Crowded Spaces and Anomalies: Chincarini, Lazo-Paz & Moneta (2026)

What this is. The paper’s core results, the crowding measures it constructs, and the empirical specifications behind the main findings: enough to know what it found and how, without reading all 17 pages. To replicate or extend it, read the full source at the original.

TL;DR

This paper investigates whether crowded equity positions, where many institutional investors hold the same stocks and may strain the liquidity available for exits, are associated with higher future returns and greater crash risk. Using Thomson/Refinitiv 13F institutional holdings from 1980 to 2021, the authors construct a Days-ADV crowding measure (the days of average daily trading volume needed for all institutions to exit a position). They find that more crowded anomaly stocks deliver significantly higher risk-adjusted returns across all 11 anomalies studied by Stambaugh, Yu, and Yuan (2012), that non-crowded anomaly portfolios have no positive FF3 alpha (and have negative alphas in some liquidity-augmented models), that the result persists after the anomaly publication dates identified by Mclean and Pontiff (2016), and that crowding is associated with greater institutional exposure to stock price crash risk. The paper extends the hedge-fund crowding-return result of Brown, Howard, and Lundblad (2021) to all 13F institutions and contradicts the mutual-fund finding of Zhong, Ding, and Tay (2017), and frames crowding as an additional channel within limits-to-arbitrage theory.

Core results

Magnitudes and significance are as reported; \*\*/\*\*\* = 5%/1%. Locators point into the source PDF.

#	Result	Locator	Magnitude
R1	Most-crowded (Q5) stocks earn higher FF3 alphas than least-crowded (Q1); the VW spread is 1.44%/month	Table 3 Panels A-B, p. 8; Table 4, p. 9	VW Q5: FF3 alpha = 0.54%/month (t=8.87); Q1: -0.90%/month (t=-7.86); Q5-Q1 spread: 1.44%/month (t=9.67); EW spread: 1.57%/month (t=12.23)
R2	Bivariate double-sort aggregate anomaly portfolio (long crowded long-leg, short least-crowded short-leg) earns large EW alpha	Table 6 Panel B, p. 10-11	EW FF3 alpha = 1.69%/month (t=11.09) full sample; 1.96%/month in-sample; 1.61%/month post-publication (t=7.67)
R3	Non-crowded anomaly portfolios have near-zero FF3 alphas and negative alphas in liquidity-augmented models	Table 7, p. 11	EW non-crowded portfolio FF3 alpha = 0.009%/month (t=0.18); VW = 0.008%/month (t=0.09); FF5P/FF5A/FF5AM alphas are negative, several significantly so
R4	Fama-MacBeth regressions show a positive LADV-return association, stronger in anomaly legs; post-publication interaction evidence is mixed	Table 9, p. 12	LADV = 0.546 (t=4.31), col. 1; Long x LADV = 0.287 (t=3.12), Short x LADV = 0.485 (t=4.86), col. 4; col. 5 baseline interactions = 0.206 (t=3.01) and 0.308 (t=3.83); incremental post-publication interactions = 0.810 (t=1.62) and 0.178 (t=2.92), respectively
R5	Crowding is positively associated with future stock price crash risk, measured by NCSKEW and DUVOL	Table 10, p. 14	Full-sample NCSKEW LADV = 0.011 (t=3.29) and DUVOL = 0.018 (t=5.13); post-1992 coefficients are significant, but pre-1993 estimates are not
R6	Crowded anomaly portfolios declined significantly more than uncrowded portfolios during the 2007-2009 and COVID-19 crises	Figure 3, p. 13	CAR differences statistically significant for financial crisis (t=1.97) and COVID-19 crisis (t=2.03)
R7	Liquidity-linked crowding measures predict spreads, while investor count and ownership share do not do so consistently	Table 3 Panels A-B, p. 8; §2.3.2, p. 5	ActRatio Q5-Q1 FF3 alpha: 1.26%/month (t=8.44) VW and 1.38%/month (t=11.92) EW; NI spreads: -0.01 (t=-0.10) VW and 0.09 (t=0.80) EW; PSO spreads: 0.00 (t=0.05) VW and 0.23 (t=2.08) EW
R8	Crowding-return spreads are stronger among transient, short-horizon, and investment-advisor holdings	Table 5, p. 9	FF3 spread alpha: short-horizon 0.946%/month (t=5.48), long-horizon 0.236% (t=1.79), transient 1.284% (t=5.13), quasi-indexer 0.431% (t=2.86), investment advisors 0.759% (t=4.20)
R9	The crowding premium extends across the 97-anomaly sample in four anomaly groups and persists post-publication	Table 8, p. 12	Full-sample double-sort FF3 alpha: event 1.250%/month (t=10.57), market 1.550% (t=10.66), valuation 1.306% (t=10.18), fundamental 1.408% (t=10.79); post-publication: 1.083%, 1.530%, 1.190%, 0.982%, respectively
R10	The crowding-crash-risk association appears stronger on short anomaly legs, with weaker evidence for some estimates	Table 10, p. 14	Short x LADV: NCSKEW 0.008 (t=2.16), DUVOL 0.005 (t=1.89); incremental post-publication x Short x LADV: NCSKEW 0.004 (t=1.74), DUVOL 0.006 (t=2.30)
R11	Low Days-ADV on anomaly short legs overlaps with high Days-to-Cover, which explains part of the short-leg return spread	§3.4.2, p. 15; Table A18 Panel A, cited p. 15	Mean quarterly rank correlation = -0.53; EW FF3 strategy alpha = 1.1%/month using high-DTC shorts versus 1.7%/month using low-Days-ADV shorts
R12	Crowded-portfolio returns do not show the reversal predicted by a temporary price-impact explanation	§3.4.3, p. 15 (authors cite Internet Appendix Table A8)	The performance persists over the next six quarters after portfolio sorting, with no short-run reversal
R13	Days-ADV has a declining aggregate trend early in the sample but rises among the most crowded stocks after the late 2000s	Fig. 2 and accompanying text, p. 7	Median Days-ADV among most crowded stocks increases from approximately 260 days in 2012 to over 300 days by 2020; accompanying text identifies a common break at 1992:Q4, while the Fig. 2 caption says 1995

Overall (paper’s conclusion). Crowding is positively associated with future abnormal returns across all 11 stock market anomalies studied, and the anomaly alpha is generated almost entirely by the most crowded stocks. This result is robust to different factor model specifications (FF3, FF5, FF5 augmented with liquidity and momentum), persists after publication dates, and is stronger for transient and short-horizon institutions. Crowding is also associated with greater institutional exposure to crash risk, consistent with the idea that crowded positions impose additional risk for which investors require compensation and that crowding adds a new consideration to limits-to-arbitrage arguments.

Theory / model

The paper does not specify a formal economic model. It tests whether institutional crowding predicts future returns and crash exposure. The proposed channel is that common positions can make investors’ exit decisions correlated: limited trading volume can turn a shared shock into price pressure, and limits to short selling can make correction of anomaly mispricing costly. The authors predict higher returns on crowded anomaly positions as compensation for this risk, and higher subsequent crash-risk measures. This descriptive design does not identify a causal effect.

Portfolio similarity (equations 1-2, p. 5). For two institutional portfolios, the cosine-similarity measure is

$$s_{ij}=\frac{\mathbf{w}_i'\mathbf{w}_j}{|\mathbf{w}_i||\mathbf{w}_j|} \tag{1}$$

For a group of M portfolios, the average pairwise similarity excludes each portfolio’s self-similarity:

$$C=\frac{\sum_{i=1}^{M}\sum_{j=1}^{M}s_{i,j}-M}{M^2-M} \tag{2}$$

Stock-level crowding (equations 3-4, p. 5). Days-ADV scales the total institutional dollar holdings in a security by its average daily dollar trading volume in the prior quarter:

$$\text{Days-ADV}_{i,t}=\frac{\sum_{j=1}^{N}\text{InstHold}_{i,j,t-1}}{\text{ADV}_{i,t-1}} \tag{3}$$

The alternative Activity Ratio uses shares held by active investors and average share turnover:

$$\text{ActRatio}_{i,t}=\frac{\sum_{j=1}^{N}\text{Shares}_{i,t-2}}{\text{AvgTurn}_{i,t-1}} \tag{4}$$

The paper reports that the two measures correlate at 0.99 (Internet Appendix Table A1, reported in §2.3.2, p. 6). Number of institutional investors and percentage of shares outstanding are other proxies, but they do not account for security liquidity.

Firm-specific returns and crash measures (equations 5-7, pp. 7-8). Weekly stock returns are regressed on lead, contemporaneous, and lagged market and industry returns to address nonsynchronous trading. The resulting residual is transformed as $R_{i,t}=\log(1+\epsilon_{i,t})$. The regression is:

$$r_{i,t}=\alpha_i+\beta_{1,i}r_{m,t-1}+\beta_{2,i}r_{k,t-1}+\beta_{3,i}r_{m,t}+\beta_{4,i}r_{k,t}+\beta_{5,i}r_{m,t+1}+\beta_{6,i}r_{k,t+1}+\epsilon_{i,t} \tag{5}$$

Negative conditional skewness is

$$\text{NCSKEW}_{i,t}=-\frac{n(n-1)^{3/2}\sum R_{i,t}^{3}}{(n-1)(n-2)(\sum R_{i,t}^{2})^{3/2}} \tag{6}$$

and down-to-up volatility is

$$\text{DUVOL}_{i,t}=\log\left(\frac{(n_u-1)\sum_{\text{DOWN}}R_{i,t}^{2}}{(n_d-1)\sum_{\text{UP}}R_{i,t}^{2}}\right) \tag{7}$$

Here n is the number of firm-specific weekly returns, and n_u and n_d count returns above and below the period mean. Higher values of either measure indicate more crash risk.

Days-to-Cover (equation 8, p. 15). To assess short-selling costs, the paper follows Hong et al. (2016) and defines DTC as the short ratio divided by average daily turnover:

$$\text{DTC}=\frac{\text{SR}}{\text{Average Daily Turnover}} \tag{8}$$

The short ratio is monthly short interest divided by shares outstanding. This analysis uses 2003-2021 because Nasdaq short-interest data are available from 2003.

Method

The empirical method combines quarterly portfolio sorts, factor-adjusted return tests, Fama-MacBeth cross-sectional regressions, and firm-year crash-risk regressions. Days-ADV is lagged one quarter because 13F holdings are disclosed with up to a 45-day delay. Stock and anomaly portfolios are formed from US common shares on NYSE, AMEX, and Nasdaq with prices above $5, excluding utilities and financial firms. The core window is 1980:Q1 to 2021:Q4; momentum portfolios are rebalanced quarterly. The paper uses a 1992:Q4 break to compare subsamples. No causal identification design is claimed.

For the main return tests, stocks are sorted into crowding quintiles quarterly. Equal- or value-weighted portfolio excess returns are regressed on factor returns, and the intercept is the alpha. The main specification is the Fama-French three-factor regression; the paper also uses FF5, FF5 plus Pastor-Stambaugh liquidity, FF5 plus Amihud IML, and FF5 plus IML and momentum. Newey-West standard errors are used for portfolio return alphas. Double sorts first rank stocks on anomaly characteristics and then select the top 30% of Days-ADV within long legs and the bottom 30% within short legs. For non-crowded portfolios, the middle 40% is selected. The 97-anomaly extension groups the signals into event, market, valuation, and fundamental portfolios.

The FF3 risk-adjustment benchmark follows Fama and French (1993). The Fama-MacBeth design regresses next-quarter cumulative monthly returns on LADV and controls in each quarter, then averages the cross-sectional estimates over time. Standard errors use Newey-West with four lags. The crash-risk panel uses one-year-ahead NCSKEW or DUVOL with firm and year fixed effects and firm-clustered standard errors. Controls include firm-specific returns and their kurtosis and standard deviation, market-to-book, liabilities to assets, ROA, size, turnover, analyst coverage, and the lagged dependent variable.

Empirical specifications

Factor-adjusted portfolio returns (Section 3.1, Table 3, p. 8). For portfolio p and month t, the primary alpha specification is

$$R^e_{p,t}=\alpha_p+\beta_{1,p}\text{MktRf}_t+\beta_{2,p}\text{SMB}_t+\beta_{3,p}\text{HML}_t+\varepsilon_{p,t}$$

The dependent variable is the monthly excess return of the quarterly rebalanced portfolio. The specification has no fixed effects; Newey-West standard errors are reported. The sample is 1980:Q1-2021:Q4. The expanded factor specifications add the FF5 factors and, in turn, liquidity and momentum factors. Table 5 and Table 6 use the same factor-alpha approach for investor-type and anomaly portfolios, respectively.

Fama-MacBeth return regression (Table 9, p. 12). Each quarter, the baseline cross-sectional regression is next-quarter cumulative monthly return on LADV and controls:

$$r_{i,t+1}=\alpha_t+\beta_{1,t}\text{LADV}_{i,t}+\gamma_t'X_{i,t}+\varepsilon_{i,t+1}$$

The interaction specification adds long- and short-anomaly membership and their LADV interactions:

$$r_{i,t+1}=\alpha_t+\beta_{1,t}\text{LADV}_{i,t}+\beta_{2,t}\text{Long}_{i,t}+\beta_{3,t}(\text{Long}\times\text{LADV})_{i,t}+\beta_{4,t}\text{Short}_{i,t}+\beta_{5,t}(\text{Short}\times\text{LADV})_{i,t}+\gamma_t'X_{i,t}+\varepsilon_{i,t+1}$$

Here X includes log institutional ownership, log size, age, prior return volatility, book-to-market, dividend yield, turnover, and cumulative returns over the prior three and nine months. Column 5 adds post-publication status and its long- and short-leg interactions with LADV. The sample sizes reported in Table 9 are 294,301 observations for the full sample and 79,352 and 213,299 for the two 1992-break subsamples. Regressions are quarterly cross-sections, with Newey-West standard errors using four lags on the time-series averages of coefficients; no fixed effects are specified.

Crash-risk panel regression (Table 10, p. 14). The principal panel specification is

$$\text{CrashRisk}_{i,t+1}=\alpha+\beta_1\text{LADV}_{i,t}+\gamma'\text{Controls}_{i,t}+\mu_i+\lambda_t+\varepsilon_{i,t+1}$$

where the outcome is next-year NCSKEW or DUVOL. Extended specifications include Long and Short dummies, their LADV interactions, post-publication status, and the corresponding triple interactions. The full sample has 102,940 observations; firm and year fixed effects are included, standard errors are clustered by firm, and controls are measured over the prior fiscal year. The paper also reports quarterly-frequency versions in Internet Appendix Tables A6-A7.

Crisis event study (Figure 3, p. 13). For daily portfolio returns risk-adjusted by FF3, portfolio cumulative abnormal returns are constructed over windows beginning 12 months before and ending 12 months after each crisis onset:

$$\text{CAR}_{p,[a,b]}=\sum_{t=a}^{b}(R_{p,t}-\widehat{R}^{\text{FF3}}_{p,t})$$

The analysis covers November 2006-February 2010 and March 2019-February 2021, with value-weighted portfolios under both buy-and-hold and quarterly rebalancing. The figure reports t-statistics for crowded versus uncrowded anomaly portfolios; it is an event-window comparison rather than a causal event-study design.

Datasets used

Dataset	Role in paper	Wiki page
Thomson/Refinitiv (TR) 13F institutional holdings	Primary crowding measure (Days-ADV, cosine similarity); institution-type classification into transient, dedicated, quasi-indexer, hedge fund, mutual fund	WRDS (licensed)
CRSP monthly stock data	Stock returns, prices, trading volume, shares outstanding; anomaly variable construction	WRDS (licensed)
Compustat annual fundamentals	Accounting-based anomaly variables (accruals, NOA, asset growth, profitability, etc.) and short interest data (2003-2021)	WRDS (licensed)
I/B/E/S analyst data	Number of analysts following each stock (control variable in FM regressions)	WRDS (licensed)
Kenneth French Data Library	FF3, FF5, and momentum factor returns for risk adjustment	Ken French library
Brian Bushee institution classification	Transient, dedicated, quasi-indexer institution type labels	no page yet

Sample: US common stocks on NYSE, AMEX, and Nasdaq with price above $5, excluding utilities and financial firms. Main sample: 1980:Q1 to 2021:Q4 (quarterly rebalancing). Exception: momentum anomaly portfolios are rebalanced quarterly (not annually). DTC analysis restricted to 2003-2021, when Nasdaq short interest data becomes available.

When to read the full paper

Use the original if you are: investigating the relationship between institutional crowding and anomaly returns (Table 6 provides anomaly-by-anomaly bivariate-sort alphas for all 11 anomalies across five factor models); studying crash risk as a channel linking institutional crowding to limits-to- arbitrage; extending the results to the 97 anomalies of Mclean and Pontiff (2016) (Table 8 and the Internet Appendix); or assessing whether the DTC measure of Hong et al. (2016) accounts for the short-leg alpha in crowded spaces.

Attribution and rights

Source: peer-reviewed, Journal of Banking and Finance 182 (2026) 107579. This distillation was extracted by an LLM on 2026-06-25 and is not human-verified or independently reproduced. The CC BY-NC-ND 4.0 licence permits verbatim sharing but not derivative works; the verbatim PDF is not hosted in this batch.

Attribution (CC BY-NC-ND 4.0). Chincarini, Ludwig B., Renato Lazo-Paz, and Fabio Moneta. “Crowded spaces and anomalies.” Journal of Banking and Finance 182 (2026) 107579. DOI: 10.1016/j.jbankfin.2025.107579. © 2025 The Authors. Published by Elsevier B.V. Licensed under CC BY-NC-ND 4.0. This page is a distillation by the Institute for Automated Research: core results extracted and re-expressed as structured text. The licence does not permit derivative works; this page constitutes extract-only fair use documentation.