Improving econometric forecasts by using subperiod data

"The method proposed here includes two innovations which should improve the accuracy of econometric forecasting. First, it replaces the subjective, judgmental adjustments commonly used with a more formal, objective econometric procedure. Second, it includes a methodology for testing the usefulness of subperiod data which forecasters often inspect when choosing intercept adjustments. A sample application to the MIT-Penn-SSRC Model demonstrates that the procedure is both feasible and potentially helpful in the context of a large macroeconometric model"--Federal Reserve Bank of Minneapolis web site