Asset Pricing

April 8, 2016
Adrien Verdelhan and Deborah J. Lucas, both of MIT, Organizers

Michael D. Bauer, Federal Reserve Bank of San Francisco, and James D. Hamilton, University of California at San Diego and NBER

Robust Bond Risk Premia

A consensus has recently emerged that variables beyond the level, slope, and curvature of the yield curve can help predict bond returns. This paper shows that the statistical tests underlying this evidence are subject to serious small-sample distortions. Bauer and Hamilton propose more robust tests, including a novel bootstrap procedure specifically designed to test the "spanning hypothesis." They revisit the evidence in five published studies, find most rejections of the spanning hypothesis to be spurious, and conclude that the current consensus is wrong. Only the level and the slope of the yield curve are robust predictors of bond returns.


Brian Weller, Northwestern University

Measuring Tail Risks at High Frequency

Weller develops a new methodology for measuring tail risks using the cross section of bid-ask spreads. Market makers embed tail risk information into spreads because (1) they lose to arbitrageurs when changes to asset values exceed the cost of liquidity and (2) underlying price movements and potential costs are linear in factor loadings. Using this insight, simple cross-sectional regressions relating spreads and trading volume to factor betas can recover tail risks in real time for common factors in stock returns. The methodology disentangles financial and aggregate market risks during the 2007–2008 Financial Crisis; anticipates jump risks associated with Federal Open Market Committee announcements; and quantifies a sharp, temporary increase in market tail risk before and throughout the 2010 Flash Crash. The recovered time series of implied market risks also aligns closely with both realized market jumps and the VIX.


Lars P. Hansen, University of Chicago and NBER, and Thomas J. Sargent, New York University and NBER

Sets of Models and Prices of Uncertainty (NBER Working Paper No. 22000)

A decision maker constructs a convex set of non-negative martingales to use as likelihood ratios that represent parametric alternatives to a baseline model and also non-parametric models statistically close to both the baseline model and the parametric alternatives. Max-min expected utility over that set gives rise to equilibrium prices of model uncertainty expressed as worst-case distortions to drifts in a representative investor's baseline model. Hansen and Sargent offer quantitative illustrations for baseline models of consumption dynamics that display long-run risk. They describe a set of parametric alternatives that generates countercyclical prices of uncertainty.

Nina Boyarchenko and David Lucca, Federal Reserve Bank of New York, and Laura Veldkamp, New York University and NBER

Taking Orders and Taking Notes: Dealer Information Sharing in Financial Markets

The use of order flow information by financial firms has come to the forefront of the regulatory debate. Central to this discussion is whether a dealer who acquires information by taking client orders can share that information. Boyarchenko, Lucca, and Veldkamp explore how information sharing affects dealers, clients and issuer revenues in U.S. Treasury auctions. Because one cannot observe alternative information regimes, the researchers build a model, calibrate it to auction results data, and use it to quantify counter-factuals. They estimate that yearly auction revenues with full-information sharing (with clients and between dealers) would be $5 billion higher than in a "Chinese Wall" regime in which no information is shared. When information sharing enables collusion, the collusion costs revenue, but prohibiting information sharing costs more. For investors, the welfare effects of information sharing depend on how information is shared. Surprisingly, investors benefit when dealers share information with each other, not when they share more with clients. For the market, when investors can bid directly, information sharing creates a new financial accelerator: Only investors with bad news bid through intermediaries, who then share that information with others. Thus, sharing amplifies the effect of negative news. Tests of two model predictions support its key features.


Erik P. Gilje, University of Pennsylvania; Robert C. Ready, University of Rochester; and Nikolai Roussanov, University of Pennsylvania and NBER

Fracking, Drilling, and Asset Pricing: Estimating the Economic Benefits of the Shale Revolution

Gilje, Ready, and Roussanov quantify the effect of a significant technological innovation, shale oil development, on asset prices. Using stock price changes on major news announcement days allows us to link aggregate stock price changes to shale development activity as well as other oil supply shocks. They exploit cross-sectional variation in industry portfolio returns on announcement days to construct a shale mimicking portfolio. This portfolio can help explain aggregate stock market fluctuations, but only during the time period of shale oil development. Based on the estimated effect of this mimicking portfolio on aggregate stock market returns, they find that $2.5 trillion of the increase in aggregate U.S. equity market capitalization since 2012 can be attributed to shale oil. Industries benefiting the most from the shale oil revolution, as indicated by their shale announcement day returns, added more jobs over the shale period than those unrelated to shale.


Robert Novy-Marx, University of Rochester and NBER

Testing Strategies Based on Multiple Signals

Strategies selected by combining multiple signals suffer severe overfitting biases, because underlying signals are typically signed such that each predicts positive in-sample returns. As a result, "highly significant" back-tested performance is easy to generate selecting stocks using combinations of randomly generated signals, which by construction have no true power. Novy-Marx analyzes t-statistic distributions for multi-signal strategies, both empirically and theoretically, to determine appropriate critical values, which can be several times standard levels. Overfitting bias also severely exacerbates the multiple testing bias that arises when investigators consider more results than they present. Combining the best `k` out of `n` candidate signals yields biases similar to those obtained using the single best of `n^k` candidate signals.