Belief tested · Analysts, insiders and short sellers

Do crowded trades end badly?

The belief“Crowded trades end badly.”
VerdictNo edge
50 in 100stocks beat the median stock over the next month after the signal. A coin flip gives 50.

Not in the data. Since 1995, US stocks in the top tenth of their own year's crowding, read from trading, beat the same day's median stock over the next month 50 times in 100, and over three months 50; a coin flip gives 50. By crowding score, the most crowded fifth beat the median no less often than the least crowded. Neither end showed a measurable edge.

At a glance
4648505254
Of 100
Crowding peak, next session
50.1
Crowding peak, next week
50.2
Crowding peak, next month
49.9
Crowding peak, next three months
50.0
range a coin flip produces50 = chanceno measurable edgeproven relation
Samplen = 1,763of the 1,767-stock universe
Signals counted232,890on Wednesdays since 1995
Years below 5016 / 32next month
Rank correlation0.007t 0.7, next month
Period1995–20261,634 Wednesdays

The paper

The full study: its data, method, robustness, limits and sources.

Opulence Alpha Research · Published Sep 25, 2026 · Data through Aug 21, 2026

Keywords: crowded trades, crowded trade, crowded stocks, crowding risk stocks, herding stocks, is a crowded trade dangerous·JEL classification: G11, G12, G14, C12, C58

Study design
Universe
The same for every study: 1,767 US common stocks in the 11 GICS sectors, 422 of them since delisted; S&P 500 members since 1996 plus large and mid-sized companies outside the index
Period
Every Wednesday from Jan 4, 1995 to Aug 19, 2026: 1,634 Wednesdays across 7,959 trading sessions
Sample
n = 1,763 stocks of the 1,767-stock universe; 1,889,758 stock-weeks, a median of 1,148 stocks per Wednesday
Outcome
Return over the next 1, 5, 21 and 63 trading sessions against the same day's median stock
Inference
t on non-overlapping dates; proven only when |t| ≥ 3 and the same sign in at least 60% of years

1Introduction

A crowded trade is one that many investors are in at the same time. The warning repeated on trading desks and forums is that crowded stocks end badly: once everyone who wanted in is in, no one is left to push the price up, and the rush for the exit pushes it down. Stein (2009) argues that when many investors crowd into the same trade, prices can overshoot. Yet stocks with unusually heavy trading have been found to earn higher, not lower, returns in the weeks that follow (Gervais, Kaniel and Mingelgrin 2001). This study measures crowding from trading itself and asks whether the most crowded stocks went on to trail other stocks.

2Data and method

2.1Sample design

Population and frame. The population is US common stocks listed on the NYSE and Nasdaq; funds, ETFs, trusts, preferred shares, warrants and units are excluded. The sampling frame is a fixed universe of 1,767 companies, drawn once when the platform was built and not re-sampled since, in two strata. Stratum 1 is a census of the S&P 500: every company in the index at any time since 1996 whose price history could be recovered, 1,027 companies of which 413 have since delisted; it holds 76% of the index's members in 1996 and at least 96% in every year from 2010. Stratum 2 is 740 large and mid-sized companies outside the index, selected in proportion to the market's sector weights from the stocks that passed a minimum share price of $15 and a minimum average daily trading value of $25 million; 9 of them have since delisted.

Sample. The unit of observation is a stock-week: one stock on one Wednesday. A stock enters a Wednesday's cross-section when it has a valid close that day, a value of the signal and a measured outcome; bars flagged as bad data and returns that cross a change of issuer are left out. This study's sample is n = 1,763 stocks of the 1,767: 1,889,758 stock-weeks on 1,634 Wednesdays from Jan 4, 1995 to Aug 19, 2026, a median of 1,148 stocks per Wednesday (range 820 to 1,371). 331,954 stock-weeks (17.6%) come from the 422 companies that have since delisted. Every stock and every Wednesday carries equal weight.

Representativeness. Table 1 gives the sample by GICS sector beside the S&P Composite 1500: 7.9% of companies would have to change sector for the two to match exactly. By latest market value, 55% of the active companies are large (at least $10bn), 38% mid ($2–10bn) and 6% small. Because Stratum 2 was chosen from companies listed at construction, its history carries survivorship bias. Section 4 repeats the lead result among the stocks that were S&P 500 members on each date, the part of the sample largely free of that bias.

Table 1. The sample by sector

Companies in this study's sample, of them those since delisted, the size of the active companies, the sample's share of stock-weeks and the sector's share of the S&P Composite 1500.

SectorSampleDelistedLargeMidSmallStock-weeksS&P 1500
Information Technology28267122761813.9%12.7%
Financials2566011474814.6%17.2%
Industrials25553126671115.9%17.5%
Health Care2374792811612.4%10.9%
Consumer Discretionary2093969871212.3%12.9%
Energy10530393155.9%4.7%
Consumer Staples10138401755.9%4.9%
Materials9434352425.6%5.1%
Real Estate8410403225.7%6.9%
Communication Services8031281353.9%3.3%
Utilities6013351204%4%
All sectors1,76342274051484100%100%

Sample: n = 1,763 of the 1,767 companies. Size by latest market value for the 1,338 active companies with one: large ≥ $10bn, mid $2–10bn, small < $2bn. S&P 1500 shares count the constituents of the S&P 500, MidCap 400 and SmallCap 600 (1,506 companies, lists read Sep 25, 2026). Sectors are each company's current GICS sector.

2.2Signals

Table 2. Signals studied
SignalDefinitionSinceWednesdays
Crowding peakThe crowding score in the top tenth of the stock's own previous 252 sessionsJan 4, 19951,633
Crowding scoreA blend of unusually heavy volume, faster turnover and co-movement with the market, from 0 to 1Jan 4, 19951,634

2.3Measurement

Outcome. The return from the close on the Wednesday to the close 1, 5, 21 and 63 trading sessions later, on closes adjusted for splits and dividends, compared with the same day's median stock; half of all stocks beat the median by construction, so chance is 50 in 100 on every date. No delisting return is added.

Each Wednesday, the stocks showing the reading are scored against the same day's median stock. A state counts on every Wednesday it holds; a cross or breakout counts only on the Wednesday it happens.

Each Wednesday the stocks are ranked on the signal. The study reports how often each fifth of that ranking beat the median stock, and the rank correlation (IC) between the signal and the return that followed.

The t-statistic uses non-overlapping dates only. A result is called proven when |t| is at least 3, it held in at least 60% of years and it is large enough to matter; with 1,470 tests across the studies, a looser bar would pass dozens by luck.

Full data and methods

3Results

Figure 1 reads every horizon for stocks in the top tenth of their own year's crowding. They beat the median stock 50 times in 100 over the next session, 50 over the next week, 50 over the next month and 50 over the next three months. All four sit within a point of a coin flip, and none is a measurable edge.

Figure 1. How often it beat the median stock

Share of stocks showing the reading that beat the same day's median stock, by horizon. The line at 50 is chance, the grey band the range chance alone produces; a filled square is a proven relation.

4648505254
Of 100
Crowding peak, next session
50.1
Crowding peak, next week
50.2
Crowding peak, next month
49.9
Crowding peak, next three months
50.0
range a coin flip produces50 = chanceno measurable edgeproven relation
Figure 2. Out of every 100 stocks
Crowding peak: 50 of 100 beat the median stock over the next month
Figure 3. Each fifth, against the median stock

Every Wednesday the stocks are cut into fifths on the signal, lowest to highest; bars show how often each fifth beat the same day's median stock over the next month. The dashed line is chance.

This signalNo edge
5056445049lowest50low50middle50high50highest

Next month.

What an edge looks like: short-term reversalProven
5056445051lowest50low50middle49high48highest

Next week, three-session return against the industry; the same way in 32 of 32 years.

Figure 4. How strong, and for how long

Rank correlation between the signal and the return that followed, with its 95% interval. An interval that crosses zero is no relation.

-0.030+0.03
session
+0.005
week
+0.008
month
+0.007
three months
+0.008
Figure 5. Year by year

Each square is one calendar year; filled = a year in which the average stock showing the signal lagged the median stock over the next month. Years are counted, not shown in order.

16 of 32 years below 50

A signal with an edge would fill most squares, or leave most empty.

Figure 6. Where it holds

Rank correlation inside each GICS sector over the next month, stocks ranked only against their own sector. Colour only where |t| ≥ 2.

Real Estate-0.002t 0.1
Health Care-0.001t -0.4
Financials+0.000t -0.2
Communication Services+0.001t -0.6
Energy+0.002t -0.0
Consumer Staples+0.004t -0.3
Information Technology+0.004t 0.4
Consumer Discretionary+0.005t -0.4
Utilities+0.007t 1.0
Industrials+0.009t 0.7
Materials+0.017t 1.4

Out of every 100 stocks at a crowding peak, 50 beat the median stock over the following month (Figure 2). Ranked into fifths by the crowding score, the most crowded fifth beat the median stock 50 times in 100 over the next month and the least crowded 49 (Figure 3). Average returns against the average stock tilted slightly the belief's way, -0.12 points for the most crowded fifth and 0.17 for the least, but the rank correlation between crowding and the return that followed stays close to zero at every horizon (Figure 4); over the next month it was 0.007.

4Robustness

Table 3 reads the crowding peak horizon by horizon: over the next month it lagged the median stock in 16 of 32 years and beat it in the rest. In Table 4 the crowding score's rank correlation keeps a small positive sign in every decade and in 9 of 11 sectors, yet stays too close to zero to count as an edge, also among S&P 500 members on the date (IC 0.011, t 1.0).

Table 3. The reading, horizon by horizon

Share that beat the median stock, its t-statistic on non-overlapping dates, and the years in which it pointed the same way.

HorizonBeat the mediantYears, same way
Crowding peak, session50.1-0.617 / 32
Crowding peak, week50.20.620 / 32
Crowding peak, month49.9-0.316 / 32
Crowding peak, three months50.00.214 / 32
Table 4. Robustness of the lead relations

Rank IC and its t-statistic on non-overlapping dates.

SampleICtWednesdays
Crowding score, next month
All stocks, whole period0.00690.71,630
By decade: 1995–20040.00570.3516
By decade: 2005–20140.0056-0.1518
By decade: 2015–0.00910.9596
S&P 500 members on the date0.01111.01,578
Sectors with the overall sign9 / 11

5Limitations

  • Crowding here is read from trading, not from holdings. The study does not see which funds own a stock, so crowding in hedge-fund portfolios, as reported in quarterly filings, is a different measure it does not test.
  • The study counts how often crowded stocks beat the median stock over fixed horizons; it does not measure the size of rare, sharp falls, which is where the warning about crowded exits usually points.
  • Before costs. Averages exclude trading costs, taxes and market impact.
  • Same-close timing. Returns start at the close the signal is computed from; a real trade would start later.
  • Survivors among smaller companies. The non-index names were chosen from companies listed when the universe was built; no delisting returns are added.

6Conclusion

Crowding describes how heavily a stock is being traded and how closely it moves with the crowd; it says who else is in a position. Measured against the same day's median stock, stocks at a crowding peak beat it about as often as a coin flip would, and the least crowded fifth beat it no more often than the most crowded. Being crowded describes today's trading, not tomorrow's price.

References

  1. Stein, J. C. (2009). Presidential Address: Sophisticated Investors and Market Efficiency. Journal of Finance 64(4).
  2. Gervais, S., Kaniel, R. and Mingelgrin, D. H. (2001). The High-Volume Return Premium. Journal of Finance 56(3).
  3. Benjamini, Y. and Hochberg, Y. (1995). Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society, Series B 57(1).
  4. Grinold, R. C. and Kahn, R. N. (2000). Active Portfolio Management, 2nd ed. McGraw-Hill.
  5. Harvey, C. R., Liu, Y. and Zhu, H. (2016). … and the Cross-Section of Expected Returns. Review of Financial Studies 29(1).
  6. Shumway, T. (1997). The Delisting Bias in CRSP Data. Journal of Finance 52(1).

Appendix A. Questions readers ask

Are crowded trades dangerous?
Not for the typical stock's next month, in this data. Since 1995, US stocks in the top tenth of their own year's crowding beat the same day's median stock over the next month 50 times in 100, and over three months 50 times, against 50 for a coin flip. The study counts how often stocks beat the median, not the size of rare, sharp falls.
What is a crowded trade?
A position many investors hold or trade at once, so that their exit could move the price. Here crowding is read from trading itself: unusually heavy volume, faster turnover and how closely the stock moves with the rest of the market, blended into one score.
Do the least crowded stocks do better?
Not measurably. Ranked by the crowding score, the least crowded fifth beat the median stock over the next month 49 times in 100 and the most crowded fifth 50. The least crowded fifth's average return was slightly higher, but the rank correlation between crowding and the next month's return is close to zero, so neither end stood apart from chance.
How was crowding measured?
Every Wednesday from Jan 4, 1995 to Aug 19, 2026, each stock's crowding score was compared with its own previous year; the stocks in their top tenth were scored against the same day's median stock over the next 1, 5, 21 and 63 trading sessions, on the same universe of 1,767 US stocks every study uses.

Data availability and citation

Data availability

Every figure in this paper is quoted from one frozen snapshot, published as JSON with the sample description, the robustness results and the test counts. The same snapshot feeds the research console, so the two cannot disagree.

How to cite

Opulence Alpha Research (2026). Do crowded trades end badly? Opulence Alpha Studies, Sep 25, 2026. https://opulencealpha.ai/studies/crowded-trades

Research, not advice. A measured tendency across hundreds of stocks is a nudge for any one of them, never a forecast of its price.