The Use of Asset Growth in Empirical Asset Pricing Models∗

We provide evidence that the empirical performance of the new factor models proposed by Hou, Xue, and Zhang (2015) and Fama and French (2015) depends crucially on how their investment factor is constructed. Specifically, we call attention to the fact that, in both models, the investment factor is based on the measure of growth in total assets from Cooper, Gulen, and Schill (2008) and not on what most researchers would view as traditional measures of corporate investment. For both models, we show that there are large decreases in their ability to price the cross-section of returns when the investment factor is instead constructed using the traditional investment measures, or when it is constructed using arguably more complete measures that account for investment in intangibles. Additionally, we do not find a significant decrease in performance when we replace the asset-growth factor with a factor based on growth in noncash current assets or long-term debt (which cannot be complete measures of investment). Our results challenge the idea that traditional investment models can fully account for the explanatory power of the asset-growth factor used in the Hou, Xue, and Zhang (2015) and Fama and French (2015) models. ∗This paper has benefited from comments and discussions with Daniel Andrei, Adem Atmaz, Turan Bali, Hank Bessembinder, Stefano Cassella, Scott Cederburg, Sergey Chernenko, Tarun Chordia, Kent Daniel, David Denis, Theodore Goodman, Cam Harvey, Kate Holland, Greg Kadlec, Leonid Kogan, Karl Lins, Yan Liu, Rene Stulz, Michael Woeppel, Irene Yi, and seminar participants at Arizona State University, Georgia State University, and Purdue University. We thank Mitchell Johnston for excellent research assistance.