Forecast Disagreement about Long-run Macroeconomic Relationships

Abstract

Using survey forecast data, we study if professional forecasters utilize long-run co-integration relationships among macroeconomic variables to forecast future as postulated in workhorse stochastic growth models. There exists a significant heterogeneity among forecasters, the majority of whom do not use these long-run relationships and generally make more accurate forecasts (comparing with those who use). Simple parsimonious recursive forecasting models are fitted to the data as one way to approximate the expectation formation process of the forecasters who utilize (or do not utilize) the long-run relationships.