The Lost Capital Asset Pricing Model
Coverage of the 2018 Northern Finance Association Conference
BY Caroline Cakebread | August 28, 2018
The capital asset pricing model (CAPM) is a key pillar of investing – it helps investors determine the right required rate of return and, in doing so, drives their asset allocation decisions. However, it’s not without its critics: seen through an empirical lens, CAPM repeatedly falls short, a failure which Richard Roll attributes to the fact that the composition of the market portfolio isn’t observable.
To make another case for the CAPM, a new paper, “The Lost Capital Asset Pricing Model,” digs deeper into the data in order to understand why the CAPM continues to fail from an empirical perspective. It will be presented at the upcoming Northern Finance Association Conference, where Canadian Investment Reviewis proud to be a media sponsor for the fourth year in a row.
The authors, Daniel Andrei from UCLA; Julien Cujean from the University of Maryland, and Mungo Wilson, from Oxford University, note that econometricians seeking to test the CAPM empirically are simply testing whether a market portfolio is mean-variance efficient. This has limitations they argue since investors that hold the market portfolio are looking at a wider set of data. That gives them an information edge over the econometrician and makes the case for looking at the CAPM differently. As the authors note:
Observing investors’ actions (for instance, their investment decisions) likely reveals some of this information. This approach based on “revealed preference” has caught on recently (Barber, Huang, and Odean, 2016; Berk and Van Binsbergen, 2016).
At the same time, the authors suggest their paper could be helpful for factor models in asset pricing.
Some variables may appear to the econometrician as priced factors simply because betas are mismeasured. Rather than being priced factors these variables are instruments for the measurement error in betas. Do there exist economic criteria that would allow the empiricist to distinguish variables that are economically meaningful from those that are not? This matter opens up fascinating directions for future research.
You can read the full paper here.