In this section, data set, research methodology and empirical findings are presented.
3.1 Data Set
This study looks possible relationships between economic growth and other macroeconomic factors, notably Inflation, BRVM Market Capitalization, Capital Flow and BRVM Composite Index.
Gross Domestic Product (GDP)
The economic well-being of a nation or region is frequently evaluated using Gross Domestic Product (GDP), which quantifies the total market worth of all final products and services generated and provided within a designated timeframe by one or multiple countries. It represents the economic growth in this study.
Inflation
As a widely used metric for gauging economic spending, the Consumer Price Index (CPI) serves as a valuable tool. In the context of this particular investigation, the CPI was employed to assess inflation levels.
Market Capitalization
The market capitalization of a publicly traded company represents the combined worth of all the common shares held by shareholders. This value is determined by multiplying the market price per common share by the total number of outstanding common shares.
Capital Flows
These include all transactions recorded in the capital accounts, as well as the financial account of the balance of payments, such as foreign direct investments (FDI), portfolio investments, and other investments including loans and banking capital.
BRVM Composite Index
The BRVM Composite is a stock index calculated from the value of each stock on the BRVM. The BRVM Composite is a West African stock market index created with the base 100 on September 15, 1998. Monthly data from January 2009 to December 2020 were used to conduct this study. Data were collected from International Monetary Fund, World Bank and BRVM websites. Not all data were available in monthly data; those available in daily and annual frequences, were converted into monthly frequencies.
Table 1
Macroeconomic Factors Used in Analysis
Variable
|
Indicator
|
Measurement
|
Source
|
Variable Type
|
Economic Growth
|
GDP
|
\(\:\frac{{\text{G}\text{D}\text{P}}_{\text{t}}-\:{\text{G}\text{D}\text{P}}_{\text{t}-1}}{{\text{G}\text{D}\text{P}}_{\text{t}-1}}\)
|
\(\:\text{L}\text{o}\text{g}(\frac{{\text{G}\text{D}\text{P}}_{\text{t}}}{{\text{G}\text{D}\text{P}}_{\text{t}-1}}\))
|
World Bank
|
Dependent
|
Inflation
|
Consumer Price Index
|
\(\:\frac{\text{C}\text{P}\text{I}-\:{\text{C}\text{P}\text{I}}_{\text{t}-1}}{{\text{C}\text{P}\text{I}}_{\text{t}-1}}\)
|
\(\:\text{L}\text{o}\text{g}\left(\frac{{\text{C}\text{P}\text{I}}_{\text{t}}}{{\text{C}\text{P}\text{I}}_{\text{t}-1}}\right)\)
|
International Monetary Fund
|
Independent
|
Stock Market Index
|
BRVM Composite
|
\(\:\frac{{\text{P}}_{\text{t}}-\:{\text{P}}_{\text{t}-1}}{{\text{P}}_{\text{t}-1}}\)
|
\(\:\text{L}\text{o}\text{g}\left(\frac{{\text{P}}_{\text{t}}}{{\text{P}}_{\text{t}-1}}\right)\)
|
BRVM
|
Independent
|
Market Capitalization
|
Market Capitalization
|
\(\:\frac{{\text{M}\text{K}\text{C}}_{\text{t}}-\:{\text{M}\text{K}\text{C}}_{\text{t}-1}}{{\text{M}\text{K}\text{C}}_{\text{t}-1}}\)
|
\(\:\text{L}\text{o}\text{g}\left(\frac{{\text{M}\text{K}\text{C}}_{\text{t}}}{{\text{M}\text{K}\text{C}}_{\text{t}-1}}\right)\)
|
BRVM
|
Independent
|
Capital Flows
|
Capital Flow
|
\(\:\frac{{\text{C}\text{P}\text{F}}_{\text{t}}-\:{\text{C}\text{P}\text{F}}_{\text{t}-1}}{{\text{C}\text{P}\text{F}}_{\text{t}-1}}\)
|
\(\:\text{L}\text{o}\text{g}\left(\frac{{\text{C}\text{P}\text{F}}_{\text{t}}}{{\text{C}\text{P}\text{F}}_{\text{t}-1}}\right)\)
|
World Bank
|
Independent
|
Source: Generated by Author |
3.2 Research Methodology
This study explores the existence of relationship (long or short) between GDP and other variables (BRVM Composite, Market Capitalization, Capital flow and Inflation). Eviews 10 was used to perform the analyses. The generic research model determined is the following:
\(\:{\text{G}\text{D}\text{P}}_{\text{i}\text{t}}={{\beta\:}}_{0}+\:{{\beta\:}}_{1}{\text{I}\text{N}\text{F}\text{L}}_{\text{i}\text{t}}+\:{{\beta\:}}_{2}{\text{M}\text{K}\text{C}}_{\text{i}\text{t}}+\:{{\beta\:}}_{3}{\text{C}\text{P}\text{F}}_{\text{i}\text{t}}+\:{{\beta\:}}_{4}{\text{B}\text{R}\text{V}\text{M}}_{\text{i}\text{t}}+\:{{\epsilon\:}}_{\text{i}\text{t}}\)
|
(1)
|
Where:
β0
Constant
β1
Inflation sensitivity monthly change
β2
Sensitivity of monthly change in Market Capitalization
β3
Sensitivity of monthly change in Capital Flow
β4
Sensitivity of monthly change in BRVM Composite
Ꜫit
Error term
The series normality test was first carried out. The variables logged values will be used to conduct the afterward-tests. The series stationarity test was then performed to detect a possible unit root; the test was run at both the Level and the First Difference. In the event where the series are I(0) and I(1), the Bound Test statistics and not the Johansen Test will be used for cointegration. The Bound Test is more appropriate because it is applicable on series at I(0) and I(1) and is more suitable for small and finite data samples. In the case of a Level relationship the variables, it will indicate that there are both short-run and long-run relationships between the variables. An Autoregressive Distributed Lag (ARDL) and a Vector Error Correction (VEC) models will be determined. The Error Correct model to be estimated is as follows:
\(\:\varDelta\:{\text{G}\text{D}\text{P}}_{\text{T}}=\:{{\gamma\:}}_{0}+\:\sum\:_{\text{i}=0}^{\text{p}}{{\gamma\:}}_{1}{\varDelta\:\text{I}\text{N}\text{F}\text{L}}_{\text{t}-1}+\:\sum\:_{\text{i}=0}^{\text{P}}{{\phi\:}}_{\text{i}}{CPF}_{\text{t}-1}+\:\sum\:_{\text{i}=0}^{\text{P}}{{\phi\:}}_{\text{i}}{MKC}_{\text{t}-1}+\:\sum\:_{\text{i}=0}^{\text{P}}{{\phi\:}}_{\text{i}}{BRVM}_{\text{t}-1}\:+\:{{\mu\:}\text{E}\text{C}\text{T}}_{\text{t}-1}+\:{\text{u}}_{\text{i}}\)
|
(2)
|
γ 1 and φi stand for short-term coefficients, as for ∆, it represents the difference operator, µ stands for the order of delay, ui represent the residuals and ECTt−1 signifies the term for error correction.
The ECT is expressed as follows:
$$\:{\text{G}\text{D}\text{P}}_{\text{t}-1}=\:{{\delta\:}}_{1}{\text{I}\text{N}\text{F}\text{L}}_{\text{t}-1}+\:{{\delta\:}}_{2}{\text{C}\text{P}\text{F}}_{\text{t}-1}+\:{{\delta\:}}_{3}{\text{M}\text{K}\text{C}}_{\text{t}-1}+\:{{\delta\:}}_{4}{\text{B}\text{R}\text{V}\text{M}}_{\text{t}-1}+\:{{\epsilon\:}}_{1}$$
3
The long-term association between variables is displayed by ECT. The rate at which stock returns revert to equilibrium following a long-term divergence is measured by the u coefficient. The system is considered balanced if the error correction coefficient is less than 1, and when it is negatively indicated, it suggests that, in the event of a deviation from balance, there is a movement towards equilibrium. In other words, according to Bozkurt (2007: 166), the mistake correction process functions.
3.3 Empirical Findings
The result for the normality test is as follows
Table 2
|
GDP
|
CAPITAL_FLOW
|
BRVM_COMP
|
INFLATION_CHANGE
|
MARKET_CAP
|
Mean
|
3.889118
|
2.975233
|
-0.000436
|
-0.337083
|
0.085132
|
Median
|
4.754179
|
3.111801
|
-0.001400
|
-0.041667
|
0.106027
|
Maximum
|
7.084700
|
4.971522
|
0.152600
|
3.560000
|
0.353846
|
Minimum
|
-1.235500
|
1.003936
|
-0.110900
|
-6.710000
|
-0.115385
|
Std. Dev.
|
2.225470
|
0.899308
|
0.044650
|
2.358703
|
0.101483
|
Skewness
|
-0.721538
|
-0.190079
|
0.549780
|
-0.565118
|
-0.102936
|
Kurtosis
|
2.282130
|
2.454369
|
3.828300
|
2.692086
|
2.643682
|
Jarque-Bera
|
15.58682
|
2.653401
|
11.37067
|
8.233467
|
1.016077
|
Probability
|
0.000412
|
0.265351
|
0.003395
|
0.016298
|
0.601675
|
Sum
|
560.0330
|
428.4336
|
-0.062800
|
-48.54000
|
12.25895
|
Sum Sq. Dev.
|
708.2384
|
115.6519
|
0.285092
|
795.5778
|
1.472739
|
Observations
|
144
|
144
|
144
|
144
|
144
|
Source: Generated by Author |
The skewness’ different values are all close to zero (0), which is the normal value for skewness for a normally distributed series. The Kurtosis, that displays the peak of flatness of the distribution of a series, has the value 3 as the normal Kurtosis value for a normally distributed series; the different series in this study have kurtosis values close to 3. Theses give enough idea that the series are normally distributed.
The Jarque-Bera statistics measure the difference the skewness the kurtosis of the variables. The null hypothesis states that the distribution is normal. At 1% significance level, it is safe to accept the null hypothesis. After this test, the logged values of the series were used to perform further tests.
Next, the series stationarity test was performed, and the relative results are as follow:
Table 3
Unit Root Test for Stationarity
UNIT ROOT TEST RESULTS TABLE (ADF)
|
|
|
|
Null Hypothesis: the variable has a unit root
|
|
|
|
|
At Level
|
|
|
|
|
|
|
|
LGDP
|
LBRVM
|
LCPF
|
LINFL
|
LMKC
|
With Constant
|
t-Statistic
|
-2.4177
|
-10.9836
|
-2.3408
|
-3.1960
|
-1.3025
|
|
Prob.
|
0.1387
|
0.0000
|
0.1609
|
0.0223
|
0.6272
|
|
|
n0
|
***
|
n0
|
**
|
n0
|
With Constant & Trend
|
t-Statistic
|
-2.3991
|
-11.0415
|
-2.2295
|
-3.3079
|
-0.6541
|
|
Prob.
|
0.3785
|
0.0000
|
0.4690
|
0.0691
|
0.9737
|
|
|
n0
|
***
|
n0
|
*
|
n0
|
Without Constant & Trend
|
t-Statistic
|
-0.7523
|
0.0300
|
-0.8969
|
-0.1454
|
-0.6330
|
|
Prob.
|
0.3889
|
0.6906
|
0.3259
|
0.6319
|
0.4413
|
|
|
n0
|
n0
|
n0
|
n0
|
n0
|
|
At First Difference
|
|
|
|
|
|
|
d(LGDP)
|
d(LBRVM)
|
d(LCPF)
|
d(LINFL)
|
d(LMKC)
|
With Constant
|
t-Statistic
|
-5.1972
|
-5.8691
|
-1.3978
|
-5.4405
|
-2.5570
|
|
Prob.
|
0.0000
|
0.0000
|
0.5816
|
0.0000
|
0.1047
|
|
|
***
|
***
|
n0
|
***
|
n0
|
With Constant & Trend
|
t-Statistic
|
-5.1808
|
-5.8046
|
-1.3210
|
-5.3427
|
-2.5787
|
|
Prob.
|
0.0002
|
0.0000
|
0.8783
|
0.0001
|
0.2908
|
|
|
***
|
***
|
n0
|
***
|
n0
|
Without Constant & Trend
|
t-Statistic
|
-5.2155
|
-5.8855
|
-1.3094
|
-5.4638
|
-2.5649
|
|
Prob.
|
0.0000
|
0.0000
|
0.1753
|
0.0000
|
0.0105
|
|
|
***
|
***
|
n0
|
***
|
**
|
Notes:
|
|
|
|
|
|
a: (*)Significant at the 10%; (**)Significant at the 5%; (***) Significant at the 1% and (no) Not Significant
|
b: Lag Length based on AIC
|
|
|
|
|
c: Probability based on MacKinnon (1996) one-sided p-values.
|
|
Source: Generated by Author |
The stationarity test depicts that some variables (BRVM and Inflation) are stationary at Level (I(0)) and the rests are I(1). It has been noticed that all the variables are stationary at first difference apart from Capital Flow.
The result for the Bound Test for cointegration is as displayed below;
Table 4
Bound Test for Cointegration
F-Bounds Test
|
Null Hypothesis: No levels relationship
|
Test Statistic
|
Value
|
Signif.
|
I(0)
|
I(1)
|
|
|
|
Asymptotic: n = 1000
|
|
F-statistic
|
2.993358
|
10%
|
2.2
|
3.09
|
k
|
4
|
5%
|
2.56
|
3.49
|
|
|
2.5%
|
2.88
|
3.87
|
|
|
1%
|
3.29
|
4.37
|
Actual Sample Size
|
142
|
|
Finite Sample: n = 80
|
|
|
|
10%
|
2.303
|
3.22
|
|
|
5%
|
2.688
|
3.698
|
|
|
1%
|
3.602
|
4.787
|
Source: Generated by Author |
The F-statistics value is superior to the level value (I(0)) value. This leads to the conclusion that there is enough significance (5%) to reject the null hypothesis of no long-run correlation between the variables. In other others, there is a long-run relationship between the variables.
Both the long and short run models need to be estimated; appropriate techniques are the ARDL and VEC models.
After running the ARDL models, it is necessary to determine the lag length for the afterward-analyses.
Table 5
Lag
|
LogL
|
LR
|
FPE
|
AIC
|
SC
|
HQ
|
0
|
-386.7358
|
NA
|
0.000219
|
5.760821
|
5.867904
|
5.804337
|
1
|
679.2595
|
2037.932
|
4.91e-11
|
-9.547933
|
-8.905436
|
-9.286839
|
2
|
990.2493
|
571.6725
|
7.33e-13*
|
-13.75367*
|
-12.57575*
|
-13.27499*
|
3
|
1010.195
|
35.19851
|
7.93e-13
|
-13.67934
|
-11.96601
|
-12.98309
|
4
|
1018.648
|
14.29610
|
1.02e-12
|
-13.43601
|
-11.18727
|
-12.52218
|
5
|
1053.145
|
55.80269*
|
8.96e-13
|
-13.57566
|
-10.79150
|
-12.44425
|
6
|
1065.952
|
19.77595
|
1.09e-12
|
-13.39635
|
-10.07678
|
-12.04736
|
7
|
1076.532
|
15.55843
|
1.38e-12
|
-13.18429
|
-9.329306
|
-11.61772
|
8
|
1082.882
|
8.872365
|
1.87e-12
|
-12.91004
|
-8.519637
|
-11.12589
|
Source: Generated by Author |
As the most decision criteria (including Akaike Information Criteria - AIC), the optimal lag length is 2.
The resulting ARDL model is determined below:
Table 6
ARDL Model – Short Run Relationship
Variable
|
Coefficient
|
Std. Error
|
t-Statistic
|
Prob.*
|
LGDP(-1)
|
1.561969
|
0.069857
|
22.35951
|
0.0000
|
LGDP(-2)
|
-0.579662
|
0.072438
|
-8.002134
|
0.0000
|
LMKC
|
0.358929
|
0.261641
|
1.371839
|
0.1725
|
LMKC(-1)
|
-0.547928
|
0.262494
|
-2.087395
|
0.0388
|
LINFL
|
-0.034949
|
0.015853
|
-2.204529
|
0.0292
|
LCPF
|
-1.689745
|
0.545708
|
-3.096425
|
0.0024
|
LCPF(-1)
|
3.029527
|
1.041353
|
2.909222
|
0.0043
|
LCPF(-2)
|
-1.403177
|
0.542557
|
-2.586228
|
0.0108
|
LBRVM
|
0.051809
|
0.135294
|
0.382938
|
0.7024
|
LBRVM(-1)
|
-0.118675
|
0.133456
|
-0.889243
|
0.3755
|
LBRVM(-2)
|
0.274655
|
0.135520
|
2.026680
|
0.0447
|
C
|
0.091134
|
0.034009
|
2.679693
|
0.0083
|
R-squared
|
0.986939
|
Mean dependent var
|
0.743008
|
Adjusted R-squared
|
0.985834
|
S.D. dependent var
|
0.217211
|
S.E. of regression
|
0.025853
|
Akaike info criterion
|
-4.392089
|
Sum squared resid
|
0.086886
|
Schwarz criterion
|
-4.142301
|
Log likelihood
|
323.8383
|
Hannan-Quinn criter.
|
-4.290585
|
F-statistic
|
893.0384
|
Durbin-Watson stat
|
2.066688
|
Prob(F-statistic)
|
0.000000
|
|
|
|
*Note: p-values and any subsequent tests do not account for model
|
selection.
|
|
|
Source: Generated by Author |
The results show that 98.58% of change in GDP is explained by the regressors. Inflation and Capital Flow, with enough significance level, have both negative effects on GDP. The results also show that Market Capitalization and BRVM Composite index have positive influence on GDP; however, these variables do not have their coefficients significant at 10%.
The long-run relationship results depict the following
Table 7
VEC Model – Long Run Relationship
Variable
|
Coefficient
|
Std. Error
|
t-Statistic
|
Prob.
|
D(LGDP(-1))
|
0.743704
|
0.069961
|
10.63024
|
0.0000
|
C
|
-0.000246
|
0.002341
|
-0.105022
|
0.9165
|
D(LBRVM)
|
0.002949
|
0.118500
|
0.024884
|
0.9802
|
D(LBRVM(-1))
|
-0.232597
|
0.115459
|
-2.014544
|
0.0460
|
D(LCPF)
|
-1.513056
|
0.593852
|
-2.547869
|
0.0120
|
D(LCPF(-1))
|
1.412658
|
0.593474
|
2.380318
|
0.0187
|
D(LINFL)
|
0.048705
|
0.105546
|
0.461455
|
0.6452
|
D(LINFL(-1))
|
-0.073573
|
0.099911
|
-0.736380
|
0.4628
|
D(LMKC)
|
0.144769
|
0.261334
|
0.553963
|
0.5805
|
D(LMKC(-1))
|
-0.570278
|
0.323002
|
-1.765552
|
0.0798
|
ECT(-1)
|
-0.022552
|
0.013302
|
-1.695332
|
0.0924
|
R-squared
|
0.538897
|
Mean dependent var
|
-0.001108
|
Adjusted R-squared
|
0.503698
|
S.D. dependent var
|
0.037861
|
S.E. of regression
|
0.026673
|
Akaike info criterion
|
-4.336057
|
Sum squared resid
|
0.093197
|
Schwarz criterion
|
-4.107085
|
Log likelihood
|
318.8601
|
Hannan-Quinn criter.
|
-4.243012
|
F-statistic
|
15.31015
|
Durbin-Watson stat
|
2.157282
|
Prob(F-statistic)
|
0.000000
|
|
|
|
Source: Generated by Author |
Following the results, it can be noticed that there is an ECT value of -0.0225. the negative sign indicates that shocks in the short run will be corrected in the long run at 10% significance level. In other words, the models will re-adjust itself in the long run after a shock in the long.
To understand the nature of the relationship between the variables, the Granger Causality test was conducted, and the results are as depicted below. The null hypothesis for the test states that A does not Granger cause B.
Table 8
Null Hypothesis:
|
Obs
|
F-Statistic
|
Prob.
|
Decision
|
LBRVM does not Granger Cause LGDP
|
142
|
3.56159
|
0.0311
|
Reject
|
LGDP does not Granger Cause LBRVM
|
0.32659
|
0.7219
|
Accept
|
LCPF does not Granger Cause LGDP
|
142
|
0.60815
|
0.5458
|
Accept
|
LGDP does not Granger Cause LCPF
|
5.36273
|
0.0057
|
Reject
|
LINFL does not Granger Cause LGDP
|
142
|
0.56083
|
0.5720
|
Accept
|
LGDP does not Granger Cause LINFL
|
0.39282
|
0.6759
|
Accept
|
LMKC does not Granger Cause LGDP
|
142
|
2.93939
|
0.0562
|
Reject
|
LGDP does not Granger Cause LMKC
|
9.91522
|
0.0001
|
Reject
|
LCPF does not Granger Cause LBRVM
|
142
|
4.36648
|
0.0145
|
Reject
|
LBRVM does not Granger Cause LCPF
|
0.47680
|
0.6218
|
Accept
|
LINFL does not Granger Cause LBRVM
|
142
|
2.75601
|
0.0671
|
Reject
|
LBRVM does not Granger Cause LINFL
|
0.12762
|
0.8803
|
Accept
|
LMKC does not Granger Cause LBRVM
|
142
|
2.07924
|
0.1290
|
Accept
|
LBRVM does not Granger Cause LMKC
|
3.26855
|
0.0411
|
Reject
|
LINFL does not Granger Cause LCPF
|
142
|
1.95106
|
0.1461
|
Accept
|
LCPF does not Granger Cause LINFL
|
0.08616
|
0.9175
|
Accept
|
LMKC does not Granger Cause LCPF
|
142
|
0.31692
|
0.7289
|
Accept
|
LCPF does not Granger Cause LMKC
|
2.45837
|
0.0893
|
Reject
|
LMKC does not Granger Cause LINFL
|
142
|
0.14488
|
0.8653
|
Accept
|
LINFL does not Granger Cause LMKC
|
2.94754
|
0.0558
|
Reject
|
Source: Generated by Author |
According to Granger Causality, BRVM Composite Granger causes GDP with the inverse not being true. It simply says that a good performance of BRVM Composite index may have an effect on the Malian GDP and not the other way around. The results also indicated that there is a bidirectional relationship between Market Capitalization and GDP; Market Capitalization Granger causes GDP and at the same GDP Granger causes Market Capitalization.
Moreover, the causality test unveils that GDP causes Capital Flow and not the inverse. Indeed, a high GDP indicated a good economic health and that attracts investor, or foreign direct investment. It has also been determined that Inflation has a unidirectional effect on Market Capitalization.