Abstract
We compare the forecasting performance of small and large Bayesian vector-autoregressive (BVAR) models for the United States. We do the forecast evaluation of the competing models for the sample that ends before the pandemic and for the sample that contains the pandemic period. The findings document that these models can be used for structural analysis and generate credible impulse response functions. Furthermore, the results indicate that there are only small gains from the application of a large BVAR model compared to a small BVAR model.