The sensitivity of estimated DSGE models to alternative data vintages

Abstract

This paper considers the estimation of a structural DSGE model with three alternative macroeconomic data vintages corresponding to the first-release (real-time data), the third-release data, and the highly revised data to assess the sensitivity of the estimation arising from the data revision process. The empirical evidence based on a structural econometric approach suggests that some structural and shock process parameters are only identified whenever highly revised data become available. More generally, several parameters are highly sensitive to data vintage in the estimation procedure. Data revisions also affect the estimated properties of the economic agents’ expectations that determine their decisions. Its empirical validation is assessed through the corresponding observable counterparts reported in the Survey of Professional Forecasters.