Dynamic analyses of the relationship between non-performing loans and macroeconomic indicators in the Turkish financial sector

This study applies vector autoregressive (VAR) methodology to analyze the dynamic relationship between Turkish banking sector's non-performing loans (NPL) and the macroeconomic indicators during the period of 2004-2010. The vector autoregressive model is one of the most flexible and easy to use model for the analysis of multivariate time series. The VAR models have been used especially for describing the dynamic behavior of financial time series and also for forecasting. Besides data description and forecasting, the VAR models can also be used for structural inference and policy analysis. On the other hand, restrictions (exogeneity of some of the variables and etc.) may be imposed on VAR models regarding statistical procedures. Variables used in the VAR model are Non-performing loans to total banking sector loans ratio, real effective exchange rate, gross domestic product, nominal interest rates, consumer price index and foreign trade import volume index. All of the historical series consist of quarterly data covering the period of 1997-2003 and 2004-2010. Our impulse response analyses show that it takes at least 8 quarters to see the impact of unexpected changes in the gross domestic product growth rate, the real exchange rate, and the import volume on the non-performing loans while unexpected changes in the nominal interest rate (the policy rate) impact the non-performing loans within six quarters. The results from our variance decomposition analyses suggest that stress tests solely based on macroeconomic indicators, may not give sufficiently early warning signals about the potential upcoming credit crisis and that most powerful predictor of the banking sector non-performing loan ratio is the previous values of the ratio itself. Thus, policies only dealing with macroeconomic factors may not be sufficient to fend off any likely financial crisis in Turkey.