摘要：We show how fan charts generated from Bayesian vector autoregression models can be useful for assessing (i) the effect of the zero-lower-bound constraint on forecasting uncertainty and (ii) the credibility of stress tests conducted to evaluate financial stability. To illustrate these issues, we use a data set for the Czech Republic and macroeconomic scenarios that are used by the Czech National Bank (CNB) in stress tests of the banking sector. Our results demonstrate how different modeling approaches to the zero lower bound affect the resulting fan charts. The pros and cons of the considered methods are discussed; ignoring the zero-lower-bound constraint represents the worst approach. Next, using our fan charts, we propose a method for evaluating whether the assumptions that are employed in the bank’s stress tests regarding the macroeconomic outlook are sufficiently adverse and consistent with past cross-correlations observed in the data. We find that CNB stress tests are sufficiently conservative in this respect
