The impact of agency on time and risk preferences

The paper investigates whether and how personal agency—defined as having a larger set of available options—causally shapes core economic preferences: patience in intertemporal choice and tolerance for risk. Prior arguments emphasize both instrumental and intrinsic value of agency, but its role in shaping preferences remains underexplored. The authors posit that greater agency increases risk tolerance by reducing perceived risk, which in turn leads to greater willingness to wait for larger, later rewards given the inherent uncertainty of delayed outcomes. Using correlational evidence from the World Values Survey (WVS; 2010–2014; N=86,272; 61 countries), they show higher self-reported agency is associated with greater reported savings, even after controlling for income, employment, marital status, and education; agency also explains residual variation in savings beyond standard determinants. Motivated by these observations, they conduct experiments to causally test whether endowing agency in adverse states (resource scarcity; environmental stressors) increases patience and risk tolerance, even when agency is not exercised.

The authors situate their work within literature on agency, control, and adverse states. Philosophers, political scientists, and economists have discussed agency’s intrinsic and instrumental value. Psychology and behavioral economics findings show diminished agency elicits punishment of those responsible and increases well-being when restored. Perceived control affects risk perception and tolerance (e.g., illusion of control; anticipated regret mediates control–risk links), implying higher agency may reduce perceived risk and increase risk-taking. Intertemporal choice theory links uncertainty to hyperbolic discounting, suggesting that risk attitudes influence patience; greater risk aversion favors smaller–sooner rewards. Research on scarcity documents that experiencing resource scarcity increases risk aversion and impatience, but often conflates adversity with lack of agency, making mechanisms unclear. The authors propose that personal agency is a key pathway linking adverse states to impatience, and that increasing agency can mitigate these effects. They also discuss policy-relevant evidence on unconditional cash transfers, which increase recipients’ agency and are associated with improved outcomes, suggesting agency could foster more forward-looking choices.

Motivating correlational analysis (WVS): Using the 2010–2014 World Values Survey multinational dataset (N=86,272; 61 countries), the authors measure reported savings (saved money, just got by, spent savings, or spent savings and borrowed) and perceived agency (1–10 scale of freedom of choice and control over life outcomes). They estimate OLS regressions with robust standard errors clustered at the country level, controlling for income, employment, marital status, education, and other covariates. They also regress residuals from a model with standard determinants on agency to test whether agency explains unexplained variation in savings. Results are correlational and used as motivation, not causal claims. Study 1 (time scarcity and agency; MTurk; N=217): Random assignment to Scarcity–Agency (Agency; N=69), Scarcity–No Agency (No Agency; N=76), or Control (no scarcity; N=72). Task 1: 15 true/false cognitive aptitude questions (Wonderlic-adapted). Scarcity manipulation via time limits: Scarcity groups had 10 seconds per question; Control had unlimited time. Task 2: another 15 questions with a threshold-based bonus (doubling payment if correct ≥ random threshold 1–15). Scarcity increased for Scarcity groups to 6 seconds per item; Control still no limit. Agency manipulation: In Task 2, the Agency group could add 4 seconds per question at any time by checking a box, incurring a cost of forfeiting 80% of base payment; No Agency had no such option. This cost discouraged exercising agency to keep experienced adversity comparable across scarcity groups. Four Agency participants who exercised agency were routed out before outcome measures. Measures and checks: After Task 2, participants reported perceived agency and affect (anger, sadness, happiness; 1–5). Scarcity check: felt time availability (1 = not enough time; 7 = too much). Time preferences: two incentivized token allocation decisions between earlier and later dates (today vs one week; one week vs two weeks); tokens allocated to later dates paid 50% more than earlier, one decision randomly selected for payment. Patience measured as tokens allocated to later dates; analyses collapse across the two decisions. Risk preferences: four hypothetical binary choices between safer and riskier financial gambles. Analyses: OLS with robust SE clustered at participant level; compare conditions on patience, test mediation via risk tolerance (proportion choosing safe options; higher values = greater risk aversion). Lee bounds computed to address potential selection from excluding those who exercised agency. Robustness checks include alternate risk operationalizations and comprehension controls. Study 1b (manipulation check): Participants randomized to the same three conditions; behavioral measure of subsequent demand for agency, plus affect and trust in the experimenter. Aim: confirm manipulation affected agency without changing affect or trust. Study 2 (environmental stressor and agency; lab undergraduates; N=115): Participants solved 30 anagrams in 5 minutes while exposed to aversive, unpredictable noise (3000 Hz, 90 dB tone at random intervals) via headphones. Random assignment to Agency (N=61) or No Agency (N=54). No Agency: removing headphones disqualifies from study/bonus. Agency: option to remove headphones anytime at a cost of 50% of potential $20 bonus (1 in 10 chance), discouraging use to equalize experienced noise. Four practice anagrams without noise preceded the task. Time preferences: After the anagram task, participants answered 27 hypothetical choices between smaller–sooner and larger–later monetary rewards (amounts and delays varied). Primary analysis: percent of smaller–sooner choices (nonparametric). Parametric analysis: estimate individual hyperbolic discounting parameter k from indifference points using V = A / (1 + kD); larger k indicates greater impatience. Exclusions: main analysis excludes two Agency participants who removed headphones; k-estimation excludes participants choosing all sooner or all later options (non-rationalizable). Robustness: intent-to-treat including those who removed headphones; regression controls include anagrams solved. Study 2b (manipulation check): Behavioral measure of demand for agency after headphone-removal option manipulation to confirm perceived agency shift.

Correlational (WVS): Higher reported agency predicts greater reported savings controlling for socioeconomic covariates (OLS with robust SE clustered by country, p = 0.001). Agency explains residual variation in savings beyond standard determinants (p < 0.0005). Study 1 (time scarcity): Scarcity manipulation effective: scarcity groups solved ≈4 fewer items correctly than Control (MScarcity 8.17 vs MControl 12.28; t(211)=11.715, p<0.0005); Agency and No Agency solved similar numbers (MAgency 8.31 vs MNoAgency 8.05; t(139)=0.686, p=0.49). Scarcity perception higher in scarcity groups (M=2.38) vs Control (M=4.59; t(209)=13.056, p<0.0005). Agency manipulation effective: perceived agency lower in No Agency (M=2.30) vs Control (M=2.74; t(144)=2.087, p=0.039); Agency vs Control not different (M=2.51 vs 2.74; t(133)=1.103, p=0.272). Affect did not differ across Agency/No Agency (ps ≥ 0.136). Only 4/69 (≈6%) in Agency exercised agency. - Patience: Agency participants allocated about 30% fewer tokens to earlier dates than No Agency (MAgency=19.6 vs MNoAgency=28.6; OLS with robust SE clustered by participant, p=0.041), indicating greater patience with agency despite equal scarcity. Agency and Control were indistinguishable (MControl=19.2; p=0.924). No Agency were less patient than Control (p=0.031). Results robust across specifications (see Table 1) and to Lee bounds (95% CI for treatment effect excluding 0 for patience: 0.32–16.91). - Mediation by risk tolerance: Risk aversion (proportion of safe choices) higher in No Agency than Agency (B=0.13, p=0.006). Risk aversion predicts more tokens to earlier dates (scarcity groups only: B=23.63, p=0.001; all groups: B=20.37, p<0.0005). The No Agency effect on impatience drops from significant (B=9.02, p=0.041) to nonsignificant (B=6.23, p=0.160) when controlling for risk aversion; ≈30% of the effect is mediated. Study 1b: Agency manipulation reduced subsequent demand for agency among those who previously had agency; no differences in negative affect or trust across groups. Study 2 (noise stressor): Excluding two who removed headphones (N=113), Agency participants chose smaller–sooner options 49% vs 58% in No Agency (OLS with robust SE, p=0.008; Mann–Whitney U, p=0.015), indicating greater patience with agency despite equal exposure. Intent-to-treat including headphone removers yields consistent results (ps ≤ 0.018). Parametric estimates: discount rate k lower in Agency (M=0.007) than No Agency (M=0.015); t(107)=2.727, p=0.008. Regression table shows Agency coefficient ≈ −9.6 to −10.2 percentage points on smaller–sooner choices across specifications (p from 0.009 to 0.001). Study 2b confirmed the manipulation increased perceived agency.

The findings show that personal agency causally shapes economic preferences under adverse states. When individuals facing identical adversity are endowed with the option to alleviate it—even at substantial cost and rarely exercised—their perceived agency increases, which elevates risk tolerance and yields more patient intertemporal choices. Agency eliminated the impatience typically induced by scarcity: behavior of Agency participants matched that of those not experiencing scarcity. Mediation analyses indicate that shifts in risk tolerance are a key pathway through which agency influences patience, consistent with theories linking uncertainty and risk attitudes to temporal discounting. Generalization to environmental stressors (aversive noise) demonstrates that the agency effect is not limited to resource scarcity. These results clarify that previously observed impatience under adverse states may be driven in part by lack of agency rather than adversity per se. Policy implications include designing programs that expand recipients’ choice sets (e.g., unconditional transfers), which may foster more forward-looking decisions beyond instrumental or intrinsic benefits of freedom of choice.

This paper establishes that endowing individuals with personal agency over adverse states increases risk tolerance and patience, even when the agency option is rarely exercised. Agency fully counteracts the impatience induced by scarcity and generalizes to other stressors. Mediation analyses identify increased risk tolerance as a mechanism. The work reframes how adversity influences preferences, highlighting agency as a central driver and offering a pathway for policy and institutional designs to promote farsighted behavior by expanding choice sets. Future research could examine boundary conditions and heterogeneity (e.g., by baseline cognitive ability, socioeconomic status, or cultural context), explore additional adverse contexts (e.g., health stressors, financial shocks), investigate long-run persistence of agency effects, and test agency-enhancing interventions in field settings with real financial outcomes.

- The WVS analysis is correlational and cannot establish causality; despite extensive controls and residual analyses, unobserved confounding may remain. - Study 1 excluded participants who exercised agency to equate experienced scarcity, which may introduce selection; Lee bounds suggest robustness, but some residual selection concerns remain. - Potential heterogeneity: The adverse-state impact and agency manipulation could be greater for individuals with lower cognitive ability; while random assignment ensures internal validity on average, differential effects cannot be ruled out. - Generalizability: Study 1 used MTurk participants and Study 2 used undergraduates; external validity to other populations and real-world decisions may be limited. - Some measures (risk preferences in Study 1; time preferences in Study 2) were hypothetical, which may differ from incentivized behavior, though Study 1’s time allocations were incentivized and results were robust across analyses. - The agency options were deliberately costly to deter use; different cost structures or types of agency could yield different effect sizes.