Real-Time Tests of the Leading Economic Index: Do Changes in the Index Composition Matter?

In an important paper, Diebold and Rudebusch (1991) find that, despite good performance for post revision historical versions, the U.S. Index of Leading Economic Indicators (LEI) fails to improve forecasts in real time out-of-sample tests. This paper revisits the issue of real-time performance of the LEI using growth rate forecast specifications. We contrast real-time out-of-sample tests of LEI forecasts using „composition-changing” or „as-published” versions of the LEI with those based on „composition-constant” indexes. The goal is ...