Academic journal article Review - Federal Reserve Bank of St. Louis

Estimating U.S. Output Growth with Vintage Data in a State-Space Framework/Commentary

Academic journal article Review - Federal Reserve Bank of St. Louis

Estimating U.S. Output Growth with Vintage Data in a State-Space Framework/Commentary

Article excerpt

(ProQuest: ... denotes formulae omitted.)

Richard G. Anderson and Charles S. Gascon

This study uses a state-space model to estimate the "true" unobserved measure of total output in the U.S. economy. The analysis uses the entire history (i.e., all vintages) of selected real-time data series to compute revisions and corresponding statistics for those series. The revision statistics, along with the most recent data vintage, are used in a state-space model to extract filtered estimates of the "true" series. Under certain assumptions, Monte Carlo simulations suggest this framework can improve published estimates by as much as 30 percent, lasting an average of 11 periods. Realtime experiments using a measure of real gross domestic product show improvement closer to 10 percent, lasting for 1 to 2 quarters. (JEL ClO, C53, E01)

Federal Reserve Bank of St. Louis Review, July/August 2009, 91(4), pp. 349-69.

Statistical agencies face a tradeoff between accuracy and timely reporting of macroeconomic data. As a result, agencies release their best estimates of the "true" unobserved series in the proceeding month, quarter, or year with some measurement error.1 As agencies collect more information, they revise their estimates, and the data are said to be more "mature." As the reported data mature, the estimates, on average, are assumed to converge toward the "true" unobserved values. This study examines a methodology in which the "true" value of an economic variable is latent in the sense of the state vector in a state-space model. In doing so, we use recent modeling suggestions by Jacobs and van Norden (2006) and Cunningham et al. (2007) regarding relationships among realtime data, measurement error as a heteroskedastic stochastic process, and the latent, "true" data for an economic variable of interest.

The importance of potential output growth in policymaking motivates our study. Forwardlooking macroeconomic models suggest that the predicted future path of the output gap should be important to policymakers. To the extent that policymakers are concerned with a Federal Reserve-style "dual mandate," an output gap equal to 1 percent of potential output maybe quite alarming if projections suggest it will continue, but relatively innocuous if the gap is expected to shrink rapidly during the next few quarters. Recent studies on inflation forecasting conclude that the output gap, when measured in real-time using vintage data, has little predictive power for inflation (e.g., Orphanides and van Norden, 2005; and Stock and Watson, 2007). It is also important to study the real-time measurement of potential output because policymakers occasionally face possible changes/breaks in the underlying growth trend of productivity and, hence, potential output.

Our objective in this study is not to assess inflation-forecasting models, although that has been a major use of potential output measures; rather, it is to estimate the "true" value of real output for use in the construction of trend-like measures of potential output. One of the larger recent studies in this vein, albeit focused on inflation prediction, is by Orphanides and van Norden (2005). The study considers, as predictive variables for inflation, both a wide range of output gap measures (which differ with respect to data vintage and the trend estimator) and lagged values of real output growth. Their conclusion regarding output gap models as predictors of inflation is straightforward - the output gap does not reliably predict inflation, although the differences in forecast performance between output-gap and outputgrowth models are not statistically significant:

[O]ur analysis suggests that a practitioner could do well by simply taking into account the information contained in real output growth without attempting to measure the level of the output gap. This model was consistently among the best performers, particularly over the post-1983 forecast sample, (p. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.