Macroeconomic forecasting during recessions and expansions in the US and the euro area

Abstract This study systematically evaluates forecasting performance of 11 Dynamic Stochastic General Equilibrium (DSGE) and 2 Bayesian Vector Autoregression (BVAR) models during recessions and expansions in the US and the euro area. Results show that no single model dominates: parsimonious models perform well in stable periods and at short horizons, while richer DSGE specifications with financial frictions, flexible inflation targeting, or labor market dynamics improve forecasts during recessions. BVARs excel in interest rate forecasting, especially in expansions. Crisis‐specific extensions, such as COVID‐related shocks, yield temporary gains. Forecast accuracy depends on the economic state, variable, horizon, and evaluation metric, underscoring the need for a diversified, context‐dependent modeling toolkit.