Predicting the Delta-Hedged Option Returns Using LASSO
Log in to collectAcademic paper
The Cross-Section of Individual Equity Option Returns
Mobina Shafaati; Don M. Chance; Robert Brooks
- GHDominion University College
- Old Dominion University
- Louisiana State University
- ?Louisiana State University, Baton Rouge - Department of Finance
- ?Louisiana State University, Baton Rouge - E.J. Ourso College of Business Administration
- University of Alabama
- ?University of Alabama - Department of Economics, Finance and Legal Studies
Strategy in a nutshell
This strategy focuses on American equity options and their underlying stocks, using data from OptionMetrics, CRSP, Compustat, and I/B/E/S. After filtering out options with non-standard settlement, early exercise, extreme prices, or arbitrage violations, 107 explanatory variables are constructed—8 option-related (including open interest, trading volume, volatility measures, and skewness/kurtosis) and 99 stock-related characteristics. Delta-hedged portfolios are formed by taking a long call option position hedged with a short position in the underlying stock. Each month, portfolios are sorted into deciles based on predicted returns for the following month, estimated via LASSO regression using the past ten years of data, with the penalty parameter selected through five-fold cross-validation. The rolling 10-year window updates coefficients monthly. The trading rule is to buy the top decile portfolios with the highest forecasted returns and sell the bottom decile with the lowest, equally weighted and rebalanced monthly.
Economic rationale
Theoretically, delta-hedged option returns should be unpredictable if options were perfectly replicable by their underlying stocks and risk-free bonds. However, empirical evidence shows systematic patterns in the cross-section of delta-hedged returns. Using LASSO regression allows the identification of a subset of option- and stock-level characteristics with predictive power, while zeroing out irrelevant variables. The rolling estimation approach captures time variation in the selected characteristics, enabling the strategy to exploit persistent cross-sectional predictability in option returns, thereby generating potential economic gains.