Volume

Reviewed 12 August 2026 · Sourced from the SEC, FINRA, S&P Dow Jones Indices, FTSE Russell, CME Group and named academic research

What volume actually counts

A stock closes down 6% on what the chart calls huge volume. The video you are watching freezes on the candle and says it: “Look at that — way more sellers than buyers.” It sounds like an explanation. It is not one. It cannot be one.

Volume is the number of shares that changed hands over a period — a minute, a day, a week. Every completed trade has exactly two sides. If 500 shares change hands, somebody sold 500 and somebody bought 500, at the same price, in the same instant. The tape records that once, as 500 shares. Not as 500 bought plus 500 sold.

So the count is always, mechanically, balanced. There has never been a session with more shares bought than sold, in any stock, ever. “More sellers than buyers” is not an exaggeration; it is a sentence with no possible referent. What people mean by it is that sellers were more urgent — willing to accept worse prices to get filled. That is a claim about price and the order book, not about volume.

So what does volume tell you? Activity. How much of the company's float actually moved, and how many people had a reason to act today rather than wait. Nothing about which way. A day where 40 million shares trade and the stock closes flat and a day where 40 million trade and it falls 9% have the same volume, because volume has no sign.

The SEC's own glossary keeps the two facts separate. The consolidated tape, it says, is “a high-speed, electronic system that reports the latest price and volume data on sales of exchange-listed stocks.” Two numbers on the same wire, and only one has a direction.

Where the number on your screen comes from

This is the part almost nobody explains, and it is the part with real sources behind it.

There is no single place where a US stock trades. There are national securities exchanges, and there is a large off-exchange market: alternative trading systems (ATSs, including the venues people call dark pools) and wholesale market makers that execute retail orders internally rather than sending them to an exchange. Reports from all of it are aggregated onto the consolidated tape, and the number your broker labels “volume” is the output of that consolidation.

How much sits off-exchange is not a rounding error. In remarks at a SIFMA conference in October 2024, then-SEC Chair Gary Gensler said that “as of 2023, approximately 45 percent of share volume was executed in the dark markets, up from 25 percent in 2009.” Close to half. A large share of that is retail order flow, which frequently never touches an exchange at all.

FINRA publishes the off-exchange side, free. Its OTC Transparency data reports over-the-counter trading on a delayed basis: ATS activity is attributed by name, and lower-volume non-ATS activity is aggregated without attribution. The quarterly ATS statistics give total shares, total trades and average trade size per ATS, split into Tier 1 NMS stocks, Tier 2 stocks and OTC securities.

Here is why that reaches you. The volume figure you read depends entirely on what is being consolidated into it. A chart drawing one exchange's prints, a chart drawing the consolidated tape, and a screener on a different vendor's feed can show three different numbers for the same session. None is lying. They count different sets of prints, and they rarely say which.

The tape's own rules have moved as well. The SEC's market structure staff note that “beginning December 9, 2013, odd lot transactions in all NMS stocks are now reported to the public tape.” Before that date, trades of fewer than 100 shares did not print publicly at all. Volume histories crossing that boundary are not measuring the same thing on both sides of it — and odd lots matter more now, because a $700 share price pushes ordinary orders into odd lots by default.

Why a raw share count tells you nothing

Forty million shares traded today. Is that a lot?

Unanswerable as asked. Forty million is enormous for one company and a slow morning for another. A raw share count has no scale attached, so it has to be put on one. There are two scales worth knowing.

Relative volume

Relative volume compares today's volume to the same stock's own recent average. It takes the company out of the comparison and leaves the change.

N is convention. Twenty and fifty sessions are the common windows, neither has an authority behind it, and both are round numbers that loosely mean “the last month” and “the last quarter.” Change N and the answer changes, which is a fair measure of how much precision it deserves.

One trap catches nearly everybody: comparing a partial day to a full day. At 10:15 a stock has traded a fraction of what it will trade by the close, so dividing that by a full-day average always looks quiet. Intraday relative volume has to be measured against volume at the same point in the session, or it is measuring the clock.

Dollar volume

The second scale is money. Dollar volume is shares traded times price, and it answers a better question: how much capital actually moved?

At the extremes it gets absurd. Hold dollar volume fixed at $50,000,000 for the day. A $5 stock needs 10,000,000 shares to get there. A $500 stock gets there on 100,000. Same money, one hundred times the share count. Any screen ranking by raw share volume puts the $5 stock near the top every day, which tells you nothing about how actively it trades.

Keep dollar volume separate from market capitalization, which gets confused with it constantly. Market cap is price times shares outstanding — a size. Dollar volume is price times shares traded — an activity. An enormous company can have a dead session and a tiny one a frantic one.

Why an ETF's volume is a different animal

Put a single stock and an ETF side by side, compare their volume, and you are comparing two mechanisms rather than two levels of interest.

A company's shares outstanding is effectively fixed day to day. If you want shares, an existing holder has to hand them over, and that prints as volume. An ETF does not work that way. Under the SEC's ETF rule, Rule 6c-11, effective 23 December 2019, an ETF may “issue (and redeem) creation units to (and from) authorized participants in exchange for a basket and a cash balancing amount (if any),” in the words of the SEC's own compliance guide for the rule. Authorized participants are large firms with a written agreement in place with the fund, and the rule also permits custom baskets that are not a pro-rata slice of the portfolio, provided the fund adopts written policies governing how baskets are constructed and accepted.

Two consequences for the volume number.

• The share count is elastic. Demand for an ETF does not have to be met by prying shares out of an existing holder. It can be met by creating new shares against a delivery of the underlying securities. So a modest volume figure on an ETF does not carry the “nobody is willing to sell” implication the same figure would carry on a single stock.
• On-screen volume understates the machinery. How much of an ETF you can actually trade depends on how easily the underlying basket can be assembled, not just on printed volume. The reverse holds too: a large ETF trade may be crossed between two investors or met with an in-kind creation, generating little or no trading in the underlying names.

So “this ETF and that stock both trade 4 million shares a day, they are equally liquid” is not a valid comparison. Relative volume against an ETF's own history still means something. Volume compared across an ETF and a single company does not.

The claim that volume confirms a move

Here is the convention, stated as fairly as it can be, because most readers will hear it from somebody who believes it.

Practitioners hold that a price move on heavy volume deserves more weight than the same move on light volume: that a breakout through a level ought to come with a visible expansion in volume, and that a rally on shrinking volume is suspect. This is taught almost everywhere technical analysis is taught, including in stage 2 of our own Technical Analysis course, Trend & Volume. The reasoning: volume is an independent series from price, so it works as a second opinion rather than a restatement.

Now the honest part. “Volume confirms” is convention. No regulator, exchange or standard-setting body defines it, publishes a threshold for it, or validates it. The figures you will see quoted — one and a half times average, twice average — are rules of thumb out of practice, and nobody publishes them as a standard.

The academic literature on the price-volume relation does exist, and it is more specific than the folklore. Jonathan M. Karpoff's survey “The Relation between Price Changes and Trading Volume: A Survey” (Journal of Financial and Quantitative Analysis 22(1), 1987, pp. 109–126) reviews the research and separates two findings: volume correlates positively with the magnitude of a price change regardless of direction, and in equity markets it also correlates positively with the signed change, meaning volume tends to run heavier when prices rise than when they fall.

Read that carefully, because it is not the trading claim. It says volume and large moves tend to occur together, on average, across big samples over long periods — not that a volume reading tells you whether the next move holds. And the second half cuts against how volume is usually invoked: if heavy volume on an advance is close to the base rate, treating it as evidence double-counts the ordinary.

Two more problems are structural rather than statistical. “High volume” can only be identified after the fact. A session's volume is not known until the session ends. During the day you are extrapolating, and the extrapolation is exactly the partial-day error described above. On a chart of last year the heavy-volume bars are obvious, and that is hindsight wearing a label.

Volume is also ambiguous on an index. The volume behind an index or a broad index ETF is the sum of thousands of unrelated decisions about hundreds of unrelated companies. Calling that aggregate “conviction” assigns one intention to a crowd that does not have one — which is why support and resistance gets hedged the same way on this site.

Volume patterns that can actually be sourced

Folklore is not the only thing on offer. Some volume behavior is large, repeatable, mechanical and documented by named sources — and it beats confirmation talk, because it tells you when a volume figure is about plumbing rather than opinion.

The closing auction

A large and growing share of the day happens in one event at the bell. Vincent Bogousslavsky and Dmitriy Muravyev measured it in “Who Trades at the Close? Implications for Price Discovery and Liquidity”: aggregate closing auction volume was 7.48% of aggregate daily dollar volume in 2018, up from 3.11% in 2010, roughly $15.2 billion on a typical day. They also found the shift is tighter than “late in the day”: volume in the 3:30 to 3:55 p.m. window fell as a share of the total, migrating into the last five minutes and the auction itself. Those figures stop in 2018; treat the level as dated and the direction as established.

The reason is mechanical: index funds are measured against an index's closing price, so they want that price, so they trade in the auction. Not sentiment — a benchmarking obligation.

Rebalancing and quarterly expirations

When an index changes what it holds, every fund tracking it has to change what it holds, on the same day, usually at the close. S&P Dow Jones Indices' equity index policies and practices put the quarterly rebalancing effective date on the third Friday of March, June, September and December. FTSE Russell's reconstitution is the largest single instance: LSEG's announcement of the June 2026 reconstitution states that roughly $12.2 trillion is benchmarked to or invested in products based on the Russell US indexes, that the event “culminates in one of the highest-volume trading days of the year across major US equity exchanges,” and that $217.2 billion traded at the close of the June 2025 reconstitution. The same release notes it is again semi-annual, in June and December, effective 2026.

Expirations pile onto the same dates. CME's equity index roll dates page anchors the futures roll to “the Monday prior to the third Friday of the expiration month,” and index futures, index options and single-stock options expirations all cluster on that third Friday, so volume balloons. The nickname, triple witching, is trader slang, not an official term.

Two more, both documented

Given the migration to the close, the modern intraday shape is more a J than a U. And the Markets · Fundamental page carries the next week of earnings dates, so you can check whether a spike is news or a calendar entry.

What happens when volume becomes an indicator

Volume gets packaged into indicators, and the most popular one contradicts the first paragraph of this page. Worth saying out loud.

On-Balance Volume (OBV) was developed by Joseph E. Granville and set out in Granville's New Key to Stock Market Profits (1963). The construction is a running total: if today closes higher than yesterday, add the whole day's volume; if it closes lower, subtract the whole day's volume.

Look at what that does. Every share in that day's volume was bought by somebody and sold by somebody. OBV assigns all of it to one side on the strength of a single number: where the stock closed. A day that closes up one cent credits 100% of its volume as buying pressure. The premise of the indicator contradicts the definitional fact that volume has no side.

That does not make it worthless, but it changes what it is. OBV is a cumulative tally of volume on up-closes minus volume on down-closes. Described that way it is honest and occasionally informative — a smoothed picture of whether the busy days have been up days. Described as “buying pressure,” it is a category error.

The same critique lands on every “buying volume vs selling volume” readout on a retail platform, and on accumulation and distribution indicators. Where those splits are not just using the close, they classify each print by whether it happened nearer the bid or the ask, and call bid-side prints seller-initiated. That is an inference about who initiated the trade, not a count of buyers, and it misclassifies plenty of prints — including much of the off-exchange flow above, which often prints between the quotes.

Two more things. VWAP, the volume-weighted average price, is arithmetic rather than a signal: total dollars traded divided by total shares traded. Institutions use it as an execution benchmark, which is the honest description of it. And volume is often smoothed with a moving average exactly the way price is, so it inherits every property of one including the lag — same for volume-based MACD-style oscillators. A smoothed volume series is not a more reliable one. It is a later one.

What is confirmed and what is convention

Claims about volume run the whole range, from a regulator's own words to trading-floor habit. Sorted honestly:

The split is easy to remember. The market structure of volume — where it prints, who reports it, what is included, which days force it — is documented and citable. The interpretation of volume is craft. Mixing the two is how a reader ends up trusting a chart annotation as much as a regulatory filing.

What trips people up

• Saying “more buyers than sellers.” There is no such state. If you mean buyers were more urgent, say that, and know you are describing price, not volume.
• Comparing raw share counts across stocks. At the same dollar value, a $5 stock out-trades a $500 stock on share count every day. Use relative or dollar volume.
• Reading intraday volume against a full-day average. At 10:15 everything looks quiet, because it is 10:15.
• Letting one block trade define a day. A single large print can double a thin stock's daily volume without anything having changed about the company. Check the print, not just the bar.
• Treating an index's volume as conviction. Thousands of unrelated prints in hundreds of unrelated companies do not add up to a shared intention.
• Naming climax volume after the fact. If the label can only be applied once the turn already happened, it described the past.
• Believing an indicator solved the two-sided problem. OBV assigns a full day's volume to one side based on the close. Bid/ask classifications infer who initiated. Neither counts buyers, because nothing can.
• Assuming your volume number is the volume number. It is whatever your provider consolidates, and providers differ on venues, odd lots, extended hours and late prints.
• Reading a calendar event as a signal. Reconstitution days, quarterly expirations and earnings dates produce enormous volume on schedule. Look at the date before you look for a story.

The figures on this page are checked against the source that publishes them, and dated. Published rates move after the release named above — the linked source always carries the current number. This page explains a term; it does not recommend a product.