SPB Git

spb/wp7_uqo Public

UQO Working Paper No. 7 — Options-implied information for cross-asset return and volatility prediction: evidence from 3.8B option contracts.

Python 66.5% TeX 32.7% Makefile 0.8%
2.1 KB · 24 lines latex
Raw Blame History
1% =============================================================================2% Author: Simon-Pierre Boucher3% Contact: contact@spboucher.ai4% =============================================================================5% Conclusion6% =============================================================================7\section{Conclusion}\label{sec:conclude}89This paper provides comprehensive evidence on the information content of the options market, drawing on 3.83~billion option contracts and 11.5~billion intraday observations spanning 2010--2025 and six asset classes. The findings can be summarized as follows.1011\textbf{Finding 1.} Implied kurtosis, put-call ratios, and implied skewness predict weekly cross-sectional stock returns. Quintile long-short portfolios sorted on implied kurtosis deliver annualized Sharpe ratios of 2.33 ($t=19.84$). Daily predictability is negligible.1213\textbf{Finding 2.} Combining the IV surface with HAR-RV improves 1-day realized-variance forecasting by 23.3\% in $R^2$, an improvement that is robust across nine subperiods.1415\textbf{Finding 3.} The implied-to-realized correlation ratio predicts market stress at 5- to 20-day horizons ($t$-statistics: 3.91--8.30).1617\textbf{Finding 4.} The information content of the Greeks for realized volatility increases with time-to-expiry ($R^2$: 3.8\% at one week $\to$ 9.5\% at three months). The price-magnet hypothesis is not supported.1819\textbf{Finding 5.} HAR-RV dominates machine-learning models and the VIX for out-of-sample SPX realized-variance forecasting (2020--2025). Short-tenor ATM implied volatility carries the highest marginal predictive content (50.8\% of Random Forest importance).2021\textbf{Cross-cutting result.} ATM implied volatility Granger-causes realized volatility in 100\% of tickers ($F=62.4$), and IV shocks explain 73.8\% of the RV forecast error variance at the 20-day horizon.2223Future research should explore regularized approaches to exploiting the high-dimensional options surface, assess the net profitability of option-implied strategies under realistic transaction costs, and extend the analysis to international markets.24