Microsoft Word Thesis Gent (1). doc

6.3 Empirical results 
In order to interpret the results that macroeconomic shocks cause the credit risk, the 
impulse response function is used. The VAR model allows doing estimation of 
macroeconomic shocks and their impact on credit risk (NPL ratio). The VAR functions 
show the period of a variable in response (NPL ratio) to a shock in macroeconomic 
variables. This shows which, shock provokes the variation in an endogenous variable 
(NPL ratio).
 
The execution is done by identifying the structural innovations from the 

---

52 
reduced form residuals, which in this case is used as Choleski-decomposition as it 
imposes zero restrictions.  
In addition, to verify that the lags of endogenous variable are significant to enhance the 
estimated performance of the other variables for the credit risk (NPL ratio), the Granger 
causality test was used. Specifically, these are pair wise tests of the null hypothesis that 
the coefficients of an endogenous variable in equation equal zero and thus Granger 
causes it. Hence, if the null hypothesis is rejected by a standard F-test, it means that 
there is inclusion of a specific variable in the equation of other variable
The Granger causality tests on my model, shows that long term interest rate (IR_SA) 
does not Grangerly causes credit risk (NPL ratio), which means that long-term interest 
rate does not help to predict the credit risk (see table 2). Therefore, null hypothesis that 
long-term interest rate does not Grangerly causes NPL ratio is not rejected. The impact 
of inflation (L_CPI_SA) to credit risk (NPL ratio), is significant on 90% of confidence 
interval with the p-value 0.0647. So, null hypothesis has been rejected. Hence, inflation 
does Granger causes credit risk, and helps on predicting it. Important to note is that 
inflation has impact on credit risk, but credit risk does not Granger causes the inflation. 
In terms of exchange rate (L_ER_SA) as impact on credit risk Granger causality test 
shows that exchange rate does not Grangerly Causes credit risk. So the exchange rate 
does not help to predict the credit risk. Therefore, null hypothesis is not rejected. 
According to GDP (L_GDP) and unemployment rate (UNEMP), results show that both 
macroeconomic variables do not Granger causes the credit risk; even credit risk do not 
Granger causes shocks on GDP and unemployment rate. Thus it means that GDP and 
Unemployment rate, referring to Granger causality test do not help to predict the credit 
risk and null hypothesis are not rejected. In a nutshell, almost all macroeconomic 
variables are insignificant and do not Granger causes the credit risk, except the 
inflation, which has a significant impact on credit risk.