Dealer Gamma and Market Impact: Pinning, Intraday Momentum, 0DTE, and the Limits of GEX

Dealer gamma is the combined gamma of the option positions held by the market makers and banks that delta hedge, and it matters because their hedging trades push on the underlying market. When dealers are long gamma, delta hedging makes them sell into rallies and buy into dips, which tends to dampen moves. When they are short gamma, hedging makes them buy as prices rise and sell as prices fall, which tends to amplify moves.

In short: the mechanism is real and documented in peer-reviewed research. Ni, Pearson and Poteshman (Journal of Financial Economics, 2005) showed that stock prices cluster at strike prices on expiration days, partly because of hedge rebalancing. Baltussen, Da, Lammers and Martens (Journal of Financial Economics, 2021) linked intraday momentum in the last half hour of trading to short-gamma hedging. Studies of 0DTE index options that use actual market-maker positions find dealer gamma is usually positive and usually reduces volatility, with occasional negative-gamma periods that raise it by a few points. Public “GEX” numbers rest on assumptions about who holds what, and those assumptions can flip the sign of the answer.

This article builds on gamma explained and the hedging mechanics in how options market makers hedge. It describes evidence and mechanisms, not trading signals.

The misconception: “GEX tells you where the market is going”

Daily commentary often reports a single gamma exposure figure and a “flip level” below which dealers are said to turn short gamma and markets become unstable. The underlying idea is sound. The precision is not. Nobody outside the exchanges and the dealers knows the dealer side of every trade, over-the-counter positions are invisible, and dealers hedge with other options as well as with futures. The research below is careful about these limits. Much of the commentary is not.

The mechanism, with numbers

A delta-hedged option position needs a hedge adjustment of roughly gamma times the price change for each option-share. For index options, the dollar amount to trade for a 1% move is:

Hedge trade for a 1% move = gamma x open interest x contract multiplier x (1% of the index level) x index level

The sign is what matters for market impact. A dealer that is long an option (positive gamma) sees its delta rise as the market rises and sells to stay neutral. A dealer that is short an option does the opposite. Whether aggregate dealer hedging leans against the market or with it depends on the net gamma across every strike and expiry, and that net depends on who bought and who sold each contract.

Evidence 1: pinning at expiration

Ni, Pearson and Poteshman (“Stock Price Clustering on Option Expiration Dates,” JFE, 2005) used OptionMetrics data on every US equity option from January 1996 to September 2002 and Cboe open interest broken down by investor type. Their findings:

A larger share of optionable stocks closed near a strike price on expiration Fridays than on the Fridays before and after, and the difference was statistically significant.
Stocks without listed options showed no such clustering, and stocks that became optionable started clustering only after they had options.
Optionable stocks’ returns were altered by at least 16.5 basis points per expiration date on average, a lower bound that implies shifts in market capitalization on the order of $9 billion per expiration date.

They tested four explanations and found support for two: delta-hedge rebalancing by traders with net purchased option positions (a mechanism modeled by Avellaneda and Lipkin, 2003), and stock price manipulation by firm proprietary traders who had written options in the week before expiration. When likely hedgers were net long the nearest-to-expiry, nearest-to-the-money options, clustering rose into expiration. When they were net short, clustering fell in the days before expiration, though it still rose on the expiration date itself.

Worked example: why long-gamma hedging pins

A delta-hedged trader is long 10,000 contracts (1,000,000 shares) of a $50 call. Volatility is 25% and the interest rate 4%. Hold the stock price fixed and let time pass:

Stock price	Days to expiry	Call delta	Hedge (short shares)
$50.50	2	0.7118	711,845
$50.50	1	0.7809	780,925
$49.50	2	0.3008	300,837
$49.50	1	0.2257	225,686

Above the strike, the call’s delta drifts toward 1 as expiry nears, so the hedger must sell 69,080 shares over a day with no change in price. Below the strike, delta drifts toward 0 and the hedger buys back 75,151 shares. Both trades push the stock toward $50. Reverse the position (a hedger short the calls) and the same arithmetic pushes the stock away from the strike. That is exactly the asymmetry Ni, Pearson and Poteshman found in the data.

Evidence 2: volatility of individual stocks

Ni, Pearson, Poteshman and White (“Does Option Trading Have a Pervasive Impact on Underlying Stock Prices?”, Review of Financial Studies, 2021) extended the idea from expiration days to every day. They report evidence consistent with the hypothesis that option market makers’ hedge rebalancing affects stock return volatility and the probability of large price moves: the more net purchased option positions held by likely hedgers, the lower the subsequent volatility of the underlying stock. This is the cross-sectional version of the claim that long-gamma dealers dampen moves.

Evidence 3: intraday momentum

Baltussen, Da, Lammers and Martens (“Hedging Demand and Market Intraday Momentum,” JFE, 2021) studied more than 60 futures on equity indexes, bonds, commodities, and currencies from December 1974 to May 2020. Across asset classes, the return over the rest of the day predicts the return over the last 30 minutes before the close, and the effect reverses over the following days. They link it to hedging by participants with short gamma, such as option market makers and leveraged ETFs.

For the S&P 500 they built a net gamma exposure (NGE) proxy from open interest: gamma times open interest times 100 times the index level for each call, the same with a minus sign for each put, summed and divided by the market value of the index. The construction assumes, in their words, that call options are sold by investors and bought by market makers, that puts are bought by investors and sold by market makers, and that market makers hedge their deltas exactly. Results:

Between 1996 and May 2020 there were 2,930 days with negative NGE and 3,158 with positive NGE.
On days after negative NGE, the regression coefficient of the last-half-hour return on the rest-of-day return was 6.63 (t-statistic 4.78, coefficients multiplied by 100), with an R-squared of 3.58%.
On days after positive NGE, the coefficient was 0.82 (t-statistic 1.03), not significant, with an R-squared of 0.05%.

The authors are explicit about the limits. Their footnote notes that the data cover only listed options, that the assumption that market makers are short all puts and long all calls is a proxy, and that the sample misses OTC positions such as institutional put buying and insurer hedging. They also report that markets with more leveraged ETF activity show stronger intraday momentum.

Evidence 4: 0DTE options

Same-day expiring S&P 500 options carry enormous gamma near the money, which raised the question of whether their hedging destabilizes the index. With the index at 6,000 and 15% volatility, and time measured in trading hours (252 days of 6.5 hours), an at-the-money 0DTE option at the open has about 4.6 times the gamma of an at-the-money option with 21 trading days left. With 3 hours left it has about 6.8 times, and with 1 hour left about 11.7 times.

The best evidence comes from studies that use trade-level data identifying the market maker side, rather than assumptions:

Amaya, Garcia-Ares, Pearson and Vasquez (“0DTE Index Options and Market Volatility: How Large is Their Impact?”, working paper dated January 25, 2025, using trade records provided by Cboe Global Markets) reconstructed the aggregate market-maker position in every SPX and SPXW series at one-minute frequency from July 2020 to June 2023. Market-maker gamma was typically positive but often negative. The typical effect was to reduce volatility (the median impact cut daily volatility by 0.08 percentage points). The maximum impact was to raise annualized daily volatility by 3.3 percentage points and annualized 30-minute volatility by 6.4 percentage points, which the authors note is well within the range of normal daily changes in realized volatility.
Dim, Eraker and Vilkov (“0DTEs: Trading, Gamma Risk and Volatility Propagation,” SSRN working paper, 2024) find that market makers’ inventory gamma is on average positive and negatively related to future intraday volatility, with positive gamma strengthening intraday reversals and negative gamma strengthening momentum.
Adams, Fontaine and Ornthanalai (2024), as summarized by Amaya et al., exploit the introduction of Tuesday and Thursday SPXW expirations (first expiring April 26 and May 19, 2022) and find market volatility is lower on days when 0DTE options are available. That paper and Dim, Eraker and Vilkov’s have since been merged into a single July 2026 draft by all six authors, which puts the reduction at about 7% of average 10-minute realized volatility.

The dedicated treatment of this market is in inside the 0DTE market.

What central bank research has said

The Bank for International Settlements has written about dealer hedging in both directions.

Dampening, March 2024. A box in the BIS Quarterly Review (March 2024) argued that the growth of yield-enhancing structured products (a covered call is the simplest example), through which clients sell options that dealers buy, leaves dealers long options. Hedging those positions means buying when the index falls and selling when it rises, which the BIS described as acting “in a contrarian way, effectively dampening the price movements of the underlying asset.” The same box reported that 0DTE options accounted for more than 50% of SPX options volume in August 2023, up from 5% in 2016, but argued that 0DTE trading was unlikely to be the main reason the VIX was low.
The August 5, 2024 VIX spike. BIS Bulletin No. 95 (Todorov and Vilkov, October 29, 2024) found that because the VIX is computed from quotes rather than trades, dealers’ quote-setting and the widening of bid-ask spreads, particularly for puts, contributed to the record one-day spike, and that VIX ETFs and dispersion trades were unlikely to have been the main driver.
Amplification, September 2026. A box in the BIS Quarterly Review (September 2026) on Korean memory-chip stocks argued that leveraged ETFs, dealers short calls sold to investors, and issuers of autocallable products all had to trade in the direction of the move. It estimated that for SK Hynix the leveraged-product rebalancing needed on a 10% price move rose to almost $5 billion at a June 2026 peak, against daily turnover under $10 billion, and could add roughly 2 to 4 percentage points to a 10% move. It also reported that the ratio of call to put open interest on SK Hynix rose about thirtyfold between March and June 2026, peaking near 100.

The pattern across these papers: whether dealer hedging dampens or amplifies depends on which side of the trade end users are on, and that differs by product and by market.

Worked example: how the sign assumption flips a GEX estimate

Take a hypothetical index at 6,000 with one expiry 21 calendar days away, a 4% rate, a 1.3% dividend yield, and this open interest (put volatilities higher than call volatilities, as skew usually has it):

Option	Strike	Implied vol	Open interest (contracts)	Dollar gamma per 1% move
Put	5,400	24%	60,000	$0.42 bn
Put	5,700	20%	80,000	$2.12 bn
Put	5,850	18%	50,000	$2.25 bn
Put	6,000	15.5%	30,000	$1.93 bn
Call	6,000	15%	40,000	$2.65 bn
Call	6,150	13%	60,000	$3.54 bn
Call	6,300	12%	70,000	$1.55 bn

Now aggregate under two assumptions. Assumption A is the one Baltussen et al. use: dealers long every call and short every put. Assumption B: dealers short everything, as they would be if end users had bought all of these options (for example retail call buying plus institutional put buying).

Index level	A: dealer hedge trade per 1% move	B: dealer hedge trade per 1% move
5,700	-$6.99 bn (short gamma)	-$9.54 bn
5,850	-$4.63 bn (short gamma)	-$12.08 bn
6,000	+$1.02 bn (long gamma)	-$14.45 bn
6,150	+$7.09 bn (long gamma)	-$15.32 bn
6,300	+$8.64 bn (long gamma)	-$12.39 bn

Same open interest, same model, same market. Under A, dealers are mildly long gamma at 6,000 and the “flip level” sits just below the current price. Under B, dealers are heavily short gamma everywhere. A positive figure means that for a 1% rise, dealers would sell that much; a negative figure means they would buy into the rise. Neither column is “the” GEX. The difference is entirely the unobservable question of who sold.

The limits of GEX claims

Professionals who use these estimates treat them as rough context, for reasons the academic papers themselves spell out:

The sign is assumed. Public estimates infer dealer positions from open interest, which counts contracts but not who holds them. Amaya et al. could measure the market-maker side only because Cboe supplied capacity-coded trade records.
OTC positions are missing. Baltussen et al. note that their measure omits OTC puts bought by institutions and insurers. Structured-product hedges (see structured products and autocallables) sit largely off exchange.
Intraday positions are missing. Open interest is reported once a day. Much 0DTE volume opens and closes within the session, so a morning GEX figure can be stale by lunch. The one-minute reconstruction in Amaya et al. exists for exactly this reason.
Dealers do not all hedge the same way. Some hedge with offsetting options rather than futures, some rebalance on bands rather than continuously, and some carry inventory deliberately. The formula assumes exact, continuous delta hedging.
Model inputs move the flip level. Gamma depends on implied volatility and time. A higher volatility assumption spreads gamma across strikes and moves the computed flip point.
Size relative to liquidity is what matters. A $5 billion hedge flow means something different in S&P 500 futures than in a single stock, which is why the BIS’s 2026 Korean example compared flows to turnover in the stock itself.
Measured effects are modest on average. The strongest data set available put the maximum 0DTE gamma effect at a few volatility points, not a crash mechanism.

What can go wrong when gamma does matter

Expiration days in single stocks. Pinning effects are concentrated in names with large open interest near the money relative to their trading volume, and they can make expiration-day closing prices less informative.
Concentrated one-sided demand. When end users pile into one side (calls in a handful of popular stocks, or leveraged long ETFs), dealer and issuer hedging all point the same way, as the BIS described for Korea in 2026.
Barrier products. Issuers of autocallables and knock-in notes can face sharp changes in delta near barrier levels, which turns a dampening hedge into an amplifying one.
Volatility feedback. Short-gamma episodes coincide with rising implied volatility, which raises margin and widens quotes at the same time, as the August 2024 episode illustrated through quote-setting rather than hedging flows. See also Volmageddon: February 2018.

The short version

Dealer hedging moves prices in a direction set by the sign of dealers’ net gamma: long gamma leans against moves, short gamma leans with them. The best research confirms the mechanism at specific moments (expiration pinning, the last half hour of trading) and finds that, in the 0DTE market where data on actual market-maker positions exist, dealers are usually long gamma and the effect on volatility is usually small. Public GEX figures depend on an assumption about who sold each option, and a change in that assumption can turn a dampening estimate into an amplifying one. Professionals treat them as a hypothesis to check, not a forecast.