Academia.eduAcademia.edu

Figure 2 – uploaded by Syden Mishi

See full PDFdownloadDownload figure
Here the effects of explanatory variables are mediated by the differences across banks. Meaning by including dummy for each bank the pure effect of exogenous factors is estimated through controlling for the unobserved heteroge- neity. Each dummy is absorbing the effects particular to each bank. The results in table 3 below justify the inclusion of dummy variable given the effect on the results, also revealed by an increase in the explanatory power of the models as shown by a sharp increase in r-squared (to over 43%) and adjusted r-squared (of 41%) from as low values as 10%. TABLE 2: LEAST SQUARE DUMMY VARIABLE MODEL (LSDV)

Table 3 Here the effects of explanatory variables are mediated by the differences across banks. Meaning by including dummy for each bank the pure effect of exogenous factors is estimated through controlling for the unobserved heteroge- neity. Each dummy is absorbing the effects particular to each bank. The results in table 3 below justify the inclusion of dummy variable given the effect on the results, also revealed by an increase in the explanatory power of the models as shown by a sharp increase in r-squared (to over 43%) and adjusted r-squared (of 41%) from as low values as 10%. TABLE 2: LEAST SQUARE DUMMY VARIABLE MODEL (LSDV)

Related Figures (5)

TABLE |: POOLED ORDINARY LEAST SQUARES ESTIMATION RESULTS From this analysis, size, industry concentration and interaction between the two are significant. With the latter positively related to credit risk. As size in- creases, credit risk exposure declines. This may be explained by availability of resources as cushion, diversification of income sources as most of the smaller banks concentrate on one business line. With concentration, as the industry be- come highly concentrated, risk taking declines in line with the findings of Levy and Micco (2007) Marquez (2002), Matutes and Vives (2000) arguing that with high competition (low concentration) borrow-specific information is dispersed
TABLE |: POOLED ORDINARY LEAST SQUARES ESTIMATION RESULTS From this analysis, size, industry concentration and interaction between the two are significant. With the latter positively related to credit risk. As size in- creases, credit risk exposure declines. This may be explained by availability of resources as cushion, diversification of income sources as most of the smaller banks concentrate on one business line. With concentration, as the industry be- come highly concentrated, risk taking declines in line with the findings of Levy and Micco (2007) Marquez (2002), Matutes and Vives (2000) arguing that with high competition (low concentration) borrow-specific information is dispersed
TABLE 3: COMPARING POLS anp LSDV According to Green (2000), in order to choose between fixed or random ef- fects a Hausman test is run with the null hypothesis that the preferred model
TABLE 3: COMPARING POLS anp LSDV According to Green (2000), in order to choose between fixed or random ef- fects a Hausman test is run with the null hypothesis that the preferred model
With the Prob>F being less than 5%, our model fits well, as all the coeffi- cients in the model are different from zero. Given the two-tail p-values: size, bhi, size*hhi and expo all have significant effect on credit risk. The same con- clusions can be drawn from the t values at 95% confidence interval. Further- more the intra-class correlation (‘rho’) tells us that 87% of the variance in credit risk is due to differences across panels. TABLE 4: FIXED EFFECTS MODEL
With the Prob>F being less than 5%, our model fits well, as all the coeffi- cients in the model are different from zero. Given the two-tail p-values: size, bhi, size*hhi and expo all have significant effect on credit risk. The same con- clusions can be drawn from the t values at 95% confidence interval. Further- more the intra-class correlation (‘rho’) tells us that 87% of the variance in credit risk is due to differences across panels. TABLE 4: FIXED EFFECTS MODEL
Finally we econometrically compare our models to see the power of each and as a robustness check mechanism to ensure that the results we presented from the chosen model can be used to draw policy conclusion. TABLE 5: AREG ESTIMATIONS: LINEAR REGRESSION, ABSORBING INDICATORS
Finally we econometrically compare our models to see the power of each and as a robustness check mechanism to ensure that the results we presented from the chosen model can be used to draw policy conclusion. TABLE 5: AREG ESTIMATIONS: LINEAR REGRESSION, ABSORBING INDICATORS
TABLE 7: COMPARING FIXED EFFECT ASSUMPTION, OLS AND AREG
TABLE 7: COMPARING FIXED EFFECT ASSUMPTION, OLS AND AREG
Related topics:
BusinessEconomicsIndustry Concentration
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved