Staff Report 78

A Use of Index Models in Macroeconomic Forecasting

Robert B. Litterman

Revised October 1, 1982

This paper illustrates the application of observable index models to the problem of macroeconomic forecasting. In this context, a Bayesian prior is used to describe a class of models which impose the index structure with more or less weight. An out-of-sample forecasting experiment is used to measure the possible benefits of this approach. In addition, impulse response functions and the decomposition of forecast variance are analyzed to suggest a possible separation of real and nominal shocks into separate channels.

Download Paper (pdf)