Impact of tax revenue on economic growth in Rwanda from 2007-2017

4.9 VECTOR ERROR CORRECTION MODEL

As the variables were non stationary at their levels integrated of order I(1),stationary at first difference and cointegrated ,we analyzed the short run relationship among them by formulating an error correction model. The logic behind that model is to recover the long run information lost by differencing variables by introducing an error correction term gives the proportion of shocks accumulated in the previous period that are corrected in the current period. The results of VECM are presented in the following table.

Table 9: Vector Error correction model Results

Dependent Variable: D(LGDP)

Method: Least Squares

Date: 11/02/19 Time: 14:20

Sample (adjusted): 2007Q3 2017Q4

Included observations: 42 after adjustments

D(LGDP) = C(1)*( LGDP(-1) - 0.798843352892*LDT(-1) + 0.060223872029

1*LTGS(-1) + 0.0334606747663*LTITT(-1) - 4.08820567175 ) + C(2)

*D(LGDP(-1)) + C(3)*D(LDT(-1)) + C(4)*D(LTGS(-1)) + C(5)*D(LTITT(

-1)) + C(6)
	Coefficient		Std. Error		t-Statistic	Prob.
C(1)	-0.142323		0.07576		-1.878597	0.0684			C(2)
0.205349	0.172688		1.189129		0.2422	C(3)
-0.043396	0.046513		-0.932985		0.357	C(4)
0.026459	0.073927		0.357907		0.7225	C(5)
-0.008059	0.030651		-0.262929		0.7941

P a g e 38 | 48

C(6)

	0.026192	0.006795 3.854348	0.0005
R-squared	0.127675	Mean dependent var	0.03194
Adjusted R-squared	0.006519	S.D. dependent var	0.027286
S.E. of regression	0.027197	Akaike info criterion	-4.239863
Sum squared resid	0.026628	Schwarz criterion	-3.991624
Log likelihood	95.03712	Hannan-Quinn criter.	-4.148873
F-statistic	1.053803	Durbin-Watson stat	1.884039
Prob(F-statistic)	0.401802

Source: Eviews 7,2019

Table 6 shows the findings of VECM and the results confirm that only the speed of adjustment of the model is 14.2% with error correction of -0.142323 and it is statistically significant at 10%. This implies that 14.2% of errors realized in the previous Quarter are corrected in the current one. This means that each quarter 14.2% of disequilibrium errors will be corrected due to any change from the equilibrium.

4.7 GRANGER CAUSALITY TESTS

Granger Causality tests clarified how the variables affect (drive) each other. The results are

presented in Table 10 below.

Table 10: Granger Causality Tests

Null Hypothesis:	Obs F-		Prob.	CONCLUSION
LDT does not Granger Cause				LDT does not Granger Cause
LGDP	43	c 2.74474	0.1054	LGDP
LGDP does not Granger Cause			1.00E-	LGDP does Granger Cause
LDT		25.5152	05	LDT
LTGS does not Granger Cause				LTGS does not Granger Cause
LGDP	43	1.81716	0.1852	LGDP
LGDP does not Granger Cause				LGDP does Granger Cause
LTGS		9.12791	0.0044	LTGS
LTITT does not Granger Cause				LTITT does not Granger Cause
LGDP	43	0.00221	0.9627	LGDP
LGDP does not Granger Cause				LGDP does Granger Cause
LTITT		5.68013	0.022	LGDP

Source: Elaborated by research using eviews 8,2019

P a g e 39 | 48

In this section, the study seeks to establish if there is evidence of a causal relationship between the variables of interest.

? Since P-value=1.00E-05 or 0.001% is less than 5%, we reject null hypothesis of LGDP does not Granger Cause LDT in order to accept alternative hypothesis of LGDP does Granger Cause LDT. therefore, the results reflect that there is evidence of uni-directional causality from LGDP to LDT.

? Since P-value=0.0044 or 0.4% is less than 10%, we reject null hypothesis of LGDP does not Granger Cause LTGS in order to accept alternative hypothesis of LGDP does Granger Cause LTGS. therefore, the results reflect that there is evidence of uni-directional causality from LGDP to LTGS.

? Since P-value=0.022or 2.2% is less than 10%, we reject null hypothesis of LGDP does not Granger Cause LTITT in order to accept alternative hypothesis of LGDP does Granger Cause LTITT, therefore, the results reflect that there is evidence of uni-directional causality from LGDP to LTITT.