Do Option Characteristics Predict the Underlying Stock Returns in the Cross-Section?

 Management Science, 2025, with Andreas Neuhierl, Rasmus Tangsgaard Varneskov and Guofu Zhou 

We provide a comprehensive analysis of options-implied information for predicting the cross-section of stock returns. Based on large sets of firm and option characteristics and using traditional portfolio sorts and modern high-dimensional methods, we find that option information matters. However, in contrast to existing studies, there are only few option characteristics that have significant incremental predictive power after controlling for the large set of firm characteristics. Further analysis reveals that the strongest option characteristics are associated with asset mispricing, future tail return realizations and short-selling costs. Our findings are consistent with models of informed trading and limits to arbitrage. 

Recovering Implied Volatility

Management Science, 2023, with Ohad Kadan and Fang Liu

We propose a methodology for estimating option-implied forward-looking variances and covariances of assets and portfolios, which may not possess actively-traded options. Our approach relies on the observation that any factor structure for stock returns naturally induces a factor structure for return volatility. We implement the methodology empirically and show that our forward-looking moment estimates provide useful implications for the prediction of jumps and for portfolio choice. 

Recovering the FOMC Risk Premium

Journal of Financial Economics, 2022, with Hong Liu and Guofu Zhou

The Federal Open Market Committee (FOMC) meetings have significant impact on market returns. We propose a methodology to recover the risk premium associated with FOMC meetings from option prices. We also predict the sizes of upward/downward market price jumps after an imminent FOMC meeting. In our empirical analysis, with observed price data for 83 meetings and with data backed out via machine learning for the remaining 109 meetings from 1996 to 2019, we find that the risk premium varies from 2 to 299 basis points (bps), with an average of 41 bps which is consistent with the average realized returns documented in the literature. The average predicted upward jump size is 101 bps, and the average predicted downward jump size is 129 bps.

A Bound on Expected Stock Returns 

Review of Financial Studies, 2020,  with Ohad Kadan.

We present a sufficient condition under which the prices of options written on a particular stock can be aggregated to calculate a lower bound on the expected returns of that stock. The sufficient condition imposes a restriction on a combination of the stock's systematic and idiosyncratic risk. The lower bound is forward-looking and can be calculated on a high-frequency basis. We estimate the lower bound empirically and study its cross-sectional properties. We find that the bound increases with beta and book-to-market ratio and decreases with size and momentum. The bound also provides an economically meaningful signal on future stock returns. 

Volatility-Managed Portfolio: Does It Really Work?

Journal of Portfolio Management, 2019,  with Guofu Zhou and Fang Liu.

In this article, the authors find that a typical application of volatility-timing strategies to the stock market suffers from a look-ahead bias, despite existing evidence on successes of the strategies at the stock level. After correcting the bias, the strategy becomes very difficult to implement in practice as its maximum drawdown is 68--93% in almost all cases. Moreover, the strategy outperforms the market only during the financial crisis period. The authors also consider three alternative volatility-timing strategies and find that they do not outperform the market either. Their results show that one cannot easily beat the market via timing the market alone. 

Out-of-sample equity premium prediction: A scenario analysis approach

Journal of Forecasting, 2018,  with Peiming Wang and Feifang Hu.

In this article, the authors find that a typical application of volatility-timing strategies to the stock market suffers from a look-ahead bias, despite existing evidence on successes of the strategies at the stock level. After correcting the bias, the strategy becomes very difficult to implement in practice as its maximum drawdown is 68--93% in almost all cases. Moreover, the strategy outperforms the market only during the financial crisis period. The authors also consider three alternative volatility-timing strategies and find that they do not outperform the market either. Their results show that one cannot easily beat the market via timing the market alone. 

Variance Asymmetry Managed Portfolios

I propose a forward-looking measure of the asymmetry in the variance of asset returns and introduce a way to estimate it from option prices. This measure is model free and it serves as a close approximation for the asset expected excess return. I provide an empirically supported sufficient condition under which the risk-neutral variance asymmetry ranks stocks based on their physical expected returns. Empirically, I find strong cross-sectional correlation between this measure and future stock returns. Variance asymmetry managed portfolios yield economically large average returns and Sharpe ratios. Crash risk and standard asset pricing factors do not explain this abnormal performance.

The Information Content of The Implied Volatility Surface: Can Option Prices Predict Jumps?

 with Yufeng Han and Fang Liu

We ask whether option prices contain information on the likelihood and direction of jumps in the underlying stock prices. Applying the partial least squares (PLS) approach to the entire surface of the implied volatilities (IV), we show that option prices can successfully predict downward jumps in stock prices, but not upward jumps. The PLS estimated downward jump factor can predict stock returns with a spread of 1.53% per month between stocks predicted to have the lowest probability of downward jumps and stocks predicted to have the highest probability of downward jumps. Both put and call option prices, and options of both short and long maturity contribute to the predictability. Furthermore, the predictability of the downward jump is robust to many firm characteristics as well as options related variables. Consistent with the notion that informed investors trade in the options markets to profit from negative information in order to circumvent the short-sale constraint, we find that stronger predictability is associated with tighter short-sale constraints in the equity market, and in periods when the market has poor performance. 

Recovering Conditional Factor Risk Premia 

 with Ohad Kadan and Fang Liu

We offer an approach for recovering option-implied time-varying forward-looking risk premia of systematic factors---even if they do not possess actively-traded options. We apply this approach to the market, size, value, and momentum factors. We find that factor premia are highly volatile. Both the market and the value premia tend to be higher during slowdowns and recessions and during turbulent times. By contrast, the momentum premium is higher during periods of high economic growth and low volatility. We use the recovered factor premia to construct trading strategies, which mitigate market and momentum crash risk and to predict returns of individual stocks even if they do not possess traded options.

Heterogeneous Responses in Financial Markets: Insights from Machine Learning

with Xiwei Tang and Guofu Zhou

We propose a machine learning framework that extends the Fama-MacBeth regression to capture heterogeneity in stock return responses to firm characteristics, allowing more flexible estimation of expected returns in the cross section. Using 15 representative firm characteristics, our method nearly doubles the Sharpe ratio of the long-short portfolio formed using Fama-MacBeth return forecasts. It also offers greater interpretability and outperforms other machine learning models, even in high-dimensional settings with 94 characteristics. Our results emphasize the importance of heterogeneity in stock return responses, especially during recessions, and challenge the traditional homogeneity assumption embedded in the Fama-MacBeth regression, with broad implications for empirical asset pricing. 

Option Expected Hedging Demand

with Guofu Zhou and Zhaoque (Chosen) Zhou

Options market makers' delta hedging has an increasing impact on underlying stock prices as both the option volume and the ratio of option volume to stock volume grow drastically in recent years. We introduce a novel approach utilizing real-time option information to calculate the spot elasticity of delta (ED) and expected hedging demand (EHD), and find that the EHD significantly predicts future stock returns in the cross section. The positive impact of EHD on stock prices lasts up to five trading days, and then a reversal follows. The empirical evidence of heterogeneous EHD-return relationship, influenced by ED, leads to varied option market maker behaviors, and is consistent with conventional economic theory. Moreover, we find that EHD has a little correlation with other popular firm characteristics, representing a new risk that is not captured by conventional factor models. 

Ex-Ante Risk Premia on Earnings Announcements: Evidence from the Options Market

with Hong Liu, Yingdong Mao, and Guofu Zhou

We provide the first estimate of the ex-ante risk premia on earnings announcements using forward-looking information from the options market. We find that the average earnings announcement risk premium is highly significant at 13 basis points, with substantial variation across firms and over time. Beyond its asset pricing implications, our study also sheds economic insight into understanding the post-earnings-announcement drift and what drives the profitability of widely analyzed straddle strategies. Moreover, our results enable the design of a market timing strategy based on earnings announcement risk premia, resulting in an improvement in the market Sharpe ratio by 37%. 

Lender Coordination and Loan Renegotiation

with Xiumin Martin, Hanmeng (Ivy) Wang, and Yifang Xie

This paper employs an unsupervised machine learning technique to develop a novel measure, "clique," representing groups with maximal interconnections among lenders in the syndicated loan market. Drawing upon network theory, the clique measure captures the informational role of networks that facilitate cooperation. We show that loan renegotiation increases in syndicates with a higher proportion of clique lenders, i.e., participant lenders from the same clique as the lead lender. Results remain robust when controlling for bilateral relationships between participant and lead lenders, lead-lender centrality, and borrower-year fixed effects, as well as through a reduced-form estimation involving a plausibly exogenous shock to the proportion of clique lenders. Loan performance improves with a higher proportion of clique lenders. However, the presence of clique lenders can be costly for borrowers with high proprietary costs. Overall, our study highlights the crucial role of lender clique in facilitating lender coordination by acting as an ex-ante mechanism. 

Inferring Trade Directions in Options via Machine Learning

with Si Gao, Guofu Zhou, and Zhaoque (Chosen) Zhou

We propose a novel machine learning method, GS-LASSO which integrates XGBoost with SHapley Additive exPlanations (SHAP) and LASSO, to infer trade directions. Using a proprietary dataset from the Philadelphia Stock Exchange (PHLX), we find that our new method GS-LASSO achieves 76.71% accuracy, whereas the existing methods can infer correctly buy orders with no greater than 60% accuracy. The improved accuracy helps not only to obtain more accurate measurement of market microstructure metrics, but also to provide new insights into wholesalers’ profitability in executing option trades of retail investors.