How far can we forecast? Statistical tests of the predictive content
Forecasts are uninformative whenever the forecast error variance fails to be smaller than the unconditional variance of the target variable. We develop two tests for the null hypothesis that forecasts become uninformative beyond some limiting forecast horizon h*. One test compares the mean-squared error of the forecast to the sample variance, the other test relies on forecast encompassing. The latter test has a normal limiting distribution and better size properties. In an empirical application to forecasts for macroeconomic key variables from the survey of Consensus Economics, the tests suggest that these forecasts are barely informative beyond 2-4 quarters ahead.