Microsoft Word Thesis Gent (1). doc

6.2 Data and variables description 
I collected time-series quarterly data covering the period from, 2002 to 2011, including 
main macroeconomic variables and credit risk in banking sector in Czech Republic. 
Data were collected from ARAD time-series database provided by CNB official web 
page and OECD online statistic library. In order to quantify the credit risk as a key 
measure for quality of loan, I used the NPL ratio. As main macroeconomics variables 
causing the NPL ratio, I used variables such as real GDP, the long-term interest rate 
(IR), the exchange rate of CZK/UDS (ER), the inflation (CPI), and unemployment rate 
(UNEMP).

---

48 
The VAR methodology is used to estimate the empirical model over the quarter period 
of 2002 to 2011. In order to assess the impact of macroeconomic variables in credit risk 
(NPL ratio) the impulse response analysis is used (response of NPL ratio to 
macroeconomic shocks). Data that I collected are expressed in percentage (the long-
term interest rate and unemployment rate) except real GDP, exchange rate and CPI they 
were not expressed in percentage, therefore I used logarithm of these variables. 
Variables were not stationary regarding Argumented Dicky-Fuller test (ADF), 
(Appendix 1). Thus, in purpose to proceed further and to estimate my model by using 
unrestricted VAR method, and to observe the impact of macroeconomic shocks reaction 
on credit risk, data has to be stationary. I used the level of first log difference in order to 
transform data into stationary. Unit root test for cointegration specifies that the 
variables are level stationary
series (see Appendix 1) and that all variables are I (1). 
While, in terms of long-term interest rate, exchange rate and inflation, I used the 
seasonally adjusted (using Eviews software) in order to remove the seasonality.