Research · Real data
Short Volume as a Share of All Trading
Measured from FINRA's daily files across 43 sessions: short sales ran between 45.07 and 51.52 per cent of all consolidated US equity volume, averaging 49 per cent. The number is real, it is stable, and it means almost nothing like what it appears to mean.
The measurement
Every trading day FINRA publishes a file listing, for each US equity symbol, the volume that was marked short and the total consolidated volume. It is free, it needs no registration, and it is one of the few genuinely public windows onto how the tape is composed.
Aggregating 43 sessions from 2026-06-08 to 2026-08-07, covering an average of 11,980 symbols per session and around 8.29 billion shares a day, gives a strikingly narrow result: short-marked executions were 49% of all volume on average, and never left the band between 45.07% and 51.52%.
What it is not measuring
A reader meeting "half of all trading is short selling" for the first time reasonably concludes that the market is heavily bet against. The figure above rules that out on its own: total volume nearly doubles across these sessions, from around 6.59 to 12.11 billion shares, while the short share stays inside a seven-point band. Sentiment does not behave like that. Market structure does.
The mechanism is mundane. When a market maker fills your buy order from inventory it does not hold, that execution is marked short, and it will be flattened within seconds. Add hedging against options positions and the short legs of index arbitrage, and a tape on which about half of all executions are marked short is what continuous two-sided liquidity provision looks like. None of it carries a view.
The corollary matters more than the finding: because the baseline is roughly 50 per cent, a single symbol showing "60 per cent short volume" is a far weaker signal than it sounds. The question is never the level but the deviation from that symbol's own normal, and even then it is a statement about how the flow was executed, not about who is betting what.
The distribution
An average alone hides whether a series is stable or merely centred. Across these 43 sessions the daily short share has a median of 49.07% against a mean of 49%, almost no skew, and a standard deviation of only 1.52 percentage points.
That tightness is the finding. Very little that is published daily about the American tape holds within a point and a half of itself for two months: not volume, not volatility, not breadth, not advancing issues. A quantity that does is describing the mechanics of how trades are executed rather than what anyone thinks, and it sets the bar for what would count as an unusual reading. Two standard deviations here is a move of about three points, which is why "60 per cent short volume" in a market-wide context would be extraordinary and, in a single symbol, is ordinary.
Does the share move with activity?
The obvious hypothesis is that busy days are busy because short sellers are active, which would put the share at its highest when volume is at its highest. The data says the opposite. Across this window the correlation between the daily short share and total consolidated volume is -0.54, moderately negative.
Splitting the sessions into three groups by total volume makes the relationship legible without the statistic.
| Sessions | Mean volume | Mean short share |
|---|---|---|
| Quietest third (14) | 7.21bn | 50.17% |
| Middle third (14) | 7.99bn | 49.43% |
| Busiest third (15) | 9.59bn | 47.51% |
The busiest third of sessions traded substantially more shares than the quietest third and carried a lower proportion of short-marked volume. The likeliest reading is dilution rather than any change in short-selling behaviour: market-making flow is close to a constant fraction of ordinary activity, and a heavy session is heavy because directional participants arrived in size and were executed long. Their volume swells the denominator while adding comparatively little to the numerator.
Two caveats belong with that reading. Forty-three sessions is a short window, and a correlation of this size on this sample is suggestive rather than settled. And the mechanism proposed here is an explanation, not a measurement, the file cannot identify participants, so nothing in it proves who executed the extra volume.
A test that found nothing
Grouping the same sessions by day of the week is the sort of check that produces a publishable pattern if you go looking for one. Here it does not: the range between the lowest weekday mean and the highest is 0.63 percentage points, well inside the daily variation, on seven to nine observations per day.
| Weekday | Sessions | Mean short share |
|---|---|---|
| Monday | 9 | 48.9% |
| Tuesday | 9 | 48.65% |
| Wednesday | 9 | 49.28% |
| Thursday | 9 | 49.04% |
| Friday | 7 | 49.17% |
The result is reported because it was run. A study that publishes only the tests that worked is not a study, and a reader deciding whether to trust the correlation in the previous section is entitled to know that the same data was also asked an easier question and answered no.
Short-exempt volume
The file carries a separate column for short-exempt volume, short sales not subject to the Reg SHO price test, which applies only while a circuit-breaker restriction is in force after a 10 per cent decline, and which bona fide market-making and certain arbitrage transactions can be marked against. Across this window it was 2.03% of all consolidated volume and 4.15% of short volume.
Small, but large enough to matter to anyone publishing a figure. The convention followed here is to include exempt volume in the short totals, because the question being asked is how many executions were marked short at all. Excluding it is equally defensible if the question is about price-test-constrained selling, and it lowers every percentage on this page by roughly three points, which is two standard deviations of the daily series. Two analysts can therefore disagree by more than the entire natural variation of the measure without either being wrong, purely on a definitional choice.
A second finding, from the same file
The most-traded symbols by share count are not the largest companies. On 2026-08-07 the highest consolidated volumes belonged to low-priced and leveraged tickers rather than to mega-caps, a consequence of counting shares rather than dollars, since a two-dollar stock produces fifty times the share volume of a hundred-dollar stock for the same money committed.
| Symbol | Total volume | Marked short |
|---|---|---|
| HCWC | 336.5M | 53.75% |
| MSTU | 138.8M | 52.93% |
| SPCX | 113.2M | 56.97% |
| HTZ | 103.8M | 57.6% |
| QNME | 71.1M | 55.57% |
| WWR | 67.7M | 60.44% |
| VIVS | 64.4M | 57.13% |
| TTD | 56.5M | 46.79% |
| ONDS | 51.9M | 46.1% |
| BYND | 51.2M | 80.42% |
This is the practical reason every volume comparison on this site is made against an instrument's own history rather than across instruments. Share volume is not a common currency, and any ranking built from it is mostly a ranking of share prices.
How to reproduce this
Source: https://cdn.finra.org/equity/regsho/daily/, one file per trading day, named for the date. The aggregation used here sums total and short volume across every symbol with positive volume in each file, then divides. The files are public and the steps above are the whole method, so the study can be rerun over any window; the figures here are fixed to the range stated in the caption and are not a live feed.
Two limits worth stating. Same-day round trips are counted twice, once as a short sale and once as the covering purchase, so the file overstates directional activity by construction. And short-exempt volume (a separate column, small in absolute terms) is included in the short figures here, which is the convention but not the only defensible choice.
Frequently asked questions
Does half of all volume being short mean the market is bearish?
No, and this is the reading the number invites and does not support. Most of that short volume is mechanical. When a market maker sells stock it does not hold in order to fill a buy order, that execution is marked short even though the firm has no directional opinion and will flatten within seconds. Add hedging against options inventory and index-arbitrage legs, and roughly half of all executions being marked short is simply what continuous two-sided liquidity provision looks like on a modern tape.
Is short volume the same as short interest?
No, and confusing them is the single most common error with this data. Short volume counts executions marked short during a session, a position opened and closed the same day appears in it twice and leaves no trace afterwards. Short interest is the stock of open short positions, reported twice monthly. A symbol can show heavy short volume every day while its short interest barely moves, because the same shares are being sold and bought back continuously.
What does the file actually cover?
FINRA’s daily Reg SHO file reports consolidated volume for every US equity symbol, aggregating trades reported to FINRA facilities and the exchange tapes. It is published each trading day and is free to download. It does not break volume down by venue in this form, does not separate retail from institutional, and does not include options or futures volume.
Why is this page dated rather than live?
Because the site is a static build and an honest snapshot is better than a stale feed. The figures above cover a specific range of sessions and say so. A page that presents a fixed number as though it updates is the kind of quiet decay that makes a reference untrustworthy. The source files are public and the method is stated in full below, so the study can be rerun over any window.
Why is the short share lower on the busiest days?
That is the clearest finding in this window: the correlation between the short share and total volume is -0.54, so heavy sessions carry proportionally less short-marked volume rather than more. The likeliest explanation is dilution. Market-making flow is close to a constant fraction of ordinary activity, and a high-volume session is high-volume because directional participants arrived in size, real buyers and sellers moving positions, executed long. Their volume enters the denominator without adding much to the short numerator.
How stable is the number over time?
Over these sessions the standard deviation of the daily share is 1.52 percentage points around a mean of 49 and a median of 49.07. For a market statistic that is extraordinarily tight, tighter than almost anything else published daily about the tape. It is the strongest available evidence that the measure is describing plumbing rather than opinion, because opinion does not hold to within a couple of points for two months.
Is there a day-of-week effect?
Not a usable one. Grouping the same sessions by weekday, the mean short share spans only 0.63 percentage points from the lowest day to the highest, well inside the daily variation, on samples of seven to nine sessions each. Anyone reporting a weekday pattern in short volume from a sample this size is reporting noise, and the honest result of running the test is that there is nothing there.
What is short-exempt volume?
A subset of short sales exempt from the Reg SHO price test, which applies only when a stock has fallen 10 per cent in a day and a circuit-breaker restriction is in force. Bona fide market-making and certain arbitrage transactions can be marked exempt. In this window it was 2.03 per cent of all volume and 4.15 per cent of short volume, small, but not negligible, and whether you include it changes any figure you publish.
Can I use this data on a single stock?
Yes, and it is more useful there than in aggregate, provided the baseline is that symbol’s own. Because the market-wide share sits near half, a symbol at 60 per cent is not "heavily shorted" in any meaningful sense; a symbol whose share has run at 40 per cent for a year and prints 65 per cent is showing something worth a look. The comparison is always against its own distribution, never against the market number or against another symbol.
Does it show who is selling short?
No. The file reports volume by symbol and nothing about the participants, no split between retail and institutional, no venue breakdown in this form, no positions. Any narrative about hedge funds attacking a particular stock that cites this data is adding the actors from imagination. The data supports statements about how executions were marked, and stops there.
Why are the most-traded symbols not the biggest companies?
Because the file counts shares, not money. A two-dollar stock produces fifty times the share volume of a hundred-dollar stock for the same capital committed, so any ranking by share volume is substantially a ranking by low price. The practical consequence, which applies to every volume comparison anywhere on this site, is that share volume is not a common currency between instruments.
How far back does the FINRA file go?
Daily files are available for many years, one per trading day, published free and without registration. Coverage and formatting have changed over time — the fields and the consolidation of the reported facilities are not identical throughout — so a very long history needs the file layout checked at each break rather than assumed. For most questions a few months is plenty, because the series is stable.
Does a rising short share predict a decline?
Nothing in this window supports that, and the negative correlation with volume points the other way: the share tends to be lowest exactly when activity is heaviest, which is often when prices are moving most. A test worth running on a longer sample is whether the deviation of a symbol from its own baseline has any forward relationship with returns. Publishing the aggregate number as a sentiment gauge, which is common, is not supported by its own behaviour.
How does this relate to up/down volume?
They are complementary halves of one question. Up/down volume splits the day’s volume by whether the issue closed higher or lower, a measure of where activity went. Short volume splits the same total by how the execution was marked, a measure of how it was transacted. Neither reveals intent, and using both keeps the difference between participation and execution visible.
Can these figures be reproduced exactly?
Yes, and that is the point of stating the method. Take the daily Reg SHO consolidated files for 2026-06-08 to 2026-08-07, sum total and short volume across every symbol with positive volume in each file, and divide. If your numbers differ, the usual causes are including symbols with zero volume, mixing the consolidated file with a single-facility file, or excluding short-exempt volume from the short total.