A macro stress test model of credit risk for the Brazilian banking sector

Outline

• Abstract
• Jel Classification
• Keywords
• 1. Introduction
• 2. Literature Review
• 3. Methodology
• 3.1. Overview of the Methodology
• 3.2. the Macro Model
• 3.3. Microeconomic Model
• 4. Stress Tests
• 4.1. Simulation of Npls Under Alternative Scenarios
• 4.2. Portfolio Aggregation Bias
• 4.3. Credit Var
• 5. Final Considerations
• References

5. Final considerations

The econometric estimations presented in this paper provide strong evidence of a cyclical behavior of loan quality. The estimations substantiate the existence of a robust inverse relationship between GDP growth and NPLs, with the effects operating with up to three quarter lags. The results also indicate differences in the persistence of NPLs across credit types, and in their sensitivity to economic activity. Loan quality in Brazil appears to be more sensitive to GDP growth for small consumer loans, credit to agriculture, sugar and alcohol, livestock, and textile. In addition, credit for vehicle acquisition and electric and electronic equipment displayed high level of NPLs under distressed macroeconomic scenarios.

Banks with relatively higher exposures to these sectors are likely to experience larger credit losses under a macroeconomic downturn. While intuitive, the modeling of differences in the sensitivity of loan quality to macroeconomic conditions at the level of individual banks is novel to the macro stress testing literature. Existing models based on bank-level data do not allow for a differential response of credit quality to macroeconomic conditions across credit types, possibly due to lack of data availability. On the other hand, existing macro stress test models that exploit variations in the sensitivity of loan quality to macroeconomic conditions across credit types are based on aggregated data, and are therefore less suited to assess the adequacy of bank capital at the level of individual institutions.

The results presented in this paper show that the lack of sufficiently granular data on the composition of bank credit portfolios can bias the results in a way that is contrary to a prudent criterion.

Totheextentthatthesensitivityofcreditqualitytomacroeconomic conditions varies between different credit types, the lack of differentiation would tend to underestimate the deterioration of credit quality for the highly procyclical credit types and sectors under a distressed macroeconomic environment (and overestimate the deterioration of credit quality for the relatively safer credit types).

These biases would translate in a systematic way into the assessment of bank risk profiles, leading to an underestimation of tail losses in the more vulnerable institutions in the banking system under analysis.

The model presented in this paper represents an improvement over existing literature but is still subject to several important caveats. First, the model assumes a linear relationship between loan quality and macroeconomic conditions which may fail to capture potential non-linear relationships during periods of severe macroeconomic distress. Second, the model assumes that historic correlations between loan quality and macroeconomic conditions are symmetric during the upturn and the downturn of the economic cycle, and remain valid during periods of severe distress. Third, the model fails to capture potential feedback effects between credit quality and economic growth, as it does not fully integrate the macro and microeconomic modules. In particular, the macroeconomic module allows credit volumes to vary over time, while the microeconomic module assumes that individual banks maintain constant credit portfolios. To the extent that credit quality tends to deteriorate during periods of slow credit growth, the model presented in this paper may underestimate banks’ loan losses. All these caveats may likely bias the results in the same direction during periods of financial distress, causing a potential underestimation of bank credit losses. Further analysis is needed to address these shortcomings.