A Macro Stress Test Model of Credit Risk for the Brazilian Banking Sector

Abstract:     
This paper proposes a model to conduct macro stress test of credit risk for the banking system based on scenario analysis. We employ an original bank-level data set that splits bank credit portfolios in 21 granular categories, encompassing household and corporate loans. The results corroborate the presence of a strong procyclical behavior of credit quality, and show a robust negative relationship between (the logistic transformation of non-performing loans (NPLs) and GDP growth, with a lag response up to three quarters. The results also indicate that the procyclical behavior of loan quality varies across credit types. The latter result, which is novel in the literature, suggests that banks with larger exposures to highly procyclical credit types and economic sectors would tend to undergo sharper deterioration in the quality of their credit portfolios during the economic downturn. Lack of sufficient portfolio granularity in macro stress testing fails to capture these effects and thus introduces a source of bias that tends to underestimate the tail losses stemming from the riskier banks in a system.

Number of Pages in PDF File: 36

Keywords: banking system, stress tests, financial crisis, credit risk

JEL Classification: G1, G15, G32

working papers series