Crime, Punishment, and Social Order

One Harm, Two Legal Systems

During an argument outside a restaurant, one person deliberately punches another and breaks his jaw. The victim misses work, incurs medical bills, and suffers pain. He may bring a civil claim seeking compensation. But the assault can also lead to a criminal prosecution brought by the government, even if the victim would prefer to drop the matter.

Why use two legal systems for one event?

The civil case resembles the tort system studied in Chapter 7. The injured person controls the claim, usually faces a civil standard of proof, and seeks damages. The criminal case uses public prosecutors, a higher standard of proof, punishment, stigma, and the coercive authority of the state. A civil settlement can ordinarily end the victim’s claim. It does not necessarily end the government’s prosecution.

The difference cannot be that torts cause harm while crimes do not. Both do. Nor can the boundary be reduced to intent. Many crimes require a culpable mental state, but criminal law also reaches reckless conduct and some regulatory offenses. Civil law sometimes addresses intentional torts such as assault, fraud, and conversion. The same conduct can produce both civil and criminal proceedings.

Crime and tort are better understood as institutional bundles. Each bundle combines rules about who initiates enforcement, who pays for it, what information can be compelled, what must be proved, who controls settlement, what sanction follows, and who receives any payment. The bundles overlap because no single enforcement institution handles every harmful act well.

Chapter 10 asked whether common-law rules tend to reduce the number and cost of future injuries rather than merely determine whose ox was gored in a dispute that already occurred. The economic analysis of crime makes the same forward-looking move. It asks how legal rules change future offenses and what society must spend to obtain that change.

This perspective differs sharply from a view of criminal law concerned only with identifying guilt and imposing deserved punishment. Economics does not prove that justice, proportionality, rights, or condemnation are irrelevant. Those values help determine which policies are legitimate. Economic analysis adds another set of questions:

• How does the probability of punishment affect behavior?
• How does sanction severity affect behavior?
• Which stage of enforcement is the weak link?
• What does it cost to prevent another offense?
• Which sanctions transfer resources, and which destroy resources?
• How do false convictions and false acquittals affect social welfare?
• What does imprisonment accomplish through deterrence, incapacitation, or rehabilitation?
• Why does enforcement depend on victims, witnesses, communities, firms, and platforms as well as police and courts?

The answers begin with a simple model. The model is powerful because it forces the analyst to think forward and at the margin. It is limited because real people, crimes, and institutions are more complicated than one equation.

Modern criminal law is largely statutory, even in countries whose private law developed through common-law decisions. Legislatures define offenses and authorized sanctions; agencies, police, prosecutors, juries, judges, corrections systems, and appellate courts implement those rules. Exact doctrines vary across jurisdictions. This chapter therefore stays at the principles level. Its object is not to teach a criminal code. It is to show how the economic framework organizes recurring problems of offending, enforcement, and punishment.

The Economic Choice to Offend

Gary Becker’s famous 1968 article treated crime as a subject for ordinary economic reasoning. The move was provocative. It did not assume that offenders are admirable or that all crimes are carefully planned. It asked whether choices to violate law respond to expected benefits and costs.

Consider a person deciding whether to steal a television worth $1,000. If the only consequence of being caught is returning the television, and only one thief in four is caught, theft remains attractive in the stripped model. The thief obtains $1,000 if not caught and merely gives back what was taken if caught. Returning stolen property is necessary, but it is not enough to make the expected gain negative when detection is uncertain.

A fuller choice includes several elements:

• $B$ is the perceived private benefit from the offense.
• $O$ is the lawful opportunity cost and other private cost of offending.
• $p$ is the perceived probability that a formal sanction is actually imposed.
• $S$ is the severity of the sanction if imposed.

The stripped model writes expected private gain as:

In words, subtract lawful opportunity cost and expected formal punishment from the perceived benefit. When $G$ is positive, the baseline predicts that offending is privately attractive. When $G$ is negative, the baseline predicts deterrence.

The symbols are not mysterious. They are a checklist. What can the offender gain? What legitimate opportunities would be sacrificed? What is the chance that punishment actually occurs? How severe would it be?

Opportunity cost matters because offending uses time, skill, and access that could be used elsewhere. A person with attractive lawful work may give up more by risking a job, license, reputation, or future earnings. A person facing weak lawful opportunities may give up less. This does not imply that poverty mechanically causes crime or excuses it. It means that legal opportunities belong in a model of choice alongside police and punishment.

Benefits also need not be monetary. Anger, status, revenge, excitement, group approval, or saving time can enter the perceived payoff. Costs can include shame, family disapproval, expulsion, job loss, immigration consequences, and lost future relationships. Formal law operates inside a larger system of incentives.

The model does not claim that an offender opens a spreadsheet before acting. Consumers do not calculate demand curves before responding to a sale, and drivers do not solve equations before reacting to traffic. The model claims that behavior can respond systematically when expected costs or benefits change.

That prediction is about direction before magnitude. Chapter 2 called the relationship between expected punishment and offending the crime curve. Raising the expected cost of an offense should generally reduce its frequency, but responsiveness can be steep or flat. A planned tax-evasion scheme may respond strongly to audit probability and financial penalties. A violent act committed during panic, intoxication, or rage may respond less to a remote sentence increase. The empirical question is not merely whether incentives matter, but how much they matter for this offense, population, and institution.

The baseline must therefore be qualified:

• People may misunderstand the probability of punishment.
• A vivid arrest may receive more weight than a remote statistical risk.
• Present-biased people may heavily discount punishment that occurs months or years later.
• Risk attitudes differ.
• Addiction, emotion, impairment, or impulsivity may narrow attention.
• Social norms may make an act unthinkable for most people even without a formal sanction.
• Some offenders may value criminal status or accept risks that most people avoid.

These complications do not require abandoning economic analysis. They identify which assumptions need to be changed and which policy margins might matter.

The Enforcement Chain

The symbol $p$ can conceal most of the legal system.

An offense may have to be reported or detected. A suspect may have to be identified and apprehended. A prosecutor may have to bring a viable case. Evidence may have to satisfy the proof standard. A conviction may have to survive review. A fine may have to be collected or a custodial sentence actually imposed.

A simplified sequence is:

offense -> detection -> apprehension -> prosecution and conviction -> sanction imposed or collected

Suppose, only for illustration, that an offense has:

• a 50 percent conditional probability of detection
• an 80 percent probability of apprehension after detection
• a 75 percent probability of conviction after apprehension
• a 90 percent probability that the announced sanction is imposed or collected after conviction

The final probability is the product of the conditional stages:

Only 27 percent of offenses produce the realized sanction under these teaching assumptions. If the sanction is equivalent to $10,000, expected formal punishment is:

The expected formal punishment is $2,700. That is not a fine a court would impose. It is a probability-weighted measure used before uncertainty is resolved.

The numbers are invented to demonstrate the logic. Real stages are more complicated and often correlated. Strong evidence may raise both apprehension and conviction probabilities. Plea bargaining can replace trial. Charges and sanctions can vary. Victims may not report. Appeals can delay punishment. A fine can be partly collected. The model is still useful because it exposes weak links.

Suppose detection rises from 50 to 60 percent while every other stage remains unchanged. The final probability becomes 32.4 percent. By contrast, raising the final conditional stage from 90 to 100 percent produces a final probability of only 30 percent. Improving a weaker early stage can matter more than perfecting an already-strong late stage.

This is why “more police” and “longer sentences” are not interchangeable policies. Police deployment may change visibility, detection, and perceived certainty. Prosecutorial resources may change case selection and conviction. Court capacity may change delay. Corrections policy may change the sanction actually served. A statute announcing a severe sentence can have little effect if people do not expect to be caught or do not know about the change.

Research reviews generally find more consistent evidence that crime responds to the perceived certainty of apprehension and visible police presence than to increases in already substantial sentence severity. That conclusion does not mean severity never matters. It means the policy margin and the institution producing $p$ deserve at least as much attention as the number written in a sentencing statute.

How Much Deterrence?

If higher expected punishment reduces crime, why not keep raising it until crime disappears?

The answer begins with scarcity. Enforcement consumes labor, technology, court time, prison capacity, public attention, and trust. Punishment can destroy liberty, earnings, health, relationships, and human capital. Errors impose costs on innocent people and weaken legitimacy. Private precautions consume resources too. A society choosing crime-control policy faces trade-offs, not a free dial marked “deterrence.”

A simple social-loss framework includes:

• harm from offenses that occur
• private prevention and avoidance costs
• public detection and adjudication costs
• punishment and corrections costs
• error and abuse costs
• collateral effects on families, communities, firms, and future opportunities

The efficient policy minimizes the total, subject to legal and moral constraints. Preventing one more offense is worthwhile when the marginal social benefit of doing so exceeds the marginal social cost. That is the same marginal logic used for accident precaution, pollution control, and contract performance.

The phrase “optimal amount of crime” can sound callous. It does not mean murder or robbery becomes desirable. It means that eliminating the final offense may require unlimited surveillance, indiscriminate searches, enormous taxes, extreme punishment, or unacceptable restrictions on innocent conduct. A society can prefer zero murders as an outcome while rationally refusing methods whose additional costs and abuses exceed the harm they prevent.

Two distinct cases should not be confused.

First, some valuable activities create incidental risk. Driving produces mobility, employment, and exchange, but it also creates accident risk. The aim of a well-designed traffic sanction is not necessarily to eliminate driving, speeding by every amount, or every crash at any cost. It is to make drivers take more of the social cost into account when choosing speed, route, precaution, and whether to make a trip.

Second, some prohibited conduct has no comparable social benefit. Murder does not become socially valuable because the offender benefits. Yet achieving zero murders can still be prohibitively costly. The reason some offenses remain is then enforcement scarcity, not an efficient positive activity level.

Speeding illustrates the distinction. A sanction may operate partly like a price by making the driver bear more of the expected accident cost imposed on others. But a speeding ticket is not literally a market price purchasing permission to violate the law. Legal rules can also express categorical obligations, protect rights, simplify enforcement, and coordinate expectations. The economic analogy identifies an incentive effect; it does not erase the legal command.

Necessity supplies a different pressure test. Imagine a lost hiker facing a deadly winter storm who breaks into an empty cabin, uses a blanket, and calls for rescue. The entry is technically an invasion, but maximum deterrence would be perverse. The emergency benefit vastly exceeds the owner’s limited loss, and prior bargaining was impossible. A legal system can preserve the owner’s claim for damage while declining to treat the emergency like ordinary burglary. The example shows why legal categories need exceptions and why deterrence is not simply “make every violation prohibitively expensive.”

Marginal Deterrence

Deterrence operates across offenses as well as between offending and not offending.

Suppose armed robbery and armed robbery followed by killing a witness receive the same punishment. Once the robbery begins, the offender has no additional legal incentive to spare the witness. A sanction schedule should preserve a meaningful difference between lesser and greater harm.

This is marginal deterrence. The law should maintain incentives to stop, retreat, reduce harm, or choose the less serious offense after some violation has already occurred.

The idea applies beyond violent crime. A tax penalty that is already maximal for a small underreporting error can weaken the incentive to avoid larger evasion. A cybersecurity rule that imposes the same sanction for late disclosure and permanent concealment can weaken incentives to report after a violation. A firm that has already breached one requirement should still gain by cooperating, mitigating loss, and providing accurate information.

Marginal deterrence creates tension with a simple strategy of raising every penalty. If the maximum feasible punishment is used for minor offenses, little room remains to distinguish greater ones. Proportionality can therefore support deterrence as well as justice. A graduated sanction schedule tells people not only “do not offend” but also “if you have begun, do not make the harm worse.”

What Is the Sanction Supposed to Do?

Legal responses are often discussed as if every payment performs the same function. They do not.

Table 11.1. Different legal responses perform different functions. A single remedy can perform more than one function, but compensation, restitution, disgorgement, and punishment should not be treated as synonyms.

Suppose a thief breaks a $300 window and steals a laptop worth $1,000 to its owner. Returning the laptop does not repair the window. Paying $300 repairs the physical loss but may not remove every gain from theft. If only a small fraction of thieves are caught, making the caught thief return what was taken leaves a positive expected payoff. Deterrence may require a sanction beyond restitution or disgorgement.

The distinction also explains why civil and criminal cases can coexist. A victim’s compensation does not by itself answer the public question of deterrence, incapacitation, or condemnation. A criminal fine paid to the state does not by itself make the victim whole. Combining proceedings can serve multiple functions, although it also adds cost and creates coordination problems.

Fines, Imprisonment, and Punishment Cost

Economists often begin sanction design with a strong presumption in favor of fines. A collected fine harms the offender but transfers money to the government. Imprisonment harms the offender and requires society to spend additional resources operating a corrections system.

That comparison is useful but incomplete.

Table 11.2. Fines and imprisonment create different costs and capabilities. Neither sanction is universally superior. The comparison depends on collectability, risk, incapacitation, institutional quality, collateral effects, and the policy objective.

A fine is not costless. Collection and adjudication use resources. A fine can push a family into hardship, reduce lawful employment, or create incentives to hide income. An erroneous fine can cause losses before repayment. If enforcement agencies retain the proceeds, revenue can distort their priorities. Still, the transfer feature gives fines an important advantage over sanctions that destroy resources without producing an offsetting receipt.

Fines face a wealth constraint. An offender cannot pay unlimited amounts. Hidden assets and unstable income make collection difficult. Once the desired monetary sanction exceeds ability to pay, the legal system must accept weaker deterrence, use another sanction, improve collection, or combine sanctions.

Imprisonment escapes the narrow collectability limit because liberty can be taken from a person with little wealth. It also incapacitates. But confinement is expensive in a much broader sense. It requires public expenditure and imposes lost liberty, time, earnings, family contact, health risks, and post-release consequences. Those losses help make prison a punishment; they are also real social costs.

The correct comparison is therefore not “fines are cheap, prison is bad.” It is: which feasible sanction produces the desired deterrence, incapacitation, restitution, condemnation, and legitimacy at the lowest total cost and error risk?

Probability, Severity, and Risk

Becker’s stripped model creates a famous puzzle. Suppose a 20 percent chance of a $10,000 fine deters an offense. Why not use a 10 percent chance of a $20,000 fine? Expected monetary punishment remains $2,000, but the government needs to catch and process only half as many offenders. Repeating the logic seems to recommend an extremely severe penalty imposed with an extremely small probability.

The conclusion fails once the assumptions are relaxed.

First, offenders may be unable to pay the larger fine. Second, risk attitudes matter. Third, errors become more destructive as sanctions grow. Fourth, severe penalties may be disproportionate, unconstitutional, inhumane, or politically illegitimate. Fifth, people may misperceive very small probabilities. Sixth, rare punishment can weaken the visibility and credibility of enforcement. Seventh, punishment itself can create collateral harm.

Return to the risk example from Chapter 2. Compare:

• a $200 sanction imposed with certainty
• a one-in-one-thousand chance of a $200,000 sanction

Both have expected monetary value of $200:

Most risk-averse people would strongly prefer the certain $200 loss to the tiny chance of catastrophe. The rare severe sanction may therefore deter more than its expected monetary value suggests. But the additional deterrence comes partly from forcing people to bear risk. Risk is a social cost, not free punishment.

Do not assume every offender is risk-averse. Some may be risk-neutral or risk-seeking. Some may focus on the immediate benefit and heavily discount future punishment. Others may believe, accurately or not, that they are unusually unlikely to be caught. Policy must distinguish preferences over risk, perceptions of probability, and discounting of delay.

A sentencing lottery exposes the same problem without using money. Suppose 100 otherwise similar convicted offenders must collectively serve 100 years. One policy gives every offender a one-year sentence. Another randomly gives 90 offenders no prison and 10 offenders ten years each. The average sentence is one year under either policy, but the policies are not equivalent. The lottery concentrates risk and error, incapacitates a different group for much longer, and can undermine equal treatment and proportionality. Potential offenders may also respond differently to a small chance of a long sentence than to a certain short one. Holding average punishment constant does not hold the institutional consequences constant.

The theoretical probability-severity trade-off remains useful because it identifies a real allocation problem. Resources spent raising detection cannot be spent elsewhere. Severity can sometimes be increased at lower administrative cost. The optimal combination depends on behavioral response, punishment cost, wealth, error, proportionality, and credibility.

Should a Rich Driver Pay More?

Finland supplies a provocative sanction-design example. Current official materials distinguish fixed traffic penalty fees for lower-level violations from income-based day fines for specified traffic crimes, including sufficiently serious speeding. The number of day fines reflects the seriousness of the offense, while the monetary amount of each day fine depends on the driver’s income after specified adjustments.

At first, the policy appears to conflict with the externality framework. Holding speed, road, vehicle, traffic, and weather fixed, a wealthy driver’s car does not create more physical accident harm merely because the driver earns more. If the fine is only a Pigouvian price for expected external damage, scaling it by income is difficult to justify.

But that is only one possible objective.

Table 11.3. Income-based fines can pursue different objectives. Economic analysis cannot evaluate the rule until the analyst identifies whether the fine is functioning as an externality price, behavioral deterrent, proportionate punishment, revenue instrument, or some combination.

A fixed fine may barely affect a wealthy driver’s behavior while imposing severe hardship on a poorer driver. An income-based fine can seek a more equal deterrent effect or a more equal punitive burden. The wealthy driver may also place a higher money value on time saved, which changes the private benefit of speeding even though it does not change physical external harm.

Administration complicates the analysis. Income must be measured and updated. People with volatile earnings, assets but little current income, foreign income, or family obligations can be difficult to classify. Very large fines can raise proportionality and equality concerns of their own. Enforcement priorities can change if fines produce revenue.

The correct lesson is not that income-based fines are plainly efficient or plainly irrational. It is that “economic logic” does not select a fine until the analyst identifies the objective, behavioral margin, information requirement, and realistic alternative.

Intent, Attempts, and Outcome Luck

Criminal law often cares about mental state in ways that ordinary accident liability does not. A careful driver, a negligent driver, a reckless driver, and a person who deliberately uses a car to injure someone may cause the same physical loss. Their behavior creates different concerns.

Intent can matter economically for several reasons.

An intentional offender may conceal the act, choose a vulnerable victim, disable precautions, destroy evidence, or repeat the conduct when another opportunity appears. The act may have an obvious market substitute. A person who values someone else’s laptop can offer to buy it; theft bypasses consent and triggers defensive expenditures. Intent can therefore signal that simple compensation will not create adequate deterrence.

Intent also affects information and condemnation. Deliberate harm can reveal something useful to future employers, trading partners, neighbors, or victims. Criminal conviction and stigma transmit information, although stigma can overshoot, persist after formal punishment ends, and burden reintegration. Mental state also matters to proportionality: accidental, negligent, reckless, knowing, and purposeful conduct need not deserve identical sanctions.

The economic model should not pretend these distinctions are easy to observe. Intent is private information. Legal systems infer it from actions, statements, preparation, knowledge, and surrounding facts. More demanding mental-state requirements can protect innocent or careless actors from severe punishment, but they can also make intentional wrongdoing harder to prove.

Attempts reveal another role for intent. Suppose a person deliberately shoots at another person and misses. Ex post, the intended victim is uninjured. If punishment depended only on realized physical harm, the failed shooter might face little sanction while an otherwise identical shooter whose bullet happened to strike would face severe punishment.

That difference would make liability depend heavily on luck.

From an ex ante perspective, both shots created a serious risk of death. Punishing the attempt changes incentives before the next trigger is pulled. It treats intentional dangerous conduct as a proper object of deterrence even when chance prevents the intended result.

The argument does not imply that thoughts should be criminal. Legal tests require conduct beyond mere imagination or remote preparation, and the exact threshold varies across jurisdictions. The principles-level point is narrower: attempt liability can respond to dangerous choice and intent rather than waiting for the worst possible outcome.

The distinction connects crime to Chapter 7. Speed limits, drunk-driving rules, and safety requirements regulate risky conduct before an accident. Attempt law can operate similarly, although the conduct is often discovered only after the attempt fails. Ex ante and ex post rules can be combined because they use different information and reach different margins.

Proof, Error, and State Power

Criminal punishment is coercive, stigmatizing, and often difficult to reverse. A false conviction can take liberty, employment, reputation, family contact, and sometimes life. A false acquittal can leave harm unpunished, reduce deterrence, expose future victims, and weaken confidence in law.

The criminal proof standard allocates these error risks. Raising the standard reduces false convictions but can increase false acquittals. Lowering it can make enforcement easier while increasing the chance that innocent people are punished.

This is not a formula for converting “beyond a reasonable doubt” into a numerical probability. Chapter 3 warned against that shortcut. Evidence differs across cases, and legal standards communicate institutional commitments rather than actuarial thresholds alone.

Error cost also interacts with sanction severity. The expected harm from a false conviction rises when the punishment becomes more destructive. A rare catastrophic sanction does not merely create more risk for guilty offenders. It makes mistakes more costly and can increase the incentive to fabricate evidence, coerce pleas, corrupt officials, or use prosecution strategically.

The state possesses powers that private plaintiffs generally do not: investigation, arrest, compulsory process, confinement, and force. Those powers can solve serious enforcement failures. They also create a reason for procedural constraint. An efficient enforcement system is not one that maximizes convictions. It is one that reduces the total costs of offending, enforcement, punishment, and error while remaining within rights and legitimacy constraints.

Valuing Reductions in Death Risk

Crime policy sometimes asks how much society should spend to reduce small mortality risks. Should a city add officers at dangerous locations? Should a corrections system invest in violence prevention? Should a road rule reduce fatal-crash risk if compliance and enforcement are costly?

Economists use value of statistical life reasoning to discipline such choices. The term is easily misunderstood. It is not the price of an identified person’s life.

Suppose 10,000 people would each pay $100 for a policy that reduces each person’s annual probability of death by one in 10,000. Across the group, the expected reduction equals one statistical death. Aggregate willingness to pay is $1 million. Under the invented assumptions, the implied value of a statistical life is $1 million.

The calculation aggregates willingness to pay for small risk changes. It does not say that an identifiable person can be sacrificed for $1 million. Actual estimates raise difficult questions about income, information, involuntary risk, distribution, age, baseline risk, and whose preferences count.

For this chapter, the concept has one limited role. Crime prevention uses scarce resources. Avoided deaths and injuries must count as benefits even though they do not arrive with market prices. Risk valuation makes the comparison explicit without pretending to settle every moral or distributional issue.

Does Prison Work? First Define “Work.”

Few policy questions generate more slogans than imprisonment. One side says prison works because a confined offender cannot victimize the public and severe punishment deters. Another says prison does not work because recidivism remains high, long sentences have weak deterrent effects, and confinement damages future opportunities.

Both claims are incomplete. Prison can affect crime through several different channels.

Table 11.4. Four ways prison can affect crime. The mechanisms operate on different people and time horizons. A study of one channel does not automatically establish the others.

General deterrence asks how the threat of imprisonment affects people who have not yet offended. A sentence enhancement may reduce current offenses before the newly lengthened terms have kept anyone confined longer. That timing can help identify deterrence.

Specific deterrence asks how experiencing punishment changes the same person’s later behavior. Prison could make future punishment more salient and discourage reoffending. It could also weaken lawful employment, expose people to criminal networks, damage health, or make another prison term less frightening. The sign is an empirical question.

Incapacitation asks what offenses do not occur outside prison because the person is confined. It does not require the prisoner to change preferences. Its size depends on what the person would have done under the realistic alternative: immediate release, probation, electronic monitoring, treatment, or a shorter sentence. A high-rate repeat offender can generate a large incapacitation benefit. Keeping an aging low-risk prisoner for one additional year may generate little.

Rehabilitation asks whether incarceration changes post-release behavior by improving skills, treating addiction or mental illness, creating work habits, or otherwise expanding lawful opportunities. Rehabilitation is not synonymous with imprisonment. It depends on what happens inside the institution and during reentry.

Prison can also express condemnation, impose deserved punishment, and reassure the public. Those may be legitimate purposes, but they should not be mislabeled as evidence that crime fell. Conversely, a high recidivism rate does not prove that incarceration produced no incapacitation while the person was confined.

What the Evidence Can and Cannot Say

Empirical work on incarceration faces a basic identification problem. Places with more crime often imprison more people. A simple correlation between prison populations and crime rates cannot reveal whether prison reduced crime, crime increased prison, or another force changed both.

Researchers therefore look for comparisons that alter incarceration or sentence severity for reasons less directly connected to current offending: court orders, policy thresholds, staggered adoption, differences in judge assignment, or targeted rules. Each design estimates a particular margin.

Table 11.5. Selected incarceration studies estimate different margins. The rows should not be treated as competing estimates of one universal prison effect. They study different people, institutions, policies, outcomes, and alternatives.

The early DiIulio-Piehl estimate and Levitt’s overcrowding study make an important point that students sometimes do not hear: incarcerating active offenders can prevent a substantial number of offenses. It is too strong to say that prison does not reduce crime.

The later studies deepen rather than erase that conclusion. Kessler and Levitt used timing to distinguish immediate deterrence from incapacitation that appears later. Vollaard found large benefits from targeting prolific offenders but also found diminishing returns. Bhuller and coauthors found that a Norwegian prison system emphasizing rehabilitation reduced reoffending for a significant group, demonstrating that institutional design is part of the treatment.

The counterevidence matters too. Lee and McCrary found little behavioral change when Florida offenders crossed into much harsher adult sanctions at age eighteen. The National Research Council concluded that the large U.S. expansion of incarceration did not clearly produce comparably large crime reductions and found especially weak evidence for extending already-long sentences.

These findings can coexist. The first prison bed can be used for a high-rate dangerous offender; an additional bed in a much larger system may be used for someone with a lower expected offense rate. The first years of confinement may cover a high-risk period; later years may hold a person after age and circumstances have reduced offending. A targeted rehabilitative institution differs from an overcrowded institution with little programming. A threat to people still deciding whether to offend differs from holding someone who would already be imprisoned.

The economic question is always marginal:

What changes if this person is incarcerated, released, diverted, or held one year longer under this institution and this realistic alternative?

The benefits include offenses prevented through all relevant channels. The costs include facilities, staff, lost liberty, foregone work, family effects, health, error, prison crime, and post-release consequences. The estimate must identify the population, policy change, institution, and time horizon before it can guide policy.

The balanced conclusion is not indecision. Prison plainly can prevent crime, and the benefits can be large for some offenders and margins. It does not follow that every incarceration, every sentence extension, or every prison system passes a benefit-cost test. Both statements are necessary for serious analysis.

Why Enforcement Becomes Public

If a harmful act creates a victim with a legal claim, why not let the victim enforce it as a tort?

Private enforcement can work well when the victim is identifiable, the claim is large enough to finance litigation, evidence is accessible, the defendant can pay, and a civil remedy can create adequate deterrence. Chapter 9 showed why those conditions often fail.

Some victims are dead, missing, frightened, or unable to finance prosecution. Some offenses produce diffuse harm or many small claims. Some offenders are judgment-proof. Some crimes are deliberately concealed. Investigating them may require warrants, forensic capacity, cross-jurisdiction cooperation, or coercive authority unavailable to private parties. A victim may accept a settlement that leaves other potential victims exposed. A civil payment may be too small to deter when detection is rare.

Public prosecution pools resources and can pursue sanctions independent of a victim’s wealth or recovery. It can also produce public information and condemnation. But public enforcement creates its own agency problems. Officials allocate limited budgets, choose cases, bargain over charges, and respond to career, political, organizational, and revenue incentives. Victims can lose control. Public power can be abused.

The tort-crime boundary therefore compares imperfect institutions. Public enforcement becomes more attractive when private incentives underproduce investigation or deterrence. Private enforcement becomes more attractive when victims have strong information and incentives and when state intervention would be slow, politicized, or excessive. Many legal systems combine the two.

Tax evasion illustrates the public rationale. The loss is spread across taxpayers and public services. An individual taxpayer has little incentive or authority to investigate another person’s return. The state can aggregate the public stake, audit records, and impose sanctions. Yet audit selection, privacy, administrative burden, settlement, and enforcement equality remain institutional questions.

Private Precaution, Displacement, and Cooperation

Public law is only one input into social order. Potential victims install locks, alarms, cameras, passwords, fraud filters, and cybersecurity systems. Firms screen employees and transactions. Neighbors report suspicious activity. Witnesses cooperate. Communities transmit norms. Insurers and platforms impose private conditions.

Private precautions can have three different effects.

A visible deadbolt may redirect a burglar toward another house. It creates crime displacement rather than reducing total crime. A loud alarm may protect one house and increase the chance that a burglar is caught, deterring offenses against nearby houses as well. It creates a positive spillover. An unobservable interior lock may change the expected payoff after entry without warning the offender beforehand. Its effect differs again.

The distinction matters for policy. A homeowner chooses precaution based primarily on personal benefit. The homeowner may invest too much in a device that shifts crime to neighbors or too little in a device that produces public deterrence. Subsidies, building codes, information sharing, policing, or collective arrangements can address those spillovers, but they bring administrative and privacy costs.

Social order also depends on cooperation. Police solve fewer cases when victims do not report, witnesses expect retaliation, or communities distrust officials. Prosecutors need credible testimony. Courts need jurors. Corrections systems need lawful administration. Enforcement probability is partly produced by ordinary citizens.

Cooperation can display multiple equilibria. If people expect others to report and expect officials to respond fairly, cooperation may be individually sensible and crime less attractive. If people expect silence, retaliation, corruption, or indifference, reporting can be dangerous and ineffective. The same formal law can produce different outcomes under different expectations.

This creates a deeper reason legitimacy matters. Fair and competent enforcement can increase cooperation, which raises detection and deterrence. Abusive or arbitrary enforcement can destroy information and trust, reducing effective capacity even when formal powers expand. Legitimacy is not merely an external value added after an efficiency calculation; it can be one of the inputs into successful enforcement.

Coercion, Capacity, and Constraint

Modern states generally claim authority to use legitimate coercion: arrest, detention, compulsory process, and punishment. Concentrating this power can prevent private retaliation, vendettas, and competing systems of force. It can make rights enforceable against powerful offenders and provide a common process for disputes that private bargaining cannot resolve.

The same concentration creates danger. A weak state may be unable to protect rights. An unconstrained state may violate them. Effective criminal law therefore requires both capacity and constraint.

Capacity includes trained personnel, information, credible courts, corrections resources, cross-jurisdiction cooperation, and the ability to carry out lawful sanctions. Constraint includes defined offenses, proof requirements, defense rights, review, transparency, proportionality, independent adjudication, and political accountability.

The law-and-economics comparison is not state versus no state in the abstract. It is among realistic enforcement arrangements. A private platform may act quickly but lack independent review. A local police force may have information but face conflicts. A national agency may have technical capacity but weak local knowledge. Courts may protect procedure but act slowly. Institutions can complement one another.

Digital Crime and Layered Enforcement

Consider a ransomware attack on a regional business. An attacker gains access through a stolen password, encrypts operating records, and demands payment through a digital channel. The event brings the chapter’s mechanisms together.

The business can invest in private precaution: authentication, backups, software updates, employee training, network separation, and incident response. Some precautions reduce total offending by making attacks less profitable. Others redirect attackers toward weaker targets. Security information shared across firms can create public benefits, while secrecy can protect reputation and sensitive systems.

A platform or payment intermediary may observe accounts, transactions, device identifiers, or patterns that no individual victim can see. It can freeze transfers, suspend accounts, preserve evidence, and impose access sanctions quickly. Those capabilities lower enforcement cost. They also concentrate rulemaking, investigation, adjudication, and punishment in one private institution.

Public authorities can aggregate reports across victims, compel information under law, coordinate across jurisdictions, and impose criminal sanctions. Yet attribution can be difficult, offenders can operate abroad, and investigation can require technical expertise. Victims may avoid reporting because they fear disruption, liability, or reputational loss.

Automation can improve detection by screening large volumes of activity. False positives can also freeze innocent users, misdirect investigators, or burden groups whose conduct resembles a risk pattern. More sensitive detection can reduce false negatives while increasing false positives. Proof and review remain necessary even when an algorithm produces a confident score.

The right institutional response is layered:

• users and firms take precautions based on information close to the activity
• platforms use transaction-specific data and technical control
• insurers and vendors price and monitor risk
• public agencies investigate organized and cross-border offenses
• courts authorize coercive measures and review disputed sanctions
• appeal and correction procedures address error

No layer is sufficient by itself. Private systems can move quickly but may externalize harm or lack legitimacy. Public systems possess coercive authority but may lack speed or technical information. Automation scales detection and error together. Digital crime changes tools and geography, but it does not replace the chapter’s economic logic.

The recurring questions remain: Who has information? Who has incentives to act? Who bears the cost of error? What sanction is feasible? Can the decision be reviewed? Which form of enforcement reduces total harm without creating greater abuse?

Big Picture

Crime and tort are overlapping institutional bundles. Tort usually relies on private claims and compensation. Criminal law uses public prosecution, heightened proof, punishment, stigma, and coercion. The boundary reflects enforcement problems, not merely a clean moral classification of harmful acts.

Becker’s model begins with expected choice. A potential offender compares perceived benefit, lawful opportunity cost, formal and informal consequences, and the probability and severity of punishment. The model does not require perfect calculation. It identifies the margins to investigate.

Probability is produced by an enforcement chain. Detection, apprehension, prosecution, conviction, and realized sanction are distinct stages. A weak link can make an announced punishment largely irrelevant. Certainty, delay, information, and institutional capacity therefore matter alongside severity.

Optimal enforcement does not mean maximum enforcement. Crime causes harm, but prevention, punishment, error, and surveillance also cost resources and can violate rights. Internalizing the cost of valuable risky activity differs from trying to eliminate conduct with no offsetting social value. Even then, eliminating the final offense can be excessively costly.

Marginal deterrence preserves incentives to avoid greater harm after a lesser violation has occurred. Compensation, restitution, disgorgement, fines, and imprisonment perform different functions. Fines often conserve resources but face wealth and collection limits. Imprisonment is costly but can incapacitate and sometimes rehabilitate.

Attempts show why law responds to ex ante danger rather than realized harm alone. Proof standards allocate error risk under coercive state power. Value-of-statistical-life reasoning helps compare small mortality-risk reductions without pricing an identified person.

Prison affects crime through general deterrence, specific deterrence, incapacitation, and rehabilitation. Evidence shows that incarceration can prevent substantial crime at some margins. It also shows heterogeneity, diminishing returns, institutional dependence, and weak deterrence from some sentence extensions. The correct question concerns the marginal offender, sentence, institution, and alternative.

Public enforcement addresses diffuse harm, weak private claims, difficult investigation, and judgment-proof offenders. It depends on private precautions, reporting, civic cooperation, and legitimacy. Digital systems add platforms, automation, and cross-border problems, but the underlying framework survives.

Criminal law is ultimately a technology for social order. It changes expected payoffs, produces public information, coordinates cooperation, and concentrates coercive authority. Its success depends not only on the power to punish, but also on information, restraint, error correction, and public trust.

Chapter Study Map

• Core ideas: crime/tort institutional bundles, Becker’s choice model, lawful opportunity cost, expected punishment, enforcement chain, optimal enforcement, internalization versus elimination, marginal deterrence, punishment cost, fines, imprisonment, day fines, intent, attempts, proof standards, value of statistical life, general and specific deterrence, incapacitation, rehabilitation, public enforcement, private precaution, displacement, cooperation, state capacity, and legitimate coercion.
• Equations and tables: explain $G = B - O - pS$ in words; calculate the overall enforcement probability from conditional stages; use Table 11.1 to distinguish legal responses; use Table 11.2 to compare fines and imprisonment; use Table 11.3 to separate day-fine objectives; use Tables 11.4 and 11.5 to distinguish prison mechanisms and evidence.
• Reasoning tasks: identify the relevant behavioral margin, distinguish probability from severity, locate an enforcement weak link, compare marginal benefit and cost, preserve marginal deterrence, separate sanction objectives, distinguish incarceration mechanisms, interpret an empirical design, and compare public with private enforcement.
• Common mistakes: assuming offenders must perform explicit calculations, treating an announced sentence as the probability of punishment, equating internalization with permission, assuming optimal enforcement means zero crime, treating fines as costless, treating every prison effect as deterrence, transporting one study across populations, or assuming automated enforcement eliminates error.
• Required applications: assault and the crime/tort overlap, theft and low detection, tax evasion, speeding, Finland’s day fines, missed-shot attempt, necessity, fines versus prison, incarceration evidence, locks and alarms, community reporting, and ransomware enforcement.
• Optional enrichment: impossible attempts, formal optimal-enforcement models, drugs, guns, capital punishment, historical crime trends, current crime statistics, a sentencing calculator, or a detailed jurisdictional survey.

Review Questions

1. Why can one assault produce both a tort claim and a criminal prosecution?
2. What does it mean to call tort and crime institutional bundles?
3. How does the forward-looking economic question differ from asking only who deserves punishment?
4. Define $B$, $O$, $p$, and $S$ in the Becker model.
5. Explain $G = B - O - pS$ entirely in words.
6. Why does lawful opportunity cost belong in a model of offending?
7. Give three nonmonetary benefits or costs that can affect an offense decision.
8. Why does the model not require offenders to perform explicit arithmetic?
9. Distinguish the direction of a deterrence response from its magnitude.
10. List the stages in the simplified enforcement chain.
11. Why can a weak link sharply reduce the final probability of punishment?
12. Why are longer sentences and higher detection probabilities not interchangeable policies?
13. Explain why optimal crime control does not necessarily mean eliminating every offense.
14. Distinguish internalizing the external cost of driving from trying to eliminate murder.
15. Why is a traffic fine not literally a price purchasing permission to speed?
16. How does necessity challenge a policy of maximum deterrence?
17. Define marginal deterrence and give an example.
18. Distinguish compensation, restitution, disgorgement, and punishment.
19. Why can returning stolen property fail to deter theft?
20. Why do economists often prefer fines to imprisonment in a stripped model?
21. Identify four limitations of fines.
22. Identify four social costs or capabilities of imprisonment that fines do not share.
23. Why can equal expected monetary penalties create different deterrence and welfare effects?
24. Separate risk aversion, probability misperception, and present bias.
25. What different objectives might justify or challenge Finland’s income-based day fines?
26. Why does attempt liability focus on ex ante danger rather than outcome alone?
27. Why are thoughts or remote preparation not equivalent to an attempt?
28. How do proof standards allocate the risks of false conviction and false acquittal?
29. Why should a verbal proof standard not be converted into an exact probability?
30. What does value-of-statistical-life reasoning measure?
31. Distinguish general deterrence, specific deterrence, incapacitation, and rehabilitation.
32. Why does a high recidivism rate not prove that prison produced no incapacitation?
33. Why do the incarceration studies in Table 11.5 not estimate one universal prison effect?
34. What is meant by the marginal prisoner or marginal sentence year?
35. Why can incarceration produce diminishing crime-reduction returns?
36. Identify five reasons private enforcement may underprovide deterrence.
37. Distinguish private deterrence, public deterrence, and crime displacement.
38. How can legitimacy affect the practical probability of enforcement?
39. Why does public coercive capacity require institutional constraint?
40. What information and enforcement advantages can a digital platform possess, and what risks accompany them?

Economic Reasoning Questions

1. A person expects a $4,000 benefit from an offense, gives up $600 in lawful earnings and other opportunities, faces a 20 percent chance of a $12,000-equivalent sanction, and expects $500 in informal and reputational costs. Calculate the expected private gain. State what the model predicts and identify two omitted considerations.
2. An enforcement system detects 40 percent of offenses, apprehends 75 percent of detected suspects, convicts 80 percent of apprehended suspects, and realizes the sanction in 90 percent of convictions. Calculate the final punishment probability. Which stage would you investigate first, and why is the answer not determined by arithmetic alone?
3. Using the previous probabilities and a $20,000 sanction, calculate expected formal punishment. Recalculate it if detection rises to 60 percent. Then recalculate the original case if sanction severity rises to $30,000. Compare the changes while identifying costs the calculation omits.
4. A city can spend $1 million on additional detection expected to prevent 100 burglaries, or use the same funds to lengthen sentences expected to prevent 30 burglaries through deterrence and incapacitation. What information is needed before choosing? Include displacement, error, and collateral effects.
5. A fine is $500 for both shoplifting and armed robbery. Explain the marginal-deterrence problem and design a graduated schedule without assuming that severity is the only policy instrument.
6. A lost hiker breaks into an empty cabin during a deadly storm and causes $800 in damage. Compare compensation, restitution, disgorgement, and punishment. Which economic facts support a necessity exception?
7. Two drivers create the same expected external accident harm, but one earns ten times as much as the other. Analyze equal fixed fines and income-based fines under externality pricing, equal deterrence, proportional punishment, redistribution, and administration.
8. Compare a certain $500 fine with a 1 percent chance of a $50,000 fine. They have the same expected monetary value. Explain how risk attitudes, wealth, error, legitimacy, and probability perception can produce different conclusions.
9. A defendant cannot pay more than $5,000. Compare a fine, probation, electronic monitoring, community service, and imprisonment as possible sanctions. Identify deterrence, resource cost, collectability, incapacitation, and collateral effects.
10. Two people intentionally fire identical shots at intended victims. One misses; the other kills. Explain the ex ante and ex post differences. What arguments support punishing both, and what arguments support different sanctions?
11. A proposal lowers the criminal proof standard to increase conviction. Trace likely effects on false convictions, false acquittals, deterrence, cooperation, guilty pleas, legitimacy, and enforcement cost.
12. Ten thousand people would each pay $250 for a policy that reduces each person’s annual mortality risk by one in 10,000. Calculate aggregate willingness to pay and the implied value per statistical life. Explain what the result does not mean.
13. A study finds that adding one year to sentences reduces offending among people who have not yet been convicted. Which prison channel does it estimate? A second study finds fewer offenses only while people remain confined. Which channel does that estimate?
14. A prison education program reduces post-release offending and increases employment. Explain why comparing participants with nonparticipants may be misleading. What kind of research design would improve the comparison?
15. A targeted incarceration policy produces large initial crime reductions but smaller reductions as it expands. Explain diminishing returns using offender heterogeneity, age, offense rates, and alternative sanctions.
16. A homeowner chooses among visible window bars, a loud alarm, and a hidden interior lock. For each, explain whether it is most likely to deter total crime, protect only the homeowner, or displace crime. What evidence would distinguish the effects?
17. Residents stop reporting offenses because they expect retaliation and distrust officials. Use the enforcement chain and multiple-equilibrium logic to explain how crime can rise even though statutory sanctions remain unchanged.
18. A platform freezes transactions whenever an automated system assigns a fraud score above a threshold. Explain the trade-off between false positives and false negatives, identify relevant private information, and design a meaningful review process.

Law and Economics Lab

The Crime-Control Design and Evidence Audit

Choose one offense approved by your instructor, such as theft, tax evasion, reckless driving, fraud, assault, ransomware, or illegal dumping. Analyze one concrete enforcement change rather than discussing crime policy in general.

1. Define the conduct and institution. Identify the prohibited act, intended or likely harm, relevant jurisdiction, current enforcement institution, and whether civil, criminal, regulatory, platform, or mixed enforcement applies.
2. Build the choice model. Identify perceived benefit, lawful opportunity cost, formal punishment, informal consequences, risk, delay, information limits, and behavioral qualifications. Use invented teaching values only when clearly labeled.
3. Map the enforcement chain. Separate reporting or detection, identification, apprehension, prosecution, conviction, review, and realized sanction. State which probabilities are conditional and identify the likely weak link.
4. Specify the behavioral margin. Decide whether the proposed change affects participation in the offense, offense frequency, seriousness, victim choice, concealment, cooperation, precaution, or displacement.
5. Compare sanctions and institutions. Evaluate at least three feasible responses. Include a monetary response and a nonmonetary response. Compare deterrence, incapacitation, rehabilitation, compensation, collectability, administrative cost, error, proportionality, distribution, and legitimacy.
6. Preserve marginal deterrence. Explain how the rule rewards stopping, reporting, mitigating harm, cooperating, or choosing a less serious action after a violation begins.
7. Audit private and public enforcement. Identify victim incentives, private precautions, spillovers, reporting, public investigation, platform control, and realistic coordination problems.
8. Find evidence. Locate at least one primary or official source describing the rule and two credible empirical studies or evidence reviews relevant to the proposed change. Record population, period, comparison, outcome, mechanism, and limitations.
9. Separate prison channels when relevant. State whether each claim concerns general deterrence, specific deterrence, incapacitation, rehabilitation, or another social purpose. Reject evidence that does not identify the claimed channel.
10. Use AI adversarially. Ask an AI system to make the strongest economic case for the proposed change and then the strongest case against it. Require citations. Verify every legal and empirical claim against current primary, official, or scholarly sources. Identify invented authority, transported estimates, omitted alternatives, and hidden value judgments.
11. Recommend and qualify. Compare the proposed rule with the strongest realistic alternative. State expected benefits, enforcement and punishment costs, error risks, distributional effects, rights or legitimacy constraints, and the evidence that would change your conclusion.

Submit the enforcement-chain model, evidence table, corrected AI claims, and recommendation. The final analysis must distinguish a fluent argument from a verified one and must state whose behavior changes, through which mechanism, at which margin.