Forecast Accuracy and Efficiency: An Evaluation of Ex Ante Substate Long-Term Forecasts

Abstract

This article identifies regional characteristics that account for differential forecast accuracy. It is found that population, income, employment, and housing start forecast accuracy depend significantly on county size, growth rate, and elderly population; industrial specialization is irrelevant. This finding has implications for making generalizations from a subset of regions. The informational efficiency of regional forecasts is also examined. It is found that employment forecasts are efficient, income forecasts tend to be overly reliant on previous period growth, and naïve population projections ignore information exploited by structural economic-demographic forecasts. These findings have important implications for the optimal combination of forecasts and the improvement of forecasting methodologies.