Title: An empirical investigation of direct and iterated multistep approaches to producing conditional forecasts
Authors: Michael McCracken - Federal Reserve Bank of St. Louis (United States) [presenting]
Abstract: When constructing an unconditional point forecast, both direct and iterated multistep (DMS and IMS) approaches are common. However, in the context of producing conditional forecasts, IMS approaches based on vector autoregressions (VAR) are far more common than simpler DMS, horizon-specific autoregressive-distributed lag (ARDL) models. This is despite the fact that there are theoretical reasons to believe that DMS methods are more robust to misspecification than are IMS methods. In the context of unconditional forecasts, the empirical relevance of these theories has been investigated. We extend that work to conditional forecasts. We do so based on linear bivariate and trivariate VARs/ARDLs estimated using a large dataset of macroeconomic time series.