More on the Performance of Higher Order Moment Estimators in Investment Equations

Abstract

Almeida, Campello, and Galvao (2010) [ACG] use Monte Carlo simulations and real data to assess the performance of estimators that deal with measurement errors in investment models. ACG are the first to provide an independent assessment of alternative methods, showing when they work properly and discussing the assumptions embedded in them. Erickson and Whited (2010) review ACG's study focusing exclusively on tests involving the Erickson and Whited (2000, 2002) [EW] estimator. While casting doubt on the usefulness of the ACG analysis, Erickson and Whited (2010) develop a number of ex-post fixes for the problems uncovered by ACG. The authors argue that the ACG tests would place the EW estimator in a more positive light had they used those fixes. This paper evaluates the new fixes proposed by Erickson and Whited and clarifies their implications for the debate about measurement error. The analysis provides further support for ACG's main conclusion: the presence of measurement error does not justify the use of the EW estimator in lieu of more robust, simpler alternatives.

Keywords: Investment equations, measurement error, Monte Carlo simulations, instrumental variables, GMM

JEL Classification: G31, C23