A Probability Model of the Coincident Economic Indicators

Abstract

The Index of Coincident Economic Indicators, currently compiled by the U.S. Department of Commerce, is designed to measure the state of overall economic activity. The index is constructed as a weighted average of four key macroeconomic time series, where the weights are obtained using rules that dare to the early days of business cycle analysis. This paper presents an explicit rime series model (formally, a dynamic factor analysis or "single index" model) that implicitly defines a variable that can be thought of as the overall state of the economy. Upon estimating this model using data from 1959-1987, the estimate of this unobserved variable is found to be highly correlated with the official Commerce Department series, particularly over business cycle horizons. Thus this model provides a formal rationalization for the traditional methodology used to develop the Coincident Index. Initial exploratory exercises indicate that traditional leading variables can prove useful in forecasting the short-run growth in this series.