Modeling the Loss Given Default of Retail Contracts

Abstract

The presented thesis consists of four essays dealing with the modeling and estimation of the LGD for retail contracts. In brief, four concepts are presented. First, the recommendation for unfinished defaults inclusion in the modeling sample is determined, as an inevitable part of the process not well developed in the literature so far. Second, the inclusion of new risk drivers connected to client behavior after granting credit is analyzed. Third, a new form of LGD decomposition is proposed, based not directly on the LGD distribution but rather on events that leads to the bi-modal shape. At last, forecast averaging way of macroeconomic variables inclusion in the LGD model is presented as a possible technique to combine idiosyncratic bank data with systematic factors related to macroeconomics.