Trying to think: An experimental study of the impact of cognitive load on financial risk taking by groups

The study examines how cognitive load influences financial risk-taking in group decision-making contexts common in corporates (e.g., boards, committees, management teams). In the information-rich environment, decision-makers frequently face cognitive load that can impair processing. Prior evidence on cognitive load’s effect on risk attitudes is mixed, with some studies finding reduced risk-taking or no effect. This experiment manipulates cognitive load during an incentivized investment task across individual, group, and gender-heterogeneous group settings to test whether cognitive load, group membership, and prior outcomes affect risk-taking. The work also explores whether effects of previous losses display recency and whether cognitive load weakens the influence of past losses due to working memory constraints.

The paper situates its research within several strands: (1) Group versus individual decision-making in corporate and professional settings, showing groups often differ from individuals in risk behaviors (e.g., auditors, investment committees, social settings). (2) Cognitive load and information overload in modern contexts, with evidence that load can impair cognitive processing; prior findings on load and risk attitudes are contradictory (some show reduced risk-taking; others no significant relationship). (3) Behavioral responses to prior outcomes: realization and house-money effects indicate that past losses or their realization can shape current risk-taking; the recency effect suggests recent outcomes weigh more heavily than distant ones. (4) Gender and risk: while many studies find gender differences in risk attitudes, the paper notes that among economics/accounting students with similar financial knowledge, gender effects may be attenuated, aligning with research that education and sophistication can reduce gender-based differences in financial decisions.

Participants: 108 first-year undergraduate students (economics/accounting majors) from the College of Management in Israel; gender-balanced (Mage=24.04; 50% female). Recruitment halted due to budget constraints. Design: Incentivized, computer-based investment game programmed in HTML, 27 rounds divided into three phases of nine rounds each: Phase 1 (individual decisions), Phase 2 (gender-homogeneous groups of three), Phase 3 (gender-heterogeneous groups of three). A demo round preceded Phase 1. Each round: participants received 100 tokens and chose how many to invest in a risky asset with a 33.33% chance to win 3.5× invested tokens and a 66.67% chance to lose invested tokens. Outcomes were randomized by computer each round; investing more tokens indicates greater risk-taking. Feedback after each round included outcome and cumulative token balance. Payments: ILS payout equal to final token balance divided by 50. Timing: Phase 1 individual decisions had a 20-second limit per decision; Phases 2 and 3 groups had 40 seconds for pre-decision discussion; one randomly selected group member entered the group’s decision. Profits/losses were allocated individually in all phases. Cognitive load manipulation: Half of participants were randomly assigned to memorize a number before each investment round and to recall/input it after making the investment, earning 75 tokens for correct recall. In Phase 1, individuals under load memorized 6-digit numbers; in Phases 2 and 3, groups under load memorized 12-digit numbers. Data structure: Unit of analysis is Participant (individual or group): 108 individuals, 36 gender-homogeneous groups, 36 gender-heterogeneous groups (total 180 participants), each with 9 rounds, yielding 1,620 Participant×Round observations. Variables: Sum (0–100 tokens invested, dependent variable), Cognitive (1 under load; 0 otherwise), Single (1 individual; 0 group), Gender (1 female; 0 male; for groups, majority gender), Female Ratio (0, 0.333, 0.666, 1), Heterogenous (1 for gender-heterogeneous groups; 0 otherwise), Round (1–9), Round2 (1–81), Loss (1 if the round’s outcome was a loss; 0 if win). Analysis: Random-effects GLS regressions with robust standard errors clustered by participants. Models estimated on full sample and on subsamples with and without cognitive load. Additional models include interaction terms to test moderation by cognitive load and lags of Loss (one- and two-period) to assess effects of prior outcomes on current risk-taking.

Descriptive statistics (N=1,620): Mean risky investment per round was 69.195 tokens (SD=32.136). Under cognitive load, participants invested significantly more (M=78.7, SD=27.1) than without load (M=59.7, SD=33.9; t=12.5, p<0.001). Table 3 (Random-effects GLS): • Cognitive load significantly increases risky investment (coef≈19.017; p<0.001). Overall R-squared rises from 0.043 to 0.130 when Cognitive Load is added; Between R-squared from 0.071 to 0.257. • Group decision-making yields higher risk-taking than individual decision-making (Single negative: coef≈-10.203; p≤0.010 across models). • Gender, Female Ratio, and group gender heterogeneity show no significant effects on risk-taking. • Experience exhibits a nonlinear pattern: Round positive (≈3.291; p=0.003), Round2 negative (≈-0.223; p=0.035). In subsamples, experience effects become insignificant under cognitive load. • Intraclass correlation (rho) indicates substantial variance across participants (≈0.288–0.425). Interaction robustness checks (Supplement S2) show cognitive load effects are similar across individuals vs groups and across gender compositions (no significant interactions). Table 4 (with lagged Loss): • Loss in previous round (lag 1) increases risky investment (full sample: coef≈11.504; p<0.001). • Loss two rounds prior (lag 2) also increases investment but less so (coef≈5.259; p<0.001), consistent with recency. • Cognitive load moderates these effects: interactions Cognitive*Loss(lag1)≈-13.393 (p<0.001) and Cognitive*Loss(lag2)≈-6.538 (p=0.020), implying that cognitive load weakens the impact of past losses on current risk-taking. • In subsamples: without cognitive load, Loss(lag1)≈18.288 (p<0.001) and Loss(lag2)≈8.569 (p<0.001); with cognitive load, Loss(lag1)≈4.593 (p=0.018) and Loss(lag2)≈1.904 (p=0.227, ns). • Cognitive load remains positively significant in all models (e.g., Table 4 full sample coef≈18.090; p<0.001; cognitive-load subsample coef≈29.164; p<0.001). Overall, cognitive load, group membership, and prior losses increase risk-taking; the effect of losses diminishes over time and is attenuated under cognitive load.

The findings directly address the research question: cognitive load increases financial risk-taking among both individuals and groups, contrary to some prior studies reporting reduced risk-taking or no relationship. Group membership is associated with higher risk appetite relative to individuals, consistent with prior evidence on group choice shifts toward risk. Gender and group gender composition do not influence risk behavior in this sample, possibly reflecting comparable financial knowledge among economics/accounting students. Experience shows an initial rise in risk-taking followed by a decline, but these learning effects are muted under cognitive load, consistent with limited cognitive resources being diverted to the memorization task. Prior losses elevate subsequent risk-taking, with a stronger effect for more recent losses, aligning with realization and recency effects. Cognitive load weakens the influence of past losses, consistent with working memory saturation reducing retention of prior outcomes. For corporates, these results suggest that cognitively loaded group decision-makers may exhibit greater risk appetite and be less responsive to recent negative information, underscoring the importance of managing cognitive load in investment and governance settings.

The study contributes by experimentally demonstrating that cognitive load increases risk-taking in both individual and group financial decisions, that group decision-making is more risk-prone than individual decision-making, and that past losses raise risk appetite with a recency pattern, though this effect is weakened under cognitive load. Practical implications include the need for firms to mitigate cognitive overload during group investment and board decisions. Suggested strategies include internal measures such as business model visualizations and mindfulness or breathing practices at the outset of meetings, and external measures such as regulatory reforms aimed at simplifying tasks for busy directors and managers. Future research should further investigate determinants of cognitive overload, its mechanisms and impacts on group financial decisions, and evaluate the effectiveness of interventions to reduce cognitive load.