Welfare implication of determinants affecting aggregate consumption expenditures in Rwanda

3.4 Diagnostic tests

After testing the short run equation, the supplementary tests are necessary to verify if the hypothesis of classical regression are confirmed.

3.4.1 Jarque-bera test (Normality test)

The result of this test arises on the below mentioned graph:

^{Series: Residuals}

^{Sample 1995 2015}

^{Observations 21}

^Mean ^Median ^Maximum ^Minimum ^Skewness ^Kurtosis ^Jarque-Bera ^Probability
^6 ^5 ^4 ^3 ^2 ^1 ^0
	^-0.06 -0.04 -0.02 0.00 0.02 0.04

Source: World Bank indicators1995-2015 and author's computation. Figure 3: Jarque-bera Test output

59

The assumptions of these tests are below mentioned:

H0 (Null hypothesis): The residuals are normally distributed

^{Std. Dev.}

H1 (The alternative hypothesis): The residuals are not normally distributed.

The null hypothesis is not rejected because the probability of 33% is greater than 10%.

The null hypothesis is not rejected means that residuals are normally distributed.

60

3.4.2 Breusch-Godfrey test (Serial correlation LM test) The E-views 8 estimation output is the following:

Breusch-Godfrey Serial Correlation LM Test:

F-statistic 0.424899 Prob. F(2,14) 0.6620

Obs*R-squared 1.201752 Prob. Chi-Square(2) 0.5483

Source: World Bank indicators1995-2015 and author's computation Table 8: Serial correlation tests

The assumptions for this test are the following: H0: no serial correlation (errors are not correlated). H1: There is serial correlation.

The null hypothesis is not rejected when the probability is less than 10%.

The probability of obs* R-squared is 54% greater than 10% which means that the model has not the errors of residuals autocorrelation.

3.4.3 Heteroscedasticity Test (Breusch Pagan Godfrey)

Heteroscedasticity Test: Breusch-Pagan-Godfrey

F-statistic	0.135075	Prob. F(4,16)	0.9670
Obs*R-squared	0.685978	Prob. Chi-Square(4)	0.9530
Scaled explained SS	0.308383	Prob. Chi-Square(4)	0.9893

Source: World Bank indicators1995-2015 and author's computation. Table 9: Heteroscedasticity Test

The assumptions for this test are the following: Ho: the model is not homoscedastic

H1: the model is heteroscedastic

61

The probability of Scaled explained SS (Chi- square) 98% is greater than 10% level of significance which means that the model is homoscedastic.

3.5 Stability tests

3.5.1 Ramsey reset test

The test above mentioned indicates whether the model is well specified or not. The assumptions for this test are as follows;

H0= the model is specific.

H1= the model is not specific.

Ramsey RESET Test Equation: UNTITLED

Specification: LNGCE LNGDP INT INF LNEXCH C

Omitted Variables: Squares of fitted values

	Value	df	Probability
t-statistic	1.271153	15	0.2230
F-statistic	1.615829	(1, 15)	0.2230
Likelihood ratio	2.148417	1	0.1427

Source: World Bank indicators1995-2015 and author's computation Table 10: Ramsey reset Test

From the above finding, the probability of 22% is greater than 10% level of significance which means that we accept the null hypothesis therefore the specification of the model is good.in other words, the model is BLUE ( Best Linear Unbiased Estimator).