[Paper Review] Option Volume Imbalance as a predictor for equity market returns

Nikolas Michael, Mihai Cucuringu+1 more|arXiv (Cornell University)|Jan 23, 2022

Capital Investment and Risk Analysis4 citations

TL;DR

This paper introduces the Option Volume Imbalance (OVI) as a predictor for equity market returns, using normalized differences in call and put option volumes across market participant classes. It finds that Market Maker OVI yields the strongest predictive power, with annualized Sharpe Ratios up to 4.5, particularly for high-implied-volatility options and put options, using a novel P&L regression approach that outperforms traditional linear models.

ABSTRACT

We investigate the use of the normalized imbalance between option volumes corresponding to positive and negative market views, as a predictor for directional price movements in the spot market. Via a nonlinear analysis, and using a decomposition of aggregated volumes into five distinct market participant classes, we find strong signs of predictability of excess market overnight returns. The strongest signals come from Market-Maker volumes. Among other findings, we demonstrate that most of the predictability stems from high-implied-volatility option contracts, and that the informational content of put option volumes is greater than that of call options.

Motivation & Objective

• To investigate whether the imbalance between call and put option volumes, categorized by market participant class, predicts future equity price movements.
• To assess the predictive power of OVI across different option characteristics such as implied volatility, delta, and rho.
• To develop and validate a novel P&L regression method tailored for directional return prediction, avoiding assumptions of linear relationships.
• To examine cross-sectional and cross-impact effects across assets using network analysis of OVI signals.
• To compare predictive performance across two major option exchanges—PHLX and NOM—using high-frequency, 10-minute bucketed data.

Proposed method

• The Option Volume Imbalance (OVI) is computed as the normalized difference between total call and put option volumes for each market participant class (MPC), with separate analysis per MPC and option type.
• A novel P&L regression framework is introduced, which directly maximizes cumulative profit and loss rather than relying on R² or t-statistics, enabling better performance under sparse data and nonlinear relationships.
• The analysis decomposes option volumes into five MPCs—Market Makers, Customers, Brokers, Firm Proprietary, and Professional Customers—using data from PHLX and NOM exchanges.
• Predictive power is evaluated using annualized Sharpe Ratios and cumulative P&L, with robustness tested via quantile ranking and alternative weighting schemes.
• Cross-impact analysis is conducted using directed networks of asset interactions, with statistical significance assessed via Bonferonni-corrected t-tests on node degree differences.
• Intraday patterns and transaction timing (opening vs. closing) are analyzed to assess temporal variation in OVI signals.

Experimental results

Research questions

• RQ1 Does the Option Volume Imbalance (OVI) derived from different market participant classes predict future equity market returns?
• RQ2 How does the predictive power of OVI vary with option characteristics such as implied volatility, delta, and rho?
• RQ3 Can a P&L regression approach outperform standard linear models in predicting directional returns from OVI signals?
• RQ4 Are there significant cross-impact effects between assets, and do they dominate direct impact effects in the network of OVI signals?
• RQ5 How do OVI signals differ across exchanges (PHLX vs. NOM), and are findings consistent across data sets?

Key findings

• Market Maker OVI delivers the highest predictive power, achieving annualized Sharpe Ratios of up to 4.5 in a simple betting scheme, even before accounting for transaction costs.
• The strongest signals emerge from high-implied-volatility options, which contribute the majority of the predictability in OVI.
• Put options carry greater informational content than call options, with negative delta and negative rho values consistently associated with higher performance.
• The 2nd to 4th quantile rank groups of OVI magnitude yield the best-performing portfolios, indicating that moderate imbalance levels are most informative.
• Cross-impact effects are significantly more prevalent than direct impact effects, with 41–58% of network degree differences rejecting the null hypothesis under uncorrected tests.
• Results from the NOM exchange (NOTO data set) are consistent with PHLX (PHOTO data set), though weaker, with Market Maker OVI still showing annualized Sharpe Ratios of 2.5–3.