Internal Model Validation in Brazil: Analysis of VaR Backtesting Methodologies

Abstract

The purpose of this paper is to analyze backtesting methodologies of VaR, focusing on aspects as suitability to volatile markets and limited data set. We verify, from regulatory standpoint, tests to complement the Basel traffic light results, using simulated and real data. The results indicate that tests based on failures proportion are not adequate for small samples even fro 1,000 observations. The Basel criterion is conservative and has low power, which does not invalidate its application, as the criterion is only one of the procedures adopted in internal model validation process. Thus, it is suggested using tests that capture the shape of returns distribution, as the Kuiper test, in addition to the Basel criterion.