Results and interpretation

People used at least basic drinking water. Based on GLS results, CAR has negative effect on ROA. When CAR increases by one unit it causes to decrease ROA by 0.0186824 units. This impact is significant, because p value is equal to 0.6% which is the significant in 1% level. This result supports results of Acaracvi and Calim (2013) who studied bank profitability determinants by testing three models which dependent variables were different and they showed that negative results could be obtained due to ownership type or centralized market.
Urban population of total population. According to results CIR variable has negative and very significant relationship with bank profitability and impact of cost income ratio was the highest among other variables. If CIR increases by 1 unit, bank profitability will increase by 0.0332123 units due to the impact of CIR. Works of Bourke (1989) and Smirlock and Marshall (1983) served as the proof of our studies, because they were also found negative relationship and the reason of this result is due to banks located in monopolistic markets.
Now we can’t conclude our work without testing unit root test. If there is unit root our whole work will be useless. Figure 13 illustrates the result of unit root test and our hypothesis is following:
H_0: Panels contain unit roots
H_A: At least one panel is stationary

Figure 15. Source: research findings
According to the result we found that probability was 0 and less than 5 % and we have to reject null hypothesis states that panels have unit roots and accept alternative hypothesis which presents one panel as stationary one. So our model does not have any unit roots and we can accept GLS model.
Before conclusion we decided to test our model by using robust standard errors for panel regressions with cross-sectional dependence or heteroscedastic, autocorrelated with MA(q), and cross-sectionally dependence test.

In order to ensure we have valid statistical inference when some of the underlying regression model’s assumptions are violated, we used this test. Our results provided that the residuals are independently distributed, standard errors which are obtained by aid of this estimator are consistent even if the residuals are heteroscedastic. In other words, we ensure again that we do not have heteroscedastic problem.
Overall, we used 4 independent variables in order to define mortality rate under five year old children per thousand live births and three of them showed very significant results except unemployment rate of countries. Also, these correlations were found based according to the previous research papers and this research paper support their outcomes. All regression analyses were done by “Stata14” software and results were copied without any corrections.