By Matt Connor
Do Stock Splits Change Short Interest?
A 4-for-1 split can quadruple reported short interest while the position behind it never changes. Here is the arithmetic, with the FINRA reports that show it.
A stock split changes the short interest number without changing the short position behind it. After a 4-for-1 split, the next report can show roughly four times as many shares sold short, each of them worth about a quarter of the old share price. Nobody sold, nobody borrowed, and the dollar value at risk is the same as the day before.
Why short interest jumps after a stock split
Short interest is a count of shares: the ones that have been sold short and not yet bought back, tallied stock by stock. Broker-dealers in the United States report their tally to FINRA twice a month, and FINRA publishes the file for every listed name. Our guide to FINRA short interest data walks through the fields.
A split rewrites the unit that count is expressed in. In a 4-for-1 split, each old share becomes four new shares and the price per share is divided by four, so the company is the same size and every holder owns the same slice of it. When a stock split takes effect covers the record date and the first session the new shares trade.
A borrowed share is still a share. When the stock a short seller borrowed splits, the obligation splits with it. An account that borrowed and sold 1,000 shares before a 4-for-1 split owes 4,000 shares after it, and the lender who lent 1,000 is owed 4,000. No trade takes place. Four things move at once:
- The borrowed position is redenominated. One thousand shares owed becomes four thousand shares owed, at a quarter of the price each.
- The published count follows it, so raw short interest steps up on the first report tallied after the effective date.
- Average daily volume is redenominated the same day, which keeps the days to cover ratio near where it was.
- Shares outstanding and the float scale by the same ratio, so short interest as a percentage of float is unchanged.
The panel below follows one large split through the reports themselves. Nvidia's stock split 10-for-1, effective Jun 10, 2024. These are the twice-monthly short interest reports for the four months around that date, each read as it was filed.
The exact SQL behind every number
WITH
(
SELECT min(execution_date)
FROM global_markets.stocks_splits
WHERE ticker = 'NVDA'
AND execution_date BETWEEN '2024-06-01' AND '2024-06-30'
) AS effective_date,
(
SELECT concat(
toString(toUInt32(argMin(split_to, execution_date))), '-for-',
toString(toUInt32(argMin(split_from, execution_date))))
FROM global_markets.stocks_splits
WHERE ticker = 'NVDA'
AND execution_date BETWEEN '2024-06-01' AND '2024-06-30'
) AS declared_ratio
SELECT
toString(d) AS settlement_date,
round(si_shares / 1e6, 1) AS short_interest_millions,
round(adv_shares / 1e6, 1) AS avg_daily_volume_millions,
if(d < effective_date, 'before', 'after') AS split_side,
declared_ratio AS declared_split,
formatDateTime(effective_date, '%b %e, %Y') AS took_effect
FROM
(
SELECT
si.settlement_date AS d,
max(si.short_interest) AS si_shares,
max(si.avg_daily_volume) AS adv_shares
FROM global_markets.stocks_short_interest AS si
WHERE si.ticker = 'NVDA'
AND si.settlement_date BETWEEN '2024-04-10' AND '2024-08-20'
GROUP BY si.settlement_date
)
ORDER BY dThe window opens on 2024-04-15, with 29.1 million shares reported short against average daily volume of 41.6 million shares. It closes on 2024-08-15, at 293.5 million shares short and 393.6 million shares a day. The split_side column marks which side of the effective date each of the 9 reports sits on. Read the short interest line alone and the position looks like it multiplied overnight. Read it next to the volume line and both series step by about the same factor, which is what a change of unit looks like.
Does a stock split change days to cover?
Days to cover, also called the short interest ratio, divides short interest by average daily trading volume. It answers one question: at the recent pace of trading, how many sessions of volume would it take to buy back every borrowed share? Our explainer on days to cover works through the arithmetic and where it misleads.
Shares divided by shares per day is a ratio, and a split multiplies both sides of it by the same number. One million shares short over ten million shares a day is 0.1 days. After a 4-for-1 split, four million over forty million is still 0.1 days. The unit changed; the answer did not.
The next panel runs days to cover for the same stock across two years of reports, computed from the two raw share counts in each file, with the split sitting in the middle of the window.
The exact SQL behind every number
WITH
(
SELECT min(execution_date)
FROM global_markets.stocks_splits
WHERE ticker = 'NVDA'
AND execution_date BETWEEN '2024-06-01' AND '2024-06-30'
) AS effective_date
SELECT
toString(d) AS settlement_date,
round(si_shares / adv_shares, 2) AS days_to_cover,
if(d < effective_date, 'before', 'after') AS split_side
FROM
(
SELECT
si.settlement_date AS d,
max(si.short_interest) AS si_shares,
max(si.avg_daily_volume) AS adv_shares
FROM global_markets.stocks_short_interest AS si
WHERE si.ticker = 'NVDA'
AND si.settlement_date BETWEEN '2023-06-01' AND '2025-06-30'
GROUP BY si.settlement_date
HAVING max(si.avg_daily_volume) > 0
)
ORDER BY dAcross 50 reports between 2023-06-15 and 2025-06-30, the ratio reads 0.54 days at the start of the window and 1.22 days at the end. It wanders in between as the short position and trading volume each change. What it does not do is step up by the split ratio at the effective date, the way the raw share count in the first panel does. One report deserves care, the one whose period straddles the effective date: its volume average covers sessions counted in old shares and sessions counted in new ones, and that single ratio blends two units.
Short interest as a percentage of float
Float is the share count actually available to trade, meaning shares outstanding less the blocks held by insiders and other holders who rarely turn them over. Short interest as a percentage of float divides one share count by another.
Take a hypothetical company with a float of 100 million shares and 8 million shares sold short, or 8 percent of float. A 4-for-1 split turns the float into 400 million shares and the short position into 32 million shares. Thirty-two over four hundred is 8 percent again. The percentage survives the change of unit where the raw count does not, and it is the better field for comparing a stock with its own history. For how names stack up against each other, see the most shorted stocks.
What the reports show across real splits
One stock could be a coincidence. The panel below takes a fixed list of household names that split two for one or larger since 2020 and puts the declared ratio next to what the reports did. The second bar is the short interest in the first report tallied after the effective date, divided by the short interest in the last report before it.
The exact SQL behind every number
WITH big_splits AS
(
SELECT
ticker,
max(execution_date) AS effective_date,
round(argMax(split_to, execution_date) / argMax(split_from, execution_date), 1) AS split_multiplier,
concat(
toString(toUInt32(argMax(split_to, execution_date))), '-for-',
toString(toUInt32(argMax(split_from, execution_date)))) AS declared_split
FROM global_markets.stocks_splits
WHERE execution_date BETWEEN '2020-01-01' AND '2026-06-30'
AND split_to >= split_from * 2
AND ticker IN ('AAPL','AMZN','AVGO','CMG','DXCM','GOOGL','LRCX','MNST','NFLX','NVDA','PANW','SHOP','TSLA','WMT')
GROUP BY ticker
),
reports AS
(
SELECT
ticker,
settlement_date,
max(short_interest) AS short_interest
FROM global_markets.stocks_short_interest
WHERE ticker IN ('AAPL','AMZN','AVGO','CMG','DXCM','GOOGL','LRCX','MNST','NFLX','NVDA','PANW','SHOP','TSLA','WMT')
AND settlement_date BETWEEN '2019-11-01' AND '2026-08-31'
GROUP BY ticker, settlement_date
)
SELECT
b.ticker AS symbol,
b.split_multiplier AS split_multiplier,
round(
argMinIf(r.short_interest, r.settlement_date, r.settlement_date >= b.effective_date)
/ nullIf(argMaxIf(r.short_interest, r.settlement_date, r.settlement_date < b.effective_date), 0),
1) AS reported_share_multiplier,
b.declared_split AS declared_split,
formatDateTime(b.effective_date, '%b %e, %Y') AS took_effect
FROM big_splits AS b
INNER JOIN reports AS r ON r.ticker = b.ticker
WHERE r.settlement_date BETWEEN b.effective_date - 40 AND b.effective_date + 40
GROUP BY b.ticker, b.split_multiplier, b.declared_split, b.effective_date
HAVING countIf(r.settlement_date < b.effective_date) > 0
AND countIf(r.settlement_date >= b.effective_date) > 0
ORDER BY split_multiplier DESC, symbol ASC
LIMIT 14CMG carries the largest ratio in the group, a 50-for-1 split effective Jun 26, 2024, and its first post-split report counted 42 times the shares of its last pre-split one. Across all 14 events, the two bars for each name land at close to the same height. They are not identical, and there is no reason for them to be: two consecutive reports sit about two weeks apart, and positions open and close in between.
The same events, read through days to cover. The first column is the ratio in the last report before the effective date, the second is the ratio in the first report after it, and the third divides the second by the first.
The exact SQL behind every number
WITH big_splits AS
(
SELECT
ticker,
max(execution_date) AS effective_date
FROM global_markets.stocks_splits
WHERE execution_date BETWEEN '2020-01-01' AND '2026-06-30'
AND split_to >= split_from * 2
AND ticker IN ('AAPL','AMZN','AVGO','CMG','DXCM','GOOGL','LRCX','MNST','NFLX','NVDA','PANW','SHOP','TSLA','WMT')
GROUP BY ticker
),
reports AS
(
SELECT
ticker,
settlement_date,
max(short_interest) / max(avg_daily_volume) AS days_to_cover
FROM global_markets.stocks_short_interest
WHERE ticker IN ('AAPL','AMZN','AVGO','CMG','DXCM','GOOGL','LRCX','MNST','NFLX','NVDA','PANW','SHOP','TSLA','WMT')
AND settlement_date BETWEEN '2019-11-01' AND '2026-08-31'
GROUP BY ticker, settlement_date
HAVING max(avg_daily_volume) > 0
)
SELECT
b.ticker AS symbol,
round(argMaxIf(r.days_to_cover, r.settlement_date, r.settlement_date < b.effective_date), 2) AS days_to_cover_before,
round(argMinIf(r.days_to_cover, r.settlement_date, r.settlement_date >= b.effective_date), 2) AS days_to_cover_after,
round(
argMinIf(r.days_to_cover, r.settlement_date, r.settlement_date >= b.effective_date)
/ nullIf(argMaxIf(r.days_to_cover, r.settlement_date, r.settlement_date < b.effective_date), 0),
2) AS cover_change,
formatDateTime(b.effective_date, '%b %e, %Y') AS took_effect
FROM big_splits AS b
INNER JOIN reports AS r ON r.ticker = b.ticker
WHERE r.settlement_date BETWEEN b.effective_date - 40 AND b.effective_date + 40
GROUP BY b.ticker, b.effective_date
HAVING countIf(r.settlement_date < b.effective_date) > 0
AND countIf(r.settlement_date >= b.effective_date) > 0
ORDER BY days_to_cover_before DESC, symbol ASC
LIMIT 14PANW carries the highest pre-split reading in the group at 5.51 days, against 5.28 days in its first report after the effective date, or 0.96 times the earlier figure. Put that change column next to the split ratios in the panel above. The share counts moved by the ratio. The cover ratios stayed in the same range, since the numerator and the denominator were rewritten in the same unit on the same day.
The timing trap: settlement dates and the publication lag
FINRA short interest is a snapshot, not a running total. Each report is tallied as of a settlement date and reaches the public roughly eight business days later, so the figure on your screen describes a position from about two weeks ago. Why short interest is two weeks old lays out that calendar.
Two things follow around a split. The first report published after the effective date can still be a pre-split snapshot, tallied while the old shares were the unit and released after the new ones started trading. And the report whose period straddles the effective date carries a volume average drawn from sessions in both units, which makes it the one row in the file to read slowly.
One question is worth answering from the publisher rather than from intuition: whether the archive you are reading restates earlier reports in post-split shares. That convention belongs to the file, not to the split itself. The panels here read each report as it was stored, with no adjustment applied on our side. Confirm the convention in the publisher's own documentation before lining a pre-split report up against a post-split one. Short interest is also a different measurement from daily short volume, and short interest versus short volume separates the two.
FAQ
Does a stock split change short interest?
A split changes the reported share count while the position behind it stays the same size. The count of shares sold short is multiplied by the split ratio, and the price per share and the float are restated by the same ratio.
Do short sellers have to buy back shares before a split?
No. A split does not close or call any position. The borrowed share count is restated in the new shares, and the borrower's obligation and the lender's claim are restated by the same ratio.
Does a stock split change days to cover?
Not by itself. Days to cover divides short interest by average daily volume, and a split multiplies both share counts by the same ratio. Readings still drift from report to report as volume and positions change.
What happens to short interest in a reverse split?
The arithmetic runs the other way. In a 1-for-10 reverse split, 10 million shares sold short become 1 million shares sold short, each worth about ten times as much, and percent of float is unchanged.
How soon after a split does the change show up in the data?
On the first report tallied on or after the effective date, which reaches the public about eight business days after its settlement date. A report published after a split can still describe a settlement date before it.
Every panel here ships with the SQL that produced it, expand one to see how each count was taken. To pull the reports around any name's own split, ask for it in plain English on the Strasmore terminal.