An Application to Model Uncertainty in Modelling Non-Performing Loans

[EN] The asset quality of a banking system has the utmost importance not only for the soundness of the banking sector, but also for other major components of an economy. The lack of consensus on a certain set of variables for modeling asset quality leads to a problem, which is also known as “model uncertainty”. In this study, we investigate how non-performing loans in the Turkish banking system respond to changes in macroeconomic and bank-specific variables. To address model uncertainty, we employ a model averaging approach. Our results confirm the main findings in the literature with regards to the nexus between asset quality and macroeconomic and bank-specific variables. In addition, parameter estimates obtained from 1,023 models using combinations of 10 variables suggest that even under extreme shocks, the NPL ratio of the Turkish banking sector remains within reasonable limits. [TR] Bankacilik sisteminin aktif kalitesi, sadece bankacilik sektorunun saglamligi acisindan degil, ekonominin diger temel bilesenleri acisindan da buyuk onem arz etmektedir. Aktif kalitesinin modellenmesinde belirli bir dizi degisken uzerinde fikir birliginin bulunmamasi, “model belirsizligi” olarak da bilinen bir soruna yol acmaktadir. Bu calismada, Turk bankacilik sistemindeki takipteki kredilerin makroekonomik ve bankaya ozgu degiskenlerdeki degisimlere nasil tepki verdigi incelenmektedir. Model belirsizligi sorununu gidermek icin model ortalamasi yaklasimi kullanilmaktadir. Sonuclarimiz, aktif kalitesi ile makroekonomik ve bankaya ozgu degiskenler arasindaki baglantiya iliskin literaturdeki ana bulgulari dogrulamaktadir. Ayrica, 10 degiskenin kombinasyonu kullanilarak 1023 modelden elde edilen katsayi tahminleri, asiri soklarda bile takipteki kredi oraninin makul sinirlar icinde kaldigini ortaya koymaktadir.