How heavily shorted are biotech stocks going into an FDA decision?

It depends almost entirely on company size. Across 1,753 FDA decisions (2020-2026), small-caps carried a median 6.7 days to cover and mid-caps 6.8, versus just 2.2 for large-caps. Nano-caps were also low at 2.5 days, largely because they are hard and expensive to borrow.

Do heavily shorted biotechs move more on the FDA decision?

Not because they are shorted. The most-shorted quintile did move about 3x more than the least (3.05% vs 1.05% median absolute move), but that quintile is 63% small- and mid-cap while the low quintile is 82% large-cap. Once you hold company size constant, the correlation between short interest and the size of the move collapses to roughly zero (-0.00 for micro, -0.03 for small, -0.02 for large). It is a size effect, not a short-interest effect.

Is this investment advice?

No. These are historical statistics only. We do not report the direction of any move, we make no forecast for any company or decision, and we give no entry, exit or position-sizing guidance.

pdufa.bio

Calendar Conferences AdComm Decisions Research

Home › Research › Short interest into FDA decisions

Short interest into FDA decisions

How heavily shorted are biotechs when the FDA decides — and does it tell you anything? Built from FINRA short-interest filings matched to 1,753 FDA decisions (2020–2026), using only the reading that was public before each decision. Facts, not advice.

The obvious conclusion is wrong, and we can show you exactly why.

Sort FDA decisions by short interest and the most-shorted names move 3× more on the decision than the least-shorted. It looks like a signal. It isn’t. The most-shorted bucket is 63% small- and mid-cap; the quiet bucket is 82% large-cap. You haven’t sorted by short interest — you’ve sorted by company size. Hold size constant and the relationship vanishes.

Short interest into FDA decisions by market-cap tier, and why the apparent link to move size is a market-cap artifact — Left: how shorted each cap tier is going in. Right: the trap, and the debunk.

1. How shorted is biotech going into an FDA decision?

This is the part nobody publishes. Days to cover = short interest ÷ average daily volume: how many days of normal trading it would take shorts to buy back.

Cap tier	n	Median days to cover	25th pct	75th pct
Nano	115	2.5d	1.0d	5.3d
Micro	311	4.3d	2.0d	7.9d
Small	283	6.7d	2.5d	10.4d
Mid	217	6.8d	4.8d	8.4d
Large	827	2.2d	1.6d	3.0d

Small- and mid-caps carry by far the heaviest short interest into an FDA decision (≈6.7 days) — roughly 3× a large-cap. Nano-caps are counter-intuitively low at 2.5 days: they are hard and expensive to borrow, so shorts often simply can’t build a position.

2. The trap

Now bucket every decision into five groups by short interest, and look at how big the decision-day move was:

Quintile	n	Median days to cover	Median \|move\|	% large-cap	% small+mid
Q1	356	1.0d	1.73%	54%	15%
Q2	346	2.0d	1.05%	82%	9%
Q3	350	2.9d	1.18%	66%	15%
Q4	350	5.1d	2.23%	30%	40%
Q5	351	10.6d	3.05%	3%	63%

Q5 moves 3.05% versus 1.05% for Q2. Across all 1,753 events the rank correlation is +0.160. A tempting headline — “heavily shorted biotechs move 3× more into FDA decisions” — is sitting right there.

3. Why it’s wrong

Look at the last two columns of that table. The quintiles aren’t sorting by short interest — they’re sorting by size. Small- and mid-caps are both more shorted and naturally more volatile. So we ran the only test that matters: hold company size constant.

Within cap tier	Correlation: short interest vs \|move\|
All events (mixed sizes)	+0.160
Micro	-0.000
Small	-0.031
Mid	-0.105
Large	-0.021

Inside a single size tier, short interest tells you nothing about how big the move will be — the correlation is zero, and if anything slightly negative. The apparent signal was a Simpson’s paradox: a real pattern in the pooled data that reverses or disappears once you condition on the confounder.

Why we published a negative result.

We could have shipped panel 2 on its own. It is true, it is dramatic, and it would have been shared. It is also useless — and anyone who traded on it would have been trading market-cap while believing they were trading short interest. We would rather publish the debunk than the headline.

Method

Data. FINRA bi-monthly short-interest filings, matched to each FDA decision by ticker. Point-in-time. For every event we use only the short-interest reading already published before that decision (median lag 9 days, 95th percentile 15) — no lookahead. Measure. Days to cover = short interest ÷ average daily volume. Move. The absolute decision-day reaction; we deliberately do not report direction. Reporting which way heavily-shorted stocks went would be an approval-odds claim in disguise, and we don’t do those. Statistics. Median, interquartile range, and Spearman rank correlation. Rows with stale short interest (>45 days old) or implausible days-to-cover (>60) were excluded.

Caveats.

1. Short interest is stale by construction. FINRA publishes twice a month with a settlement lag. The number we use was true up to ~9 days before the decision — positioning can change materially in that window.

2. Days to cover embeds volume. It is a ratio, so a volume spike shrinks it without a single share being covered.

3. Absence of evidence. We found no within-tier relationship. That is not proof none exists — only that, at this sample size and with this measure, there isn’t one worth acting on.

Cite this dataset. Free to reference with attribution (CC BY 4.0).
“pdufa.bio, Short interest into FDA decisions (2020–2026), n=1,753, https://www.pdufa.bio/research/short-interest-fda”
Journalists & researchers: data@pdufa.bio.

Not affiliated with or endorsed by the FDA or FINRA. pdufa.bio is an independent service. Informational and educational only — not investment advice. These are historical statistics; they are not a forecast for any company or decision, they are not a trading strategy, and they carry no entry, exit or position-sizing guidance. Nothing here is a short-squeeze thesis. Past behaviour does not predict future outcomes. © 2026 pdufa.bio