How Robust is the Var-Based Expectation of the Federal Reserve Board (FRB/US) to Model-Selection Uncertainty?

In the spring of 1996, the first working version of the FRB/US model replaced the venerable Monetary Policy System (MPS) model for forecasting and policy analysis in the US. The agents in the economy use a VAR-based expectation model, auxiliary to the large and complicated FRB/US model. This paper focuses on examining the robustness of the VAR-based expectations of the FRB/US model to the model-selection uncertainty -- that is, the uncertainty about the true lag-order of the autoregressive processes. The paper estimates the degree of model-selection uncertainty in this VAR-based expectation model and examines its effect on the estimated impulse responses. It also examines the sensitivity of the monetary policy recommendations to changes in the lag order of the VAR-based expectations of the FRB/US model. The analysis found that this model is very robust to model-selection uncertainty. The analysis suggests that the agents should use VAR (1) representation as the VAR (1) forecasts better. Using the variance decompositions, we show that the policy inferences are not sensitive to the changes in the lag order of the VAR-based expectation of the FRB/US model