This website uses cookies.

By clicking the "Accept" button or continuing to browse our site, you agree to first-party and session-only cookies being stored on your device to enhance site navigation and analyze site performance and traffic. For more information on our use of cookies, please see our Privacy Policy.

« Back to Results

Taming and Explaining Anomalies

Paper Session

Sunday, Jan. 7, 2018 8:00 AM - 10:00 AM

Loews Philadelphia, Commonwealth Hall C
Hosted By: American Finance Association
  • Chair: Jules van Binsbergen, University of Pennsylvania

The Lost Capital Asset Pricing Model

Daniel Andrei
,
University of California-Los Angeles
Julien Cujean
,
University of Maryland
Mungo Wilson
,
University of Oxford

Abstract

A flat Securities Market Line is not evidence against the CAPM. Under the Roll (1977) critique, the CAPM is a "lost city of Atlantis," empirically invisible. In a noisy rational-expectations economy, there exists an information gap between the average investor who holds the market and the empiricist who does not observe the market portfolio. The CAPM holds for the investor, but appears
flat to the empiricist. This distortion is empirically substantial and explains, for instance, why "Betting Against Beta" works; BAB really bets on true beta. Macroeconomic announcements reduce the distortion---for a fleeting moment the empiricist catches a glimpse of the CAPM.

The Relation Between Idiosyncratic Volatility and Expected Returns: A Statistical Artifact of Temporary Changes in Idiosyncratic Volatility

Hongyan Li
,
Virginia Tech
Raman Kumar
,
Virginia Tech

Abstract

We document a systematic pattern of temporary increases in the estimated idiosyncratic volatility for the quintile of stocks with the highest estimated idiosyncratic volatility in a given month. A large portion of this temporary increase in the estimated idiosyncratic volatility is reversed in the subsequent month, which suggests the possibility of relatively large positive estimation errors. This temporary increase in the idiosyncratic volatility for the quintile of stocks with the highest estimated idiosyncratic volatility is associated with relatively large positive returns (positive abnormal returns) in the estimation month and relatively low returns (negative abnorma returns) in the subsequent month. Our evidence shows that these temporary changes in the estimated idiosyncratic volatility and the related positive and negative abnormal returns in the estimation and subsequent months, respectively, create a negative relation between the estimated idiosyncratic volatility and subsequent month returns documented in the prior literature (Ang et al. 2006). After controlling for the (negative) relation with the past month’s return, there is no significant relation between idiosyncratic volatility and subsequent month’s returns as predicted by traditional asset pricing models. Moreover, we find no significant relation between idiosyncratic volatility and subsequent returns for subsets of stocks that do not exhibit any significant changes in idiosyncratic volatility despite large differences in the levels of their idiosyncratic volatility. Finally, there is no relation between the estimated idiosyncratic volatility and subsequent returns after a lag of 3 months when the abnormal returns associated with temporary changes are no longer present. Overall, our results are consistent with the notion that there is no relation between the true underlying idiosyncratic volatility and expected returns, and that the previously documented negative relation between estimated idiosyncratic volatility and subsequent month’s returns is being driven by estimation errors (temporary one-month changes) in the estimated idiosyncratic volatility and the associated abnormal returns.

Taming the Factor Zoo

Guanhao Feng
,
City University of Hong Kong
Stefano Giglio
,
University of Chicago
Dacheng Xiu
,
University of Chicago

Abstract

The asset pricing literature has produced hundreds of potential risk factors. Organizing this “zoo of factors” and distinguishing between useful, useless, and redundant factors require econometric techniques that can deal with the curse of dimensionality. We propose a model-selection method that allows us to systematically evaluate the contribution to asset pricing of any new factor, above and beyond what is explained by a high-dimensional set of existing factors. Our procedure selects the best parsimonious model out of the large set of existing factors, and uses it as the control in making statistical inference about the contribution of new factors. Our inference allows for model selection mistakes, and is therefore more reliable in finite sample. We derive the asymptotic properties of our test and apply it to a large set of factors proposed in the literature. We show that despite the fact that hundreds of factors have been proposed in the last 30 years, some recent factors – like profitability – have statistically significant explanatory power in addition to existing ones. We confirm the effectiveness of our procedure to discriminate factors in a recursive and out-of-sample experiment, and show that it results in a parsimonious model with a small number of factors and high cross-sectional explanatory power, even as the pool of candidate factors has expanded dramatically.

Term Structure of Recession Probabilities and the Cross Section of Asset Returns

Ti Zhou
,
Southern University of Science and Technology

Abstract

The duration of business cycles changes over time, generating time-varying investor concern about recessions. I study a new macro-factor model that directly links assets' risk premia to such concern, measured by the term structure of recession probabilities from professional forecasters. The innovation to the slope of the term structure is a negatively priced risk factor with an economically large and significant risk premium in a wide range of tests assets, consistent with how the slope predicts long-run economic activity and labor income growth. A linear factor model, including market excess return and the innovation to the slope, explains at least more than half of the cross-sectional variation of average returns on portfolios sorted on size, book-to-market, past long term return and asset growth. The factor mimicking portfolios of the model help reconcile the joint cross section of returns on equities, equity index options, and currencies and have pricing performance comparable to several multi-factor benchmarks. My evidence suggests that the slope of the term structure is a recession state variable (Cochrane, 2005), and an economic source of risk premia on test assets can be attributed to time-varying investor concern over future recessions that is priced.
Discussant(s)
Laura Veldkamp
,
New York University
Christopher Polk
,
London School of Economics
Victor Chernozhukov
,
Massachusetts Institute of Technology
Ralph Koijen
,
New York University
JEL Classifications
  • G1 - General Financial Markets