Broken trend output in a model of stock returns and economic activity

This paper investigates the interaction between real stock returns and economic activity using a new econometric technique that suggests postwar US industrial production is best characterized as broken trend stationary rather than first difference stationary. This result, however, has little impact on the Granger causal relationship between real stock returns and economic activity. Results from a small forecasting experiment indicate that determining which model forecasts the best depends upon the criteria being used to evaluate the forecast summary statistics.