Dealer Gamma Exposure and Its Measured Effect on Realized Volatility

The retail narrative around dealer gamma exposure (GEX) treats it as a near-deterministic lever: positive GEX compresses volatility, negative GEX explodes it, and a chart of strike-level gamma tells you where price will pin. The mechanism behind that narrative is real. The magnitude is not what the narrative implies. A 2025 working paper using proprietary Cboe trade-level data puts an actual number on dealer gamma's contribution to S&P 500 realized volatility, and the number is far smaller than the GEX-chart industry suggests.

The mechanism, precisely

When a customer buys a call, the options market maker (OMM) on the other side is short that call and delta-hedges by buying the underlying. As the underlying moves, the option's delta changes, and the dealer must trade more to stay delta-neutral. The size of that rebalancing trade is governed by gamma: summed across every option series in the dealer's book, the aggregate position's net gamma determines how much the dealer must trade for a given move in the underlying. When aggregate dealer gamma is negative, dealers sell into declines and buy into rallies, a pro-cyclical flow that amplifies the underlying move. When it is positive, the flow runs counter-cyclical and dampens it.

This is the entire theoretical basis for GEX. The open question, and the one most retail GEX dashboards skip, is how large the effect actually is in the data, and whether public open-interest figures are even the right input to measure it.

Why open interest isn't dealer positioning

Most retail-facing GEX tools compute gamma exposure from end-of-day open interest, under the assumption that all customer option buying is matched by dealer selling (and vice versa). Amaya, Garcia-Ares, Pearson, and Vasquez (2025) test this assumption directly using Cboe's complete trade-level records for SPX and SPXW options from 2020 to 2023, which tag every trade by counterparty capacity (customer, market maker, firm, professional customer, broker-dealer). The result: on the average day, only about 69% of SPX/SPXW volume is a customer trading directly against a market maker. Roughly 9% is market-maker-to-market-maker, and the remainder runs through firms, professional customers, and broker-dealers, categories a public OI-based GEX model has no way to net out correctly. Treating gross customer volume as a clean proxy for dealer inventory therefore introduces a structural error before any volatility analysis even begins.

Measuring the actual relationship

The paper's contribution is reconstructing the OMM's true net gamma position at one-minute frequency by cumulating every signed market-maker trade from each option series' inception, then computing the gamma of that net position using Black-Scholes-implied volatilities updated minute-by-minute. It then relates one-minute S&P 500 futures returns to lagged OMM gamma using a GARCH-style model: today's volatility depends on yesterday's volatility plus a separate term that pulls in recent dealer gamma readings, with more recent gamma values weighted more heavily than older ones. In plain terms, the model lets recent dealer positioning nudge the volatility forecast up or down, on top of whatever volatility momentum is already in the data, so that the effect of gamma can be isolated from ordinary day-to-day clustering in volatility.

The coefficient measuring that gamma effect comes out negative and statistically significant across all 36 months tested, confirming the theoretical sign: more positive dealer gamma is associated with lower volatility, and more negative dealer gamma with higher volatility. But the magnitude is the part worth sitting with.

The actual numbers

Two findings cut against the popular GEX narrative. First, aggregate OMM gamma is typically positive, not short, as the common retail framing assumes ("dealers are always short gamma" is not what the trade data shows), though negative readings became substantially more frequent after Cboe introduced Tuesday/Thursday SPX expirations in 2022, consistent with dealers absorbing more short-dated, higher-gamma inventory as 0DTE volume grew.

Second, and more importantly: the median effect of OMM gamma is to reduce daily realized volatility by roughly 0.08 to 0.2 percentage points, simply because gamma is usually positive. Using the model to simulate a counterfactual world with no gamma-driven hedge rebalancing at all, the maximum impact of OMM gamma found anywhere in the three-year sample is an increase of 3.3 percentage points in annualized daily realized volatility, and 6.4 percentage points at the 30-minute horizon. Those sound large in isolation. They are not, relative to baseline variability: the standard deviation of day-to-day changes in annualized realized volatility is 4.5 percentage points, and changes of 3 points or more happen on roughly 20% of trading days, about once a week, for reasons that have nothing to do with dealer gamma. By the paper's own benchmark, a gamma-induced swing of this size "is not large."

What this means for how GEX should actually be used

This doesn't mean GEX is meaningless. The sign of the relationship is confirmed with real statistical power, and the mechanism is mechanically sound. What it means is that GEX functions as a modest, usually-stabilizing background force, not the dominant, mechanically-predictable lever depicted in most GEX dashboards and chart overlays. Three implications follow directly:

• A strike-level GEX chart built from public open interest is an approximation of an approximation. It assumes a customer/dealer matching rate the trade data shows is closer to 69% than 100%, and it cannot see market-maker-to-market-maker or broker-dealer flow at all.
• The predictive content of GEX is concentrated in the tails, not the average day. On most days, dealer gamma's contribution to realized volatility is a fraction of a percentage point, well within the noise of ordinary day-to-day vol fluctuation. The cases where it matters are the negative-gamma tail events, which is exactly where public OI-based models are least reliable, since extreme moves are also when intraday position changes diverge furthest from the prior day's OI snapshot.
• Backtesting a GEX-conditioned strategy on EOD open interest will overstate both the strength and the precision of the underlying edge, because the input itself is a noisy proxy for the thing the academic literature actually measured using proprietary, trade-tagged data.

For anyone building a systematic strategy around dealer positioning, premium selling in positive-GEX regimes, long-gamma exposure when GEX turns negative, the rigorous version of this question isn't "what does the GEX chart show today," but "how sensitive is my backtested edge to the gap between public OI-based gamma estimates and the actual net dealer position." That's a question that requires testing against historical options data with enough granularity to model the difference, not a daily snapshot.

Quantropy is built to run exactly that kind of test: translate a plain-English dealer-positioning strategy into a backtest against tick-level historical options data, with full Greeks rather than a static end-of-day approximation. If you want to see how a GEX-conditioned strategy holds up once it's tested properly, join the waitlist at quantropy.ai.

Sources

1. Amaya, D., Garcia-Ares, P. A., Pearson, N. D., & Vasquez, A. (2025). "0DTE Index Options and Market Volatility: How Large is Their Impact?" Working paper, hosted by Cboe Global Markets.
2. Ni, S. X., Pearson, N. D., Poteshman, A. M., & White, J. (2021). "Does Option Trading Have a Pervasive Impact on Underlying Stock Prices?" Review of Financial Studies, 34(4), 1952-1986.
3. Baltussen, G., Da, Z., Lammers, S., & Martens, M. (2021). "Hedging Demand and Market Intraday Momentum." Journal of Financial Economics, 142, 377-403.
4. Cboe Global Markets. "Cboe to Add Tuesday and Thursday Expirations for SPX Weeklys Options" (2022 announcement, providing the regulatory timeline referenced in the working paper above).