On the Harm that Pretesting Does

D.L. Danilov, J.R. Magnus

Research output: Working paperDiscussion paperOther research output

317 Downloads (Pure)

Abstract

Data in econometrics are, as a rule, non-experimental and hence we have to use the same data set to select the model and also to estimate the parameters in the selected model.In standard applied econometrics practice, however, one reports zero bias and some variance of the (pretest) estimators conditional on the selected model.In this paper we find the unconditional moments of the pretest estimator, taking full account of the fact that model selection and estimation are an integrated procedure.We derive the bias, variance, and mean squared error of the pretest estimator, and show what the error is in not reporting the correct moments.This error can be very substantial.We also show that there can be large differences in underreporting between different model selection procedures.Finally, we ask how the underreporting error increases when the number of auxiliary regressors increases.
Original languageEnglish
Place of PublicationTilburg
PublisherEconometrics
Number of pages36
Volume2001-37
Publication statusPublished - 2001

Publication series

NameCentER Discussion Paper
Volume2001-37

Keywords

  • econometric models
  • testing
  • estimation

Fingerprint

Dive into the research topics of 'On the Harm that Pretesting Does'. Together they form a unique fingerprint.

Cite this