Economic theory has long relied on two great regulators of behavior: the price system and the state. Yet much of human conduct is governed by neither. Think of tipping in restaurants we will never revisit, returning lost wallets, queueing without supervision, or contributing to public goods that would survive our defection. This conduct is governed by social norms: shared expectations about behavior, sustained without central authority and enforced by the very people they constrain.

For decades, norms sat awkwardly at the margins of economic analysis. They were treated as residual noise or folded into an undifferentiated taste parameter. Experimental economics changed that. Laboratory work by Ernst Fehr and colleagues, alongside field experiments and neuroimaging studies, has revealed norms as structured mechanisms with identifiable components. These include expectations about what others do, beliefs about what others approve, and a willingness to sanction deviation even at personal cost.

For anyone designing policy or institutions, this matters enormously. Norms are not a fixed backdrop against which interventions operate. They are endogenous variables that interventions reshape, sometimes amplifying the intended effect and sometimes reversing it. Understanding that responsiveness requires three distinctions. The first is between norms that describe behavior and norms that prescribe it. The second is between the benefits of enforcement and the people who pay for it. The third is between law that crowds norms out and law that pulls them forward.

Robert Cialdini and colleagues formalised a distinction that remains among the most consequential in applied behavioral science. Descriptive norms convey what most people do. Injunctive norms convey what most people approve or disapprove of. Cristina Bicchieri's framework maps closely onto this: empirical expectations about others' behavior versus normative expectations about what others believe one ought to do. In daily life the two usually align, which is precisely why their divergence is so instructive.

Cialdini's littering experiments made the divergence visible. Participants were far more likely to litter in an environment already strewn with rubbish than in a clean one. Watching a confederate drop litter into a clean environment actually reduced littering, because the single violation made the prevailing norm of cleanliness salient. Behavior responded less to the environment itself than to whichever norm the environment brought into focus.

This creates a recurring trap for well-intentioned communication. Campaigns that dramatise the prevalence of a problem, such as “many visitors remove petrified wood” or “most students drink heavily,” wrap an injunctive message of disapproval inside a descriptive message of normality. In the Petrified Forest field study, signage emphasising widespread theft increased theft relative to signage focused on disapproval of individual thieves. The descriptive signal overwhelmed the injunctive one.

The energy study by Schultz and colleagues exposes the mechanics with particular precision. Households told how their consumption compared with their neighbors' converged toward the mean. Heavy users cut back, but light users increased consumption, producing a boomerang effect. Adding a simple injunctive cue, a smiling or frowning emoticon, eliminated the boomerang. Light users received approval for their below-average consumption, which separated the descriptive anchor from any implied permission to drift upward.

Neuroeconomic evidence suggests why descriptive information is so potent. Klucharev and colleagues found that discovering one's judgment diverged from group consensus produced signals in the rostral cingulate zone and ventral striatum resembling reinforcement-learning prediction errors, and these signals predicted subsequent conformity. Deviation appears to be registered as an error to be corrected. Injunctive norms, by contrast, typically require activation through salience. Once activated, however, they generalise across contexts more robustly, reducing littering in both clean and dirty environments.

Takeaway

Every statistic about how common a behavior is doubles as a signal about whether it is acceptable. Designers must decide which message they are actually sending.

A norm without enforcement is merely a regularity, vulnerable to anyone who calculates that deviation pays. Yet most norms lack a police force. They are sustained instead by decentralised sanctions such as disapproval, exclusion, gossip, and direct punishment. These sanctions are delivered by peers who gain little personally from administering them. Explaining why individuals bear these costs became one of the central problems of experimental economics.

Fehr and Gächter's public goods experiments supplied a striking answer. In repeated games without sanctioning options, contributions decay steadily toward free-riding as conditional cooperators withdraw in response to defectors. When participants can pay to reduce others' earnings, contributions rise and stay high, even in stranger-matching designs where punishers cannot reap future benefits. Negative emotions toward free-riders predicted punishment intensity, pointing to altruistic punishment as a proximate mechanism.

Enforcement extends beyond injured parties. Fehr and Fischbacher showed that third parties, unaffected by a dictator's allocation, will sacrifice their own earnings to punish unfair splits. De Quervain and colleagues found that effective punishment activates the dorsal striatum, a region associated with reward anticipation. Stronger striatal activation predicted willingness to incur higher punishment costs. Enforcement appears to be partly self-rewarding, which helps explain how a costly act persists without external compensation.

The distribution of costs remains the system's vulnerability. The benefits of enforcement are diffuse and shared by the whole group, while the costs fall on whoever chooses to act. This creates a second-order free-rider problem. Punishment is not always prosocial, either. Herrmann, Thöni, and Gächter documented substantial antisocial punishment across sixteen societies, in which low contributors retaliated against cooperators. It was most prevalent where norms of civic cooperation and the rule of law were weak, and it eroded the gains from cooperation.

Institutional designers have therefore sought to reduce and channel enforcement costs. Gossip lowers them dramatically, because reputational information spreads at minimal cost to the sender. Experiments show that the threat of being discussed and excluded sustains cooperation. Elinor Ostrom's field research identified graduated sanctions and accountable monitors as hallmarks of durable commons governance. When given the choice, participants in Gürerk, Irlenbusch, and Rockenbach's experiment migrated progressively toward sanctioning institutions. People value enforcement even when they hesitate to provide it.

Takeaway

Norms are only as stable as ordinary people's willingness to enforce them. Good design asks not just whether sanctions exist, but who pays for them and whether they can be turned against cooperators.

The interaction between formal law and informal norms is where behavioral insight most directly confronts policy design. Standard economic reasoning treats sanctions as additive. A fine raises the cost of an undesired behavior, so the behavior declines. The behavioral evidence shows that sanctions also carry information. That information can either reinforce the moral meaning of an act or quietly rewrite it.

Gneezy and Rustichini's study of Israeli daycare centers remains the canonical demonstration. Introducing a modest fine for late pickup increased lateness, and lateness stayed elevated after the fine was removed. The fine turned an obligation governed by reciprocity toward teachers into a market transaction with a posted price. Once reframed, the norm did not spontaneously recover. Crowding out can be durable.

Fehr and Rockenbach found a related effect in trust games. When investors could threaten sanctions against untrustworthy trustees, those who used the threat received less reciprocity than those who explicitly gave it up. The sanction signaled distrust, and trustees responded in kind. Samuel Bowles has generalised the lesson: policies designed for self-interested citizens can undermine the moral sentiments on which cooperation depends. Incentives and norms then become substitutes rather than complements.

Yet law can also strengthen norms through its expressive function. Legal scholars such as Robert Cooter and Richard McAdams argue that law coordinates beliefs by publicly signaling what society approves. This shifts perceived injunctive norms even when enforcement is sparse. Smoking bans achieved high compliance with minimal policing, partly because they licensed nonsmokers to voice objections, effectively recruiting decentralised enforcers. Tankard and Paluck found that the U.S. Supreme Court's same-sex marriage ruling shifted perceived social norms even where personal attitudes had not yet moved.

The decisive variable appears to be the gap between law and prevailing norms. Dan Kahan's analysis of sticky norms shows that laws condemning behavior far more harshly than public sentiment supports provoke resistance from the police, prosecutors, and juries who must apply them. That resistance can entrench the very norm the law sought to change. Gentle nudges, meaning incremental increases in condemnation, let enforcers and citizens adjust together. Law leverages norms best when it frames compliance as moral rather than transactional and moves a step ahead of opinion rather than a leap.

Takeaway

A sanction is also a message about what kind of behavior this is. Policies succeed when that message reinforces the moral meaning of compliance rather than putting a price on violation.

Social norms are best understood not as a soft supplement to markets and law but as a regulatory system in their own right. They have their own information channels, enforcement technology, and failure modes. Descriptive and injunctive signals shape what people believe is expected. Decentralised punishment determines whether those beliefs hold. Formal institutions determine whether the system is reinforced or displaced.

For policy design, the implication is that every intervention operates on two margins at once. It changes material incentives, and it changes the social meaning of behavior. Interventions that ignore the second margin risk boomerang effects, enforcement collapse, or durable crowding out. Standard cost-benefit models do not capture these outcomes.

The most effective choice architecture therefore treats norms as levers to be engineered with care. That means pairing descriptive feedback with injunctive cues, lowering the cost of prosocial enforcement, and pacing legal change so that it pulls norms forward rather than provoking resistance.