Are Bayesian Fan Charts Useful for Central Banks? Uncertainty, Forecasting, and Financial Stability Stress Tests

This paper shows how fan charts generated from Bayesian vector autoregression (BVAR) models can be useful for assessing 1) the forecasting accuracy of central banks’ prediction models and 2) the credibility of stress tests carried out to evaluate financial stability. Using unique data from the Czech National Bank (CNB), we compare our BVAR fan charts for inflation, GDP growth, interest rate and the exchange rate to those of the CNB, which are based on past forecasting errors. Our results suggest that in terms of the Kullback-Leibler Information Criterion, BVAR fan charts typically do not outperform those of the CNB, providing a useful cross-check of their accuracy. However, we show how BVAR fan charts can rigorously deal with the non-negativity constraint on the nominal interest rate and usefully complement the official fan charts. Finally, we put forward how BVAR fan charts can be useful for assessing financial stability and propose a simple method for evaluating whether the assumptions of banks’ stress tests about the macroeconomic outlook are sufficiently adverse.