Forecasting the U.S. Unemployment Rate: The Impact of Stock Market Volatility

This paper presents a comparison of forecasting performance for a variety of linear time series models using the U.S. unemployment rate. The main emphasis is on measuring forecast performance in bi-variate vector auto-regressive models that include either a measure of stock market uncertainty, industrial production, or capacity utilization and the U.S. unemployment rate. Accounting for a regime shift in monetary policy in the late 1970s and isolating measures of macroeconomic well being in different bivariate vector autoregressive models allows for commonsensical comparisons between models that should intuitively improve forecasts over existing linear methods. Comparisons are also made with the consensus median forecasts from the Survey of Professional Forecasters. The results indicate that models incorporating a measure of stock market volatility improved the accuracy of the forecasts over other bivariate and univariate linear models.