Estimating Production Functions When Productivity Change Is Endogenous∗

Production function estimation on micro data suffers from persistent unobserved shocks that vary within firms and cause bias. This paper presents an estimation model where the firm chooses capital investment and productivityrelevant intangible assets in response to market conditions under partly fixed adjustment costs. Estimation on Brazilian manufacturing firm data suggests that (i) firms’ unobserved shocks are associated with productivity responses to competitive conditions and that (ii) a suitable productivity proxy is investment interacted with firm-specific competition variables. Identification does not rely on timing assumptions, and non-positive investment observations can be retained. Bootstraps show that the new proxy yields less dispersed coefficient estimates than alternatives while remedying bias. JEL C51, D24 ∗I thank Daniel Ackerberg, Andrew Bernard, Gordon Hanson, Ann Harrison, Pinar KaracaMandic, Daniel McFadden, Marc Melitz, Aviv Nevo, Maurice Obstfeld, James Powell, David Romer, Thomas Rothenberg and two anonymous referees who made important suggestions. Any remaining mistakes are mine. I thank the team at IBGE Rio de Janeiro for renewed data access in 2005 and 2007. Financial support from the Social Science Research Council and the American Council of Learned Societies with an International Predissertation Fellowship (funds from the Ford Foundation) and from IICAS at UCSD is gratefully acknowledged. ¶[email protected] (www.econ.ucsd.edu/muendler). University of California San Diego, Dept. of Economics, 9500 Gilman Dr. #0508, La Jolla, CA 92093-0508, USA