Why people follow rules

The paper investigates why people follow rules—laws, regulations, and social norms—that structure social life and support large-scale cooperation. Competing explanations include obedience to authority or tradition, intrinsic respect for rules as deontic constraints, self-interest driven by sanctions and reputational incentives, conformity to social expectations, and pro-social motives. To disentangle these motives, the authors introduce CRISP, a framework positing rule-conformity (C) as a function of intrinsic respect for rules (R), extrinsic incentives (I), social expectations (S), and social preferences (P). The study aims to identify the proximate behavioural channels of rule-following and assess their relative importance in controlled experiments that systematically manipulate incentives, social expectations, and externalities.

The study synthesizes perspectives from law, economics, sociology, psychology and cultural evolution on rule-following and social order. Classical theories emphasize norms, institutions and obedience (Hobbes, Durkheim, Weber, Hart, Hayek, Tyler). Economic accounts highlight deterrence and sanctions (Becker) and reputational incentives. Social norm theories distinguish descriptive and injunctive norms (Cialdini; Bicchieri), focusing on social expectations and appropriateness judgements. Prior experimental paradigms show norm compliance and conditional cooperation (Fehr & Schurtenberger; Fischbacher & Gächter) and methods to elicit social appropriateness (Krupka & Weber). Rule-following tasks (Kimbrough & Vostroknutov) documented substantial compliance with arbitrary rules; related tasks (ball-division) and honesty paradigms benchmark prosocial costs. Research on peer effects and norm dynamics indicates contagion of violations and tipping points (Keizer; Centola; Andreoni et al.). The CRISP framework integrates these strands, proposing a unified approach to quantify intrinsic respect, social expectations and the roles of incentives and social preferences.

Design: Four sets of online experiments (plus replications) implement the CRISP framework using a minimalist rule-following task. Participants: 14,034 English-speaking participants (primarily MTurk US workers; mean age 34.6 years; 49% female). Additional lab replications at University of Nottingham (n=103 students) and other platforms (Prolific, Qualtrics). Tasks: A ‘traffic light’ task and an abstract variant. Participants control a circle, start with 20 MU (US$1), endowment decreases by 1 MU/s. The explicit rule: wait until the light turns green (or the cross disappears). Rule-following is costly (12 s wait costs ~12 MU), violations affect nobody unless externalities are introduced. Control questions: Some conditions included compulsory questions to reinforce understanding of payoff consequences, making incentives more salient. Experiment 1 (n=8,983): Baseline with negative extrinsic incentives (I<0), no externalities (X=0), no peers. Two versions: traffic light (with/without control questions), abstract task (with control questions). Experiment 2 (n=511): Measures social expectations in separate spectator tasks—normative beliefs (b^N, social appropriateness ratings) and descriptive beliefs (b^D, estimated compliance rates)—and elicits conditional rule-conformity functions n(b^N) and d(b^D) via the strategy method across five quintiles of beliefs (0–20% to 81–100%). Incentive-compatible elicitation rewards matching or accuracy. Experiment 3 (n=7,732): Peer observation manipulation of descriptive beliefs. Three iterations: A (alone), B (treatments with 0–6 peers whose movements were sampled from prior participants; peers either conformed or violated in varying proportions), C (alone). Negative incentives (I<0), no externalities (X=0). Assess immediate and carryover peer effects. Experiment 4 (n=4,045): Nested treatments adding motives cumulatively, all with control questions. BL (baseline): I<0, X=0. EX (externality): I<0, X<0 (rule violation cancels a provisional US$1 Red Cross donation). WP (weak punishment): I<0, X<0, 10% probability of losing all MU if violating. SP (strong punishment): I>0, X<0, 90% probability of losing all MU if violating. Additional between-subjects elicitation of social expectations and conditional conformity for each treatment. Outcomes: Rule-following rates and 95% CIs; conditional conformity thresholds; normative and descriptive beliefs; peer contagion parameters; earnings forfeited. Analysis: Linear mixed models for conditional conformity (participant random intercepts), THSD for treatment comparisons, peer effect regressions (change in own violations per 1% increase in observed peer violations).

Baseline rule-following is substantial despite costs and absence of sanctions or externalities: 55–70% conform in baseline experiments. Experiment 1: Traffic light task—overall 65.6% [64.5–66.7%]; without control questions 70.0% [68.7–71.2%] (n=4,970); with control questions 57.8% [55.9–59.6%] (n=2,762). Abstract task (with control questions): 59.7% [57.0–62.4%] (n=1,251). Lab replication (Nottingham): 60.2% [50.7–69.6%] (n=103). Additional replications: Prolific (58%), Qualtrics (55%), ball-division task across five countries avg 56%; individual rule-following propensities stable over time (~61–62%). Participants forgo 48.0% [46.9–49.1%] of possible earnings by following the rule. Predictors: Little demographic or Big Five variability; positive associations with patience and shame-proneness; rule-followers more likely conditional cooperators. Experiment 2: Social expectations are strong. Spectators rate rule-following as (very) socially appropriate (~80% very appropriate), and rule-breaking as socially inappropriate (~90% very/somewhat inappropriate). Spectators’ median descriptive belief: 60% conform, aligning with observed rates. Conditional conformity increases with beliefs: normative n(b^N) from 35.2% [27.8–42.6%] at 0–20% disapproval to 56.0% at 81–100%; descriptive d(b^D) from 28.5% [21.4–35.5%] at 0–20% others conform to 56% [48.6–64.1%] at 81–100%. Individual profiles: ~19–20% unconditional followers, ~29–34% unconditional violators, ~30–31% threshold-based conditional followers, remainder non-monotonic. Experiment 3: Peer violations are contagious; observing one violator reduces own conformity by ~8 pp; increases similar across group sizes; conformity decreases with higher share of violators. Yet conformity remains high: with six violating peers, 55% [49.4–61.2%] still conform; with control questions, never below 40%. Carryover: A 1% increase in observed peer violations in iteration 2 increases own rule-breaking in iteration 3 by 0.12%; lowest iteration-3 conformity across treatments is 58%. Observing violations makes violations more socially acceptable (normative beliefs shift), while appropriateness of conformity remains high. Experiment 4: Externalities and incentives increase conformity but unconditional respect and social expectations explain much of it. BL: 54.6% [50.3–58.9%]. EX (externality): +6.8 pp to 61%. WP (10% punishment): minimal additional increase (+0–1 pp over EX). SP (90% punishment): +23.1 pp to 77.8% [74.0–81.5%] (THSD P<0.001). Unconditional conformity (proxy for intrinsic respect) is sizable: BL 26.9% [19.9–34.0%]; unconditional violators ~6% in SP. Overall, unconditional rule-following (~22–23%) and conditional conformity driven by social expectations (20–25 pp shifts) account for most observed conformity; extrinsic incentives add but are not necessary for high compliance.

The findings show that people often follow rules even when rule-following is costly, anonymous, and lacks consequences for others, indicating a central role for intrinsic respect for rules and social expectations. CRISP’s four channels collectively explain conformity, with unconditional rule-following and conditional responses to normative and descriptive beliefs accounting for most behaviour. Peer violations reduce compliance and can shift normative beliefs to make violations more acceptable, yet many participants still conform even when surrounded by violators, suggesting robustness of intrinsic and expectation-driven motives. Externalities and sanctions enhance compliance, particularly strong punishment, but their incremental effects are smaller than the baseline compliance attributable to respect for rules and social expectations. These results help explain pervasive adherence to laws and social norms, including in contexts where extrinsic incentives are weak or absent, and highlight the importance of managing social expectations and visible compliance in sustaining rule-following.

The paper introduces CRISP, an integrative framework for understanding rule-conformity as a function of intrinsic respect for rules, extrinsic incentives, social expectations, and social preferences, and provides extensive experimental evidence for each channel. The key contribution is demonstrating substantial unconditional rule-following and strong conditional conformity to social expectations, even when sanctions and externalities are absent, indicating that respect for rules and expectations are foundational to social order. Policy and research implications include leveraging social expectations, visibility of compliance, and targeted incentives to sustain rule-following. Future research directions: test CRISP across diverse rules and contexts (including tasks not hinging on patience), cross-cultural generalizability and ecological drivers, within-subject designs to compare motives at the individual level, dynamic multi-shot environments to study fragility and tipping points, and the roles of rule-setting by authorities versus emergent norms.

The study uses highly stylized tasks (traffic light and abstract tasks) that demand patience and may evoke real-world associations; quantitative effects may differ for other rule types (e.g., ambiguity, observability). Samples are primarily MTurk US workers and one university student pool, limiting generalizability across populations and cultures. Between-subjects designs do not identify the relative importance of CRISP motives at the individual level. Most decisions are single-shot without repeated feedback on others’ behaviour, limiting insights into dynamic interactions and norm evolution. The rule is set by the experimenter (potential experimenter-demand effects), although this is central to the phenomenon studied; nevertheless, real-world rules vary in origin (authority-set vs. emergent), warranting further investigation.