The impact of the banking sector development on the financial performance of the communication sector in sierra leone

4.10.2 Test for Serial Correlation 
The Breusch-Godfrey test proposed by (Godfrey, 1978) for serial correlation in 
case different lags of residuals are correlated. Serial correlation do affects the 
efficiency of the regression estimators but does not affects the unbiasedness, 

---

61 
that is the estimator are not BLUE. Serial correlation may affect the regression 
standard error which may invalidate significance test. In such cases it is 
however possible that wrong inferences could be made as to how the 
independent variables are determinants of the dependent variable variations. 
The Breusch-Godfrey test general hypothesis are as follows;
H0: No serial correlation in the model 
H1: serial correlation in the model 
4.10.3 Test for Heteroscedasticity 
This test, test for constant variance in all residuals. ARDL model and OLS 
estimation assumed that the residuals have a constant variance that is 
homoscedasticity. On the other hand, in cases where the model does not have 
a constant variance that is, heteroscedasticity in the residuals in this case the 
estimated coefficients will no longer be BLUE and it will not have minimum 
variance of the unbiased estimators. This thesis will make use of the Whites 
test for heteroscedasticity with the following hypothesis;
H0: Residuals with Constant variance 
– Homoscedasticity 
H1: Residuals with Non-constant variance - Heteroscedasticity 
4.10.4 Regression 
Specification Error Test 
The test for misspecification in this thesis is conducted using the Ramsey 
Regression Specification Error Test (RESET). The Ramsey (1969) test for 
functional form, that is, if non-
linear combinations of the fitted values can well 
describe the explanatory variable. The model is said to be misspecified if non-
linear combinations of the fitted values have power in describing the 
independent variables and needs to be adjusted. The null and alternative 
hypothesis is as follows: 
H0: Non-linear combinations with no power - 
No misspecification
H1: Non-linear combinations with power - 
Misspecification.