Risk Assessment of Non-Performing Loans (NPL) Using Extreme Value Theory

This paper answers two research questions: what is the appropriate modeling tool for NPL study? and whether the NPL rates in Thailand show improving or deteriorating trend? NPL is of interests to management decision makers because it serves as an indicator for assessing risk in commercial loans. The methodologies employed in this research are system analysis under Extreme Value Theory (EVT) and relevant tests suggested by the Tail index. Conventional time series models, such as AR, MA, ARIMA and ARIMAV are not adequate to deal with short term risk assessment due to restrictive assumptions in autoregressive modeling and the requirement of longer observation period. This research fills the gap in the literature left by conventional time series modeling. NPL data of ten industries in Thailand for a period of eight operating quarters from 2013 to 2014 were used. It was found that there are two types of distribution among NPL rates. Eight industries are distributed Weibull and two industries are distributed Fréchet. This finding further reinforces the use of EVT as an appropriate modeling tool for NPL studies. Under EVT, it was found that the risk threshold for Thailand's NPL is 10.91 where three industries showing NPL near or beyond this threshold. The proposed method of NPL analysis in this paper provides a practical tool for decision makers in NPL management