Teams and Belief Overreaction: Barahona, Cassella, Jansen & Pezone (2026)

What this is. The paper’s core results, the measurement framework for belief overreaction, the decomposition of the team effect into three channels, and the key estimating equations: enough to understand what was found and how, without reading the full 17-page article. To replicate or extend it, read the original at https://doi.org/10.1016/j.jfineco.2025.104219.

TL;DR

The paper addresses a fundamental question in behavioral finance: does moving from individual to team decision-making amplify or attenuate belief overreaction to recent asset returns? Using preregistered randomized experiments on the Labvanced platform with 1,512 Prolific participants, plus a within-subject field study of US equity mutual fund managers (1980-2018), the paper finds that two-person teams reduce individual overreaction by 30 to 55 percent. A quantitative decomposition, following the approach of Enke et al. (2023), partitions the lab team effect into three channels: internal reflection (the act of pre-team deliberation), self-selection (the tendency of the less-biased member to lead), and external screening (the group interaction itself). Self-selection accounts for roughly 70 percent of the reduction. LLM analysis of roughly 18,000 chat exchanges in the Group treatment corroborates this, and dynamic evidence shows that participants reduce their leadership role after making larger forecast errors. The field results, based on the approach of Bordalo et al. (2020) for identifying overreaction, are consistent: mutual fund teams attenuate extrapolative overreaction by about 55 percent relative to the individual behavior of the same managers, and this attenuation coincides with better investment performance.

Core results

Magnitudes and significance are as reported; \* / \*\* / \*\*\* = 10% / 5% / 1%. Standard errors in brackets, clustered at the team level (equivalent to individual-level for the Individual treatment). Locators reference the source PDF.

#	Result	Locator	Magnitude
R1	Teams reduce overreaction by ~30%: Group treatment dummy lowers the individual overreaction coefficient by 0.092, significant at 1%, robust across demographics, financial sophistication, and income fixed-effects controls	Table 2, p. 6	Group = -0.092*** [0.035] (Col 1, n = 704); stable at -0.099*** to -0.101*** in Cols 2-4; Individual mean beta = 0.311, SD = 0.424
R2	Self-selection explains ~70% of the team effect: the three-channel decomposition shows internal reflection contributes essentially nothing (-0.001, n.s.) and external screening contributes -0.025 to -0.031 (n.s.), while self-selection contributes -0.067 to -0.090 (significant in most specifications)	Table 3/Eq. (5), p. 8	Most conservative spec (Col 4): IR = -0.001 (n.s.); SS = -0.068* [0.039]; ES = -0.031 (n.s.); total team effect = -0.098 [0.035]
R3	Teams show a weaker recency effect: fitting the exponential return-extrapolation model separately for each treatment, the recency weight (1 minus the decay parameter) is about 2.4x smaller for Group participants than Individual participants	Fig. 3/Eq. (7), p. 11	Group lambda2 = 0.953 (recency = 0.047) vs Individual lambda2 = 0.886 (recency = 0.114); Group places more equal weight across all 40 past returns
R4	Self-selection is dynamic and driven by past errors: participants who led the team prediction in round t-1 are significantly less likely to lead in round t after the team made a large forecast error in round t-1	Table 4/Eq. (6), p. 10	MostVotes_{t-1} x |Error_{t-1}| coefficient = -4.440*** [0.645] (Votes, Col 1); -0.137*** [0.018] (MostVotes dummy, Col 3); n = 7,657 round observations
R5	Teams achieve higher prediction accuracy: Group participants generate lower mean-squared error, lower mean-absolute error, and earn a higher experimental bonus than Individual participants	Table 5, p. 10	Group vs Individual: MSE -157.722*** [30.394]; MAE -2.015*** [0.408]; Bonus +$0.059** [0.023]; full controls, n = 703
R6	Mutual fund teams attenuate extrapolative overreaction by ~55%: within-subject comparison of team overreaction and statistical counterfactual individual overreaction shows teams transmit only about 45% of extrapolative behavior, while contrarian (non-overreacting) behavior is fully transmitted	Table 8/Eq. (10-11), p. 15	IV sum delta0 + delta1 = 0.45 (Col 7); null of full transmission (= 1) rejected at IV p = 0.015 (Col 7) and IV p = 0.018 (Col 8); contrarian-only delta0 not significantly below 1 (IV p = 0.568)

Overall (paper’s conclusion). Both in the lab and in the field, teams reduce belief overreaction relative to individuals. The dominant mechanism is self-selection: in two-person teams, the less-biased member tends to take on decision authority. This process is dynamic (driven by past forecast errors and feedback) and is confirmed by LLM-based analysis of chat exchanges. In the field, the attenuation of extrapolative trading by mutual fund teams is associated with better subsequent fund performance, while contrarian (non-overreacting) behavior is preserved.

Theory / model

The paper has no formal structural model. It motivates belief overreaction with the representativeness-heuristic framework of Barberis (2018), which predicts that investors overextrapolate recent returns, and tests team effects on this well-documented bias.

Cognitive task and AR(1) process. The hypothetical stock used in the experiment follows an AR(1) process (p. 4):

$$x_t = \rho x_{t-1} + \varepsilon_t, \qquad \varepsilon_t \sim \mathcal{N}(0, \sigma^2), \quad \rho = 0.5, \quad \sigma = 20$$

Given that past returns have only weak predictive power for future returns, the paper sets the rational benchmark at $\beta_i = 0$ (the best response when $\rho$ is difficult to infer). Any positive $\beta_i$ signals overreaction: the participant over-weights the most recent return realization.

Return-extrapolation model (recency channel). Following Greenwood and Shleifer (2014), an exponentially-weighted extrapolation model is estimated to decompose overreaction into a level (attribute substitution) and a recency component (Eq. 7, p. 11):

$$\hat{E}_i x_{i,t+1} = \lambda_0 + \lambda_1 \frac{\sum_{j=0}^{N} \lambda_2^j \, x_{t-j}}{\sum_{j=0}^{N} \lambda_2^j} + \varepsilon_{i,t} \tag{7}$$

Here $\lambda_1$ captures the overall level of attribute substitution (sensitivity to past returns in general) and $\lambda_2$ governs the relative importance of more versus less recent returns: as $\lambda_2 \to 0$, the most recent observation receives disproportionately more weight (stronger recency effect); as $\lambda_2 \to 1$, all past returns receive equal weight. The quantity $1 - \lambda_2$ is used as the recency-effect measure.

Team-effect decomposition. The aggregate team effect is $\Delta\beta_G = \bar{\beta}_G - \bar{\beta}_I$. It is partitioned into three additive channels (Eq. 3, p. 6):

$$\Delta\beta_G = \underbrace{(\bar{\beta}_{IR} - \bar{\beta}_I)}_{\Delta\beta_{IR}\,(\text{internal reflection})} + \underbrace{(\bar{\beta}_{SS} - \bar{\beta}_{IR})}_{\Delta\beta_{SS}\,(\text{self-selection})} + \underbrace{(\bar{\beta}_G - \bar{\beta}_{SS})}_{\Delta\beta_{ES}\,(\text{external screening})} \tag{3}$$

where $\bar{\beta}_{IR}$ is average overreaction in the Internal Reflection treatment (participants forecast individually before seeing their partner’s prediction, no team interaction yet), $\bar{\beta}_{SS}$ is average overreaction in the Self-Selection treatment (voting mechanism picks the team forecast), and $\bar{\beta}_G$ is the average in the actual Group treatment. The design isolates each mechanism: IR captures the effect of deliberate pre-team reasoning; the gap between SS and IR isolates the self-selection mechanism; the residual of G vs SS measures external screening via actual discussion.

Method

Lab experiment. The experiment runs on the Labvanced browser-based platform with subjects recruited via Prolific (US-based, pre-screened for 98% approval rate). Participants observe a 40-period AR(1) return series and use a vertical slider to predict the next-period return; the prediction task repeats for 20 rounds per session. Compensation uses a Brier-style scoring rule following Dwyer et al. (1993) and Afrouzi et al. (2023): $S_t = 100 \times \max(0,\, 1 - |FE_t|/\sigma)$, paid as a dollar bonus (mean approximately $6.09). Two key treatments:

• Individual (I): each participant forecasts independently in each round.
• Group (G): two randomly matched participants communicate via a live chat box and must agree on a joint forecast before advancing; both earn the group score.

Two additional preregistered treatments isolate mechanism channels:

• Internal Reflection (IR, RCT-Id AEARCTR-0013710): participants forecast individually first, see their partner’s forecast, then make a joint prediction; eliminates the actual chat discussion.
• Self-Selection (SS, RCT-Id AEARCTR-0014914): participants each independently forecast and then allocate 100 votes across the two predictions; the prediction with the most votes becomes the team forecast.

Final sample: 1,512 participants (248 Individual, 456 Group, 405 IR, 403 SS) after quality filtering for abnormally high rates of exactly-correct predictions.

LLM analysis of chat content. To quantify self-selection patterns in the Group treatment, the paper uses GPT-4o-mini (07/18/2024) on the roughly 18,000 chat exchanges collected (Fig. 2, p. 9). In a “supervised” pass, the LLM records: (i) the first numeric proposal made and its author, (ii) whether the other participant accepted or counter-proposed. The number of rounds where the first proposal was accepted without counter-proposal (uncontested rounds, mean 16.37 per team) proxies for self-selection intensity. In an “unsupervised” pass, the LLM rates each team’s self-selection score on a 0-10 scale (mean 6.56). The two measures correlate at 0.29 (Spearman, p < 0.01, Panel C), validating the LLM-based approach. This LLM analysis is conducted by Enke et al. (2023)-inspired methods adapted to team financial decisions.

Within-subject field design. The paper identifies 308 mutual fund teams in which at least one member has also managed a fund individually at some point. The statistical counterfactual $\hat{\beta}_j^{CF}$ is the equal-weighted average overreaction of the team’s members measured when they manage individually, observing them at the same point in time as the team observation. This within-subject design isolates the team effect from compositional differences between solo- and team-managed funds. An IV strategy based on Jegadeesh et al. (2019) uses disjoint subsamples of the data to construct an instrument for $\hat{\beta}_j^{CF}$ that is free of measurement error.

Empirical specifications

Overreaction measurement (R1-R5, lab). For each participant $i$, an individual overreaction coefficient $\hat{\beta}_i$ is estimated from the individual-level time-series regression (Eq. 1, p. 4):

$$\hat{E}_i x_{i,t+1} = \alpha_i + \beta_i x_t + \varepsilon_{i,t} \tag{1}$$

where $\hat{E}_i x_{i,t+1}$ is participant $i$‘s stated prediction for the next-period return and $x_t$ is the most recent return. Rational expectations benchmark: $\beta_i = 0$. Larger $\hat{\beta}_i > 0$ signals stronger overreaction.

Team effect regression (R1). The participant-level overreaction coefficients $\hat{\beta}_i$ serve as the dependent variable in the cross-sectional regression (Eq. 2, p. 5):

$$\hat{\beta}_i = \alpha_2 + \gamma G_i + \delta' X_i + \eta_i \tag{2}$$

where $G_i = 1$ for Group treatment participants and $X_i$ is a vector of demographic and financial sophistication controls. Standard errors are clustered at the team level. The key estimate is $\hat{\gamma} = -0.092$ (Table 2, Col 1), stable from -0.099 to -0.101 with controls (Cols 2-4), corresponding to a 30 percent reduction relative to the Individual mean of 0.311.

Mechanism decomposition regression (R2). The sample is expanded to include all four treatments and the regression becomes (Eq. 4, p. 8):

$$\hat{\beta}_i = \alpha_3 + \gamma_{IR} IR_i + \gamma_{SS} SS_i + \gamma_G G_i + \delta' X_i + \varepsilon_i \tag{4}$$

where $IR_i$, $SS_i$, and $G_i$ are treatment dummies. The estimated treatment differences map to the three channels in Eq. (3). The most conservative four-control specification gives (Eq. 5, p. 8):

$$\underbrace{-0.098}_{\Delta\hat{\beta}_G\;[\text{se}=0.035]} = \underbrace{-0.001}_{\Delta\hat{\beta}_{IR}} \underbrace{-0.067}_{\Delta\hat{\beta}_{SS}} \underbrace{-0.031}_{\Delta\hat{\beta}_{ES}} \tag{5}$$

Dynamic self-selection (R4). The round-level panel regression identifies how past decision-making errors affect subsequent voting behavior in the Self-Selection treatment (Eq. 6, p. 9):

$$Y_{i,t} = \alpha + \beta \, \text{MostVotes}_{i,t-1} + \gamma |\text{Error}_{t-1}| + \delta \, \text{MostVotes}_{i,t-1} \times |\text{Error}_{t-1}| + \varepsilon_t \tag{6}$$

where $\text{MostVotes}_{i,t-1} = 1$ if participant $i$ cast the most votes (and hence made the team decision) in round $t-1$, and $|\text{Error}_{t-1}|$ is the absolute team forecast error in round $t-1$, demeaned and standardized. The dependent variable $Y_{i,t}$ is either the number of votes cast in round $t$ (Cols 1-2) or an indicator for being the decision maker in round $t$ (Cols 3-4). The interaction term $\hat{\delta}$ captures the feedback loop: the decision maker in round $t-1$ reduces their vote count in round $t$ by 4.44 votes per standard-deviation increase in the team forecast error (Table 4, Col 1).

Field overreaction measurement (R6). For each fund $j$, the fund’s sensitivity of trades to past returns is estimated by a panel regression (Eqs. 8-9, p. 12):

$$\text{trade}_{s,j,t+1} = \alpha_j + \beta_j^X r_{s,t-4\to t} + \gamma_j^T C_{s,t} + \theta_{jt} + \varepsilon_{s,j,t+1} \tag{8}$$ $$\text{trade}_{s,j,t+1} \;\equiv\; \frac{(\text{shares}_{s,j,t+1} - \text{shares}_{s,j,t+1}^{\text{sp(t-adj)}}) P_{s,t+1}}{TNA_{j,t+1}} \tag{9}$$

where $r_{s,t-4\to t} = \sum_{l=0}^{3} w_l r_{s,t-l}$ is the exponentially-weighted four-quarter past return of stock $s$ (weights from Greenwood and Shleifer (2014), $\lambda = 0.56$), $C_{s,t}$ is a vector of stock controls (momentum, stock characteristics), $\theta_{jt}$ is a fund-quarter fixed effect, and $TNA_{j,t+1}$ is fund net assets. A positive $\hat{\beta}_j^X$ characterizes extrapolators; a negative $\hat{\beta}_j^X$ characterizes contrarians.

Team transmission (R6). The team overreaction $\hat{\beta}_j^{TM}$ is regressed on the statistical counterfactual $\hat{\beta}_j^{CF}$ (Eq. 10, p. 13):

$$\hat{\beta}_j^{TM} = \alpha + \delta_0 \hat{\beta}_j^{CF} + \delta_1 \hat{\beta}_j^{CF} \times D_j^E + \delta_2 D_j^E + \delta_3 C_j + \varepsilon_j \tag{10}$$

where $D_j^E = 1$ for extrapolative teams ($\hat{\beta}_j^{CF} > 0$) and $C_j$ are fund controls. Full transmission of overreaction for extrapolative teams implies $\delta_0 + \delta_1 = 1$; attenuation implies $\delta_0 + \delta_1 < 1$. Contrarian behavior is fully transmitted if $\delta_0 = 1$. The IV estimate (Table 8, Col 7) gives $\delta_0 + \delta_1 \approx 0.45$; the null $\delta_0 + \delta_1 = 1$ is rejected at $p = 0.015$ (IV Col 7) and $p = 0.018$ (IV Col 8).

Datasets used

Dataset	Role in paper	Wiki page
Lab experiment data (Labvanced / Prolific, 2024)	Primary data for Sections 2-2.5: 1,512 participants, four treatments, 20 prediction rounds each; chat transcripts analyzed by LLM	No page yet (original data; replication package on Mendeley Data)
CRSP monthly stock returns (via WRDS)	Quarterly stock-level returns for the fund trading regression (Eq. 8); past four-quarter return predictor	WRDS (licensed)
Thomson Reuters Mutual Fund Holdings (via WRDS)	Quarterly holdings of US stocks per fund; used to construct the split-adjusted trade measure (Eq. 9)	WRDS (licensed)
Morningstar	Fund investment objectives, fund family, expense ratios, fund age; used as controls in the field analysis	Morningstar (licensed)
SEC mandatory fund filings	Fund managerial structure (team vs individual management identification); fund-level panel 1980-2018	EDGAR

Sample (field): 467 unique managers, 847 unique funds, 308 unique team observations, quarterly 1980-2018. Sample (lab): 1,512 participants across four treatments, run June-November 2024.

When to read the full paper

Read the original at https://doi.org/10.1016/j.jfineco.2025.104219 if you are studying team effects on belief formation and behavioral biases (Section 2 for the experimental design and Tables 2-5 for the core results); implementing the three-channel decomposition of team effects (Section 2.3 and Eq. 3-5 for the framework); analyzing the dynamic feedback between forecast errors and team leadership roles (Section 2.4 and Table 4); studying how organizational structure (solo vs team management) affects fund manager trading behavior and fund performance in the field (Section 3 and Tables 6-8); or looking for evidence linking overreaction to investment underperformance (Section 3.6.1 and Fig. 4). The Internet Appendix contains preregistration documents, the LLM prompts (Appendix IA3), and robustness checks.

Attribution and rights

Source: peer-reviewed, Journal of Financial Economics 176 (2026), article 104219. This distillation was extracted by an LLM on 2026-06-24 and is not human-verified or independently reproduced. All rights reserved; this page is extract-only.

Barahona, Ricardo, Stefano Cassella, Kristy A.E. Jansen, and Vincenzo Pezone. “Do teams alleviate or exacerbate overreaction in beliefs?” Journal of Financial Economics 176 (2026): 104219. DOI: 10.1016/j.jfineco.2025.104219. © 2025 Elsevier B.V. All rights reserved.