The Time-Series Relatedness of State and National Indexes of Leading Indicators and Implications for Regional Forecasting

Abstract

This study investigates the time-series relatedness of state and national indexes of leading indicators and the implications of these results regarding state employment forecasting. Composite indexes of leading indicators are constructed for the United States and each of the fifty states based on housing permits, initial claims for state unemployment compensation, and average weekly hours in manufacturing. The three-component U.S. index (USLI3 ) is shown to reflect much of the cyclical variation in the U.S. composite of eleven leading indicators (USLI11 ). Final prediction error (FPE) causality tests and cointegration/long-memory components statistics show thatUSLI3 generally causes or “drives” the state indexes. These results suggest that national leading indicators may be more useful than similar state indicators in predicting state-level activity. Experimental forecasts confirm that bothUSLI3 andUSLI11 are generally more effective in improving forecasts of state nonfarm employment than the respective fifty state composite indexes.