Abstract
Several recent studies have focused on the predictive power of the yield spread for future economic activity. The current paper reformulates the work of Estrella and Mishkin (Citation1998) by focusing on the usefulness of monetary variables for generating probability predictions of rising or falling real GDP growth and inflation. Besides redefining the dependent variables, the independent monetary variables are allowed to include lagged information. Also, the current paper considers the usefulness of the Divisia monetary aggregates in the context of probit models for predicting the probability that real GDP growth or inflation will be increasing
Notes
1 High-yield spread data are only available since the mid- to late-1980s. The probit models in this study are used to generate a large set of out-of-sample probability forecasts, and therefore require a longer history in order to both estimate the models and evaluate the out-of-sample forecasts.
2 The out-of-sample forecasts were also evaluated with Estrella's (Citation1998) pseudo R2 with no substantive changes in the results. In general, cases where the QPS is less than the QPS from the constant model correspond closely with cases where pseudo R 2 is negative.