The results of the diagnostic test in Table 6 for model ROA indicate the presence of heteroskedasticity. There is no first order serial correlation for the model. In terms of the ROE model, the study finds first order serial correlation but no heteroskedasticity. Thus, in order to get the optimal result, the Panel-Corrected Standard Errors (PCSE) model is chosen. According to Beck and Katz (1995) and Tas et al. (2013), this model is the best option in rectifying the problem of heteroskedasticity, first order serial correlation and cross sectional correlation. Table 7 presents the results of the PCSE model for both ROA and ROE. It is found that the results are robust to heteroscedasticity and first order serial correlation and are classified as the Best Linear Unbiased Estimates (BLUE). The next section discusses the summary of findings for this study.