Results and interpretation

Figure 13
1	Dependent variable is MOR5	GLS
2	Independent variables
3	UNMPL	-.6475793 (-1.53)
4	LNPERCAPGDP	13.35622 (2.82)
5	lgPUD	-133.6972 (-14.93)
6	lgURPOP	57.21655 (5.97)
7	Constant	325.7052 (13.00)
8	Number of observations	126
9	t statistics in parentheses
10	* p<0.05, ** p<0.01, *** p<0.001

GLS regression analysis was done and the results are illustrated in Figure 13 and 14. GLS model relaxed heteroskedasticity and serial correlation models and gave final outcomes of our model. As it is given in figure 14 Wald chi2 is 280.7 higher than probability which means our model is significant. Also, our p value is low except p value of Unemployment rate which means our variables are also significant.

Figure 14. Source: research findings
Finally we found following results for our model:
MOR= 325.7052 – 0.6475793*UNMPL + 0.13.35622* LNPERCAPGDP –133.6972*LGPUD + 57.216555*LGURPOP+µ
Unemployment rate. According regression analysis if bank size increase by 1 % it leads to decrease ROA by 0.0023825 units and this impact is significant at 0.3% level. Many researchers stated that bank size had positive and significant impact on bank profitability, but Goldberg and Rai (1996) stated that if concentration level was higher than average it led to negative correlation of bank size and ROA. In the case of Uzbekistan, the result became different
GDP Per Capita. Refinancing rate effects on ROA negatively, but not significant, because p value is high. Overall if refinancing rate increases by 1 unit it leads to decrease ROA by 0.0267108 units. Our results reject several previous works such as published papers by Molyneux and Thornton (1992), Jiang et al. (2003) and Alessandri & Nelson (2015), but follows the results of Simiyu and Ngile (2015) who analyzed the effect of GDP, exchange rate and interest rate on profitability in Kenya for the period between 2001 and 2012 and found negative and strong relationship.