Personal sense of power predicts financial risk-taking propensity: But only when risk-related decisions are made without cognitive load

The study investigates whether cognitive load moderates the established positive relationship between personal sense of power (SOP) and financial risk-taking in gambling and investing. Power, conceptualized as asymmetric control over valued resources, can be a state or a trait (SOP). Prior work shows power increases financial risk-taking, but mechanisms and boundary conditions are underexplored. Given links between power and executive functions (EFs) and between EFs and financial risk-taking, the authors hypothesize that momentary cognitive load (impairing EFs) weakens the SOP–risk link. The study tests three hypotheses: H1 SOP positively predicts risky financial choices; H2 cognitive load negatively predicts risky financial choices; H3 cognitive load moderates the SOP–risk relationship, attenuating it under load.

The paper reviews evidence linking power to cognition and executive functions: higher power enhances attentional inhibition, working memory, and cognitive flexibility, while low power is associated with EF impairments (Keltner et al., 2003; Smith et al., 2008; Yin & Smith, 2020). SOP and induced power relate positively to cognitive flexibility and creativity. Financial risk-taking relates to cognition, with inhibitory control and working memory predicting gambling performance, and cognitive flexibility predicting both investing and gambling risk-taking. Cognitive load typically increases risk aversion in lottery tasks, leading to safer choices and faster, more risk-averse decisions (Benjamin et al., 2013; Deck & Jahedi, 2015; Whitney et al., 2008). However, effects on investment risk-taking are underexplored, and risk-taking may be domain-specific (Vlaev et al., 2010), necessitating investigation across gambling and investing.

Design: Two-wave, incentivised online experimental study using CAWI via the Polish ARIADNA panel. Wave 1 measured SOP; Wave 2 randomly assigned participants to cognitive load (experimental) or control, then administered risk-taking measures in random order. Participants: N=192 Polish working adults (93 female, 99 male), age 21–66 (M=42.98, SD=12.08). A priori power analysis (α=.05, power=.80, effect size=.075) suggested N=150; recruited ~15% extra anticipating exclusions in Holt-Laury tasks. Groups were comparable on demographics. Measures: SOP assessed with the 8-item Sense of Power Scale (1–7 Likert; four reverse-coded; mean score). Financial risk propensity measured with DOSPERT subscales for gambling and investing (three items each; 1–7; averaged). Gambling risky choices measured via incentivised Holt & Laury (2002) lottery task (10 binary choices between Lottery A and B; risk index = number of B choices; rationality checks excluded participants with dominated choices or multiple switches; 2 excluded). Investment risky choice measured via incentivised Gneezy-Potters task: allocate 10,000 PLN between a risky investment (50% success yields 2.5×, 50% lose invested amount) and an interest-free account; outcome determined incentive points (0–3). Cognitive load manipulation: Experimental group completed a Stroop-inspired word–colour memorisation task designed to deplete inhibitory control, working memory, and cognitive flexibility. Participants memorised three colour words presented in incongruent font colours, retaining both word and ink colour, and answered control recognition questions with feedback; they completed four rounds (before each risk measure and after the 5th Holt-Laury decision). Control group had no additional task. Procedure: SOP scale in Wave 1; in Wave 2, random assignment to conditions, cognitive load tasks (if applicable), then risk tasks in random order; incentives paid based on task outcomes. Analysis: Multiple regression using Hayes’ PROCESS v4.2 Model 1 in SPSS. Cognitive load dummy-coded (1=load, 0=control); variables mean-centered. Step 1 included SOP and load; Step 2 added SOP×load interaction. Descriptive statistics and correlations reported.

Correlations: SOP positively correlated with all financial propensity and behavior measures (weak to medium), e.g., with gambling propensity, investment propensity, risky gambling choices, and risky investment choice (Table 1). Effects were significant in the control group but not under cognitive load (Table 2). Regression results: Gambling propensity (DOSPERT): Step 1 model significant, F(2,189)=25.14, p<.001; SOP positive predictor (b=1.31, SE=.34, p<.001), cognitive load negative predictor (b=-3.43, SE=.34, p<.001). Step 2 interaction model significant, F(3,188)=18.32, p<.001; SOP (b=1.94, SE=.47, p<.001) and SOP×load interaction (b=-1.35, SE=.68, p<.05) significant; ΔR²=.02, p<.05. Simple slopes: control β=1.94, p<.001; load β=.59, p=.23. Risky gambling choices (Holt-Laury): Step 1 significant, F(2,187)=18.84, p<.001; SOP positive (b=.90, SE=.34, p<.001), cognitive load negative (b=-.95, SE=.34, p<.01). Step 2 significant, F(3,186)=14.47, p<.001; SOP (b=1.27, SE=.25, p<.001) and SOP×load (b=-.80, SE=.36, p<.05) significant; ΔR²=.02, p<.05. Simple slopes: control β=1.27, p<.001; load β=.48, p=.07. Investment propensity (DOSPERT): Step 1 significant, F(2,189)=19.80, p<.001; SOP positive (b=1.62, SE=.32, p<.001), cognitive load negative (b=-1.76, SE=.60, p<.001). Step 2 significant, F(3,188)=16.24, p<.001; SOP (b=2.44, SE=.43, p<.001), cognitive load (b=5.54, SE=2.69, p<.05), and SOP×load (b=-1.75, SE=.63, p<.01) significant; ΔR²=.033, p<.001. Simple slopes: control β=2.43, p<.001; load β=.69, p=.13. Risky investment choice (Gneezy-Potters allocation): Step 1 significant, F(2,189)=9.18, p<.001; SOP positive (b=791.29, SE=209.76, p<.001), cognitive load nonsignificant (b=-569.25, SE=391.51). Step 2 significant, F(3,188)=8.22, p<.001; SOP (b=1,262.55, SE=284.65, p<.001), cognitive load (b=3,615.56, SE=1,776.43, p<.05), and SOP×load (b=-1,001.61, SE=414.99, p<.05) significant; ΔR²=.03, p<.05. Simple slopes: control β=1,262.55, p<.001; load β=260.94, p=.39. Overall: H1 supported (SOP positively predicts financial risk propensity and behavior); H2 supported (cognitive load reduces risk-taking propensity and risky choices, notably in gambling and investment propensity); H3 supported (cognitive load attenuates SOP’s positive effect; SOP–risk link significant only without load).

Findings demonstrate that personal sense of power reliably increases both propensity and actual financial risk-taking in gambling and investing, but this effect depends on cognitive resources. Under cognitive load, which impairs executive functions, the positive SOP–risk link becomes nonsignificant, indicating that access to EFs is a boundary condition for power’s influence on financial risk behavior. This extends prior work on power and risk, and on cognitive load increasing risk aversion, by showing moderation across subdomains and both propensity and behavior. Results have practical relevance in modern contexts of information overload and multitasking: cognitive strain not only shifts risk preferences toward safety but also alters how psychological traits like SOP translate into behavior. Implications include designing financial education, policies, or interventions that reduce cognitive load or bolster cognitive resources to foster responsible financial decisions, and considering the role of powerful individuals’ risk tendencies in entrepreneurship and macroeconomic outcomes.

The study contributes novel evidence that cognitive load moderates the positive relationship between personal sense of power and financial risk-taking across gambling and investing, with SOP effects present only when cognitive resources are intact. It confirms that SOP increases financial risk propensity and choices, and that cognitive load generally reduces them. Future research should employ more ecologically valid, higher-stakes financial decision paradigms (e.g., investment simulators), better differentiate investment versus gambling tasks, examine experimentally induced power states alongside SOP, manipulate graded levels of cognitive load, assess perceptions of risk and benefit (additional DOSPERT components), and test interventions (e.g., mindfulness, cognitive training) to alleviate cognitive load and their interactions with individual differences.

Monetary incentives were lower than real market stakes, potentially limiting external validity. Commonly used financial risk measures (Holt-Laury and Gneezy-Potters) may oversimplify complex real-world decision-making; tasks are structurally similar, possibly blurring investing versus gambling distinctions. Online panel participants may have attempted note-taking during cognitive load tasks, though the design aimed to make this difficult; nonetheless, differences were robust. Power was treated as a trait (SOP); induced power states and their interaction with cognitive load warrant examination. The study focuses on willingness to engage; adding DOSPERT risk/benefit perception items and varying cognitive load intensity could refine insights.