Journal of Agriculture and Social Research, Vol. 15, No.

2, 2015

CASUAL RELATIONSHIP BETWEEN GROSS DOMESTIC SAVING AND

ECONOMIC GROWTH IN EAST AFRICA: EVIDENCE FROM ETHIOPIA,
UGANDA AND KENYA

SAMUEL ELIAS AND ABEBE WORKU

Department of Agricultural Economics,
College of Agriculture and Natural Sciences, Dilla University, Ethiopia
Email: eliboy2013@gmail.com

ABSTRACT
The paper aimed to analyze the causal relationship between economic growth and
savings in East Africa (1981-2014) using Vector Error Correction (VEC) method and
Johnson's approach. All statistical data used throughout this paper came from World
Bank database. The empirical study confirmed that a significant relationship between
domestic savings and economic growth in the case of Ethiopia and Uganda. However,
there is no significant relationship obtained in the case of Kenya over the study period by
Johnson co-integration approach. The results of Granger Causality between economic
growth (GDP) and gross domestic savings indicated the presence of unidirectional
causality between economic growth and gross domestic savings in the case of Ethiopia
and Uganda. Gross domestic product does Granger cause gross domestic savings; this
means that economic growth accelerates gross domestic savings in the case of Ethiopia
and Uganda. It is recommended that the countries needs to design a policy which
enhances higher economic growth through increasing total factor productivity and,
which ultimately increases the country domestic saving level. Moreover, to achieve
sustainable growth the government needs to embark on policy measures, which increase
saving and investment into the country due to its dual effect.

Key words: Causation, Domestic Savings, Economic Growth and East Africa

INTRODUCTION
The relationship between the saving and economic growth has received greater attention
from a different researcher in the world. Economic growth is important for one country
to achieve a higher living standard for citizens. Among many factors, impacting
economic growth in a given society is the level of domestic savings. According to (Saltz,
1999), (Bacha, 1990), (DeGregorio, 1992) and (Stern, 1991) have explained that
increases in savings will facilitate more rapid expansion of the capital stock and;
therefore, higher rates of investment that should lead to higher rates of economic growth.
Moreover, ample empirical studies indicated that economic growth would contribute to
growth in the personal income as the result per capita consumption expenditure and
saving rate also increase. According to the theory of marginal propensity to save, the rate
of saving expands from the increasing of income. As a result, it can be plainly understood
that when there is economic growth, the amount of savings also increase.

31
Journal of Agriculture and Social Research, Vol. 15, No. 2, 2015

Tinaromm (2005) studied the relationship between savings and economic growth
in North Africa using a Vector Error Correction Model for 1946-1992. He concluded that
private saving has both direct and indirect effects on economic growth. In his view, the
direct effect of savings is through private investment. He also showed that economic
growth has a positive effect on the private savings rate. Dipendra (2009) studied the
relation between savings and economic growth in India. The goal of this study was to
check the long-run relationship between GDP and savings. An Engel-Granger Co-
Integrated method was used, and the results showed that gross savings of the private
sector have a bigger impact on GDP than gross domestic savings. And another study by
Sothan (2014) analyzed the relationship between domestic Saving and economic growth
in Cambodia and the study found that domestic saving does not Granger cause economic
growth. This finding is contrary to the conventional wisdom that causality runs from
saving to economic growth. Sothan concluded that domestic saving and economic growth
are independent of each other in Cambodia.
Needless to say, every nation is trying to achieve sustainable economic growth
aiming to the betterment of its citizen’s standard of living. This can be achieved by
sustainable economic growth through increasing the rate of investment which is a
function of increased saving rate. To this end, every government may implement various
kinds of policy strategies such as encourage saving, stimulating investment and
productivity in their countries.
The magnitude of economic growth in any country is dependent upon the level of
investment. The rate of investment, on the other hand, is highly dependent upon the level
of gross domestic savings of a country. It is believed that increased saving will increase
the national capacity for investment and production, while a serious constraint to
sustainable economic growth can result from the low rate of saving. According to World
Bank’s study on average savings, East Asia saves more than 30 percent of Gross National
Disposable Income (GNDI), while Sub-Saharan Africa saves less than 15 percent (World
Bank, 2000).
Likewise, a study by (Kibet et al., 2009) found that saving rate in Africa has
perpetually been the lowest compared to other regions. It also faces serious credit
constraints; and this, coupled with low income could greatly reduce any little incentive to
save. Development economists have been concerned for decades about the crucial role of
domestic saving mobilization in the sustenance and reinforcement of the saving
investment-growth chain in developing economies (Nwachukwu & Egwaikhide, 2007).
Increased savings rates are therefore, of crucial importance for achieving sustainable
development and poverty-reducing growth in African countries. (Keho, 2011)
Generally, it is possible to accept that increasing gross savings contribute to higher
investment, and this leads to the higher GDP growth in the short run. It means that the
higher saving rate leads to less consumption, which could also result in the larger amount
of capital investment and finally a higher rate of economic growth. Despite the fact that,
the issue of causality between saving and growth is unsettled because of the wide
variation in results among the studies conducted on causality. The direction of causality
between saving and growth may vary because of differences in the methodology used.

32
Journal of Agriculture and Social Research, Vol. 15, No. 2, 2015

Another reason could be the choice of variable specifications for causality analysis, and
the definition of the variables used. The causal relationship may also vary from country
to country and between periods of time. Therefore, this study intended to analyze the
causal relationship between domestic saving and economic growth in east Africa (i.e.
Ethiopia, Kenya and Uganda) with the objectives of determining empirically the
existence of long run relationship between gross domestic savings and economic growth
in East Africa and to provide empirical evidence whether causal relationship exists
between gross domestic savings and economic growth and the particular direction of
causality between them.

METHODOLOGY
Sample Data and Data Sources
A time series data on gross domestic savings and per capita real GDP as a measure of
economic growth in east Africa in the case of Ethiopia, Uganda, and Kenya for the period
1981 to 2014 are used for this study. All data from 1981 to 2014 are from World
Development Indicators of the World Bank. All computations are performed using the
Eviews8 software.
Unit Root Test
The first and for most issues in the testing procedure is to determine whether the data
contain unit roots indicating that data is non-stationary or not. Most commonly used type
test employed in this study was the Augmented Dickey–Fuller (ADF) test which has
developed by Dickey and Fuller. The test is used for checking whether variables such that
GDP growth rate, Gross Saving growth rate have a unit root or not. If parameter α is
equal to Zero, it means the variable contains unit root which means the data is
nonstationary.
The Augmented Duckey-Fuller test is in two forms: one with only intercept and
another with intercept and trend. The one that is chosen depends on the nature of the
curvature of the variable being tested for a unit root. If the curvature of a time series
variable exhibit trend, then the Augmented Dickey-Fuller test is conducted with intercept
and trend. On the other hand, if the curvature of a variable exhibits no trend, then the
ADF Test is performed with the only intercept.
The ADF test equation is stated as:
---------------------------------------------- 1

The target of this study is to test existing of the relationship between economic growth
and savings in East Africa. If this kind of relationship exists, the next is to test the
direction of causality between these countries.
After careful analysis for stationary, the next step is to examine co-integration
analysis (Johnson, 1988) for each of the two variables at the first difference I (1) in East
Africa evidence from Ethiopia, Kenya, and Uganda. To estimate the co-integration rank
and vector, the following and statistics test are used where
----------------------------------------------------------- 2

33
Journal of Agriculture and Social Research, Vol. 15, No. 2, 2015

---------------------------------------------------------- 3

For the statistics, the null hypothesis is that the number of co-integrating vectors is
less than or equal to r against r =1, 2, 3, 4…, while in case of statistics, the null
hypothesis is that the number of co-integrating vectors is less than or equal to r against r
= r + 1.

The presence of a co-integrating relation forms the basis of the VEC (Vector Error
Correction) specification. Additionally, standard Granger or Sims tests may provide
invalid causal information due to the omission of error correction terms from the tests
(Doyle, 2001).

------------------------- 4

-------------------------- 5

The finding that much macro time series may contain a unit root has spurred the
development of the theory of non-stationary time series analysis. Empirical studies have
shown that the existence of non-stationary in the time series considered can lead to
spurious regression results and invalidate the conclusions reached using Granger
Causality. (Toda and Phillips, 1993) have led the methods to deal with Granger causality
in I (1) systems of variables. A causal long run relationship between non-stationary time
series when they are co-integrated could be inferred. Therefore, if co-integration analysis
is omitted, causality tests present evidence of simultaneous correlations rather than causal
relations between variables. The presence of a co-integrating relation forms the basis of
the VEC (Vector Error Correction) specification. Additionally, standard Granger or Sims
tests may provide invalid causal information due to the omission of error correction terms
from the tests (Doyle, 2001).
The simple Granger’s causality test becomes inappropriate when co-integrating
vectors are obtained in the series. According to Granger’s representation theorem (1988),
the results of co-integration imply that series have the following error-correction
representations. These are necessary to augment the simple Granger causality test with
the ECM (Error Correction Mechanism), derived from the residuals of the appropriate co-
integration relationship to test for causality:
Where, Y and X= are the variables under consideration, =the adjustment
coefficient while ECTt-1 expresses the error correction term of growth equation,
=indicates the first difference operator. In equation (4), X Granger causes Y if and
are significantly different from zero. In equation (5), Y Granger causes X if and are
significantly different from zero. F-statistic is used to test the joint null hypothesis of ,
= 0 and t-test is employed to estimate the significance of the error coefficient.

34
Journal of Agriculture and Social Research, Vol. 15, No. 2, 2015

RESULTS AND DISCUSSIONS

Summary of Augmented Duckey -Fuller (ADF) Unit Root Test
The first concern of co-integration and causality analysis is to test stationary of the
variables. And for this purpose commonly used, ADF test has been used. As indicated in
Table 1, both LNGDP and LNGDS series have unit roots in the level data at 1% and 5%
level of significance for all countries, the statistics except a null hypothesis which
strongly suggests that both series at levels contain a unit root in case all countries as
indicated in Table 1. After the variable transformed into their first differences and
running the ADF test again, the variables become stationary. In both cases when the
intercept and a trend have been included for ADF model analysis; the variables are
stationary. From this, it can be generalized that the variables are integrated of order one
i.e. I (1) for all countries. In most causality and co-integration analysis the variable
become stationary after first difference, non-stationary variables were further tested to
ascertain whether they were co-integrated and causation. In this case, one should proceed
to test for the direction of causation of the variables. Of course, this is after proving the
existence of the co-integration relationship between its variables. In other words, the null
hypotheses were tested about the rank of the co-integrating relationships that existed
among the variables.

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

Countries Variables Levels 1st difference
Intercept intercept Intercept Intercept Conclusion
+trend +trend
Ethiopia LnGDP 0.424710 -0.687243 -3.31806 -3.708328 I(1)
(p-value) (0.9810) ( 0.9656) (0.0223) ( 0.0363)
LnGDS 0.080117 -0.036102 - -8.01100 I(1)
(p-value) (0.9593) (0.9938) 7.450719 (0.0000)
(0.0000)
Kenya LnGDP 1.243255 -1.664205 - -4.374571 I(1)
(p-value) (0.9978) (0.7444) 4.164889 (0.0078)
(0.0027)
LnGDS - -2.373555 - -6.930477 I(1)
(p-value) 1.457958 (0.3856) 7.057744 (0.0000)
( 0.5420) (0.0000)
Uganda LnGDP - -1.992273 - -3.252290 I(1)
(p-value) 0.146273 ( 0.5833) 3.272799 (0.0944)
(0.9348) (0.0257)
LnGDS - -6.288135 - -5.965519 I(1)
(p-value) 1.117081 (0.0004) 7.105338 (0.0008)
( 0.6878) (0.0000)
Note: r indicates the number of co-integrating relationships. The optimal lag structure of the
VAR was selected by minimizing the AIC criterion. Critical values are taken from (Johansen and
Juselius, 1990). ** indicates rejection at the 95% critical values.

35
Journal of Agriculture and Social Research, Vol. 15, No. 2, 2015

According to Johansen test procedure, both the Trace statistic criterion and the Maximum
Eigen value criterion were used to reject the null hypothesis and to draw conclusions
about the hypotheses of the rank of the co-integrating relationships. The decision
criterion is that when the Trace Statistic is greater than the 5% critical value, the null
hypothesis is rejected and it is concluded that there is a co-integrating relationship. The
testing was continued in an iterative manner until the null hypotheses were no longer
rejected to indicate no co-integrating relationship.

Table 2: Results of Johansen’s test for multiple co integration vectors of Ethiopia,

Uganda, and Kenya
Countrie Hypothesize No. of Test statistics Critical values (95%)
s d relationship
co s
integrating
H: 0 H:1 Max. Trace Max. Trace
eigenvalue eigenvalu
e
Ethiopia r = 0 r >0 16.54351* 16.54352* 14.26460 15.4947
* * 1
r =1 r>1 0.0000004 0.000012 0.000012 3.84146
6
Uganda r = 0 r >0 20.60044* 20.60044* 15.49471 14.2646
* * 0
r =1 r>1 0.068556 0.068556 3.841466 3.84146
6
Kenya r=0 r >0 3.563465 3.495975 15.49471 14.2646
0
r =1 r>1 0.067490 0.067490 3.841466 3.84146
6
Note: r indicates the number of co-integrating relationships. The optimal lag structure of
the VAR was selected by minimizing the AIC criterion. Critical values are taken from
(Johansen and Juselius, 1990). ** indicates rejection at the 95% critical values.

Table 2 presents the results of Johansen’s test for multiple co integration vectors of
Ethiopia, Uganda, and Kenya. According to both maximal Eigen value and trace statistic
tests, the results indicate the existence of one co-integration vector in the case of Ethiopia
and Uganda in a range of study periods. Thus, the Johansen co-integration test suggests
that there is a long-run relationship between domestic savings and economic growth in
Ethiopia and Uganda. But, in the case of Kenya, there is no evidence indicating any long
run relationship between domestic saving and economic growth in the study period.
Therefore, the null hypothesis no co-integration at the 1% level in case of Kenya is
accepted by rejecting the alternative hypothesis. The same result found by (Sothan, 2014)
where domestic saving and economic growth is independent of each other in case of
36
Journal of Agriculture and Social Research, Vol. 15, No. 2, 2015

Cambodia. The long-run relationship between economic growth and savings is found to
be positive in each co-integration vector in the case of Ethiopia and Uganda. This
suggests the existence of causality in at least one direction.

Granger Causality test

Granger causality test is conducted to know the direction of the causality. In Table 3. the
results of Pair-wise Granger Causality between economic growth (GDP) and gross
domestic saving indicated the presence of unidirectional causality between economic
growth and gross domestic saving in the case of Ethiopia and Uganda. The study revealed
that for the null hypothesis of “LNGDS does not Granger Cause LNGDP”, it is not
possible to reject the null hypothesis since the F-statistics value is small and the
probability is value is high. Therefore, we accept the null hypothesis and conclude that
LNGDS does not Granger Cause LNGDP, but for the null hypothesis of “LNGDP does
not Granger Cause LNGDS," it is possible to reject the null hypothesis and conclude that
LNGDP does Granger Cause LNGDS. This means that economic growth accelerates and
augments gross domestic saving in Ethiopia and Uganda. Moreover, a study revealed that
existence of unidirectional causality (causality runs from economic growth to gross
domestic saving) in the case of Uganda and Ethiopia.
The conversional view of theoretical economics is that higher savings leads to
higher investment and higher economic growth is not supported by East African countries
case. Instead, the result of this study indicated that, the causality is from economic
growth rate to gross domestic saving. This result is supported with many studies (Salz,
999), and (Baharumshah et al., 2003).

Table 3: Granger Causality test

Ethiopia Uganda
Null Hypothesis F- P-Value Decision F- P-Value Decision
statistics statistics
GDS does not 0.43858 0.7275 Accept 1.25809 0.3039 Accept
Granger Cause GDP
GDP does not 5.01332 0.0077 Reject 4.69622 0.0200 Reject
Granger Cause GDS

CONCLUSIONS AND RECOMMENDATIONS

The co-integration test confirmed that economic growth and domestic gross savings are
co integrated into the case of Ethiopia and Uganda, indicating an existence of a long-run
relationship between the two as confirmed by the Johansen co integration test results.
Although many theoretical and empirical studies confirm the positive impact of saving on
economic growth, results of the present study do not found any long-run relationship
between domestic saving and economic growth in case of Kenya as confirmed by
Johnson co-integration test. However, the Granger causality test finally confirmed the
presence of unidirectional causality, which runs from economic growth to gross domestic
saving in case of Ethiopia and Uganda. Based on the empirical result of the study
37
Journal of Agriculture and Social Research, Vol. 15, No. 2, 2015

Granger causality test the study favored the hypothesis which suggests that the causality
is from economic growth rate to growth rate of domestic saving in the country in the
range of study period. However, the gross domestic saving growth does not Granger
Cause GDP per capita growth. The result of the empirical test clearly points toward a
positive impact of economic growth on saving, or it can say that income of the country
does play an important role to lead the saving in case of Ethiopia and Uganda. It is
recommended that the countries needs to design a policy which enhances higher
economic growth through increasing total factor productivity and which ultimately
increases the country domestic saving level. Moreover, if domestic savings are invested
efficiently and are therefore an important factor of economic growth, the main objective
of national economic policy should be to encourage the people to save by using different
mobilization techniques.

REFERENCES
Alomar, I. (2013). Economic Growth and Savings in GCC: A Co-integration and Causal
Relationship Analysis, International Journal of Humanities and Social Science, 3(9).
Bacha, E.L. (1990). A Three-Gap Model of Foreign Transfers and the GDP Growth Rate
in Developing Countries”, Journal of Development Economics, 32: 279-96.
Baharumshah et al., (2003).Savings Dynamic in Asian Countries. Journal of Asian
Economics, 13, 827-845.
DeGregorio, J. (1992). Economic Growth in Latin America, Journal of Development
Economics, 39: 59-84.
Domer, E.D. (1946). Capital Expansion, Rate of Growth, and Employment,
Econometrica, 14:137-47.
Doyle, E. (2001). Export-output causality and the role of exports in Irish growth: 1950-
1997, International Economic Journal, 15(2): 147-161.
Harrod, R. (1939). An Essay in Dynamic Theory, Economic Journal, 49:14-33.
Houthakker, H.S. (1965). On Some Determinants of Saving in Developed and
Underdeveloped Countries, in E.A.G. Robinson, (ed.). Problems in Economic
Development, London: Macmillan.
Jappelli, T., and Pagano, M. (1994). The Determinants of Saving: Lessons from Italy,
Paper presented at the Inter-America Development Bank Conference on
Determinants of Domestic Savings in Latin America, Bogota, Colombia.
Johansen, S. (1988). Statistical Analysis of Co-integrating Vectors, J Econ. Dyn. Control
12(2):231 254.
Johansen, S., and Juselius, K. (1990). Maximum Likelihood Estimation and Inference on
Co-integration with Applications to the Demand for Money, Oxford Bulletin of
Economics and Statistics, 52:169-210.
Kibet, L. K., Mutai, B. K., Ouma, D. E., Ouma, S. A. and Owuor, G. (2009).
Determinants of household saving: Case study of smallholder farmers,
entrepreneurs and teachers in rural areas of Kenya, Journal of Development and
Agricultural Economics, 1:137-143.

38
Journal of Agriculture and Social Research, Vol. 15, No. 2, 2015

Krieckhaus, J. (2002). Re conceptualizing the Development state: Public Savings and

Economic Growth, World Development, 30(10):1697-1712.
Levine,R. E. and David, R. (1992). A Sensitivity Analysis of Cross-Country Growth
Regressions, American Economic Review, 82:942–963.
Lewis,W. A. (1955). The Theory of Economic Growth, Homewood, III: Irvin.
Nwachukwu, T.E. and Egwaikhide, F.O. (2007). An Error-Correction Model of the
Determinants of Private Saving in Nigeria, A Paper presented at the African
Economic Society (AES) Conference, Cape Town, South Africa.
Olajide, S. (2010). Does Saving Really Matter For Growth In Developing Countries? The
Case of A Small Open Economy”, International Business & Economics Research
Journal, 9(4).
Romer, P. (1987). Growth Based on Increasing Returns to Specialization, The American
Economic Review, 77(2):56-62.
Saltz, I.S. (1999). An Examination of the Causal Relationship between Savings and
Growth in the Third World, Journal of Economics and Finance, 23:90-98.
Solow, R. M. (1956). A Contribution to the Theory of Economic Growth”, The Quarterly
Journal of Economics, 70:65-94.
Sothan, S. (2014). Causal Relationship between Domestic Saving and Economic
Growth: Evidence from Cambodia, International Journal of Economics and
Finance; 6(9).
Stern, N. (1991). The Determinants of Growth, Economic Journal, 101:122-133.
World Bank (2000). Saving in Developing Countries: An Overview, The World Bank
economic review, 14: 393–414.

39