Pan & Poteshman — The Information in Option Volume for Future Stock Prices (Review of Financial Studies 19(3), 2006)

T1 — peer-reviewed, Review of Financial Studies 19(3), 2006, pp. 871–908 (review-of-financial-studies). jun-pan (MIT Sloan and NBER, massachusetts-institute-of-technology) and allen-poteshman (University of Illinois at Urbana-Champaign). Read from the authors' PDF; also at https://academic.oup.com/rfs/article-abstract/19/3/871/1646711.

This is the academic form of what retail platforms sell as unusual options activity (unusual-options-activity): does a burst of option buying tell you where the stock goes next? The paper answers yes, and then spends most of its length on the part the retail version leaves out — which volume, seen how.

The signal

From a proprietary chicago-board-options-exchange dataset covering all equity options from January 1990 through December 2001. Its value is that every trade carries three flags the public tape does not have: whether a buyer or a seller initiated it, whether the trade opened or closed a position, and which class of investor was on it (firm proprietary trader, or a public customer of a discount, full-service or other broker).

The signal is a put-call ratio built only from open-buy volume — puts divided by puts plus calls, counting only contracts bought to open a new position. Stocks need at least 50 contracts of open-buy volume to enter that day's cross-section, which averages 242 stocks.

What it predicts

Regressing the next day's four-factor risk-adjusted return on the ratio gives a slope of −53 basis points, t = −32.92. In quintiles, the low-ratio stocks beat the high-ratio stocks by more than 40bp the next day and more than 1% over the following week, and the effect decays over several weeks — the profile of information being absorbed, not of a mispricing.

Predictability is stronger where information-based models say it should be: in stocks with a higher PIN (more informed traders, and this survives a size control), and in more levered contracts. By moneyness, options more than 10% out of the money give a slope of −44.67 while options more than 10% in the money flip sign to +7.95; by maturity, under 30 days gives −34.83 and beyond 179 days gives −6.91. Among the investor classes, customers of full-service brokers carry the signal, and firm proprietary traders carry none at all — which the authors read as those desks being in the option market to hedge rather than to bet.

The finding that matters most for a screener

The predictive volume is not publicly observable. When the authors rebuild the signal the way an outsider must — inferring buyer-initiated volume from the public trade-and-quote record with the Lee and Ready (1991) algorithm, which cannot see open versus close — the public signal predicts for one or two days and then reverses, which looks like price pressure rather than information. Put both signals in one regression and the nonpublic one keeps its full effect while the public one has no predictability at all.

Two more constraints on the retail version. Open-sell volume carries information in the right direction but much less of it, and close-buy volume carries none, so the "open" flag is doing real work. And applied to index options (SPX, OEX, NDX) the whole test yields no evidence of informed trading, which fits informed traders holding firm-specific rather than market-wide information.

How to read it here

The paper does not claim a tradeable strategy and does not test one net of costs. Its own reading of the result is that predictability comes from private information gradually entering prices — evidence about price discovery, not an edge on offer. Set beside backtest-overfitting and pseudo-mathematics-financial-charlatanism, this is the rare result in this corpus where the statistical work was done by someone with no product to sell, and it still stops short of the claim a trading-bot-platform makes casually.

Related

unusual-options-activity · options-greeks · dealer-gamma-exposure · option-pricing · pervasive-impact-of-option-trading · gamma-fragility · jun-pan · allen-poteshman · synthesis