Modeling Data Revisions: Measurement Error and Dynamics

Realistic modeling of data revisions can play an impor-tant role in policy formulation. A common way to model data revisions is to set up a state-space model with sep-arate blocks for measurement errors and the dynamics of ”true ” values. However, empirical work suggests that measurement errors typically have much more complex dynamics than such models allow. This paper describes a state-space model with richer dynamics in these mea-surement errors, including the noise, news and spillover effects documented in this literature. The result is a uni-fied and flexible framework that allows for more realism in the model of data revision and optimal real-time es-timation of trends and cycles in real time. We illustrate the application of this framework with an analysis of real-time data on U.S. real output. KEY WORDS: real-time analysis, data revisions, state-space model