Title: A semiparametric test for measurement error in panel data
Authors: Laura Spierdijk - University of Groningen (Netherlands) [presenting]
Erik Meijer - University of Southern California (United States)
Tom Wansbeek - University of Groningen (Netherlands)
Roger Moon - University of Southern California (United States)
Abstract: Although measurement error is a common phenomenon, most applied regression analyses do not take it into account. One potential explanation is that there is not a standard test for the presence of measurement error. We develop such a test for linear static panel data regressions, based on the insight that under weak assumptions the bias of the OLS estimator increases by first differencing, but less so by taking differences more periods apart. The test is easy to implement and apply and has desirable statistical properties. We apply the test to models for medical expenditures and productivity and show that there is strong evidence for measurement error in some of the regressors.