Table 3 Assumption of classical linear regression model

Assumptions

One

The error have zero mean (E (Ɛ) = 0)

Checked by

Adding a constant term (β₀) to the model

Result

The assumption is not violated

Two

Homoscedasticity (variance of the errors term is constant)

Checked by

Breusch–Pagan / Cook–Weisberg test for heteroskedasticity

Result

Since the P value of test statistic is highly insignificant,, there is no heteroskedasticity problem

Three

Multicollinearity

Checked by

Variance inflation factor (VIF)

Result

The finding revealed the values of variance inflation factor (VIF) on each variables is less than 10 and values 1 / VIF is greater than 0.1 or 10%. As a result multicollinearity is not a serious problem in the model

Four

Autocorrelation test

Checked by

Wooldridge test for autocorrelation in panel data

Result

The P value of test static is highly insignificant. Hence, there is no evidence for the existence of autocorrelation problem in the model

Five

Normality test (residuals are normally distributed)

Checked by

Shapiro–Wilk

Result

P- values of the residual is highly insignificant, the residuals has normal distribution pattern

Six

Model specification test

Checked by

Ramsey RESET test

Result

Prob > F = 0.8723 which is highly greater than 0.05. This shows that the null hypothesis of the model, which says no omitted variable

Reason for test

Parameter estimation technique among fixed and random effect model

Types of test done

Hausman specification test

STATA result

Since the P-values is highly insignificant (> 0.05) the researcher is decided that the random effect model is suitable for this data set