Using regional economic indexes to forecast tax bases: Evidence from New York

Abstract

This paper evaluates the use of measures of regional economic activity to forecast tax revenues for New York State and New York City at 3-, 6-, and 12-month horizons. We construct sales- and withholding-tax base series and then apply the methodology of Stock and Watson (1989, 1991) to estimate regional indexes of coincident economic indicators. Employing an out-of-sample forecasting framework, we find that the use of the coincident indexes leads to statistically and economically significant improvements in tax base forecasts compared to those generated from univariate autoregressions. In addition, the coincident indexes produce forecasts that are generally more accurate than forecasts that rely on the use of the coincident indicators separately. Though our analysis focuses on forecasting movements in tax revenue at the state or local level, it is also intended to draw attention to the value the indexes may provide in other applications.