An Early Warning System for Less Significant Italian Banks

Abstract

This paper presents a statistical early warning system for less significant institutions (LSIs) under the direct supervision of the Bank of Italy. The model is calibrated on the basis of a wider definition of possible distress events, using the universe of Italian LSIs active in the period 2008-2016 as a reference. We selected an extensive list of variables that might give early warnings of a crisis in relation both to the overall banking system and Italy’s macro-financial situation, and to individual banks’ performances. A logit model is used to calculate the probability of default (PD) for each bank, with time horizon estimates set at four and six quarters. The empirical specifications proposed are tested using several statistical indices; the ex-post analysis of the estimated PDs shows a percentage of correct bank classification in the range of 80-90 per cent, with better results the nearer the moment of prediction is to when a crisis situation actually occurs.

Keywords: early warning system, default, less significant institutions, CAMELS, banking supervision

JEL Classification: G21, G28