edwards (1993) used cross-sectional data while defending their case on the effects of trade on economic growth
for different time periods. Studies by Musila and Yiheyis (2015), Lin (2000), and Trejos and Barboza (2015)
used time-series data while studies by Zahonogo (2016) and Eriṣ and Ulaṣan (2013) used panel data for their
analysis. Digging from various studies, as mentioned earlier, some have identified a positive relationship
between trade and economic growth. For example, Chang et al. (2009) investigated the effects of trade openness
on economic growth for the period of 1960 to 2000 for the sample size of 82 countries (22 developed and 60
developing countries). They employed a simple Harris-Todaro Model and used a non-linear growth regression
on empirical analysis and came up with positive results. Lin (2000) examining the association between trade and
economic growth in China for the period of 1952 to 1997 employed a regression on the Econometric Model. He
found out that export and import growth, together with the growth rate of the volume of trade, are positively
correlated to the growth rate of the GDP per capita. Were (2015) examined the differential effects of trade on
economic growth and investment based on cross-country data from 1991 to 2011 and a sample of 85 countries
(developed, developing and least developing countries) coupled with standard growth regression. He found out
that trade is largely consistent with the positive impact on economic growth though in LDCs, especially those in
Africa, the results become insignificant. Contributing to positivity, Kim (2011) used instrumental variable
threshold regressions while investigating whether trade contributes to the welfare of an individual in the long-
run or not. Nevertheless, he concluded that trade openness has a strong positive effect on growth, especially for
developing countries. Also, Jouini (2015) observed positive results while analysing the link between economic
growth and openness to international trade, between 1980 to 2010.

On the contrary, other studies have shown a negative relationship between trade and economic growth. Musila
and Yiheyis (2015) investigated the impact of trade openness on economic growth in Kenya from 1982 to 2009
using OLS regression on estimates. The study found that the aggregate trade openness negatively impacts
economic growth. Zahonogo (2016) examined the relationship between trade and economic growth in 42
countries within Sub-Saharan Africa (SSA) from 1980 to 2012. The study employed a pooled mean group
estimation technique and concluded that trade openness and economic growth do not move together in SSA.
Trejos and Barboza (2015) wrote a paper in regard to the dynamic estimation of the relationship between trade
openness and output growth in Asia, from 1950 to 2010. Coupled with a sample size of 23 Asian countries and
using both a static OLS and a dynamic ECM estimation model, the paper concluded that Asia’s economic-
growth miracle is inversely proportional to trade openness. Eriṣ and Ulaṣan (2013) using a Bayesian Model

76
 Petro Sauti Magai

averaged the estimate of cross-country growth regressions from 1960 to 2000, and found no evidence that trade
openness is directly correlated with economic growth in the long-run. Further, Ulaşan (2015) came up with
negative results while investigating the correlation between trade openness and economic growth.

METHODOLOGY, ANALYSIS AND FINDINGS

Data and Variables
To analyse trade and economic growth in Tanzania, six variables and time series data for the period between
1970 and 2016 from the UNCTAD database were employed. As far as empirical analysis of the data was
concerned, the following model was created and expressed in log form:

lnGDPt = α + β1lnEXPt + β2lnIMPt + β3lnPOPt + β4lnFDIt + β5lnEXRt + εt....................(1)

Whereby, economic growth (GDP) is represented by GDP total expressed in US dollars at constant prices in
millions. EXP stands for exports while IMP stands for imports and both are measured in US dollars at current
price, in millions. POP is the average growth in population, while FDI is the inward flow of foreign direct
investment expressed in US dollars at current price in millions. Lastly EXR is the exchange rate, α is the
intercept, β1 to β5 are the coefficients of the respective variables, while εt is the random error term.

Descriptive Data Analysis

The descriptive statistics in Table 1 exhibits the presence of low standard deviation for all variables in this
article. This signifies that most of the numbers in these variables are very close to the mean. It also appears that
three variables - lnGDP, lnEXP and lnIMP - are right-skewed, while lnPOP, lnFDI and lnEXR are negatively
skewed. Observing the Kurtosis of the data, the analysis exhibits that all variables are platykurtic (short-tailed)
except for lnPOP and lnFDI which are leptokurtic (long-tailed). A Jarque-Bera test of normality shows that the
residuals of lnGDP, lnIMP and lnFDI are normally distributed, while the remaining three variables are not
normally distributed. Also, the correlation coefficients show that some of the selected variables are positively
correlated and some are negatively correlated with each other. Nevertheless, some are highly correlated and
others are weakly correlated. For example, lnGDP is positive and strongly correlated with all variables except
for lnPOP which is negatively correlated. The details are provided in Table 1.

Table 1: Statistical Analysis of Selected Variables

lnGDP lnEXP lnIMP lnPOP lnFDI lnEXR
Mean 9.563674 6.670292 7.470440 1.116504 3.657141 4.990557
Median 9.460340 6.252624 7.281083 1.136610 2.995732 6.004563
Maximum 10.76724 8.621054 9.400256 1.217828 7.643627 7.685744
Minimum 8.705012 5.507257 5.763183 0.935356 -4.605170 1.948817
Std. Dev. 0.598603 0.986773 1.002667 0.066631 3.047306 2.225127
Skewness 0.481105 0.813783 0.629626 -1.346781 -0.654901 -0.341818
Kurtosis 2.009635 2.258488 2.431633 4.194852 3.178447 1.375556
Jarque-Bera 3.733897 6.264343 3.737985 17.00410 3.422038 6.082931
Probability 0.154595 0.043623 0.154279 0.000203 0.180682 0.047765
Correlation
lnGDP 1.000000
lnEXP 0.944868 1.000000
lnIMP 0.963065 0.957590 1.000000
lnPOP -0.183100 -0.053877 -0.014208 1.000000
lnFDI 0.796519 0.814960 0.728093 -0.373173 1.000000
lnEXR 0.906589 0.752270 0.809163 -0.354820 0.705014 1.000000
Source: Author's own computation

VAR Lag Order Selection Criteria

The lag length was chosen based on VAR Lag Order Selection Criteria method where lag length n was selected.
All five (5) lag selection criteria confirmed the selection of the lag length of 1 as shown in Table 2. It should be
noted that the lag order is chosen to avoid autocorrelation in the residual.

77
 Petro Sauti Magai

Table 2: VAR Lag Order Selection Criteria

Lag LogL LR FPE AIC SC HQ
0 -16.63450 NA 1.80e-06 0.961533 1.162273 1.036367
1 410.1153 132.3015* 1.01e-13* -15.78290* -13.57476* -14.95973*
2 322.5629 587.9421 1.56e-12 -13.00279 -11.79835 -12.55379
* lag order selected by the criterion
LR: sequential modified LR test statistic (each test at 5% level)
FPE: Final prediction error
AIC: Akaike information criterion
SC: Schwarz information criterion
HQ: Hannan-Quinn information criterion

Estimation and Testing Procedures

This sub-section discusses all relevant methods employed in this study. These methods or testing procedures
include unit root tests and ARDL approach (bounds test) to cointegration developed by Pesaran et al. (2001),
with the condition of the regression being purely I(0), purely I(1), or mutually co-integrated. A general to
specific method is employed only to determine long-run effects.

Unit root test

This article employs the Augmented Dickey-Fuller (ADF) unit root testing procedure (Dickey & Fuller, 1979)
and the Phillips Peron (PP) test (Phillips & Perron, 1988) to test the stationarity of the series for each variable.
This is the first step in examining the time series properties of the data by looking at the patterns and trend. For
both ADF and PP tests, the interest lies in determining the size of the coefficient β as observed in equation (2 &
3).
n
∆ Yt = α + β K t −1 +  δ i ∆ K t −1 + µ t + ε t .............................................(2)
j =1
The standard Dickey-Fuller Model has been augmented by ΔKt-i, where Yt represents a linear time trend, Δ is
the first difference operator, while β, δ and μ are parameters to be estimated. Based on VAR Lag Order
Selection Criteria method, the lag length 1 was chosen to avoid autocorrelation in the residual as shown in Table
2. The reason of using the PP test lies on its advantages of correcting for serial correlation and heteroskedasticity
in error terms, as the test is applied without including number of lags (Enders, 2015). Therefore, the equations
and hypothesis to be tested are similar to those of ADF; the only difference is that the PP test ignores a number
of lags and takes the following form:
∆ Yt = α + β K t −1 + µ t + ε t ......................................................(3)

Results for the unit root

The ADF and PP test results in Table 3 indicate that the lnGDP variable whose null hypothesis of the presence
of unit root at level, was rejected indicating that lnGDP is stationary at 1% level of significance. The remaining
variables become stationary at first difference. Backed up with these results, no other method of co-integration
was upheld rather than engulfing the ARDL approach which fulfils the basic two requirements. First, the
approach does not require all the variables to be integrated in the same order (Pesaran & Pesaran, 1997).
Second, the approach requires that no variables should be integrated into order two I(2) (Ouattara, 2004).
Therefore, the ARDL Model suffices for this article. Pesaran et al. (2001) concluded that the applicability of
ARDL is possible only if some variables are purely I(0) and purely I(1) or mutually integrated. Against this
background, the ARDL Model of co-integration is legitimately employed.

Table 3: Results for Unit Root Tests

Intercept (t) Trend and intercept (t) Intercept (t) Trend and intercept (t)
Variables Augmented Dickey-Fuller (ADF) test
At levels At first difference
lnGDP -2.397336 -5.698388*** -8.704786*** -8.734846***
lnEXP 1.522989 -0.485072 -6.039820*** -6.750824***
lnIMP 0.324183 -1.163867 -4.738450*** -4.768295***
lnPOP -2.361489 -2.382887 -2.977786 -4.008856***
lnFDI 2.160272 1.423978 -12.25043*** -4.720551***
lnEXR 4.526530 -0.209088 -3.767118*** -5.309890***
Phillips Perron (PP) test
lnGDP -3.478537** -4.647726*** -13.43961*** -14.78003***

78
 Petro Sauti Magai

lnEXP 1.533738 -0.488681 -6.039820*** -6.754633***

lnIMP -0.072145 -1.380615 -4.874390*** -4.835002***
lnPOP -2.105487 -2.130879 -2.656874 -4.175640***
lnFDI -0.753243 -2.689200 -12.40846*** -13.28669***
lnEXR 4.237014 0.489438 -3.775112*** -5.243159***
Source: Author's own computation
Note: MacKinnon's (1996) critical values used in the rejection of the null hypothesis of the unit root,
where ***, ** and * represent 1%, 5% and 10% respectively

Diagnostic Stability Test

It is crucial to perform an appropriate model diagnostic test before embarking on further econometric analysis.
The diagnostic checks performed in this analysis passed four major tests: serial correlation, heteroskedasticity,
normality, recursive residuals and the CUSUMSQ (cumulative sum of recursive residuals of square). These tests
were suggested by Pesaran and Pesaran (1997). This allows one to explore whether the assumptions of the
regression model are valid and will assist in deciding whether the subsequent inference results can be trusted.

Table 4: Results of the Diagnostic Test

Test statistic Prob Remarks
Breusch-Godfrey serial correlation LM test 0.925866 0.4054 Do not reject H0
Heteroskedasticity test: Breusch-Pagan-Godfrey 0.596425 0.7312 Do not reject H0
Normality (Jarque-Bera test) 0.044442 0.1323 Do not reject H0
Source: Author's own computation.

20

15

10

-5

-10

-15

-20
1980 1985 1990 1995 2000 2005 2010 2015

CUSUM 5% Significance
1.4

1.2

1.0

0.8

0.6

0.4

0.2

0.0

-0.2

-0.4
1980 1985 1990 1995 2000 2005 2010 2015

CUSUM of Squares 5% Significance

Figure 2: Plot of Cumulative Sum and Squares of Recursive Residuals

Much of the evidence from the diagnostic test results presented in Table 4 shows that there is no indication of
heteroskedasticity and misspecification in the model. Using the Breusch-Pagan-Godfrey test the hypothesis of
the presence of heteroskedasticity is rejected. Since the Jarque-Bera statistics and its corresponding probability
is more than 0.05, this confirms that the residuals are normally distributed. Also, the model is free from serial
correlation, as confirmed by Breusch-Godfrey serial correlation LM test. Nevertheless, the presence of the
cumulative sum inside two critical lines at 5% significant level, as reflected in Figure 2, signifies the stability of
the model. This gives the go-ahead for further analysis.

Co-integration Testing Using ARDL Approach

The ARDL Model (the bound test) for co-integration is employed in order to test the relationship between
variables. Using the dynamic model (eqn 4), the ordinary least square (OLS) method on estimation is used and
the results are presented in Table 5.

79
 Petro Sauti Magai

∆ ln GDPt = µ + λ ln GDPt −1 + δ 1 ln EXPt −1 + δ 2 ln IMPt −1 + δ 3 ln POPt −1 + δ 4 ln FDI t −1

p p p P
+δ 5 ln EXRt −1 +  α i ∆ ln GDPt − i +  β1i ∆ ln EXPt − i +  β 2 i ∆ ln IMPt − i +  β 3i ∆ ln POPt − i
i =1 i=0 i =0 i=0
p P
+  β 4 i ∆ ln FDI t − i +  β 5 i ∆ ln EXRt − i + ε i .........................................................................(4)
i =0 i =0

Whereby, δ1 to δ5 correspond to the long-run relationship, while β1 to β5 correspond to short-run dynamics of the
model; whilst subscripts t and t-i represent time periods. The re-parameterised results are presented in Table 5.

Table 5: Re-parameterised Results - ΔlnGDPt

Method: LS, Sample (adjusted): 1972 - 2016
Variable Coefficient Std. Error t-Statistic Prob.
C 2.201929 6.943164 0.317136 0.7532
lnGDPt-1 -0.901149 0.232025 -3.883847 0.0005
lnEXPt-1 0.000266 0.001906 0.139615 0.8898
lnIMPt-1 2.476175 0.941856 2.629037 0.0128
lnPOPt-1 0.356005 2.322405 0.153292 0.8791
lnFDIt-1 0.516092 0.141806 3.639421 0.0009
lnEXRt-1 0.004289 0.001208 3.551875 0.0010
ΔlnGDPt-1 0.094115 0.182819 0.514798 0.6102
ΔlnEXPt-1 0.000793 0.001834 0.432155 0.6685
ΔlnIMPt-1 -1.12E-05 0.000631 -0.017731 0.9860
ΔlnPOPt-1 4.987235 4.671759 1.067528 0.2937
ΔlnFDIt-1 1.137751 0.461086 2.467545 0.0188
ΔlnEXRt-1 -0.048389 0.023614 -2.049185 0.0482
R-squared 0.739030 Mean dependent var 0.061930
Adjusted R-squared 0.228667 S.D. dependent var 2.203841
S.E. of regression 1.935536 Akaike info criterion 4.395497
Sum squared resid 119.8816 Schwarz criterion 4.917422
Log-likelihood -85.89868 Hannan-Quinn criter. 4.590065
F-statistic 2.087006 Durbin-Watson stat 1.921641
Prob(F-statistic) 0.048100

From the re-parameterised results presented in Table 5, the general to specific technique to drop or maintain
some variables is applied. Studies by Katrakilidis and Trachanas (2012) and Fousekis et al. (2016) also used this
technique on econometric analysis. Nevertheless, the decision to maintain or drop some variables lies on the
decision made by t-statistics, whereby the bigger the value of the t-statistic the better the model and vice versa.
Therefore, for the variables to be maintained and their corresponding t-statistics have to be greater than 1.96,
otherwise, the variables have to be dropped. Applying the stipulated method, four differenced variables of
lnGDP, lnEXP, lnIMP and lnPOP have to be dropped because the corresponding t-statistic was found to be less
than 1.96. A reduce model in equation (5) was also run to come up with the reduced results which are presented
in Table 6. These results are to be subjected to further econometric analysis.

∆ ln GDPt = µ + λ ln GDPt −1 + δ1 ln EXPt −1 + δ 2 ln IMPt −1 + δ 3 ln POPt −1 + δ 4 ln FDI t −1

p p P
+δ 5 ln EXRt −1 +  α i ∆ ln GDPt −i +  β 4 ∆ ln FDI t −1 +  β 5 ∆ ln EXRt −1 + ε i .........................(5)
i =1 i =1 i =1

80
 Petro Sauti Magai

Table 6: Reduced Results - ΔlnGDPt

Method: LS, Sample (adjusted): 1971 - 2016
Variable Coefficient Std. Error t-Statistic Prob.
C 2.201929 6.943164 0.317136 0.7532
lnGDPt-1 -0.343617 0.103535 -3.318839 0.0022
lnEXPt-1 0.009773 0.004445 2.198681 0.0348
lnIMPt-1 0.356005 2.322405 0.153292 0.8791
lnPOPt-1 0.417682 0.150215 2.780551 0.0088
lnFDIt-1 0.339438 0.150890 2.249577 0.0311
lnEXRt-1 0.020269 0.008793 2.305189 0.0266
ΔlnFDIt-1 0.435840 0.137650 3.166283 0.0030
ΔlnEXRt-1 -1.567204 0.589363 -2.659147 0.0119
R-squared 0.685305 Mean dependent var 4.605164
Adjusted R-squared 0.388878 S.D. dependent var 2.323495
S.E. of regression 1.816376 Akaike info criterion 4.150307
Sum squared resid 135.2681 Schwarz criterion 4.386496
Log-likelihood -91.53221 Hannan-Quinn criter. 4.239187
F-statistic 6.854284 Durbin-Watson stat 1.909584
Prob (F-statistic) 0.000099

From the reduced results presented in Table 6, the long-run relationship can be computed using the Wald Test
(the F-test). Therefore, the lower and upper bound values are employed basing on 1% significance level for the
unrestricted intercept and no trend in the model as proposed by Pesaran et al., (2001). To accept the long-run
relationship between variables, the computed value of F-statistics has to be greater than that of the upper bound
value; this will enable the rejection of the null hypothesis and accept the alternative hypothesis. If the computed
F-statistic falls below the lower value, then it means that there is no co-integration between variables. But if the
computed value of F-statistic falls between two bounds, the results are inconclusive and a different technique of
co-integration has to be applied (Ghildiyal et al., 2015). Below are the hypotheses used to assist to arrive at a
decision:

H0: λ = δ1 = δ2 = δ3 = δ4 = δ5 =0 (the long-run relationship does not exist)

H1: λ ≠ δ1 ≠ δ2 ≠ δ3 ≠ δ4 ≠ δ5 ≠ 0 (the long-run relationship does exist)

Table 7: ARDL Long-run Relationships (bounds F-test)

F- Statistic (computed) Probability Remarks
Bounds test 5.8639 0.0022 Reject Ho

Critical value bounds 1% 5% 10%

I(0) I(1) I(0) I(1) I(0) I(1)
Pesaran et al. (2001, p. 300), 3.41 4.68 2.62 3.79 2.26 3.35
Table CI(iii) Case III
Source: Author's own computation

The analysis of the ARDL bounds testing approach to co-integration results presented in Table 7 shows that the
calculated F-statistic (5.8639) is greater than that of Pesaran et al. (2001) at 1%, 5% and 10% levels of
significance. This indicates that all variables are co-integrated in the case of Tanzania, from 1970 to 2016. In
other words, all variables move together in the long-run.
δ1 δ δ δ δ
δ ln EXP ; − = 0, δ ln IMP ; − 2 = 0, δ ln POP ; − 3 = 0, δ ln FDI ; − 4 = 0, δ ln EXR ; − 5 = 0
λ λ λ λ λ
Using the reduced results, long-run coefficients are calculated and this is ultimately useful in the determination
of long-run effects. The notation above is used to compute F-statistic and its corresponding p-values, as shown
in Table 8.

81
 Petro Sauti Magai

Table 8: Estimated Results for Long-run Coefficients

Regressor Long-run coefficients F-statistic P-value
lnEXPt-1 0.0284 18.244 0.0000***
lnIMPt-1 1.0360 3.0079 0.0046***
lnPOPt-1 1.2155 1.8241 0.0760
lnFDIt-1 0.9878 3.5666 0.0010***
lnEXRt-1 0.0589 4.0196 0.0003***
Source: Author's own computation
Note: *, **, *** denote significant level at 10%, 5% and 1%, respectively

DISCUSSION OF EMPIRICAL FINDINGS

This article has analysed the relationships between trade (exports and imports) and economic growth in
Tanzania. It has also included foreign direct investment, population growth and exchange rates as explanatory
variables. Based on the empirical findings of the study and the results presented in Table 6, a positive
contribution of almost all variables to economic growth in Tanzania is observed at 1% significant level; albeit,
population growth happens to be insignificant. Nevertheless, the long-run results for trade indicate that an
increase of one million US dollars of exports and imports will proportionately increase the total GDP of the
country by 0.02% and 103.6%, respectively. Though the percentage increase of imports is so higher than that of
exports, this is not a favourable situation as trade deficit is bound to rise. These findings are consistent with
theoretical and empirical literature which suggests that for most developing countries, Tanzania included, the
level of economic performance is positively affected by increase in trade. In addition, the empirical findings of
this study coincide with those of Chang et al. (2009), Lin (2000), and Were (2015) who hold that trade has a
positive impact on economic growth.

The empirical findings also suggest that FDI inflows exert a positive influence on economic growth in the long-
run. That is an increase of one million US dollars of FDI will proportionately increase the GDP total of the
country by 98.78%. These findings also coincide with the findings of De Mello (1997), Alfaro et al. (2004),
Gui-Diby (2014) and Hong (2014) who observed that there was positive contribution to economic growth
against FDI flows. However, Agbloyor et al. (2014) found there was negative association between FDI and
economic growth in 14 African countries. Further, looking at the empirical results of the exchange rate, it is the
case that any increased rates will positively influence the economic growth by 5.89%. These results are more
less the same as those of MacDonald (2000) and Korkmaz (2013), both of whom concluded that the exchange
rate is likely to arouse economic growth in European countries. Also, the findings of this study reveal a positive
contribution of population growth on economic growth, though its long-run coefficient is not significant;
consequently, it has no impact on economic growth in Tanzania.

CONCLUSION AND POLICY RECOMMENDATIONS

The study analysed the subject of trade and economic growth to find the kind of relationships that exist between
these two aspects. Despite the fact that this subject is not new in the world of economics as the subject has been
tackled by many authors, in Tanzania, it is still a subject that needs a lot of attention. It is from this background
that this study set to investigate the question of trade and economic growth, and come with appropriate
recommendations to policy makers. Nonetheless, the article employs the general-to-specific technique only to
come up with long-run effects. Much of the evidence from the econometric analysis shows that exports and
imports are directly correlated with economic growth in the long-run. Further assessment of the individual
variables of foreign direct investment, exchange rates and population is as well directly related to economic
growth; however, the latter is not significant. The latter part leads to a conclusion that population growth has no
significant impact on economic growth in Tanzania. In as much as the contribution of imports to economic
growth is higher than that of exports, this makes an alarming call to policy makers to check the import-export
policy measures for the realisation of the full potential of trade on economic growth. This article therefore,
recommends that policy makers should pay more attention to export promotion policy and import measures that
will allow the importation of raw materials for the products to be exported; in doing so, trade deficit will be
minimised.

REFERENCES
Agbloyor, E.K., Abor, J.Y., Adjasi, C.K.D., Yawson, A., 2014. Private capital flows and economic growth in
Africa: The role of domestic financial markets. J. Int. Financ. Mark. Inst. Money Vol 30, 137–152.
Alfaro, L., Chanda, A., Kalemli-Ozcan, S., Sayek, S., 2004. FDI and economic growth: the role of local
financial markets. Journal. Int. Econ. 64, 89–112. [Link]

82
 Petro Sauti Magai

BoT, 2017. Bank of Tanzania monthly economic review, Dar es Salaam, Tanzania.
Brzezinski, M., Dzielinski, M., 2009. Is endogenous growth theory degenerating? Another look at Lakatosian
appraisal of growth theories. Journal. Econ. Methodol. Vol 16, 243–263.
Busse, M., Königer, J., 2012. Trade and Economic Growth: A Re-examination of the Empirical Evidence.
HWWI Research Paper # 123. Ruhr-University Bochum.
Chang, R., Kaltani, L., Loayza, N.V., 2009. Openness can be good for growth: The role of policy
complementarities. Journal. Dev. Econ. Vol 90, 33–49.
Colombatto, E., 1990. An Analysis of Exports and Growth in LDCs. Kyklos Vol 43, 579–597.
Daniel, C., 2000. The question of international competitiveness. Int. Adv. Econ. Res. Vol 6, 417–426.
De Mello, L.R., 1997. Foreign direct investment in developing countries and growth: A selective survey.
Journal. Dev. Stud. 34, 1–34. [Link]
Dickey, D.A., Fuller, W.A., 1979. Distribution of the estimators for autoregressive time series with a unit root.
Journal. Am. Stat. Assoc. Vol 76, 427–431.
EAC, 2015. East African Community Trade and Investment Report. Arusha, Tanzania.
Edwards, S., 1993. Openness, Trade Liberalization, and Growth in Developing Countries. Journal. Econ. Lit.
Vol 31, 1358–93.
Enders, W., 2015. Applied econometric time series. John Wiley and Sons, United Kingdom.
Eriṣ, M.N., Ulaṣan, B., 2013. Trade openness and economic growth: Bayesian model averaging estimate of
cross-country growth regressions. Econ. Model. 33, 867–883.
[Link]
Fousekis, P., Katrakilidis, C., Trachanas, E., 2016. Vertical price transmission in the US beef sector: Evidence
from the nonlinear ARDL model. Econ. Model. Vol 52, 499–506.
Frankel, J., Romer, D., 1999. Does Trade Causes Growth? [Link]
Ghildiyal, V., Pokhriyal, A.K., Mohan, A., 2015. Impact of financial deepening on economic growth in Indian
perspective: ARDL Bound testing approach to cointegration. Asian Dev. Policy Rev. Vol 3, 49–60.
Gui-Diby, S.L., 2014. Impact of foreign direct investments on economic growth in Africa: Evidence from three
decades of panel data analyses. Res. Econ. 68, 248–256. [Link]
Hong, L., 2014. Does and How does FDI Promote the Economic Growth? Evidence from Dynamic Panel Data
of Prefecture City in China. IERI Procedia 6, 57–62. [Link]
Jouini, J., 2015. Linkage between international trade and economic growth in GCC countries: Empirical
evidence from PMG estimation approach. Journal. Int. Trade Econ. Dev. Vol 24, 341–372.
Katrakilidis, C., Trachanas, E., 2012. What drives housing price dynamics in Greece: New evidence from
asymmetric ARDL cointegration. Econ. Model. Vol 29, 1064–1069.
Kim, D.-H., 2011. Trade, growth and income. Journal. Int. Trade Econ. Dev. Vol 20, 677–709.
Korkmaz, S., 2013. The Effect of Exchange Rate on Economic Growth: 2nd International M-Sphere Conference.
Dubrovnik.
Lin, S., 2000. Foreign Trade and China’s Economic Development: A Time-Series Analysis. Journal. Econ. Dev.
Vol 25.
Lucas, R.E., 1988. On the mechanics of economic development. Journal. Monet. Econ. 22, 3–42.
MacDonald, R., 2000. The Role of the Exchange Rate in Economic Growth: A Euro-Zone Perspective: National
Bank of Belgium Working Paper No. 9. University of Strathclyde in Glasgow.
Makki, S.S., Somwaru, A., 2004. Impact of foreign direct investment and trade on economic growth: Evidence
from developing countries. Am. Journal. Agric. Econ. 86, 795–801.
Mankiw, G., Romer, D., Weil, D.N., 1992. A Contribution to the Empirics of Economic Growth. Q. J. Econ.
Vol 107, 407–437.
Moussiegt, L., Beltramello, A., De Backer, K., 2012. The Export Performance of Countries within Global Value
Chains (GVCs) (OECD Science, Technology and Industry Working Papers No. 2012/02).
[Link]
Musila, J.W., Yiheyis, Z., 2015. The impact of trade openness on growth: The case of Kenya. Journal. Policy
Model. 37, 342–354. [Link]
Ouattara, B., 2004. Modelling the long run determinants of private investment in Senegal, CREDIT Research
Paper, No. 04/05, Centre for Research in Economic Development and Internat. Trade, Univ. of
Nottingham, Nottingham.
Pesaran, M.H., Pesaran, B., 1997. Working with microfit 4.0: Interactive econometric analysis. Oxford: Oxford
University Press.
Pesaran, M.H., Shin, Y., Smith, R.J., 2001. Bound testing approaches to the analysis of level relationships.
Journal. Appl. Econ. Vol 16, 289–326.
Phillips, P.C.B., Perron, P., 1988. Testing for a unit root in time series regression. Biometrika Vol 75, 335–346.
Reinert, E.S., 2008. How rich countries got rich and why poor countries stay poor. Public Affairs, New York.

83
 Petro Sauti Magai

Rodriguez, F., Rodrik, D., 2000. Trade policy and economic growth: A skeptic’s guide to the cross-national
evidence. NBER working paper No. Cambridge, MA.
Romer, P.M., 1986. Increasing Returns and Long-Run Growth. Journal. Polit. Econ. Vol 94, 1002–1037.
Rybczynski, T.M., 1955. Factor Endowment and Relative Commodity Prices. Economica Vol 22, 336–341.
Solow, R.M., 1956. A Contribution to the Theory of Economic Growth. Q. J. Econ. Vol 70, 65–94.
Trejos, S., Barboza, G., 2015. Dynamic estimation of the relationship between trade openness and output growth
in Asia. Journal. Asian Econ. 36, 110–125. [Link]
Ulaşan, B., 2015. Trade openness and economic growth: panel evidence. Appl. Econ. Lett. Vol 22, 163–167.
Wastyn, A., Sleuwaegen, L., 2013. Study on the export and FDI potential of Belgian companies.
Were, M., 2015. Differential effects of trade on economic growth and investment: A cross-country empirical
investigation. J. Afr. Trade 2, 71–85. [Link]
Zahonogo, P., 2016. Trade and economic growth in developing countries: Evidence from sub-Saharan Africa.
Journal. Afr. Trade 3, 41–56. [Link]

84