Time-series efficient factors
with Sina Ehsani
March 2020
Factors in prominent asset pricing models are positively serially correlated. We derive the optimal allocation that transforms an auto-correlated factor to a "time-series efficient" factor. The key determinant of the value of factor timing is the ratio of a factor's auto-correlation to its Sharpe ratio. Time-series efficient factors earn significantly higher Sharpe ratios than the original factors and contain all the information found in the original factors. Momentum strategies profit by timing auto-correlated factors; they pick up factor "inefficiencies." We show that, rather than augmenting models with the momentum factor, each factor can instead be made time-series efficient. An asset pricing model with time-series efficient factors, such as an efficient Fama-French five-factor model, prices momentum. Time-series efficient factors also explain more of the co-variance structure of returns; they describe the cross section better than the standard factors and align more closely with the true SDF.
Factor momentum
with Robert Arnott, Mark Clements, and Vitali Kalesnik
March 2021
Past industry returns predict future industry returns, and this predictability is at its strongest at the one-month horizon. We show that the cross section of factor returns shares this property and that industry momentum stems from factor momentum. Factor momentum transmits into the cross section of industry returns through variation in industries’ factor loadings. We show that momentum in “systematic industries,” mimicking portfolios built from factors, subsumes industry momentum as does momentum in industry-neutral factors. Industry momentum is therefore a byproduct of factor momentum, not vice versa. Momentum concentrates in its entirety in the first few highest-eigenvalue factors.
The earnings announcement return cycle
with Conson Zhang
January 2019
Stocks earn significantly negative abnormal returns before earnings announcements and positive after them. This “earnings announcement return cycle” (EARC) is unrelated to the earnings announcement premium, and it is a feature of stocks
widely covered by analysts. Analysts' forecasts follow the same pattern as returns: analysts' forecasts become more optimistic after an earnings announcement and more pessimistic as the next one draws near. We attribute one-half of the earnings announcement return cycle to this optimism cycle. The EARC may stem from mispricing: both the return and optimism patterns are stronger among high-uncertainty and difficult-to-arbitrage stocks, and the EARC strategy is more profitable on days when it would accommodate larger amounts of arbitrage capital.
Award: Alpha Letters / CQA Prize Winner at CQA Spring 2019
Informed traders, long-dated options, and the cross section of stock returns
with Mark Clements and Vitali Kalesnik
September 2017
Option prices predict the cross section of equity returns.
We show that, unconditionally, the prices of long-dated
options contain all the information relevant for predicting
returns. Information, however, shifts towards short-dated
options when an earnings announcement is imminent and when
options are cheap to trade. The difference between short-
and long-dated options also predicts the timing of merger
announcements. Our results are consistent with option
prices reflecting the actions of informed traders, and with
these traders optimally choosing option maturities to
maximize the value of their information.