The Financial Review 45 (2010) 277–285 On Model Testing in Financial Economics Robert A. Jarrow ∗ Johnson Graduate School of Management, Cornell University Abstract This paper discusses the two different contradicting philosophies for testing models in financial economics (asset pricing, corporate finance, and market-microstructure) using linear regression. We synthesize these two contradicting approaches, document the errors that may occur in the existing estimation methodologies, and suggest a modified procedure that avoids these errors. Keywords: linear regression, omitted variables, control variables JEL Classifications: C10, G10 1. Introduction In financial economics, linear regression is used in two contradicting ways to test the validity of a financial model. 1 Both methods can be illustrated with the following simple structure. Let Z i = f (x i ) be a model of a financial variable Z i as a function of x i . In addition, let y i be another financial variable that potentially impacts Z i , but that is excluded from the model. To test the model we run the linear regression Z i = α + βf (x i ) + γy i + ǫ i ∗ Corresponding author: Johnson Graduate School of Management, Cornell University, 451 Sage Hall, Ithaca, NY 14853; Phone: (607) 255-4729; E-mail: raj15@cornell.edu. 1 Although this paper emphasizes the use of linear regression in financial economics, this same issue applies more generally to model testing in economics. C  2010, The Eastern Finance Association 277