Real-Time Datasets Really Do Make a Difference: Definitional Change, Data Release, and Forecasting

Abstract

In this paper, the authors empirically assess the extent to which early release inefficiency and definitional change affect prediction precision. In particular, they carry out a series of ex-ante prediction experiments in order to examine: the marginal predictive content of the revision process, the trade-offs associated with predicting different releases of a variable, the importance of particular forms of definitional change, which the authors call "definitional breaks," and the rationality of early releases of economic variables. An important feature of our rationality tests is that they are based solely on the examination of ex-ante predictions, rather than being based on in-sample regression analysis, as are many tests in the extant literature. Their findings point to the importance of making real-time datasets available to forecasters, as the revision process has marginal predictive content, and because predictive accuracy increases when multiple releases of data are used when specifying and estimating prediction models. The authors also present new evidence that early releases of money are rational, whereas prices and output are irrational. Moreover, they find that regardless of which release of our price variable one specifies as the "target" variable to be predicted, using only "first release" data in model estimation and prediction construction yields mean square forecast error (MSFE) "best" predictions. On the other hand, models estimated and implemented using "latest available release" data are MSFE-best for predicting all releases of money. The authors argue that these contradictory findings are due to the relevance of definitional breaks in the data generating processes of the variables that they examine. In an empirical analysis, they examine the real-time predictive content of money for income, and they find that vector autoregressions with money do not perform significantly worse than autoregressions, when predicting output during the last 20 years.

Keywords: Bias, efficiency, generically comprehensive tests, rationality, preliminary, final, real-time data

JEL Classification: C32, C53, E01, E37, E47