Bootstrap VAR forecasts: The effect of model uncertainties

VAR models are popular to forecast macroeconomic time series. However, the model, parameters, and error distribution rarely known without uncertainty, so bootstrap methods applied deal with these sources of uncertainties. In this paper, performance Bonferroni cubes based on Gaussian method variants procedure that incorporate distribution, parameter bias correction, lag order uncertainty compared. Monte Carlo simulations suggest best obtained when is considered, being model uncertainties important for long-run regions in persistent models. Similar conclusions found an empirical application a four variate system containing US monthly percent changes industrial production index, S&P500 stock market its dividend yield, unemployment rate.