Forecasting how credit quality evolves in response to macroeconomic conditions is of utmost importance to assess the resilience of the banking system, particularly in countries such as Italy where traditional credit provision is by far the most significant activity performed by banks. Furthermore, stress tests, a useful tool in assessing the stability of financial intermediaries, often require conditional forecasts for credit quality measures under one or more predetermined scenarios for a set of macroeconomic variables.
This note describes an empirical model that links a measure of credit quality, the new nonperforming loans rate (NNPL rate), to macroeconomic and financial variables, and shows the results for Italy. A Bayesian approach is adopted to select econometric specifications and obtain Bayesian averages of models. Our main results show that the evolution of the NNPL rate over the past several years is tracked by two equations that include a small set of predictors, and that model uncertainty is limited. The selected models and the Bayesian averages of models can be used for short-term forecasts, scenario analysis and stress testing.
The variables that are significant in predicting the NNPL rate of nonfinancial firms are: the interest rate on loans, the growth rates of credit and GDP, firms' financial leverage, and the change in unemployment. The rise in the NNPL rate of Italian firms between 2008 and 2009 is consistent with the severe recession, the credit crunch, and the high level of leverage of Italian firms at the onset of the downturn. The further increase in the NNPL rate after 2011 is consistent with the observed increase in the cost of credit and the contraction in credit. The evolution of the NNPL rate of households is predicted quite accurately by the dynamics of GDP and credit for most of the examined period; after 2013, low interest rates and renegotiations of mortgage contracts helped to keep the NNPL rate of households below what would have been expected given the prevailing macroeconomic conditions.
