Criminology

The big picture

Ask ten people why crime happens and you will get ten confident answers: bad parenting, poverty, drugs, guns, video games, too few police, too many police, a decline in morals. Each answer arrives with total certainty and almost no evidence. Criminology is the discipline that refuses to stop there. It takes those claims, turns them into testable statements, and checks them against data.

That is the whole promise of this course. You are not going to memorize a list of theories. You are going to learn how to ask, of any claim about crime, three questions: what exactly is being asserted, what evidence would settle it, and how good is the evidence we actually have? Those questions will serve you long after the names in this course have faded.

A warning up front. Crime is one of the most politically loaded subjects in social science. People bring their politics to it, and researchers are people. The evidence in this field is often weaker than advocates on any side claim, and this course will tell you when that is true, including when the finding is one you might personally like.

Key idea: Criminology is the scientific study of crime, criminals, and society's response to them, and its central skill is judging the quality of evidence behind claims that people usually assert without any.

What criminology studies, and what it does not

The standard definition comes from Edwin Sutherland, one of the field's founders, who described criminology as the body of knowledge regarding crime as a social phenomenon, including the making of laws, the breaking of laws, and society's reaction to the breaking of laws. That three-part sentence is worth slowing down over, because each part is a separate research programme.

The making of laws asks why some harmful acts get criminalized and others do not. Why is selling a small amount of a plant a felony while a marketing campaign that kills far more people is a regulatory matter? That question is not rhetorical; it is empirical, and Module 4 takes it seriously.

The breaking of laws is the part most people mean by criminology: why individuals and groups offend, and why offending clusters in certain people, places, and times. Modules 2 through 5 live here.

Society's reaction covers punishment, policing, treatment, and prevention. Module 6 asks the hard question about that reaction: does it work, and how would we know?

Criminology is not the same as criminal justice. Criminal justice studies the institutions, the police departments, courts, and prisons, and how a case moves through them. Atlas Academy's CRIJ 1301 covers exactly that ground, and this course deliberately does not repeat it. Criminology is also not criminal law, which asks what the rules are, nor forensic science, which identifies who did it, nor true crime, which tells stories about particular horrors. Criminology asks about patterns.

Key idea: Sutherland's definition splits criminology into the making of laws, the breaking of laws, and the societal reaction, and the discipline is distinct from criminal justice, criminal law, forensics, and true crime.

Defining crime is harder than it looks

You cannot study crime without deciding what counts, and there are two main answers.

The legalistic definition says a crime is an act that violates the criminal law and is punishable by the state. It is clean, it is what official statistics measure, and it is what courts use. Its weakness is that it makes the object of study depend entirely on the legislature. When a state legalizes cannabis, crime falls by tens of thousands of offenses overnight without a single behavior changing.

The social harm definition says criminology should study serious avoidable harms whether or not they are illegal. On this view, a factory that knowingly exposes workers to a carcinogen belongs in a criminology course even if it is only fined. This definition captures more of what we intuitively care about but loses the crisp boundary; reasonable people disagree about which harms qualify.

Most working criminologists use the legalistic definition for data and the social harm critique as a running check on it. That is what this course does. When we report crime rates we mean legally defined crimes recorded by an agency. When we ask why corporate offending is punished lightly, we are using the harm lens.

Related and often confused is deviance, behavior that violates social norms. Deviance is broader than crime and only partly overlaps it. Talking loudly during a funeral is deviant, not criminal. Driving 80 in a 65 zone is criminal, and in many places barely deviant. The two categories drift apart and back together across time and place, which is itself a finding worth explaining rather than an inconvenience.

Key idea: The legalistic definition ties crime to what the law says, the social harm definition ties it to serious avoidable injury, and deviance, meaning norm violation, overlaps crime only partially.

Where the field came from

Criminology's history is short and instructive. In the eighteenth century, Cesare Beccaria and Jeremy Bentham argued that people choose crime by weighing costs and benefits, and that punishment should therefore be swift, certain, and no more severe than necessary. This is the classical school, and it is still the intellectual basis of deterrence policy.

In the nineteenth century, the positivist school replaced choice with causes, arguing that offending is determined by factors in the body, the mind, or the environment, and can be studied like any natural phenomenon. Cesare Lombroso is the emblem of this turn, and Lesson 5 will show you both why his method was a genuine step toward science and why his conclusions were badly wrong and did real harm.

In the early twentieth century, sociologists at the University of Chicago shifted attention from bodies to places, mapping delinquency across the city and finding that certain neighborhoods produced high rates decade after decade even as the ethnic groups living in them changed entirely. That finding launched most of modern criminology.

Since the 1960s the field has fragmented productively into competing traditions: learning, control, labeling, conflict, feminist, life-course, and biosocial. This course walks through them because each one earned its place by explaining something the others could not.

Key idea: Criminology moved from the classical emphasis on choice, to positivist searches for causes in the body, to the Chicago School's focus on places, and then into the plural set of competing traditions we have today.

Levels of explanation

Theories in this course answer different questions, and confusing them causes needless argument. It helps to keep three levels straight.

Level	Question it answers	Example theory
Individual	Why does this person offend and that one not?	Self-control theory
Situational	Why did this event happen here, now?	Routine activity theory
Structural	Why does this society or neighborhood have a high rate?	Social disorganization

A theory that explains individual differences well may say nothing about why a country's rate fell by half. A theory of neighborhoods may not predict which resident offends. When two criminologists appear to contradict each other, the first thing to check is whether they are even answering the same question.

The classic version of this mistake is the ecological fallacy: inferring something about individuals from a pattern across groups. Neighborhoods with more immigrants have, on average, lower crime rates in the United States. That is a fact about neighborhoods. It does not by itself tell you the offending rate of any individual, and treating it as if it did is an error whichever direction it points.

Key idea: Criminological explanations operate at individual, situational, and structural levels, and mixing them up, especially by reading group patterns as facts about individuals, produces the ecological fallacy.

Correlation, causation, and the habit of doubt

This is the most important section in the lesson. Nearly every strong claim you will hear about crime rests on a correlation, and correlations mislead in three specific ways.

Confounding means a third factor drives both. Ice cream sales correlate with assault. Neither causes the other; hot weather raises both. In criminology the confounders are rarely so obvious. Neighborhood poverty, family structure, schooling, and policing intensity all travel together, so a study that controls for only one of them may credit it with the whole effect.

Reverse causation means the arrow points the other way. Delinquent teenagers have delinquent friends. Does having those friends cause delinquency, or do delinquent teenagers seek each other out? Both are true, and untangling them took the field decades and better data.

Selection means the people in the two groups differed before the treatment. Compare people sent to prison with people given probation for the same charge and you are not comparing like with like; judges chose who went where, using information the researcher cannot see. Any difference in later offending may reflect the judge's judgment rather than the sentence.

Criminologists fight back with research designs. A randomized controlled trial assigns the intervention by lottery, which balances everything known and unknown; these are rare in criminology but they exist, and they are the strongest evidence in the field. Natural experiments exploit something arbitrary in the real world, such as a court ruling that abruptly changes a policy in one state but not its neighbor. Longitudinal studies follow the same people for years, so you can see which came first. Within-individual designs compare a person to himself at another time, which removes every stable difference between people at a stroke.

Learn to notice the design before the finding. A headline claim built on a cross-sectional survey of 300 students and a claim built on a national randomized trial deserve wildly different amounts of confidence, even when they agree.

Key idea: Confounding, reverse causation, and selection are the three ways correlational crime research misleads, and randomized trials, natural experiments, longitudinal data, and within-individual designs are the tools that push back.

Why crime research gets distorted

Even good studies get mangled on the way to the public. Four distortions recur so often that you should watch for them by reflex.

The first is percentage change on a small base. A town with two homicides last year and three this year has a 50 percent increase, and the number is meaningless noise. Small places and short periods produce dramatic percentages that mean nothing.

The second is selective baselines. Choose your starting year and you can show that American violent crime has soared or plummeted, both truthfully. The honest move is to show the long series and let the reader see the shape.

The third is counts versus rates. A growing city can have more crimes and less crime per person. Always ask for the rate, usually offenses per 100,000 residents.

The fourth is the availability heuristic. Vivid, rare crimes dominate coverage, so people systematically overestimate their frequency. Gallup polling has found for decades that a majority of Americans usually say crime is rising nationally even in years when both major measures show it falling. That gap between perception and measurement is itself a criminological topic.

Key idea: Percentage changes on small bases, cherry-picked baselines, counts reported instead of rates, and the vividness of rare crimes systematically distort public understanding of crime trends.

What criminology is good for

None of this is academic in the dismissive sense. Criminological research has changed what governments do. Randomized and quasi-experimental evaluations showed that Scared Straight programs, in which teenagers visit prisons to be frightened out of crime, made participants more likely to offend, not less; the finding survived multiple reviews and eventually changed policy in several jurisdictions. Careful place-based studies showed that concentrating patrol on the small number of street corners that generate most crime reduces offending without simply pushing it around the block.

Equally, criminology's job includes saying no. It has told legislatures that longer sentences buy little additional deterrence, that mass programs with warm names often do nothing, and that some popular reforms have effects too small to detect. A field that only confirmed what its audience wanted would be worthless.

Two public resources built for exactly this purpose appear throughout the course. CrimeSolutions, run by the National Institute of Justice, rates individual programs and practices as effective, promising, or showing no effects, based on the strength of the underlying evaluations. The Campbell Collaboration publishes systematic reviews in crime and justice using explicit, transparent methods. When you want to know whether something works, those are better starting points than any headline.

Key idea: Criminology earns its keep by testing popular programs, and its most valuable results are often negative findings that stop governments from spending money on interventions that do not work.

Common misconceptions

• Criminology is the study of catching criminals. That is forensic science and investigation; criminology studies patterns, causes, and responses.
• Crime is a fixed natural category. What counts as crime is set by law and shifts with legislation, which is why definitions matter before any counting begins.
• All crime is deviant and all deviance is criminal. The categories overlap only partly; speeding is criminal but normal, and many deviant acts are lawful.
• If two things move together, one causes the other. Confounding, reverse causation, and selection explain a great many crime correlations.
• Theories compete to be the single true explanation. They often answer different questions at different levels, so several can be right at once.

Recap

• Criminology is the scientific study of lawmaking, lawbreaking, and society's reaction, distinct from criminal justice and from criminal law.
• Crime can be defined legally, which is what statistics measure, or by social harm, which is a useful critical check; deviance is a broader, overlapping category.
• The field moved from classical choice theory, through positivist searches for causes, to the Chicago School's places, and then to competing modern traditions.
• Explanations sit at individual, situational, and structural levels, and reading group patterns as individual facts is the ecological fallacy.
• Confounding, reverse causation, and selection are the standard traps, and research design is how criminologists escape them.

Sources

1. Britannica. (n.d.). Criminology. Encyclopaedia Britannica. britannica.com
2. National Institute of Justice. (n.d.). CrimeSolutions. Office of Justice Programs. crimesolutions.ojp.gov
3. Campbell Collaboration. (n.d.). Crime and justice systematic reviews. campbellcollaboration.org
4. Bureau of Justice Statistics. (n.d.). Bureau of Justice Statistics. Office of Justice Programs. bjs.ojp.gov
5. Wikipedia contributors. (n.d.). Criminology. Wikipedia. en.wikipedia.org

The big picture

Here is a question with no honest single answer: how much crime happened in the United States last year? Not how much was reported. How much happened. Nobody knows, and the reason is not laziness. Crime is a private event with a strongly motivated participant who wants it unrecorded and a victim who often does not want to talk about it.

So criminologists triangulate. Three instruments point at the same underlying phenomenon from different angles. Police-based counts ask agencies what they recorded. Victimization surveys ask the public what happened to them. Self-report studies ask people what they have done. Each sees something the others miss, and each has a characteristic blind spot.

This lesson is about the instruments. It is dull only if you skip it, and then everything afterward becomes unreliable, because every theory in this course was tested against numbers produced by one of these three systems. If you do not know how the number was made, you cannot know what it means.

Key idea: There is no direct measurement of crime, only three imperfect instruments, and understanding how each is built is the precondition for interpreting any finding in criminology.

The dark figure of crime

The gap between crimes committed and crimes recorded is called the dark figure of crime. It is large, and its size varies enormously by offense.

National Crime Victimization Survey data have shown for decades that roughly 40 to 45 percent of violent victimizations are reported to police in a typical year, and only about a third of property victimizations. Within those totals the variation is extreme. Motor vehicle theft is reported at rates around 80 percent, largely because insurance requires a police report. Sexual assault reporting has run in the range of a quarter to a half depending on the year and the definition used. Simple theft of low-value items is often reported at rates well under a quarter.

This does two things to your reasoning. First, every police-based statistic is a floor, not a total. Second, and less obvious, the size of the undercount differs by crime, so you cannot compare offenses using police data alone. Police records make motor vehicle theft look large relative to sexual assault partly because one is nearly always reported and the other often is not.

Why do victims stay silent? The NCVS asks. The recurring reasons are that the victim dealt with it another way or considered it private, that the loss was small, that the victim believed police would not or could not help, and that the victim feared reprisal or feared getting the offender in trouble. Notice that these reasons are not evenly distributed. They concentrate in offenses between people who know each other, which is exactly where much serious violence happens.

Key idea: The dark figure is the unrecorded crime that never reaches police, it is larger for property than violence and larger still for offenses between people who know each other, and it makes every police-based count a floor rather than a total.

Police-based counts: UCR and NIBRS

The FBI has collected crime data from local agencies since 1930 through the Uniform Crime Reporting program. For most of that history the data came in as summary counts of a short list of serious offenses, aggregated monthly by each department.

Two features of the old summary system shaped a century of research. The hierarchy rule recorded only the most serious offense in a multi-offense incident, so a burglary that included an assault and a car theft appeared as one offense. And the list of counted crimes, the Part I offenses, excluded almost everything now recognized as a major harm: fraud, most drug offenses, and every form of white-collar and corporate crime.

Since January 1, 2021, the national standard has been the National Incident-Based Reporting System. NIBRS records every offense in an incident along with structured detail on victims, offenders, relationships, weapons, property, and location. Analytically it is far richer. It lets you ask questions the summary system erased, such as what proportion of aggravated assaults involve intimate partners.

The transition was messy in a way that still matters when you read anything about the early 2020s. When the FBI retired summary reporting, thousands of agencies had not yet converted, including some of the largest departments in the country. Participation dipped sharply for 2021 and recovered over the following years as agencies came online. National figures for that period rest on estimation models filling gaps, and honest analysts flag the uncertainty rather than treating the series as continuous.

Beyond coverage, police data have a deeper property that is easy to forget: they measure police activity as well as crime. Recorded drug offenses rise when a department starts a drug initiative, not necessarily when drug use rises. Recorded assaults can rise when a department improves its intake of domestic violence calls. For crimes discovered through enforcement rather than reported by victims, police data are close to useless as a measure of underlying behavior.

Key idea: UCR summary counts hid detail under the hierarchy rule and covered a narrow offense list, NIBRS records full incident detail and became the standard in 2021 after a disruptive transition, and both partly measure police activity rather than crime.

Victimization surveys: the NCVS

To reach the dark figure, the Bureau of Justice Statistics runs the National Crime Victimization Survey, begun in 1973. Interviewers contact a large national sample of households, on the order of 150,000 to 250,000 interviews a year depending on the design period, and ask household members whether specific things happened to them, whether or not they called the police.

The design has several strengths worth naming. Households stay in the sample across multiple interviews, so the survey can bound reports in time and reduce the tendency to pull older events forward. It asks about concrete incidents rather than impressions. It measures the reporting rate itself, which is how we know the dark figure's size. And it collects victim characteristics, so it can describe who is victimized.

Its limits are equally real. It cannot measure homicide, since the victim cannot be interviewed; homicide is the one crime for which police data are nearly complete, because bodies are hard to hide. It covers people aged 12 and older in households, so it misses young children and people in institutions, including prisons and some shelters. It cannot measure crimes against businesses. It relies on memory and candor, and for sensitive offenses the estimate depends heavily on question wording, which is why NCVS sexual assault estimates differ from those of surveys such as the National Intimate Partner and Sexual Violence Survey that use different definitions and framing. And like any survey, it has sampling error, so a small year-to-year change may not be a change at all.

Key idea: The NCVS surveys households directly and so captures unreported crime and the reporting rate itself, but it cannot measure homicide or crimes against businesses and its estimates for sensitive offenses depend on question design.

Self-report studies

The third instrument asks offenders. In a self-report study, respondents anonymously report their own delinquent or criminal acts. This sounds unpromising and turns out to be one of the most productive methods in the field.

Self-report work transformed criminology in the 1950s and 1960s by showing that offending was far more widespread across social classes than arrest data suggested. Arrest statistics showed a strong class gradient. Self-reports showed a much weaker one for common delinquency, which forced the field to distinguish between who offends and who gets caught. That distinction is now foundational.

Major ongoing examples include Monitoring the Future, which has surveyed American secondary students annually since 1975 on substance use and related behavior, and longitudinal projects such as the National Youth Survey and the Add Health study, which follow the same young people over years. Longitudinal self-report data are what made life-course criminology possible, since only they can establish which came first.

The weaknesses are the obvious ones plus one that is not. Respondents may forget, exaggerate, or conceal. Frequent serious offenders are underrepresented, because the most prolific are disproportionately absent from schools and households when surveys are administered. And the standard instruments historically loaded up on minor acts, so a scale could rate a student who shoplifted twice similarly to one who committed a robbery. Modern instruments handle severity better, but when you read a self-report finding, check what was actually counted.

Key idea: Self-report studies revealed that offending is far more widely distributed than arrest data suggest, and longitudinal self-report projects made life-course research possible, but they undercount the most serious and most prolific offenders.

Putting the three together

Measure	Source	Best used for	Blind spot
UCR / NIBRS	Police agency records	Homicide, police workload, local geography, incident detail	Unreported crime; reflects enforcement effort
NCVS	Household interviews	Victimization levels, reporting rates, victim profiles	No homicide, no businesses, no institutions
Self-report	Offender interviews and surveys	Prevalence of offending, onset and desistance, causes	Misses prolific and serious offenders; recall and honesty

The practical rule is to match the instrument to the question. Studying homicide? Use police data, which are close to complete. Studying whether victims trust police? Only the NCVS can tell you. Studying why some teenagers start offending and others do not? You need self-report longitudinal data, because arrests measure police contact as much as behavior.

When two instruments agree, confidence rises substantially. The great American crime decline of the 1990s, the subject of the next lesson, shows up in both police data and the NCVS. That convergence is the main reason criminologists treat it as real rather than as an artifact of recording practices.

Key idea: Match the measure to the question, and treat convergence across independent instruments as the strongest available evidence that a trend is real rather than an artifact of one system.

How crime statistics get misread

Five errors account for most of the bad crime reporting you will encounter.

Confusing reporting with offending. A campaign encouraging sexual assault survivors to come forward will raise recorded offenses. Reading that as a rise in assaults inverts the meaning: it may be a sign the system is working.

Ignoring the denominator. Crime counts must be converted to rates per 100,000 residents before comparison across places or years, and even then the denominator can mislead. A downtown with 5,000 residents and 200,000 daily visitors will show an absurd rate per resident.

Treating clearance as solving. A case cleared by arrest is not a conviction, and cases can also be cleared by exceptional means. Clearance rates measure police case closure, not justice done.

Reading noise as trend. For rare events in small jurisdictions, year-to-year swings are dominated by chance. Three years of data is not a trend. Criminologists usually want a decade or a statistical model that accounts for random variation.

Comparing across incompatible systems. International comparisons are especially treacherous, since countries define offenses differently and record at different points in the process. Homicide is the least bad crime for cross-national comparison precisely because it is defined similarly everywhere and hard to conceal.

Key idea: Most misread crime statistics come from confusing reporting changes with offending changes, using counts instead of rates, treating clearances as convictions, mistaking noise for trend, or comparing incompatible recording systems.

A worked example

Suppose a city announces that reported robberies fell 12 percent while reported aggravated assaults rose 9 percent in the same year. What is going on?

Start with the measure. Both figures come from police records, so both are subject to reporting and recording effects. Ask whether the department changed anything: a new classification policy, a NIBRS conversion, a new domestic violence unit that logs incidents previously handled informally. Any of these can move assault counts without any change in violence.

Next, ask about the base numbers. If the city had 340 robberies and 1,100 assaults, a 12 percent and 9 percent change are both outside pure noise but not dramatically so. If it had 34 and 110, both numbers are close to meaningless.

Then look for an independent check. Hospital trauma admissions for assault-related injury are collected separately from police data and are a useful cross-check on serious violence. If trauma admissions are flat while recorded assaults rose, recording change becomes the leading hypothesis.

Finally, resist the urge to explain before you have established that there is something to explain. Half of crime commentary is elaborate causal storytelling about a change that never happened.

Key idea: Reading a real crime statistic means checking the instrument, the recording practices, the base size, and an independent data source before proposing any explanation for the change.

Common misconceptions

• The FBI counts all crime in America. It compiles what agencies record and submit, which excludes the dark figure entirely.
• The NCVS and UCR should produce the same numbers. They measure different things and routinely differ even when their trends agree.
• NIBRS solved the undercount. It records recorded crime in far more detail but still cannot see crime never reported.
• Self-report data are worthless because people lie. Validation studies show reasonable accuracy for most offenses, and self-report revealed the class gap between offending and arrest.
• A rise in recorded offenses always means more offending. It can equally mean more reporting, better recording, or more enforcement.

Recap

• The dark figure of crime is the unrecorded remainder, and it varies sharply by offense type.
• UCR summary counts used a hierarchy rule and a short offense list; NIBRS records full incident detail and became the standard in 2021.
• The NCVS measures victimization and reporting rates directly but cannot capture homicide, businesses, or institutions.
• Self-report studies show offending is widespread and enable longitudinal research, while undercounting the most serious offenders.
• Match the measure to the question, and trust findings most when independent instruments converge.

Sources

1. Bureau of Justice Statistics. (n.d.). National Crime Victimization Survey (NCVS). Office of Justice Programs. bjs.ojp.gov
2. Federal Bureau of Investigation. (n.d.). Uniform Crime Reporting (UCR) Program. le.fbi.gov
3. Federal Bureau of Investigation. (n.d.). Crime Data Explorer. cde.ucr.cjis.gov
4. Institute for Social Research. (n.d.). Monitoring the Future. University of Michigan. monitoringthefuture.org
5. Bureau of Justice Statistics. (n.d.). Data collections. Office of Justice Programs. bjs.ojp.gov

The big picture

Something enormous happened to American crime between 1991 and 2014, and most people do not know it happened. Violent crime rates fell by roughly half. Homicide fell from about 9.8 per 100,000 in 1991 to around 4.4 to 5.0 per 100,000 by the mid-2010s. Property crime fell even further. Both police data and the victimization survey, which are independent instruments, show it.

This is one of the most consequential social changes of the last half century, and criminology cannot fully explain it. That admission is where this lesson earns its place. You will meet perhaps a dozen serious candidate explanations, several with real evidence, none sufficient alone, some contradicting each other, and one famous one that partially collapsed under scrutiny.

The intellectual reward is bigger than the topic. A phenomenon this large, this well measured, and this poorly explained is a permanent lesson about the limits of social science, and about the difference between having an explanation and having evidence for one.

Key idea: American violent crime fell by roughly half between 1991 and the mid-2010s, the decline appears in independent measures so it is real, and no single explanation accounts for it.

The very long run

Zoom out further than the news ever does. Historical criminologists, most systematically Manuel Eisner, have assembled homicide rates from coroners' rolls, court records, and parish registers across Europe back to the medieval period. The series is imperfect, but the shape is unmistakable: homicide rates in Western Europe fell by something like a factor of thirty to fifty between the fourteenth century and the twentieth, from roughly 30 or more per 100,000 to around 1.

That decline was not smooth and it was not uniform, but it is large enough that measurement error cannot explain it away. Explanations offered include the consolidation of state authority and courts displacing private vengeance, the spread of commerce and its habits of self-restraint, and long-run changes in norms about honor and interpersonal violence. Nobody has settled it.

Keep this in view whenever a modern debate treats current rates as historically catastrophic. A contemporary American homicide rate of 5 or 6 per 100,000 is very high compared with other wealthy countries today and very low compared with almost any premodern society.

Key idea: Homicide in Western Europe declined dramatically over centuries, which shows crime rates are historically variable rather than fixed, and makes the modern American rate high by international but low by historical standards.

The American series since 1960

The modern American story has four movements.

Period	What happened	Rough magnitude
1960 to 1991	Long, steep rise in violent crime	Violent crime rate roughly quadrupled
1991 to 2014	The great crime decline	Violent crime roughly halved; property crime fell further
2015 to 2019	Small increases then plateau at historically low levels	Modest, concentrated in some cities
2020 onward	Sharp homicide spike, then decline	Homicide rose about 30 percent in 2020, then fell substantially from 2022 through 2024

Two features of this table deserve emphasis. First, the 2020 homicide increase was, by the FBI's reporting, the largest single-year percentage rise in the modern national series, yet it left homicide well below the 1991 peak. Both facts are true, and partisans routinely cite one while suppressing the other.

Second, the 2020 spike was overwhelmingly a homicide and shooting phenomenon. Robbery and most property crime did not spike, and some fell during the same period. Any explanation for 2020 has to explain that specificity, which rules out most simple stories about generalized lawlessness.

Key idea: American crime rose steeply from 1960 to 1991, fell by half to 2014, plateaued at low levels, then saw a homicide-specific spike in 2020 that receded, and the specificity of that spike constrains what can explain it.

Explaining the 1990s decline: the leading candidates

Steven Levitt's widely cited 2004 review argued that four factors explained most of the decline and that six popular explanations did not. His four were rising incarceration, more police, the receding crack cocaine epidemic, and the legalization of abortion. His six rejected candidates were the strong 1990s economy, changing demographics, innovative policing strategies, gun control laws, concealed-carry laws, and capital punishment. That paper is a superb entry point and is also now substantially contested, which makes it an ideal teaching case.

Go through the main candidates one at a time with the evidence as it stands.

Incarceration. The American imprisonment rate roughly quadrupled between the mid-1970s and 2008. Some crime reduction from incapacitation is not seriously disputed; a person in prison cannot burgle a house. The argument is about magnitude. Levitt attributed a substantial share of the 1990s decline to it. The National Research Council's 2014 consensus report concluded that the crime-prevention effect of the incarceration increase was modest at best, and that the growth in incarceration could not be the main driver. Analyses by the Brennan Center attributed only a small share, in the range of zero to seven percent, of the 1990s decline to incarceration and essentially nothing to further increases in the 2000s. The reason is diminishing returns: as prison expands, it increasingly holds lower-rate offenders, so each additional prisoner prevents less crime. Lesson 15 works through this evidence in detail.

Police numbers and tactics. Police hiring expanded in the 1990s. Careful studies that use plausibly exogenous variation, notably work by Chalfin and McCrary, estimate meaningful effects: their preferred estimates imply an elasticity of violent crime with respect to police of roughly minus 0.5 to minus 0.7, which is large. Whether specific tactics, as opposed to sheer numbers, drove the decline is a separate and murkier question taken up in Lesson 15.

The crack epidemic's decline. The rise in homicide from the mid-1980s was heavily concentrated among young men in cities where crack markets were expanding, and the fall tracked the contraction and stabilization of those markets. This explanation has good ecological support, though it pushes the question back a level: why did the markets stabilize?

The abortion hypothesis. Donohue and Levitt argued in 2001 that legalized abortion after 1973 reduced the number of unwanted births, and that this cohort effect lowered crime beginning in the early 1990s. It was politically explosive and empirically fragile. Foote and Goetz found a coding error in the original analysis, and correcting it substantially weakened the results, though the authors defended a smaller version of the finding. Treat this as genuinely unresolved and probably smaller than originally claimed, and notice how a striking finding can persist in public memory long after the evidence has softened.

Lead. Tetraethyl lead was removed from American gasoline over the 1970s and 1980s. Childhood lead exposure has well-established effects on cognitive function and impulse control. Reyes estimated that the phase-out could account for a substantial share, possibly over half, of the 1990s violent crime decline. The individual-level toxicology is solid; the aggregate crime claim rests on ecological correlations that critics argue are vulnerable to confounding with everything else that changed on the same timeline. Plausible, probably real, magnitude disputed.

Better security. Farrell and colleagues argue that the decline in property crime, especially vehicle theft, is largely explained by improved security: electronic immobilizers, better locks, alarms. Vehicle theft fell steeply in exactly the countries and years that immobilizers spread. This is one of the better-supported explanations for property crime specifically and it explains an awkward fact for other theories, namely that crime fell across many wealthy countries at roughly the same time despite wildly different incarceration and policing policies.

Community organizations. Sharkey and colleagues found that the growth of local nonprofit organizations focused on crime and community life predicted violence reductions across cities, estimating that each additional ten such organizations per 100,000 residents was associated with roughly a nine percent drop in the murder rate. It is not a randomized design, but the effect is sizable and consistent with neighborhood theories in Module 3.

Key idea: The strongest candidates for the 1990s decline are police numbers, the receding crack epidemic, improved security, lead removal, and some incarceration effect, with the abortion hypothesis substantially weakened and every estimate contested in magnitude.

Why the accounting never balances

Add up the confidently claimed contributions of all these explanations and you get well over 100 percent of the decline. That fact alone should make you cautious.

There are structural reasons the question is hard. There is only one national time series, so every explanation is fitted to the same handful of data points. Many candidate causes moved together in the same decades, which makes them nearly impossible to separate statistically. And crime fell in Canada, much of Western Europe, and Australia over a similar period without American-scale incarceration growth, which is a serious problem for explanations centered on American criminal justice policy.

The most defensible position is pluralist: several mechanisms each contributed something, their contributions overlap, and the field cannot presently apportion them precisely. That is an unsatisfying answer, and it is the honest one. Be suspicious of anyone, in any political direction, who tells you the cause of the crime decline with confidence.

Key idea: Because the candidate causes moved together in one national time series, and because crime also fell in countries with very different policies, the decline is best treated as multiply caused with contributions that cannot yet be precisely apportioned.

Perception versus measurement

Through most of the great decline, Gallup's annual polling found that a clear majority of Americans said crime was rising nationally. In some years the gap between perception and both official measures was extreme.

Several mechanisms produce this. News coverage selects for the vivid and the rare. Local perception and national perception diverge, with people typically rating their own area safer than the country, which is a clue that the national judgment is coming from media rather than experience. And crime perception is entangled with partisanship, moving with which party holds the presidency more than with the data.

This is not a footnote. Policy responds to perceived crime, so a persistent perception gap can drive spending and legislation that the measured trend does not justify. Understanding the gap is part of understanding crime policy.

Key idea: Public belief about crime trends tracks media coverage and partisanship rather than measured rates, and because policy responds to perception, the gap has real consequences.

Common misconceptions

• Crime has been rising for decades. Violent crime fell by roughly half from 1991 to 2014 and remains far below the 1991 peak.
• Mass incarceration caused the crime decline. It contributed something, but major reviews conclude the effect was modest and subject to steep diminishing returns.
• The abortion explanation is settled science. A coding error weakened the original result and the magnitude remains disputed.
• The 2020 spike was a general crime wave. It was concentrated in homicide and shootings; most property crime did not follow the same path.
• Crime trends are an American story. Crime fell across many wealthy countries in the same period, which constrains explanations based on United States policy.

Recap

• Homicide declined dramatically in Western Europe over centuries, so crime rates are historically variable.
• American crime rose from 1960 to 1991, fell by roughly half to 2014, plateaued, then saw a homicide-specific spike in 2020 that has since receded.
• Leading explanations include police numbers, the crack market's decline, improved security, lead removal, incarceration, and community organizations.
• The claimed contributions sum to more than the decline, and simultaneous declines abroad constrain any purely American explanation.
• Public perception of crime trends diverges persistently from measurement, driven by coverage and partisanship.

Sources

1. Levitt, S. D. (2004). Understanding why crime fell in the 1990s: Four factors that explain the decline and six that do not. Journal of Economic Perspectives, 18(1), 163-190. doi.org
2. National Research Council. (2014). The growth of incarceration in the United States: Exploring causes and consequences. National Academies Press. nap.nationalacademies.org
3. Roeder, O., Eisen, L.-B., & Bowling, J. (2015). What caused the crime decline? Brennan Center for Justice. brennancenter.org
4. Sharkey, P., Torrats-Espinosa, G., & Takyar, D. (2017). Community and the crime decline: The causal effect of local nonprofits on violent crime. American Sociological Review, 82(6), 1214-1240. doi.org
5. Federal Bureau of Investigation. (n.d.). Crime Data Explorer. cde.ucr.cjis.gov

The big picture

Start with the oldest idea in criminology and the one still built into almost every criminal statute: people commit crimes because, at that moment, crime looks worth it. Raise the cost enough and offending falls. That is deterrence, and it underwrites everything from parking fines to mandatory minimum sentences.

Here is what makes this lesson interesting. The theory is broadly right about mechanism and mostly wrong about the policy conclusions people draw from it. Decades of research converge on a finding with enormous practical consequences: what deters is the probability of being caught, not the size of the eventual punishment. Nearly all American sentencing policy of the last fifty years pulled the other lever.

You will finish this lesson able to say precisely why a longer sentence is a weak crime-control tool and a more visible police presence is a strong one, and to explain the situational logic that makes some crimes simply not happen.

Key idea: Deterrence theory says punishment prevents crime by raising its expected cost, but the evidence says certainty of detection does the work while severity contributes little.

Beccaria, Bentham, and the classical school

In 1764 Cesare Beccaria published a short book, On Crimes and Punishments, arguing against a European penal order of torture, secret trials, and arbitrary death sentences. His argument was not sentimental. It was that such punishment does not work.

Beccaria held that people are rational and self-interested, that punishment is justified only to prevent future harm, and that its effectiveness depends on being certain, swift, and proportionate rather than cruel. A punishment beyond what is needed to deter is, on this view, simply violence by the state. He also argued that laws should be public and clear, since a person cannot be deterred by a rule he cannot know.

Jeremy Bentham built the utilitarian machinery around this. In his hedonic calculus, individuals weigh expected pleasure against expected pain, and the state's job is to set penalties just high enough to tip the calculation. Bentham's fingerprints are everywhere in modern penal policy, including the idea that punishment should be graduated so that an offender choosing between two crimes has reason to pick the lesser.

The classical school's practical legacy is enormous: written criminal codes, proportionality, the presumption that punishment must be justified by its consequences, and abolition of some of the more grotesque punishments of the eighteenth century.

Key idea: Beccaria and Bentham founded the classical school on the claim that offenders are rational calculators, so punishment should be certain, swift, and proportionate rather than severe.

The three elements of deterrence

Modern deterrence theory decomposes the threat of punishment into three parts.

Certainty is the perceived probability of being caught and punished. Severity is the magnitude of the punishment if caught. Celerity is the swiftness with which punishment follows the act.

A second distinction cuts across these. Specific deterrence is the effect of punishment on the person punished, discouraging him from reoffending. General deterrence is the effect on everyone else, who observe the threat and adjust. The two can move in opposite directions, which is one of the field's uncomfortable findings: a sentence that generally deters the public might, through its effects on the individual, make that particular person more likely to reoffend.

Notice the word perceived in the definition of certainty. Deterrence operates through the offender's beliefs, not through the true statistics. If a burglar believes the chance of arrest is one in fifty when it is really one in twelve, his behavior responds to the fifty. This makes public awareness part of the mechanism, and it explains why deterrence-based programs often invest heavily in communication.

Key idea: Deterrence has three components, certainty, severity, and celerity, and it works through perceived risk, so what an offender believes matters more than the actual probability.

What the evidence actually shows

Daniel Nagin's authoritative review of the deterrence literature reaches conclusions that are unusually consistent for social science.

First, the certainty of apprehension is a far more powerful deterrent than the severity of punishment. This finding appears across many designs and offense types and has been stable for decades. The National Institute of Justice summarizes it plainly in its public guidance on deterrence.

Second, increases in the length of prison sentences produce, at best, very small marginal deterrent effects. Studies of three-strikes laws, mandatory minimums, and sentence enhancements generally find effects that are small, statistically fragile, or absent. The main reason is that offenders steeply discount the future and often have poor knowledge of actual penalties. A person who does not expect to be caught is largely indifferent to whether the sentence is five years or fifteen.

Third, police presence deters. Studies exploiting sudden, externally driven changes in police deployment, such as terror-alert-driven surges in Washington DC studied by Klick and Tabarrok, find crime falling in the areas that received extra police. This is certainty in action: visible officers change the perceived probability of being caught right now.

Fourth, on the death penalty specifically, a 2012 National Research Council committee reviewed three decades of studies and concluded that the research to date is not informative about whether capital punishment increases, decreases, or has no effect on homicide, and that these studies should not be used to inform policy. That is a striking verdict: not that the effect is zero, but that the research design cannot answer the question.

Key idea: Certainty of apprehension deters, longer sentences add very little deterrence, visible policing reduces crime in the short run, and the death penalty literature is judged too weak to support conclusions in either direction.

Why severity underperforms

It is worth understanding the mechanism, because the finding is counterintuitive to most people.

Offenders discount the future steeply. A punishment arriving after a year of pretrial proceedings, discounted by a low perceived chance of arriving at all, has little weight against an immediate reward. Doubling a distant, improbable cost changes very little.

Many offenses also occur under conditions hostile to calculation: intoxication, anger, group dynamics, or fear. Homicides in particular often arise from escalating disputes rather than plans, which is exactly the situation in which penalty schedules do not enter anyone's mind.

And knowledge is poor. Surveys of offenders repeatedly find that most cannot accurately state the penalties for offenses they have committed. A law that nobody knows about cannot deter through severity, though it can still incapacitate.

None of this means severity is irrelevant to justice. Proportional punishment can be defended on retributive grounds, on grounds of expressing social condemnation, or for incapacitation. The point is narrower and firmer: as a lever for reducing crime through deterrence, severity is weak, and buying it is expensive.

Key idea: Severity underperforms because offenders discount the future, many offenses occur in states hostile to calculation, and knowledge of actual penalties is poor.

Rational choice, bounded

Modern rational choice theory in criminology, developed by Derek Cornish and Ronald Clarke, softens the classical picture in an important way. Offenders are treated as reasoning agents, but with limited information, limited time, and limited cognitive effort. This is bounded rationality.

Cornish and Clarke also insisted on being crime-specific. The reasoning behind burglary of a suburban house has almost nothing in common with the reasoning behind a bar fight or a tax fraud. Asking about the rationality of crime in general is close to meaningless; asking what a burglar looks for on a street is answerable and useful.

They further separated the involvement decision, whether to be the kind of person who does this at all, from the event decision, whether to do this here and now. Involvement is shaped by background factors of the sort Modules 3 and 4 study. The event decision is where situational factors dominate, and it is the one that policy can most easily change.

Interview studies with active burglars, notably by Wright and Decker, support the bounded picture. Burglars use quick heuristics: occupancy cues, escape routes, cover from view, signs of a dog, evidence of an alarm. They rarely calculate anything resembling expected utility, but their choices are systematic and predictable, which is what matters for prevention.

Key idea: Rational choice theory treats offenders as boundedly rational and crime-specific decision makers, separating the background involvement decision from the situational event decision.

Routine activity theory

In 1979 Lawrence Cohen and Marcus Felson published an argument that reoriented the field. American crime had risen sharply through the 1960s at a time when unemployment fell and social conditions improved. If crime came from deprivation, this made no sense.

Their answer moved the question from offenders to opportunities. A criminal event, they argued, requires the convergence in time and space of three things: a motivated offender, a suitable target, and the absence of a capable guardian. Remove any one and no crime occurs, regardless of how many motivated people exist.

The postwar change that mattered was in routine activities. More women entered the paid workforce and more people spent time away from home, leaving houses empty during the day. Consumer goods became lighter, more valuable, and more portable, which is to say more suitable as targets. Cars, televisions, and later laptops and phones travel well. Opportunity expanded, and crime followed.

Felson later stressed that a capable guardian is usually not a police officer. It is a neighbor at a window, a shopkeeper, a doorman, a friend walking with you, a dog. Most guardianship is ordinary and unpaid, which is why changes in daily routine have such large effects.

Key idea: Routine activity theory explains crime as the convergence of a motivated offender, a suitable target, and no capable guardian, so changes in everyday routines and in the portability of goods change crime rates without changing motivation.

Situational crime prevention and its critics

These theories share a policy implication: change the situation rather than the person. Situational crime prevention, associated with Clarke, catalogues techniques that increase effort, increase risk, reduce reward, reduce provocation, and remove excuses.

The examples are unglamorous and often effective. Steering column locks and later electronic immobilizers cut vehicle theft. Exact-fare systems and cash controls reduced bus driver robberies. Removing coin-operated meters ended a whole category of theft. Improved street lighting shows crime reductions in meta-analyses. Better ATM siting reduces robbery.

The standard objection is displacement: block a crime here and it moves next door. This is a genuine possibility, and the empirical answer is reassuring. Reviews consistently find that displacement occurs sometimes but is usually partial rather than complete, so prevention still yields a net reduction. Studies also frequently find the opposite effect, a diffusion of benefits, in which crime falls in nearby areas that received no intervention, apparently because offenders overestimate the intervention's reach.

Deeper objections remain. Situational prevention treats symptoms rather than causes and can produce a fortified, unequal landscape in which the well-resourced protect themselves and displacement lands on those who cannot. It is largely silent about crimes of passion, about corporate offending, and about why some people are motivated in the first place. Those are the subjects of the rest of this course.

Key idea: Situational prevention reduces crime by changing opportunities, displacement is usually partial rather than total and is sometimes outweighed by diffusion of benefits, but the approach addresses opportunity rather than motivation.

Common misconceptions

• Harsher sentences are the strongest crime deterrent. Certainty of apprehension does nearly all the deterrent work; severity adds little.
• Deterrence research shows punishment does not work. It shows that one lever, severity, is weak while another, perceived certainty, is effective.
• Rational choice means offenders coolly calculate expected utility. Modern versions assume bounded, crime-specific, heuristic reasoning.
• Prevention just displaces crime elsewhere. Displacement is usually partial, and diffusion of benefits to nearby areas is commonly observed.
• A capable guardian means a police officer. Most guardianship is ordinary: neighbors, staff, companions, and visibility.

Recap

• Beccaria and Bentham founded the classical school on rational offenders and certain, swift, proportionate punishment.
• Deterrence has three elements, certainty, severity, and celerity, and operates through perceived rather than actual risk.
• Evidence strongly favors certainty over severity, and a National Research Council review found death penalty studies uninformative.
• Rational choice theory assumes bounded, crime-specific reasoning and separates involvement from event decisions.
• Routine activity theory explains events as the convergence of motivated offender, suitable target, and absent guardian, grounding situational prevention.

Sources

1. Nagin, D. S. (2013). Deterrence in the twenty-first century. Crime and Justice, 42(1), 199-263. doi.org
2. National Institute of Justice. (2016). Five things about deterrence. Office of Justice Programs. nij.ojp.gov
3. Cohen, L. E., & Felson, M. (1979). Social change and crime rate trends: A routine activity approach. American Sociological Review, 44(4), 588-608. doi.org
4. National Research Council. (2012). Deterrence and the death penalty. National Academies Press. nap.nationalacademies.org
5. Britannica. (n.d.). Cesare Beccaria. Encyclopaedia Britannica. britannica.com

The big picture

This is the most dangerous topic in criminology, and skipping it would be a mistake. The idea that criminality is written in the body produced a century of bad science and real atrocity. It also, in a chastened and very different modern form, produces findings that are replicable and worth understanding.

The lesson has two jobs. The first is historical: to show you exactly how intelligent people using apparently scientific methods reached conclusions that were both false and harmful, because the failure modes are still available today. The second is contemporary: to give you an accurate, unhysterical account of what biosocial research has and has not shown, so you are neither taken in by overclaiming nor unable to read the literature at all.

The connecting thread is a single discipline: knowing what a finding licenses you to conclude. Almost every disaster in this history came from a correlation being read as a cause, and a cause being read as a destiny.

Key idea: Biological explanations of crime have a history of catastrophic error and misuse, and modern biosocial research is only defensible when its findings are stated with the limits the evidence actually supports.

Phrenology and physiognomy

In the late eighteenth century Franz Joseph Gall proposed that mental faculties are localized in specific brain regions, that the size of a region reflects its power, and that the skull's shape reveals the brain beneath it. His follower Johann Spurzheim popularized phrenology across Europe and America, complete with charts of bumps corresponding to traits such as destructiveness and acquisitiveness.

Two of Gall's premises were correct and important: the brain is the organ of mind, and function is to some degree localized. The rest was wrong. Skull shape does not track cortical function, the trait categories were invented rather than discovered, and the method was pure confirmation bias. Practitioners examined a known criminal, found a prominent bump, and recorded a confirmation. Nobody counted the law-abiding people with identical skulls.

That is the first lesson in method, and it recurs: a study of offenders alone cannot establish what distinguishes offenders. Without a comparison group you can only ever confirm what you expected.

Key idea: Phrenology failed because it lacked comparison groups and mistook confirmation for evidence, though its premise that the brain is the organ of mind was correct.

Lombroso and the born criminal

Cesare Lombroso, an Italian physician, published L'uomo delinquente, or Criminal Man, in 1876. He argued that many offenders are atavistic, meaning evolutionary throwbacks to an earlier human type, and that they can be identified by physical stigmata: sloping foreheads, large jaws, unusual ear shapes, long arms, tattoos, insensitivity to pain.

He is often dismissed as a crank, which is too easy. Lombroso genuinely tried to bring measurement and comparison to a subject previously handled by moral philosophy. He measured thousands of prisoners. He revised his estimates downward over successive editions and added social and environmental categories of offender. He founded the positivist tradition that treats crime as something to be explained rather than merely condemned. That methodological turn was real progress.

His conclusions were nonetheless wrong, and the reasons are instructive. His comparison groups were inadequate, often soldiers, and his measurements were not blind, so he knew which group each subject belonged to while measuring. Many of the traits he catalogued track poverty, malnutrition, occupational injury, and untreated illness, which are common in prisons for reasons that have nothing to do with atavism. And the theoretical frame, applying a garbled Darwinism to human types, was itself unfounded.

The definitive refutation came in 1913 from Charles Goring, an English prison physician who compared thousands of convicts with a large non-criminal control group, including university students and soldiers, and applied statistical methods. He found no distinctive physical criminal type. That is how the question was actually settled: better controls and better statistics.

Key idea: Lombroso brought measurement to criminology but concluded that criminals are atavistic types, a claim refuted by Goring in 1913 using proper control groups and statistics.

What the idea was used for

The history matters because the conclusion did not stay in journals.

If criminality is inherited and physically marked, then preventing crime means preventing certain people from reproducing. That inference drove the eugenics movement in the early twentieth century. In the United States, more than thirty states passed compulsory sterilization laws targeting people classified as feeble-minded, criminalistic, or otherwise defective. The Supreme Court upheld such a law in Buck v. Bell in 1927, in an opinion by Justice Oliver Wendell Holmes Jr. containing one of the most notorious sentences in American law. Estimates place the number of people sterilized under these programs in the United States at roughly 60,000 or more, disproportionately poor women and members of minority groups. Some of these programs continued into the 1970s.

The same body of thinking was taken further in Nazi Germany, where hereditarian criminology contributed to the classification and mass murder of people labeled asocial and criminal.

Later episodes were smaller but similar in structure. In the 1960s the discovery of men with an extra Y chromosome in some secure institutions produced a wave of claims about an inherently violent XYY type. Systematic studies, especially a large Danish register study by Witkin and colleagues, found that XYY men were not notably violent; the modest elevation in convictions was mostly for property offenses and appeared largely explained by lower average cognitive test scores. The myth outlived the correction by decades.

Key idea: Hereditarian criminology directly supported compulsory sterilization laws upheld in Buck v. Bell and, in Germany, mass murder, and the later XYY panic shows how quickly a weak finding becomes a public certainty.

Why that history constrains today's research

None of this makes biological research about behavior illegitimate. It does impose obligations that a purely sociological study does not carry to the same degree.

The first is precision about what a finding means. Most biosocial results describe small average differences between large groups, with enormous overlap. A difference in group means is not a diagnostic test, and treating it as one is exactly the error that produced the history above.

The second is attention to how findings will be used. A researcher cannot control public use, but can refuse to overstate, can report effect sizes, and can name what the finding does not show.

The third is awareness that biological measures often proxy social conditions. Lead exposure, poor prenatal nutrition, head injury, and chronic stress are biological facts with social causes and social distributions. Finding a brain correlate of antisocial behavior does not tell you whether the origin was genetic, toxic, nutritional, or traumatic.

Key idea: Because biological differences between groups are small and heavily overlapping, and because biological measures often carry the imprint of social conditions, biosocial findings must be stated as averages rather than as diagnoses.

Modern biosocial criminology: what holds up

With those guardrails, here is a fair summary of contemporary work.

Heritability. Twin and adoption studies consistently find that antisocial behavior is moderately heritable. A large meta-analysis by Rhee and Waldman, pooling many twin and adoption studies, estimated roughly 40 percent of variance attributable to genetic factors, with the remainder split between shared and non-shared environment. That number is frequently misread, so state carefully what it means: within the studied populations, about 40 percent of the differences among individuals are statistically associated with genetic differences. It says nothing about any individual, nothing about differences between groups, and nothing about whether the behavior can be changed. A highly heritable trait can be entirely fixable; eyeglasses correct highly heritable myopia.

Candidate genes. A famous 2002 study by Caspi and colleagues reported that a variant of the MAOA gene predicted antisocial outcomes only among men who had been maltreated as children, a gene by environment interaction. It was elegant and hugely influential. It also arrived just before psychiatric genetics discovered that most single candidate gene findings do not replicate. Meta-analyses of the MAOA interaction give modest support, but the broader lesson stands: single-gene explanations of complex behavior have largely failed. Contemporary genetics uses genome-wide data and polygenic scores that aggregate thousands of tiny effects, and even these predict behavioral outcomes weakly.

Resting heart rate. One of the more robust individual-level findings is that low resting heart rate is associated with antisocial and aggressive behavior. A meta-analysis by Ortiz and Raine found an effect size around minus 0.4 in standardized terms, and a large Swedish cohort study by Latvala and colleagues found elevated later violent offending among men with the lowest resting heart rates. The leading interpretations are chronic underarousal, which may make risk-seeking rewarding, or fearlessness. The effect is real and replicated; it is also modest and useless for predicting individuals.

Neurodevelopment and injury. Prefrontal function supports planning, impulse control, and the weighing of consequences. Damage or delayed development in these systems is associated with impulsive aggression, and traumatic brain injury is markedly more prevalent among incarcerated people than in the general population, although causal direction is genuinely unclear since a violent life produces head injuries as well as following from them.

Toxins and prenatal exposure. Childhood lead exposure has well-documented effects on cognition and impulse control, and prenatal exposure to alcohol and nicotine is associated with later conduct problems. These are the least controversial biosocial findings because the mechanism is understood and the exposure is preventable by policy.

Key idea: Modern biosocial work supports moderate heritability of antisocial behavior, a replicated low resting heart rate association, prefrontal involvement in impulse control, and clear harms from lead and prenatal exposures, while single-gene explanations have largely failed.

Reading heritability without error

Because heritability is the most abused statistic in this area, four rules are worth memorizing.

Heritability is a property of a population, not a person. It describes how variance is distributed in a particular group at a particular time, and it changes when conditions change. If every child received identical parenting, the heritability of behavior would rise, not because genes became stronger but because environmental variance shrank.

Heritability within groups says nothing about differences between groups. This point is old, well established, and constantly ignored.

Heritable does not mean unchangeable. The environment can move a highly heritable trait a great deal.

Gene-environment correlation blurs the categories. Genetically influenced traits shape the environments people are selected into and choose, so what looks like a genetic effect on offending can run through schooling, peers, and treatment by others. Studies that cannot separate these mechanisms should not claim to have found a direct biological pathway.

Key idea: Heritability describes variance in a population rather than an individual, does not license claims about group differences, does not imply immutability, and is entangled with gene-environment correlation.

Common misconceptions

• Lombroso was simply a fool. His method was a real step toward empirical criminology; his controls, blinding, and theory were the failures.
• Eugenics was a fringe European phenomenon. American sterilization laws were upheld by the Supreme Court in 1927 and operated for decades.
• An extra Y chromosome causes violence. Large register studies found no meaningful link to violent offending.
• Forty percent heritability means 40 percent of a person's criminality is genetic. It describes variance across a population, not a share within any individual.
• Biological findings mean crime is fixed at birth. Heritable traits respond to environments, and the best-established biological risks, such as lead, are preventable.

Recap

• Phrenology and Lombroso's atavism failed on control groups, blinding, and confounding with poverty and injury.
• Goring's 1913 controlled study found no physical criminal type, settling the question empirically.
• Hereditarian criminology supported compulsory sterilization in the United States and worse in Nazi Germany.
• Modern biosocial research finds moderate heritability, a replicated low resting heart rate association, and clear effects of lead and prenatal exposures.
• Heritability is a population statistic that implies nothing about individuals, group differences, or immutability.

Sources

1. Britannica. (n.d.). Cesare Lombroso. Encyclopaedia Britannica. britannica.com
2. Britannica. (n.d.). Phrenology. Encyclopaedia Britannica. britannica.com
3. Rhee, S. H., & Waldman, I. D. (2002). Genetic and environmental influences on antisocial behavior: A meta-analysis of twin and adoption studies. Psychological Bulletin, 128(3), 490-529. doi.org
4. Caspi, A., McClay, J., Moffitt, T. E., Mill, J., Martin, J., Craig, I. W., Taylor, A., & Poulton, R. (2002). Role of genotype in the cycle of violence in maltreated children. Science, 297(5582), 851-854. doi.org
5. Wikipedia contributors. (n.d.). Buck v. Bell. Wikipedia. en.wikipedia.org

The big picture

Most people in high-crime neighborhoods never commit a serious crime. Most people who face severe strain never rob anyone. Structural explanations, which arrive in Module 3, are powerful for rates but silent about which particular person offends. Psychological criminology fills that gap by asking what is different about individuals who offend, especially those who offend persistently.

The honest summary is that psychological correlates of offending are real, modest, and often misused. Impulsivity, low verbal ability, and certain personality profiles predict offending at levels social scientists consider meaningful and forecasters consider almost useless for individuals. Both statements are true, and holding both at once is the skill this lesson teaches.

It also handles the topic where public belief and evidence diverge most sharply in all of criminology: the relationship between mental illness and violence.

Key idea: Psychological factors explain why some individuals in similar circumstances offend, but the associations are modest averages that cannot reliably classify any particular person.

From psychoanalysis to conditioning

Early psychological criminology was psychodynamic. Freudian accounts explained offending as a failure of the superego or as unconscious guilt seeking punishment. John Bowlby's attachment work proposed that early maternal separation produced what he called affectionless character and later delinquency.

These ideas were enormously influential and empirically weak in their original form. Bowlby's famous study of forty-four juvenile thieves lacked adequate controls and relied on retrospective reports. The general framework was largely unfalsifiable, which is a fatal property in a science.

What survived is real, though. Attachment research in a modern form, using observation and longitudinal designs, does find that insecure and especially disorganized attachment predicts later behavior problems, with small to moderate effect sizes. The mechanism proposed now is about emotion regulation and expectations of others, not about a mythologized maternal bond.

Behaviorist criminology took a different route. Hans Eysenck argued that criminal behavior reflects poor conditionability: conscience, on his account, is a set of conditioned emotional responses acquired through childhood punishment, and people who condition poorly acquire weaker restraints. He linked this to personality dimensions of extraversion, neuroticism, and psychoticism. The specific personality model has not held up well, but the underlying idea, that individuals differ in how strongly they learn from aversive consequences, remains in the literature.

Key idea: Psychodynamic accounts of crime were influential but poorly tested, while modern attachment research and Eysenck's focus on individual differences in conditioning left more durable traces.

Personality and offending

Contemporary personality research uses the five factor model. A meta-analysis by Miller and Lynam found that antisocial behavior is most consistently associated with low agreeableness and low conscientiousness. In plainer language, offending is associated with being hostile, callous, and manipulative rather than cooperative, and with being disorganized, undisciplined, and heedless of consequences rather than diligent.

Sitting beneath both is the single most reliable psychological correlate in criminology: impulsivity, the tendency to act without weighing consequences. It appears under many names, including low self-control, sensation seeking, and poor executive function, and it predicts offending across ages, countries, measures, and designs. Module 4 examines the theory built entirely on it.

A second robust finding concerns cognitive ability. Studies since Hirschi and Hindelang's 1977 review consistently find that delinquents score roughly eight points lower on IQ tests than non-delinquents, an effect concentrated in verbal rather than performance ability. Terrie Moffitt's neuropsychological research links this to executive function and language, which support self-regulation, planning, and understanding of consequences. This association survives controls for social class and holds in self-report data, so it is not merely an artifact of who gets caught. It is also modest, and the causal pathway may run partly through school failure and its social consequences rather than directly.

Key idea: Offending is associated with low agreeableness, low conscientiousness, high impulsivity, and modestly lower verbal ability, with impulsivity the most consistent psychological correlate in the field.

Psychopathy

Few concepts are more used and abused. Psychopathy, as described clinically by Hervey Cleckley and operationalized by Robert Hare, is a constellation of interpersonal and affective features, superficial charm, grandiosity, pathological lying, manipulativeness, lack of remorse or empathy, shallow emotion, and failure to accept responsibility, combined with a lifestyle of impulsivity, irresponsibility, and antisocial conduct.

Hare's Psychopathy Checklist-Revised, the PCL-R, scores twenty items from an interview and file review, producing a score out of forty. It is a research and clinical instrument requiring training, not a questionnaire.

Several distinctions matter. Psychopathy is not the same as antisocial personality disorder, the DSM diagnosis, which emphasizes behavioral violation of others' rights. Most people meeting ASPD criteria in prison do not score high on the PCL-R, so the categories overlap without coinciding. Psychopathy is also not psychosis; psychopathic individuals are not out of contact with reality. And it is not synonymous with violence, though it predicts it modestly.

On prevalence, careful estimates put high-scoring psychopathy at roughly one percent of the general adult male population and somewhere in the range of 15 to 25 percent of incarcerated populations, with estimates varying by cutoff and setting.

On prediction, the PCL-R predicts general and violent recidivism at levels usually described as moderate. Its predictive power comes disproportionately from the lifestyle and antisocial factor rather than the interpersonal and affective factor that makes the concept famous. Reviews have also found that scores can vary depending on which side retained the evaluator, an uncomfortable finding for a tool used in sentencing and civil commitment.

The ethical caution is real. The label is close to indelible and carries connotations of untreatability that the evidence does not fully support; some studies of intensive programs report reduced violent recidivism among high-scoring youth. Applying the label to juveniles is especially contested because the traits are less stable in development.

Key idea: Psychopathy is a specific constellation of affective, interpersonal, and lifestyle traits measured by the PCL-R, distinct from antisocial personality disorder and from psychosis, predictive of recidivism at moderate levels, and dangerous as a label.

Mental illness and violence

Public belief here is badly out of line with evidence, and the correction is one of the most useful things criminology has to offer.

Severe mental illness accounts for a very small share of societal violence. Studies attempting population attributable risk estimates typically place the contribution of serious mental illness to overall violence in the low single digits of percentage points. If severe mental illness were eliminated tomorrow, the great majority of violence would remain.

The MacArthur Violence Risk Assessment Study followed patients discharged from psychiatric facilities and found that those without substance use disorders had violence rates statistically indistinguishable from neighbors in the same communities. Substance use raised risk substantially in both groups. The pattern that emerges across the literature is that substance use, prior violence, and social circumstances predict violence far better than diagnosis does.

Certain narrower conditions do carry elevated risk, particularly untreated psychosis with specific threat or persecutory delusions, and risk is elevated around first episodes and periods without treatment. Acknowledging that is not in tension with the population finding; it means the risk concentrates in a small and often treatable set of circumstances.

Meanwhile, people with serious mental illness are substantially more likely than others to be victims of violent crime. And American jails and prisons hold very large numbers of people with serious mental illness, a fact driven by the collapse of community treatment capacity rather than by any special criminality.

Key idea: Serious mental illness contributes only a small share of overall violence, substance use and prior violence predict far better than diagnosis, and people with mental illness are more often victims than perpetrators.

Developmental risk factors

The most useful psychological findings for prevention come from longitudinal studies that follow children for decades. The Cambridge Study in Delinquent Development, directed for many years by David Farrington, followed several hundred South London boys from age eight into middle age, collecting data from the boys, their parents, their teachers, and official records.

The risk factors it identified for chronic offending are remarkably consistent with other longitudinal projects: high impulsivity and hyperactivity, low school attainment, poor parental supervision, harsh or erratic discipline, parental conflict or separation, a convicted parent, low family income, and large family size.

Two features of this list deserve attention. First, the factors are strongly correlated with each other and with poverty, so isolating any one is difficult; the honest interpretation is a syndrome of cumulative disadvantage rather than a set of independent causes. Second, the relationship with outcomes is cumulative and probabilistic. A child with several risk factors has substantially elevated odds of chronic offending, and most such children still do not become chronic offenders. Risk factors describe populations.

That last point governs how these findings should be used. They support universal or broadly targeted early interventions, examined in Lesson 16, far better than they support identifying individual future criminals, which they cannot do.

Key idea: Longitudinal studies identify a consistent cluster of childhood risk factors for chronic offending, but the factors are entangled, their effects are cumulative and probabilistic, and they cannot identify individual future offenders.

Risk assessment and its limits

Modern corrections uses structured risk assessment instruments that combine such factors into scores predicting reoffending. Two findings are well established. Actuarial instruments outperform unstructured clinical judgment, which is a genuinely useful result. And their accuracy, typically reported as an area under the curve around 0.65 to 0.72 for violence, is much better than chance and far short of certainty.

Then comes the arithmetic problem. When the outcome is rare, even an accurate instrument produces many false positives. If serious violence occurs in three percent of a population, a tool flagging a group as high risk will be wrong about most of the flagged individuals. This is a mathematical property of base rates, not a flaw in any particular tool.

Fairness debates follow directly. Reporting on the COMPAS instrument argued it produced racially unequal error rates; subsequent analysis showed that when base rates differ between groups, an instrument cannot simultaneously satisfy several intuitive definitions of fairness. That is a proven impossibility rather than an engineering failure. It means the choice of which fairness criterion to satisfy is a value judgment that must be made openly rather than delegated to a vendor.

Key idea: Actuarial risk tools beat clinical judgment but achieve only moderate accuracy, low base rates guarantee many false positives, and no instrument can satisfy every fairness definition at once when group base rates differ.

Common misconceptions

• Most violent crime is committed by people with mental illness. Serious mental illness accounts for a small share of overall violence.
• Psychopath and antisocial personality disorder are the same thing. They overlap but most prisoners meeting ASPD criteria do not score high on the PCL-R.
• Psychopathy is untreatable. The claim is stronger than the evidence, and some intensive programs report reduced violent recidivism.
• Childhood risk factors can identify future criminals. They shift probabilities across populations; most high-risk children never become chronic offenders.
• A good risk instrument makes accurate individual predictions. Moderate accuracy plus a low base rate guarantees many false positives.

Recap

• Psychodynamic criminology was influential but weakly evidenced; attachment and conditioning ideas survive in modern form.
• Low agreeableness, low conscientiousness, impulsivity, and modestly lower verbal ability are the most consistent psychological correlates of offending.
• Psychopathy is defined by affective, interpersonal, and lifestyle traits, measured by the PCL-R, and predicts recidivism moderately.
• Serious mental illness contributes little to overall violence, while substance use and prior violence predict much better.
• Longitudinal risk factors are cumulative and probabilistic, and risk instruments beat clinical judgment while producing many false positives.

Sources

1. Miller, J. D., & Lynam, D. (2001). Structural models of personality and their relation to antisocial behavior: A meta-analytic review. Criminology, 39(4), 765-798. doi.org
2. Wikipedia contributors. (n.d.). Psychopathy Checklist. Wikipedia. en.wikipedia.org
3. Farrington, D. P. (2003). Key results from the first forty years of the Cambridge Study in Delinquent Development. In Taking stock of delinquency (pp. 137-183). Springer. doi.org
4. National Institute of Mental Health. (n.d.). Mental illness. National Institutes of Health. nimh.nih.gov
5. National Institute of Justice. (n.d.). Risk assessment. Office of Justice Programs. nij.ojp.gov

The big picture

Take a map of any American city and shade it by violent crime rate. The result is not random and it is not evenly spread. Crime piles up in specific neighborhoods, and often in specific blocks within those neighborhoods, and it stays there. In many cities you could hand someone a crime map from 1990 and a crime map from 2020 and they would look disturbingly alike.

That stability is the puzzle this lesson addresses. If crime were simply a matter of bad individuals, places should change as their populations turn over. They largely do not. Something about the place itself persists.

The tradition that grew from this observation is the most productive in criminology. It produced the finding that crime concentrates in a tiny fraction of street segments, the concept of collective efficacy, and much of the evidence behind the policing strategies in Module 6.

Key idea: Crime concentrates in particular neighborhoods and blocks and stays concentrated there across decades of population turnover, which means something about places, not only about people, is doing causal work.

The Chicago School and Shaw and McKay

In the early twentieth century, sociologists at the University of Chicago treated their city as a laboratory. Robert Park and Ernest Burgess described urban growth as a series of concentric zones radiating from a central business district, with a zone in transition just outside the center where industry encroached, housing decayed, rents fell, and new arrivals landed first.

Clifford Shaw and Henry McKay took this map and plotted the home addresses of tens of thousands of juveniles referred to Chicago courts across several decades. Their 1942 book reported two findings.

The first was that delinquency rates were highest in the zone in transition and declined steadily with distance from the center. The second, far more important, was that this pattern held across roughly thirty years even though the ethnic composition of those neighborhoods changed completely. German, Irish, Italian, Polish, and later Black and Mexican families moved through the same districts. The rates stayed with the districts, not with the groups.

That single finding demolished the then-common explanation that particular immigrant groups were inherently criminal. If group X has high rates in the zone in transition and low rates once it moves outward, and group Y then arrives and takes on high rates in the same place, the place is doing the work.

Key idea: Shaw and McKay showed that high delinquency areas in Chicago kept their rates across decades of complete ethnic turnover, refuting explanations based on the character of particular groups.

Social disorganization

Their explanation was social disorganization: the inability of a community structure to realize the common values of its residents and maintain effective social controls.

Three structural conditions were said to produce it. Poverty deprives residents of resources and drives out institutions. Residential mobility means people move constantly, so relationships never mature and nobody knows the neighbors. Ethnic heterogeneity, in a period of language barriers and mutual suspicion, impeded communication and joint action.

Crucially, these three do not cause crime directly. They cause a breakdown in informal social control, meaning the ordinary supervision, intervention, and shared expectation that keep behavior in check without any police involvement. Delinquent traditions then take root and are transmitted to new arrivals, which is why the pattern persists.

The theory attracted serious criticism. The term disorganization was pejorative and sometimes circular, since researchers occasionally inferred disorganization from the crime rate it was supposed to explain. Ethnographies found that neighborhoods labeled disorganized often had dense social organization, just not organization oriented toward mainstream institutions. And the theory took the structural conditions as given, saying little about the political and economic decisions, including redlining, disinvestment, and highway construction, that produced concentrated disadvantage in the first place.

Key idea: Social disorganization theory holds that poverty, residential mobility, and heterogeneity weaken informal social control rather than causing crime directly, though the concept was criticized as pejorative, sometimes circular, and silent about how disadvantage was produced.

Collective efficacy

The tradition was rebuilt in the 1990s, most influentially by Robert Sampson, Stephen Raudenbush, and Felton Earls using the Project on Human Development in Chicago Neighborhoods. They surveyed thousands of residents across hundreds of Chicago neighborhood clusters and measured something more specific than disorganization.

Collective efficacy combines two things: social cohesion and trust among neighbors, and the shared willingness to intervene for the common good. The second half is the innovation. Knowing your neighbors is not enough. What matters is whether residents would actually act, telling a group of children to stop skipping school, breaking up a fight, or calling about a broken streetlight.

Their 1997 paper in Science reported that collective efficacy was strongly and negatively associated with violence, and that it accounted for a substantial share of the association between concentrated disadvantage and residential instability on one hand and violence on the other. In their analysis, the structural conditions influenced violence largely through their effect on collective efficacy.

Two clarifications keep this from being overread. First, collective efficacy is not the same as friendship density. Neighborhoods with many close ties but no willingness to intervene, or with ties that shield offenders, do not get the benefit. Second, the original design was cross-sectional, so the direction of causation is contested: violence itself frightens people indoors and erodes willingness to intervene. Later longitudinal work supports effects running in both directions, which is a feedback loop rather than a clean one-way cause.

Key idea: Collective efficacy, meaning cohesion plus a shared willingness to intervene, is strongly associated with lower neighborhood violence and appears to mediate much of the effect of concentrated disadvantage, although violence also erodes collective efficacy in return.

Crime concentrates at micro-places

Zoom in further and the pattern sharpens dramatically. In 1989 Lawrence Sherman, Patrick Gartin, and Michael Buerger analyzed a year of calls for service in Minneapolis and found that about three percent of addresses generated roughly half of all calls. Many addresses generated none at all.

David Weisburd generalized this into what he calls the law of crime concentration: across cities, a small share of street segments, typically around five percent, accounts for about half of all crime, and roughly one percent accounts for about a quarter. His longitudinal work in Seattle found these hot segments to be remarkably stable across years, and found that adjacent segments can differ enormously, so that one block is a chronic hot spot while the block behind it is quiet.

This has two large implications. Practically, it means place-based interventions can be extraordinarily targeted, which is the basis of hot spots policing in Lesson 15. Theoretically, it means neighborhood-level explanations are too coarse. Whatever produces crime operates at the scale of a corner, a bar, a bus stop, or a single building, which points back to the situational factors from Lesson 4: guardianship, sightlines, the presence of a place manager, and the routine flow of people.

Key idea: Crime concentrates so tightly that roughly five percent of street segments account for about half of it, and this concentration is stable over time and varies sharply between adjacent blocks.

Concentrated disadvantage and immigration

Two contemporary findings extend this tradition and complicate popular assumptions.

The first concerns concentrated disadvantage. William Julius Wilson argued that deindustrialization and the departure of middle-class residents left some urban neighborhoods with a combination of joblessness, poverty, and institutional withdrawal that no earlier immigrant neighborhood had faced in the same form, and that residential segregation locked the pattern in place. The criminological implication is that what matters is not poverty as such but its concentration and duration, a claim supported by findings that neighborhood effects are largest where disadvantage has persisted across generations.

The second concerns immigration. Public belief in most countries holds that immigration raises crime. The American research literature consistently finds the opposite association: neighborhoods with higher concentrations of immigrants tend to have lower rates of violence, controlling for other characteristics, and Sampson's Chicago analyses found first-generation immigrants had substantially lower odds of violence than third-generation residents. Meta-analytic work on the relationship between immigration and crime at the area level finds effects clustering around zero to slightly negative.

Be careful about what this does and does not establish. These are area-level associations with the usual confounding risks, and area findings do not translate into claims about individuals. What the evidence does support is that the strong version of the popular claim, that immigration drives up crime rates, is not what the data show.

Key idea: Concentrated and persistent disadvantage rather than poverty alone predicts neighborhood violence, and immigrant concentration is consistently associated with lower rather than higher area violence in American research.

Can you test neighborhood effects?

Everything above is correlational, and neighborhoods are not randomly assigned. People sort into places for reasons connected to their own risk of offending, which is the selection problem in its purest form.

The most serious attempt to solve it was Moving to Opportunity, a federal randomized experiment beginning in the mid-1990s that offered families in high-poverty public housing vouchers to move to lower-poverty neighborhoods, with random assignment to voucher and control groups.

The crime results were not what neighborhood theory predicted. Effects on offending were mixed and modest. Some analyses found reductions in violent crime arrests among girls, alongside increases in property crime arrests among boys. Adult economic outcomes showed little improvement in the medium term. Mental health improvements, especially for adults and girls, were substantial and durable.

Later work by Chetty, Hendren, and Katz using tax records changed the picture again by looking at timing: children who moved before roughly age thirteen showed clear long-term gains in earnings and college attendance, while those who moved as teenagers did not, and may have been harmed by disruption. Exposure duration during childhood mattered.

The honest summary is that neighborhoods matter, but moving people is not a simple intervention, effects depend heavily on age and duration, and the mechanisms are not just about the crime rate of the destination. A strong theory that survives contact with a randomized trial usually comes out more complicated than it went in.

Key idea: Moving to Opportunity found mixed and modest crime effects with clear mental health benefits, and later work showed that long-term gains depended on moving young, which complicates simple claims about neighborhood effects.

Common misconceptions

• High-crime neighborhoods are socially empty. They usually have dense social ties; what varies is willingness to intervene and connection to institutions.
• Collective efficacy means everyone is friends. It is cohesion plus willingness to act, and friendship density alone does not deliver the benefit.
• Crime is spread evenly across a bad neighborhood. It concentrates on a small number of street segments, with quiet blocks immediately adjacent.
• Immigration raises neighborhood crime. American area-level research consistently finds the association runs the other way.
• Moving families out of poor neighborhoods reliably reduces offending. The randomized evidence found mixed, age-dependent effects.

Recap

• Shaw and McKay found delinquency rates stayed with Chicago districts across complete ethnic turnover, pointing to place rather than group.
• Social disorganization theory says poverty, mobility, and heterogeneity weaken informal social control, which then permits crime.
• Collective efficacy, cohesion plus willingness to intervene, is strongly associated with lower violence and mediates much of concentrated disadvantage.
• Crime concentrates at micro-places, with roughly five percent of street segments generating about half of all crime.
• Moving to Opportunity showed mixed crime effects and clear mental health gains, with long-term benefits depending on moving in early childhood.

Sources

1. Sampson, R. J., Raudenbush, S. W., & Earls, F. (1997). Neighborhoods and violent crime: A multilevel study of collective efficacy. Science, 277(5328), 918-924. doi.org
2. Sherman, L. W., Gartin, P. R., & Buerger, M. E. (1989). Hot spots of predatory crime: Routine activities and the criminology of place. Criminology, 27(1), 27-56. doi.org
3. Weisburd, D. (2015). The law of crime concentration and the criminology of place. Criminology, 53(2), 133-157. doi.org
4. Chetty, R., Hendren, N., & Katz, L. F. (2016). The effects of exposure to better neighborhoods on children: New evidence from the Moving to Opportunity experiment. American Economic Review, 106(4), 855-902. doi.org
5. Wikipedia contributors. (n.d.). Social disorganization theory. Wikipedia. en.wikipedia.org

The big picture

The last lesson explained where crime happens. This one asks what pressure produces it. The family of theories here shares one intuition: crime is not simply a defect in individuals but a predictable response to a society that generates wants it cannot satisfy for everyone.

That intuition has two versions and they are often confused. The structural version says the gap between what a society tells people to want and the means it distributes for getting it produces pressure toward illegitimate methods. The individual version says that being treated badly, losing what you value, or failing at what matters to you produces anger, and anger produces crime.

Both have evidence. The structural version is weaker than a century of textbooks implied, and the individual version is one of the better-supported theories in the field. Sorting out which is which is the work of this lesson.

Key idea: Strain theories explain crime as a response to pressure, with a structural version about the gap between cultural goals and legitimate means and an individual version about negative experiences producing anger.

Durkheim, anomie, and Merton

Emile Durkheim introduced anomie, a condition of normlessness that arises when rapid social change loosens the regulation of human desires. His concern was that unregulated aspiration produces chronic dissatisfaction, since wants expand without limit.

Robert Merton, in a 1938 paper that became one of the most cited in sociology, redirected the concept. For Merton the problem was not weak regulation of desire but a mismatch inside the culture itself.

American society, he argued, preaches a universal goal of monetary success and personal advancement, and preaches it to everyone regardless of position. It does not distribute the institutionalized means, meaning education, credentials, capital, and connections, anywhere near as evenly. The result is structural pressure concentrated among those who accept the goal and lack the means.

Merton laid out five modes of adaptation, defined by whether a person accepts the cultural goals and the legitimate means.

Adaptation	Goals	Means	Example
Conformity	Accept	Accept	Working and saving toward success
Innovation	Accept	Reject	Theft, fraud, drug selling
Ritualism	Reject	Accept	Going through the motions with no ambition
Retreatism	Reject	Reject	Withdrawal, chronic addiction, vagrancy
Rebellion	Replace	Replace	Political radicalism proposing new goals

Innovation is the adaptation that produces most ordinary crime, and it is worth noticing that Merton's framework applies as readily to a corporate executive falsifying earnings as to a burglar. The pressure is generated by the goal, not by poverty as such.

Key idea: Merton argued that a culture pushing universal monetary success while distributing legitimate means unequally generates pressure toward innovation, one of five modes of adaptation.

How well does classic strain hold up?

Less well than its fame suggests, and the reasons are instructive.

The most direct test asks whether the gap between a person's aspirations and expectations predicts delinquency. Studies through the 1970s and 1980s repeatedly found this discrepancy measure to be a weak predictor, and sometimes found that youths with the highest aspirations offended least. That is a serious problem for the mechanism as stated.

Second, if strain is generated by blocked access to money, offending should be strongly concentrated among the poorest. Self-report data show a much weaker class gradient for common delinquency than arrest data do, although the relationship strengthens for serious and persistent offending. Poverty matters, but not in the simple dose-response way the theory implies.

Third, absolute deprivation does not track crime well across societies, while relative deprivation, the sense of unjust disadvantage compared with a relevant reference group, does better. Cross-national work generally finds income inequality more strongly associated with homicide than poverty levels alone, which is more consistent with a comparison-based mechanism than a subsistence one.

Steven Messner and Richard Rosenfeld revised the theory as institutional anomie theory, arguing that the American problem is the dominance of the economy over other institutions: family, school, and politics get subordinated to market demands, weakening their capacity to restrain. The testable implication is that societies which buffer citizens from pure market dependence, through generous welfare provision, should have lower homicide. Cross-national studies broadly support this association, with the usual warning that comparing countries introduces enormous confounding.

Key idea: Classic strain performed poorly in tests using aspiration gaps and class gradients, while relative deprivation, inequality, and institutional anomie versions of the argument have fared better.

General strain theory

Robert Agnew rebuilt strain theory in 1992 by widening it beyond money and grounding it in emotion. General strain theory identifies three sources of strain.

The first is the failure to achieve positively valued goals, which includes Merton's money gap but also fairness and respect. The second is the removal of positively valued stimuli: a parent dies, a relationship ends, a job disappears. The third is the presentation of negative stimuli: abuse, bullying, criminal victimization, discrimination, humiliation.

Strain does not cause crime directly. It produces negative emotions, above all anger, which lowers inhibition, creates a desire for retaliation, and generates pressure for corrective action. Crime is one way of coping.

Crucially, most strained people do not offend, and Agnew specified what conditions the outcome. Coping resources, problem-solving skill, social support, conventional commitments, and personality traits such as low negative emotionality and high constraint all shape whether strain converts into offending.

The evidence is comparatively good. Strains of the third type in particular, especially victimization, child abuse, discrimination, and bullying, are consistently associated with offending across many samples, and anger partially mediates the relationship. Effect sizes are modest, and much of the research is cross-sectional, but longitudinal tests generally support the ordering.

The main criticism is falsifiability: if any negative experience counts as strain, the theory cannot fail. Agnew took this seriously and specified which strains should matter most: those seen as high in magnitude, perceived as unjust, associated with low social control, and creating some pressure or incentive for criminal coping. That specification is what makes the theory testable rather than merely plausible.

Key idea: General strain theory holds that goal blockage, loss, and noxious experiences produce anger, which pressures people toward criminal coping unless conditioning factors intervene, and it is best supported for victimization, abuse, and discrimination.

Subcultural explanations

A parallel tradition asked what happens when strained people find each other. Their answer was that groups develop alternative standards under which the excluded can succeed.

Albert Cohen argued in 1955 that working-class boys are judged in school by a middle-class measuring rod they are poorly prepared to meet, producing status frustration. Some respond by inverting the standard, forming groups that value precisely what school condemns. Cohen used this to explain a fact his contemporaries found puzzling: much delinquency is not profitable at all. Vandalism, joyriding, and gratuitous fighting make sense as status-earning within an inverted system, and make no sense as income.

Richard Cloward and Lloyd Ohlin added in 1960 that illegitimate opportunity is also unequally distributed. Their differential opportunity theory holds that becoming a successful thief requires access to skills, fences, and mentors just as a legitimate career requires training. Where stable criminal organization exists, a criminal subculture develops; where it does not, a conflict subculture organized around violence and turf tends to appear; and those who fail in both form a retreatist subculture centered on drugs.

Walter Miller went further, arguing that lower-class culture had its own focal concerns, including trouble, toughness, smartness, excitement, fate, and autonomy, which produced delinquency as a byproduct of ordinary conformity to that culture. This has aged worst of the three. It treats a class as culturally uniform, relies on impressionistic ethnography, and comes close to blaming poverty on the values of the poor.

Key idea: Subcultural theories explain non-utilitarian delinquency as status-seeking within inverted standards, and differential opportunity adds that illegitimate paths are themselves unequally available.

The code of the street, and what it is not

The most influential modern subcultural work is Elijah Anderson's ethnography of inner-city Philadelphia. Anderson describes a code of the street, an informal set of rules about respect and its defense.

His central claim is widely misread, so state it precisely. Anderson distinguishes decent families, who hold mainstream values, from street families, who are more invested in the code, and he insists that decent families are the majority. Even so, young people from decent families must learn to perform the code in public, because in an environment where calling the police is not a reliable option, a reputation for being unwilling to defend yourself invites predation.

The code is therefore an adaptation to the absence of effective protection, not a rejection of mainstream values. That distinction is the whole point and it is what separates Anderson from Miller.

Quantitative work supports parts of it. Studies by Eric Stewart and Ronald Simons found that adolescents scoring higher on street code adherence were more likely to engage in violence, net of controls, and that neighborhood context amplified the effect.

The most important corrective comes from Robert Sampson and Dawn Bartusch, who found that residents of disadvantaged Chicago neighborhoods were, if anything, less tolerant of deviance than residents of advantaged ones, while being far more cynical about the law and the police. Later work by David Kirk and Andrew Papachristos developed this as legal cynicism: a shared belief that law enforcement is illegitimate, unresponsive, and unable to provide safety. Neighborhoods high in legal cynicism show elevated violence even controlling for disadvantage.

This reframes the whole subcultural tradition. What distinguishes high-violence neighborhoods is usually not different values about right and wrong. It is a well-founded belief that the official system will not protect you, which makes private violence a rational and sometimes necessary form of self-help.

Key idea: Anderson's code of the street is an adaptation to the unavailability of protection rather than a rejection of mainstream values, and research on legal cynicism shows residents of high-violence areas often hold conventional values while distrusting the law.

Common misconceptions

• Merton's theory says poverty causes crime. It says the gap between universal goals and unequally distributed means causes pressure, which applies to executives too.
• Strain theory is well confirmed in its classic form. Aspiration gap tests performed poorly; the general and institutional versions carry the evidence.
• General strain theory predicts that strained people offend. It predicts that strain plus anger plus poor coping resources raises the probability.
• Subcultural theory means poor communities hold criminal values. Research finds residents of disadvantaged neighborhoods are often less tolerant of deviance and more cynical about the law.
• The code of the street describes everyone in those neighborhoods. Anderson describes a minority strongly invested in it and a majority who must navigate it.

Recap

• Merton located crime in the mismatch between universal cultural goals and unequally distributed legitimate means, with innovation the crime-producing adaptation.
• Classic strain tests were weak, while relative deprivation, inequality, and institutional anomie versions performed better.
• Agnew's general strain theory adds loss and noxious experience, works through anger, and is conditioned by coping resources.
• Subcultural theories explain non-profitable delinquency as status-seeking, and differential opportunity notes that illegitimate paths are also unequal.
• The code of the street is an adaptation to unreliable protection, and legal cynicism rather than deviant values distinguishes many high-violence neighborhoods.

Sources

1. Merton, R. K. (1938). Social structure and anomie. American Sociological Review, 3(5), 672-682. doi.org
2. Agnew, R. (1992). Foundation for a general strain theory of crime and delinquency. Criminology, 30(1), 47-88. doi.org
3. Sampson, R. J., & Bartusch, D. J. (1998). Legal cynicism and (subcultural?) tolerance of deviance. Law & Society Review, 32(4), 777-804. doi.org
4. Britannica. (n.d.). Anomie. Encyclopaedia Britannica. britannica.com
5. Wikipedia contributors. (n.d.). Strain theory (sociology). Wikipedia. en.wikipedia.org

The big picture

Two of the most productive theories in criminology are mirror images of each other, and both were largely built by the same man.

Learning theory asks the obvious question: how does a person come to offend? Its answer is that offending is learned in ordinary social relationships, exactly like anything else people learn.

Control theory refuses the question. Travis Hirschi's move was to ask instead why most people do not offend, given that crime is often easy, profitable, and immediately rewarding. On this view offending needs no special motivation. What needs explaining is restraint.

Between them these two traditions account for a large share of what criminology can reliably predict about individuals. They also generate the field's most famous unresolved argument, about whether delinquent friends cause delinquency or delinquents simply find each other.

Key idea: Learning theory explains how offending is acquired through social relationships, while control theory asks instead why most people refrain, and the two together carry much of criminology's predictive power at the individual level.

Sutherland's differential association

Edwin Sutherland set out differential association in nine propositions, and the important ones are these.

Criminal behavior is learned, not inherited or invented. It is learned in interaction with other people, and principally within intimate personal groups, which means impersonal sources such as media play a small part compared with friends and family. What is learned includes both techniques, sometimes complicated and sometimes very simple, and the motives, drives, rationalizations, and attitudes that go with them.

The central proposition is the sixth. A person becomes delinquent because of an excess of definitions favorable to violation of law over definitions unfavorable to it. A definition here is an attitude or justification: that this particular law is stupid, that the victim deserves it, that everyone does it, that this does not really count as stealing.

Associations vary in frequency, duration, priority, and intensity. Priority means early associations weigh more. Intensity means associations with people who matter to you weigh more. This is why a distant acquaintance who offends influences you far less than a close friend.

The ninth proposition is the one people forget and is Sutherland's sharpest blow against strain theory. Criminal behavior is an expression of the same general needs and values as non-criminal behavior, so it cannot be explained by those needs. Both the thief and the honest worker want money. Wanting money explains neither.

The theory's real triumph was scope. Sutherland used it to explain white-collar crime, arguing that a young executive learns definitions favorable to price fixing from senior colleagues in exactly the way a young man learns burglary from an older one. No theory built on poverty or personal pathology can do that.

Key idea: Sutherland argued that criminal behavior is learned in intimate groups and that offending follows from an excess of definitions favorable to law violation, a mechanism that explains white-collar and street crime alike.

Akers and social learning

Sutherland described what is learned but not the mechanism. Ronald Akers supplied it by adding behavioral psychology, producing social learning theory with four components.

Differential association supplies the social context and exposure. Definitions are the attitudes and justifications. Differential reinforcement is the balance of anticipated and actual rewards and punishments for the behavior, which come mostly from peers rather than from the legal system. Imitation is modeling of observed behavior, which matters most for acquiring a new behavior rather than sustaining it.

The addition of reinforcement solved a real problem. Sutherland could not explain why a person exposed to criminal definitions might not act on them. Akers can: because the behavior was punished, socially or otherwise, or because a competing behavior was better rewarded.

A large meta-analytic review of macro and micro predictors by Travis Pratt and colleagues found social learning variables, particularly definitions and differential association with delinquent peers, among the strongest correlates of offending in the entire literature. On raw predictive power this tradition performs about as well as anything criminology has.

Key idea: Akers added differential reinforcement and imitation to Sutherland's account, and meta-analytic work finds social learning variables among the strongest correlates of offending in criminology.

The selection versus socialization problem

Now the complication, and it is a serious one. That delinquent youths have delinquent friends is one of the most reliable findings in the field. What it means is contested.

Socialization says friends influence you: exposure to their definitions and reinforcement changes your behavior. Selection says you chose them: people with existing tendencies seek each other out, so the correlation reflects sorting rather than influence. Control theorists have long argued the selection version, since on their account delinquents have weak bonds and end up together by default.

Two features of the data make this hard. Most surveys ask respondents to report their friends' delinquency, and people project their own behavior onto their friends, inflating the correlation. And friendship and behavior change together, so cross-sectional data cannot order them.

The resolution came from better designs. Longitudinal studies that measure friends' behavior from the friends themselves, and network studies that track who befriends whom over time, generally find both processes operating, with selection often somewhat larger but socialization real and non-trivial. Experimental work in developmental psychology supports influence directly: adolescents in a driving simulation took substantially more risks when peers were watching, while adults did not.

There is also a structural fact that supports the learning tradition regardless of mechanism. Juvenile offending is overwhelmingly co-offending: young people commit crimes in groups, while adult offenders act alone far more often. Whatever else groups do, they are the setting in which youth crime actually happens.

Key idea: Delinquent peers and delinquency are correlated through both selection and socialization, better designs find both processes real, and the group nature of juvenile co-offending makes peers central regardless.

Hirschi's social bond theory

Travis Hirschi's 1969 book turned the question around. If crime is often easy and rewarding, motivation is not scarce. What restrains people is their bond to conventional society, which he broke into four elements.

Attachment is emotional connection to others, especially parents, teachers, and peers, which makes their disapproval costly. Commitment is the rational stake accumulated in conventional activity: education, reputation, a career, things you would lose. Involvement is time absorbed by conventional activity, the idle hands argument. Belief is acceptance of the moral validity of the law.

Hirschi's own data supported most of this, and later research largely agrees on attachment, commitment, and belief, which show consistent negative associations with delinquency. Involvement is the weak element. Keeping teenagers busy has not proven to be much of a protective factor, and some structured activities are associated with more delinquency, presumably because they concentrate teenagers together with limited supervision.

Note also a subtlety about attachment that the theory handles well and popular versions do not. In Hirschi's account attachment to peers should reduce delinquency regardless of whether those peers are delinquent, since sensitivity to others is what restrains. Learning theory predicts the opposite for delinquent peers. This is one of the few places where the two traditions make directly opposed predictions, and the evidence broadly favors learning theory here: attachment to delinquent peers is not protective.

Key idea: Hirschi explained conformity through attachment, commitment, involvement, and belief, with the first, second, and fourth well supported while involvement performs poorly.

Self-control and the general theory

In 1990 Michael Gottfredson and Hirschi made a far bolder claim. All crime, they argued, plus analogous behaviors such as smoking, heavy drinking, accidents, and unsafe sex, results from low self-control.

Low self-control on their account is a stable trait comprising impulsivity, preference for simple tasks, risk seeking, physical rather than mental activity, self-centeredness, and quick temper. It is produced by ineffective parenting, specifically failure to monitor behavior, recognize deviance, and punish it consistently, and it is set by roughly age eight to ten and stable thereafter. Crime then occurs when a person with low self-control meets an opportunity.

Its explanatory reach is impressive. It accounts for the age-crime curve, for the versatility of offenders who rarely specialize, for the co-occurrence of offending with accidents and health problems, and for why offenders' lives are disorderly in non-criminal ways too.

The evidence, and the criticism, are both substantial. A meta-analysis by Pratt and Cullen found self-control among the strongest known correlates of crime, with an average effect around 0.26 in correlational terms, robust across measurement approaches. That is a genuine success.

But the same meta-analysis found that social learning variables retained independent effects after controlling for self-control, which directly contradicts the theory's claim to be a general and sufficient explanation. Three further criticisms bite.

The first is tautology. If self-control is measured partly by asking whether people do impulsive and deviant things, then finding that low self-control predicts crime is close to circular. Behavioral measures that avoid this, such as laboratory tasks, generally show weaker relationships than attitudinal scales.

The second is the stability thesis. Longitudinal research, including work by Callie Burt and colleagues, finds that self-control does change after age ten, that changes are predicted by later family and neighborhood conditions, and that those changes predict offending. If self-control is not fixed, a central pillar falls.

The third is parenting. The theory attributes self-control almost entirely to parental management, which understates schooling, neighborhood, genetic contributions, and the fact that difficult children shape their parents' behavior as well as the reverse.

Hirschi later reformulated self-control as the tendency to consider the full range of consequences of an act, which brings it strikingly close to his own earlier bond theory. That convergence is a fitting end for a debate between two theories written by the same author.

Key idea: Self-control is one of criminology's strongest correlates of offending, but claims that it is the sole cause, fixed by age ten, and produced only by parenting are contradicted by evidence, and its measurement risks tautology.

Common misconceptions

• Differential association means bad friends make you bad. It specifies an excess of definitions favorable to law violation, weighted by frequency, duration, priority, and intensity.
• Learning theory only explains street crime. Sutherland built it to explain white-collar offending, where it works particularly well.
• Peer effects are entirely selection. Better designs find both selection and socialization, and experiments show direct peer influence on adolescent risk taking.
• Keeping teenagers busy prevents delinquency. Involvement is the weakest element of bond theory and the idle hands argument has poor support.
• Self-control theory is confirmed because self-control predicts crime. Prediction is strong, but the sufficiency, stability, and parenting claims all face contrary evidence.

Recap

• Sutherland held that crime is learned in intimate groups through an excess of definitions favorable to law violation.
• Akers added differential reinforcement and imitation, and social learning variables are among criminology's strongest correlates.
• The peer-delinquency link involves both selection and socialization, and juvenile crime is overwhelmingly committed in groups.
• Hirschi's bond theory explains conformity through attachment, commitment, involvement, and belief, with involvement the weakest element.
• Self-control predicts offending strongly, but the general theory's claims of sufficiency, early fixation, and parental origin are contested.

Sources

1. Pratt, T. C., & Cullen, F. T. (2000). The empirical status of Gottfredson and Hirschi's general theory of crime: A meta-analysis. Criminology, 38(3), 931-964. doi.org
2. Pratt, T. C., Cullen, F. T., Sellers, C. S., Winfree, L. T., Madensen, T. D., Daigle, L. E., Fearn, N. E., & Gau, J. M. (2010). The empirical status of social learning theory: A meta-analysis. Justice Quarterly, 27(6), 765-802. doi.org
3. Gardner, M., & Steinberg, L. (2005). Peer influence on risk taking, risk preference, and risky decision making in adolescence and adulthood. Developmental Psychology, 41(4), 625-635. doi.org
4. Wikipedia contributors. (n.d.). Differential association. Wikipedia. en.wikipedia.org
5. Wikipedia contributors. (n.d.). Social control theory. Wikipedia. en.wikipedia.org

The big picture

Every theory so far has taken the criminal law as given and asked why people break it. This lesson takes the law itself as the thing to be explained.

The traditions gathered here share one move: they shift attention from the offender to the audience. Who decides what counts as crime? What happens to a person once the system has processed and labeled him? Whose harms get criminalized and whose get regulated, fined, or ignored?

These questions are easy to dismiss as politics rather than science, and some of the writing in this tradition earns that dismissal. But two of its central claims have survived hard empirical testing better than almost anything else in criminology, and one of them, that formally processing juveniles makes them more likely to offend, has been confirmed by randomized experiments. That deserves your attention regardless of where your politics sit.

Key idea: Critical traditions shift the question from why people offend to who defines and enforces crime, and their strongest claims, especially about the effects of formal processing, have survived rigorous testing.

Labeling theory

Frank Tannenbaum described in 1938 what he called the dramatization of evil: a community responds to a young person's misbehavior by defining him as a delinquent, and the definition becomes a self-fulfilling identity.

Edwin Lemert gave this its key distinction. Primary deviance is the initial rule-breaking, which is common, often trivial, and has many causes. Secondary deviance is offending that occurs in response to society's reaction to the primary deviance, once the person has been identified, treated, and begun to see himself as the kind of person the label describes.

Howard Becker sharpened it in 1963. Social groups, he argued, create deviance by making rules whose infraction constitutes deviance, and applying those rules to particular people. Deviance is therefore not a quality of the act but a consequence of the application of rules and sanctions. Becker also introduced moral entrepreneurs, the crusaders and rule enforcers who work to have particular behaviors defined as criminal, and the idea of a master status, in which the label of criminal overrides every other identity a person has.

The mechanisms are concrete. A criminal record blocks employment, housing, and licensing. Formal processing removes a young person from school and from conventional peers, and places him among others who are similarly labeled. Family and teachers adjust their expectations. And the person's own self-concept shifts.

Key idea: Labeling theory distinguishes primary deviance from secondary deviance produced by society's reaction, and identifies concrete mechanisms including record-based exclusion, changed expectations, and altered self-concept.

Does labeling actually work that way?

For decades the theory was criticized as unfalsifiable and politically motivated. Then people tested it.

The strongest evidence comes from a Campbell Collaboration systematic review by Anthony Petrosino and colleagues, which pooled randomized experiments comparing juveniles formally processed by the justice system against those diverted or released. Across those experiments, formal processing produced no crime-reduction benefit, and the pooled effect pointed toward slightly more subsequent delinquency, not less. That is a remarkable finding: randomized evidence that the intervention meant to reduce offending increases it.

Longitudinal research fills in the mechanism. Work by Jon Gunnar Bernburg and Marvin Krohn using the Rochester Youth Development Study found that official intervention in adolescence predicted reduced educational attainment and employment in early adulthood, and that this blocked-opportunity pathway accounted for much of the increased later offending. The label works through life chances, not only through identity.

The theory still has real limits. It says little about primary deviance, which is where most explanation is needed. It can portray offenders as passive recipients of labels. And deterrent effects do exist for some people in some circumstances, so reaction is not uniformly criminogenic. The defensible modern position is that formal system contact carries a real criminogenic risk that must be weighed against any deterrent or incapacitative benefit, which is exactly how diversion policy is now justified.

Key idea: Randomized evidence pooled by the Campbell Collaboration found that formally processing juveniles produced no crime reduction and slightly increased later offending, with longitudinal work tracing the effect through damaged education and employment.

Conflict and critical criminology

Conflict criminology begins from the observation that criminal law is written by someone. Where consensus theories assume law expresses shared values, conflict theories argue that law reflects the interests of groups with the power to shape it.

Willem Bonger, writing in the Marxist tradition, argued that capitalism cultivates egoism in everyone, producing crime among the poor from need and among the wealthy from greed, while only the former is systematically punished. Richard Quinney argued that crime is a definition applied by authorized agents, and that the definitions favor the interests of the powerful.

The most memorable empirical illustration is William Chambliss's study of two groups of delinquent high school boys. The Saints were middle class, drove cars, misbehaved out of town, and spoke deferentially to police; the Roughnecks were working class, visible in their own neighborhood, and hostile when stopped. Chambliss recorded similar or greater rates of delinquency among the Saints, yet the Roughnecks acquired records and reputations while the Saints went to college. As an ethnography with two small groups it proves nothing on its own, but it illustrates a selection process that larger studies have since documented.

Critical criminology is the broad family that grew from this, and it is not one thing. Peacemaking criminology treats punishment itself as a form of the violence it claims to oppose. Green criminology extends the harm lens to environmental destruction. Cultural criminology studies the meanings and pleasures of transgression. Abolitionist scholarship argues for replacing rather than reforming penal institutions.

The most useful internal correction came from left realism, developed by Jock Young, John Lea, and others in Britain during the 1980s. They argued that critical criminology had drifted into treating crime as a mere label and street crime as trivial, when in fact working-class and minority communities suffer the highest rates of victimization and want protection. Left realism insisted on taking crime seriously as a real harm falling hardest on the least powerful, while keeping the structural critique. That combination is the mature version of this tradition.

Key idea: Conflict criminology treats criminal law as an expression of power rather than consensus, critical criminology extends this in several directions, and left realism corrected it by insisting that street crime is a real harm concentrated on the least powerful.

Disparity: what the evidence supports

The central empirical claim of this tradition is that the system treats people unequally. Handling this honestly requires separating questions that are usually run together.

The first question is whether disparities exist in outcomes. They plainly do, and Bureau of Justice Statistics data document large racial differences in arrest, incarceration, and supervision rates.

The second question is how much of that gap reflects differences in offending versus differences in treatment, and the answer differs sharply by offense.

For serious violence, the evidence suggests arrest disparities are not mostly generated by police discretion. When National Crime Victimization Survey respondents describe their assailants, the racial distribution of described offenders for robbery and serious assault tracks arrest data reasonably closely. Since victim descriptions are independent of police decision making, the convergence is informative.

For drug offenses, the picture reverses. Self-report surveys have long shown broadly similar rates of illegal drug use across racial groups, while arrest rates for drug offenses have differed by large multiples. Here the gap is generated by enforcement patterns: which neighborhoods are patrolled, which markets are visible, and which transactions occur outdoors.

For police stops, some of the strongest evidence comes from a large analysis of nearly 100 million traffic stops by Emma Pierson and colleagues. Using the veil-of-darkness test, which compares stop composition before and after dark when officers cannot easily see a driver's race, they found evidence of bias in stop decisions. They also found that officers applied a lower evidentiary threshold when deciding to search Black and Hispanic drivers, who were searched more often but found with contraband at similar or lower rates.

For sentencing, research by M. Marit Rehavi and Sonja Starr found that a large share of the federal sentence gap between Black and white defendants arises earlier, at the prosecutor's charging decision, particularly decisions to file charges carrying mandatory minimums, rather than at the judge's sentencing decision. That finding matters practically: it locates the leverage point.

The honest summary is that disparity is real, that its sources differ by offense and stage, that the strongest evidence of discriminatory treatment concerns discretionary low-level enforcement and charging rather than serious violence arrests, and that criminologists continue to disagree about magnitudes.

Key idea: Criminal justice disparities are real but arise from different sources at different stages, with victim-report data suggesting limited police discretion effects for serious violence and strong evidence of enforcement-driven disparity in drug arrests, stops, searches, and charging.

Feminist criminology

Kathleen Daly and Meda Chesney-Lind posed two problems in 1988 that mainstream criminology had barely noticed.

The generalizability problem asks whether theories built and tested almost entirely on boys and men explain women's offending. Some do reasonably well; others do not.

The gender ratio problem is sharper. Sex is the single largest and most stable correlate of crime ever measured. Men account for the large majority of arrests for violent offenses and nearly all homicide arrests, across every society and period for which we have data. Yet no mainstream theory was designed to explain it, and most cannot. If low self-control or strain or peer association explained offending, why would the sex difference be so enormous and so stable?

Pathways research examines how women arrive in the system, and finds a pattern distinct from men's. Studies of incarcerated women consistently find very high rates of prior physical and sexual abuse, trauma, and substance dependence, with offending frequently connected to those histories and to economic marginality. Chesney-Lind documented how girls' responses to abuse, such as running away, were historically criminalized as status offenses, so that the system punished a survival strategy.

John Hagan's power-control theory connects family structure to the gender gap, arguing that patriarchal households supervise daughters more closely than sons, producing a larger gender gap in delinquency, while more egalitarian households show a smaller one. Evidence is mixed but the approach usefully treats the gender gap as something produced rather than natural.

James Messerschmidt's work on masculinities inverts the usual question, asking not why women offend so little but why men offend so much, and treating some crime as a resource for accomplishing masculinity when other resources are unavailable.

Key idea: Feminist criminology raised the generalizability and gender ratio problems, documented distinct pathways into offending shaped by abuse and economic marginality, and reframed the enormous sex difference in offending as something requiring explanation.

The gender symmetry debate

One controversy inside this literature is worth working through, because it is a case study in how measurement produces disagreement.

Surveys using the Conflict Tactics Scales, which ask both partners how often they performed specific acts such as pushing or slapping, repeatedly find roughly symmetrical rates of partner violence by men and women. Studies using crime victimization surveys, police data, hospital records, and shelter samples find strongly asymmetrical patterns, with women far more likely to be seriously injured, killed, or subjected to sustained coercive control.

Michael Johnson's typology reconciles much of this. He distinguishes situational couple violence, which arises from specific conflicts, is often mutual, is usually less severe, and dominates general population surveys, from intimate terrorism, a pattern of coercive control backed by violence, which is heavily male-perpetrated and dominates the samples that agencies see. Two literatures were sampling two different phenomena and arguing as though they were one.

The lesson generalizes far beyond this topic. When two bodies of research on the same subject disagree completely, the first thing to examine is not who is biased but what each one actually measured and who ended up in each sample.

Key idea: Act-count surveys find symmetry in partner violence while agency and injury data find strong asymmetry, and Johnson's distinction between situational couple violence and intimate terrorism shows the two literatures were measuring different phenomena.

Common misconceptions

• Labeling theory claims the offender did nothing wrong. It concerns the effects of the reaction on subsequent offending, not the reality of primary deviance.
• Diverting juveniles is soft on crime. Randomized evidence finds formal processing produces no crime reduction and slightly more later offending.
• Critical criminology dismisses street crime. Left realism arose precisely to correct that error within the tradition.
• All criminal justice disparity comes from one source. Sources differ by offense and stage, with the clearest treatment effects in discretionary enforcement and charging.
• Partner violence research is hopelessly contradictory. The contradiction largely dissolves once situational couple violence and coercive control are separated.

Recap

• Labeling theory separates primary deviance from secondary deviance produced by the societal reaction, operating through records, exclusion, and identity.
• Randomized evidence shows formal juvenile processing does not reduce and may increase later offending.
• Conflict criminology treats law as an expression of power, and left realism restored attention to the real harm of street crime.
• Disparities are real but arise differently by offense and stage, with the clearest evidence in stops, searches, drug arrests, and charging decisions.
• Feminist criminology posed the generalizability and gender ratio problems and documented abuse-linked pathways into women's offending.

Sources

1. Petrosino, A., Turpin-Petrosino, C., & Guckenburg, S. (2010). Formal system processing of juveniles: Effects on delinquency. Campbell Systematic Reviews, 6(1), 1-88. doi.org
2. Bernburg, J. G., & Krohn, M. D. (2003). Labeling, life chances, and adult crime. Criminology, 41(4), 1287-1318. doi.org
3. Pierson, E., Simoiu, C., Overgoor, J., Corbett-Davies, S., Jenson, D., Shoemaker, A., Ramachandran, V., Barghouty, P., Phillips, C., Shroff, R., & Goel, S. (2020). A large-scale analysis of racial disparities in police stops across the United States. Nature Human Behaviour, 4, 736-745. doi.org
4. Rehavi, M. M., & Starr, S. B. (2014). Racial disparity in federal criminal sentences. Journal of Political Economy, 122(6), 1320-1354. doi.org
5. Wikipedia contributors. (n.d.). Labeling theory. Wikipedia. en.wikipedia.org

The big picture

Homicide is the crime criminology understands best, for a grim methodological reason: bodies are hard to hide, so the dark figure is small and the data are good.

What those data show is concentration. Homicide is not spread across the population like a disease with equal exposure. It concentrates by sex, by age, by relationship, by neighborhood, by block, and most strikingly by social network. In several cities researchers have identified networks containing a few percent of the population that account for the majority of gunshot victimization.

Then there is the firearm question, which is the most politically charged empirical dispute in criminology. This lesson handles it the way the best researchers do: by naming which specific policies have decent evidence, which have contested evidence, and which have almost none, and by being clear that the honest answer for many popular proposals is that we do not know.

Key idea: Homicide is measured well and is extraordinarily concentrated by demography, relationship, place, and social network, while the evidence on most firearm policies is far thinner than the confidence of public debate suggests.

The shape of American homicide

Several patterns hold consistently across decades of American data.

Sex. Roughly four out of five homicide victims and close to nine out of ten known offenders are male. This is among the largest and most stable patterns in criminology, and it holds internationally.

Age. Both offending and victimization peak in late adolescence and the twenties, then decline sharply. Lesson 14 examines why.

Relationship. Stranger homicide is a minority of cases. Most homicides with a known relationship involve acquaintances, associates, family members, or partners. Public fear focuses on the stranger case, which is the least common.

Weapon. Firearms are used in roughly three quarters or more of American homicides in recent years, a share that has risen since the 1990s.

Circumstance. A large share of homicides arise from arguments rather than from other felonies. David Luckenbill's classic analysis described homicide as a situated transaction, a character contest in which an insult is offered, a response is demanded by onlookers, and neither party can retreat without humiliation. Alcohol is present in a large share of these events. This is not the calculating offender of Lesson 4, and it explains why penalty severity has so little purchase here.

Clearance. Homicide clearance rates in the United States have declined over decades, falling from roughly 90 percent in the 1960s to around half in recent years. Since homicide is measured well, this decline is real, and it varies enormously between cities and between victim groups.

Key idea: American homicide is overwhelmingly male, young, arising from arguments among people who know each other, committed with firearms, and increasingly unsolved.

Concentration in networks

The most consequential recent finding in violence research concerns how tightly it concentrates.

Andrew Papachristos and colleagues mapped co-arrest networks in Chicago and Boston and found that gunshot victimization traveled through them. In one Chicago analysis, a network component containing under six percent of the neighborhood population accounted for roughly seventy percent of nonfatal gunshot injuries. Being socially close to a gunshot victim substantially raised a person's own risk, in a pattern the authors compared to contagion.

Two implications follow. First, the risk of serious violence is not well described by broad demographic categories. A young man in a high-violence neighborhood who is not connected to the relevant network has a risk far below what neighborhood-level statistics imply. Second, if violence concentrates in identifiable networks, interventions can be targeted at those networks rather than at whole neighborhoods, which is precisely the logic behind focused deterrence in Lesson 15.

The finding also reframes prevention. If exposure to violence raises the risk of further violence, then treating victims, interrupting retaliation, and addressing trauma become crime-control measures rather than merely humanitarian ones.

Key idea: Gunshot victimization concentrates in small social networks that account for the large majority of shootings, making network-targeted intervention feasible and reframing victim services as violence prevention.

Weapon instrumentality

Before any policy question, one empirical matter has to be settled: does the weapon matter, or only the intent?

The instrumentality position, argued most influentially by Franklin Zimring, holds that the weapon substantially affects whether an assault ends in death. Zimring's Chicago research compared knife and gun assaults with similar circumstances and intent and found gun assaults far more likely to kill, with case fatality ratios several times higher. Subsequent work has generally supported a large lethality difference.

The competing position holds that determined offenders substitute weapons, so restricting one merely shifts the method without changing deaths. The evidence suggests substitution is partial at best. Guns kill more efficiently, act at distance, permit multiple attempts in seconds, and require less physical commitment.

Instrumentality matters because it means the same number of violent intentions can produce very different numbers of deaths depending on what is at hand. It also generalizes: the same logic underlies the well-documented effect of restricting access to lethal means on suicide rates.

Key idea: Research on weapon instrumentality finds that gun assaults are several times more likely to kill than otherwise similar knife assaults, so weapon availability affects deaths even when intent is unchanged.

Firearm policy: what the strongest research shows

The RAND Corporation's Science of Gun Policy project is the most useful entry point, because it screens the literature by methodological quality rather than by conclusion. Its central finding is itself important: of thousands of studies, only a small fraction meet basic standards for causal inference, and for most policies the evidence is inconclusive.

Here is a fair summary of where things stand, with the direction of the disagreement named.

Child-access prevention laws, which impose liability for leaving firearms accessible to minors, have among the strongest supporting evidence in the field, associated with reductions in firearm suicides and unintentional firearm injuries and deaths among young people.

Stand-your-ground laws, which remove the duty to retreat, have moderately strong evidence of increasing homicide, with several studies finding increases in the range of roughly eight to eleven percent in adopting states.

Permit-to-purchase licensing has suggestive but contested evidence. Studies using synthetic control methods found that Connecticut's 1995 permit law was associated with a large reduction in firearm homicide, and that Missouri's 2007 repeal of its permit law was associated with a substantial increase. Critics argue the synthetic comparison units are sensitive to specification choices, and results shift under alternative designs.

Right-to-carry laws are the field's longest-running fight. John Lott argued in the 1990s that permissive carry laws reduce violent crime. A National Research Council committee reviewing that work in 2004 concluded the evidence was insufficient to determine any effect, with one dissent. More recent panel analyses by John Donohue and colleagues estimate that right-to-carry laws increase violent crime by roughly thirteen to fifteen percent over ten years. Other economists dispute the specification. The reasonable position is that the early claim of large crime reductions has not held up, while the size of any increase remains contested.

The federal assault weapons ban of 1994 to 2004 was evaluated for the Justice Department by Christopher Koper, who concluded that effects on crime were too small to be reliably measured, in part because the banned weapons were used in a small share of crimes and because the ban grandfathered existing stock.

Two contextual facts belong here. First, international comparison: research by Erin Grinshteyn and David Hemenway found the United States homicide rate several times higher than that of other high-income countries, with the firearm homicide rate many times higher, while non-firearm homicide was also somewhat elevated. Second, a long funding gap: from 1996 the Dickey Amendment was widely read as barring federal firearm injury research, and dedicated federal funding largely stopped for roughly two decades before resuming in 2019. Part of why the evidence base is thin is that it was not built.

Key idea: Child-access prevention laws have the strongest supporting evidence, stand-your-ground laws show moderate evidence of increasing homicide, permit-to-purchase and right-to-carry findings are contested, the assault weapons ban's effects were too small to measure, and much of the literature is thin because federal funding lapsed for two decades.

How to argue about guns honestly

Three habits keep this discussion useful.

Separate the outcomes. Firearm suicide, which accounts for over half of American gun deaths, responds to different policies than homicide does, and evidence on means restriction and suicide is comparatively strong. Mass shootings, which dominate coverage, are a small fraction of firearm deaths and are the hardest category to study because they are rare. Everyday interpersonal shootings account for most homicide. A policy can work on one and not another.

Separate the mechanisms. A law can act on who may buy, what may be sold, how it must be stored, who may carry, or what happens after a crime. These are not interchangeable, and evidence for one implies nothing about the others.

Take seriously that most such studies are quasi-experimental, using state-level variation over time. That design is vulnerable to other things changing simultaneously, which is exactly why estimates from different research teams diverge so often. When two credible teams get opposite signs, the honest conclusion is uncertainty, not that the team you disagree with is corrupt.

Key idea: Sound firearm analysis separates suicide, mass shootings, and everyday homicide, distinguishes the specific mechanism a law uses, and treats divergent quasi-experimental estimates as genuine uncertainty.

Hate crime

A hate crime is a criminal offense motivated in whole or in part by bias against a protected characteristic such as race, ethnicity, religion, sexual orientation, gender identity, or disability. The bias is not the crime; there must be an underlying offense, and the bias motivation triggers enhanced penalties or separate charges. The Supreme Court upheld penalty enhancement against a First Amendment challenge in Wisconsin v. Mitchell in 1993, reasoning that motive has always been a legitimate sentencing consideration.

Measurement here is unusually bad, and the gap is the most important fact to know. FBI hate crime statistics rest on voluntary agency reporting and typically record on the order of ten to twelve thousand incidents a year, with thousands of participating agencies affirmatively reporting zero incidents. Bureau of Justice Statistics analyses of the National Crime Victimization Survey estimate hate crime victimizations in the hundreds of thousands annually. The discrepancy is roughly two orders of magnitude.

The explanations are structural rather than mysterious: victims often do not report, officers must recognize and record a bias motivation, agencies must submit data, and many do not. This makes year-to-year changes in FBI hate crime counts extremely difficult to interpret, since a rise can reflect improved agency participation.

Substantively, race and ethnicity bias accounts for the largest share of reported incidents, followed by religion and sexual orientation. Research also finds that hate crimes are more likely than comparable offenses to involve violence against strangers and to produce heightened psychological harm to victims and to the wider targeted community, which is the standard justification for treating them as a distinct category.

Key idea: Hate crime requires an underlying offense plus a bias motivation, penalty enhancements were upheld in Wisconsin v. Mitchell, and voluntary FBI reporting records a small fraction of the victimizations that surveys estimate.

Common misconceptions

• Most homicides are committed by strangers. Cases with a known relationship overwhelmingly involve acquaintances, associates, or family.
• Homicide risk is spread across whole neighborhoods. It concentrates in small social networks that account for most shootings.
• The weapon does not matter, only intent. Gun assaults are several times more likely to kill than comparable knife assaults.
• The evidence on gun policy is clear. Most policies have inconclusive evidence by RAND's screening, and several key findings are actively disputed.
• FBI hate crime counts measure hate crime. They rest on voluntary reporting and capture a small fraction of survey-estimated victimizations.

Recap

• Homicide concentrates by sex, age, relationship, place, and social network, and often arises from escalating arguments.
• Network research finds small components accounting for the majority of gunshot victimization, enabling targeted intervention.
• Weapon instrumentality means gun assaults kill far more often than comparable assaults with other weapons.
• Child-access prevention has the strongest evidence, stand-your-ground shows evidence of harm, and much else is contested or inconclusive.
• Hate crime requires an underlying offense plus bias motive, and official counts capture a small fraction of estimated victimizations.

Sources

1. RAND Corporation. (n.d.). The science of gun policy. rand.org
2. Papachristos, A. V., Braga, A. A., & Hureau, D. M. (2012). Social networks and the risk of gunshot injury. Journal of Urban Health, 89(6), 992-1003. doi.org
3. Grinshteyn, E., & Hemenway, D. (2016). Violent death rates: The US compared with other high-income OECD countries, 2010. American Journal of Medicine, 129(3), 266-273. doi.org
4. Federal Bureau of Investigation. (n.d.). Hate crimes. fbi.gov
5. Bureau of Justice Statistics. (n.d.). Hate crime. Office of Justice Programs. bjs.ojp.gov

The big picture

Property crime is where most crime actually is. Larceny-theft alone dwarfs every violent category in volume, and property offenses make up the overwhelming majority of what people experience as crime.

It is also where the crime decline was largest and where criminology has its most practically useful findings. Burglars turn out to be studiable, their decision rules turn out to be simple, and interventions built on those rules turn out to work.

Then the lesson turns to two forms of acquisitive crime that the standard statistics handle badly. Organized crime is a business, and understanding it means asking what service it sells. Cybercrime and fraud are now an enormous share of what victimizes ordinary people, and they are nearly invisible in the measures from Lesson 2. That invisibility should change how confidently you repeat the claim that crime has fallen.

Key idea: Property crime dominates the volume of offending, its mechanisms are unusually well understood, and the growth of fraud and cybercrime is largely missing from the statistics used to describe crime trends.

How burglars actually decide

Some of the best evidence in criminology comes from interviewing active offenders. Richard Wright and Scott Decker interviewed over a hundred active residential burglars in St. Louis, recruited on the street rather than in prison, which matters because prison samples over-represent the unsuccessful.

What they found undercuts the image of the calculating professional. Most burglaries were driven by an immediate need for cash, often connected to drugs, alcohol, or the demands of a street lifestyle, and were decided quickly. Planning was thin. Targets were usually close to home, in areas the offender knew well.

Target selection ran on a short list of cues. Is anyone home? Occupancy is the dominant consideration, and burglars invest real effort in checking it. Can I be seen? Cover from neighbors and passing traffic matters more than lock quality. Can I get out? Escape routes and the presence of a dog rank high. Is there any sign of an alarm?

Two practical implications follow. First, most burglary prevention advice that focuses on hardware is aiming at a secondary consideration. Occupancy signals and visibility do more work. Second, since offenders operate in familiar areas near home, burglary is geographically predictable in a way that supports place-based prevention.

Key idea: Interviews with active burglars show quick decisions driven by immediate cash need, target selection dominated by occupancy cues, visibility, and escape routes, and offending concentrated near the offender's own area.

Stolen goods markets

A stolen television is not money. Mike Sutton's market reduction approach makes the obvious but neglected point that acquisitive crime depends on the ability to convert goods into cash, so the market for stolen goods is part of the causal system.

Sutton documented several market types, from commercial fences, to residential fences operating from a home, to network sales among acquaintances, to direct hawking on the street. Offenders often steal to order, meaning demand pulls theft rather than supply pushing it.

The policy implication is that disrupting resale can reduce theft without any change in the offender population, and evaluations of secondhand market regulation and of targeting fences have shown promise, though the evidence base is smaller than for place-based approaches.

The clearest modern illustration runs the other way. As scrap metal prices rose in certain periods, thefts of copper piping, catalytic converters, and cable rose with them, and fell again when prices dropped or when scrap purchase regulations tightened. Theft tracked resale value, which is close to a natural experiment on the point.

Key idea: Acquisitive crime depends on stolen goods markets, offenders often steal to order, and theft of commodities tracks their resale prices closely enough to make the market part of the causal explanation.

Repeat and near-repeat victimization

One of the most replicated findings in property crime is that victimization predicts victimization. A burgled house faces a substantially elevated risk of being burgled again, and the elevation is concentrated in the weeks immediately following, decaying over the following months.

The pattern extends spatially. Near-repeat victimization means that homes close to a burgled property also face short-term elevated risk, in a pattern criminologists compare to the way disease or aftershocks spread. The leading explanation is that a successful offender returns, using knowledge already gained, and applies it to similar nearby targets.

The practical value is high, because it converts a static risk map into a short-term forecast. Cocooning strategies, which alert and harden the immediate neighbors of a recent burglary, follow directly from this finding, as do several predictive policing systems. It is worth noting that the underlying finding about repeat victimization is far better established than the effectiveness of the commercial software products built on top of it.

Key idea: Burgled properties and their immediate neighbors face sharply elevated short-term risk, a repeat and near-repeat pattern that supports targeted short-term prevention.

Why property crime fell, and what came back

Lesson 3 introduced the security hypothesis. It is worth seeing the strongest case for it, because it is the rare criminological argument with a clean test.

Graham Farrell and colleagues noted that vehicle theft fell dramatically across many countries beginning in the 1990s, and that the timing in each country tracked the mandated introduction of electronic engine immobilizers. Cars that could not be hot-wired stopped being stolen at the old rates. Because the policy arrived at different times in different countries and applied to new vehicles first, the pattern is unusually well identified.

Then the same mechanism ran in reverse and confirmed itself. In the early 2020s, American thefts of certain Kia and Hyundai models rose enormously after it became widely known, spread through social media, that those models lacked immobilizers standard on competitors. Theft concentrated on precisely the vulnerable model years. That is about as close to a demonstration as observational criminology gets: the same population, the same cities, the same era, with theft tracking a single design feature.

This is worth remembering whenever a crime trend is attributed to broad social forces. Sometimes the cause is a part.

Key idea: Vehicle theft fell as electronic immobilizers were introduced country by country, and rose again on specific American models that lacked them, providing unusually clean evidence that security design drives theft rates.

Organized crime

Organized crime refers to continuing criminal enterprises that operate for profit through illicit markets, using violence and corruption to protect themselves. The definition is contested at the edges, particularly about how much hierarchy is required.

The most illuminating analysis comes from Diego Gambetta, who asked what the Sicilian mafia actually sells. His answer: private protection. In markets where the state will not enforce contracts, because the goods are illegal, or because state institutions are weak or corrupt, participants still need someone to guarantee deals, settle disputes, and punish cheats. Mafias supply that service and charge for it.

This reframes several puzzles. It explains why organized crime flourishes around prohibited goods, since illegal markets cannot use courts. It explains the emphasis on reputation and on visible violence, which are the advertising and enforcement of a protection business. And it explains why mafias appear in legal sectors such as construction and waste hauling, where they sell cartel enforcement to firms that want competition suppressed.

The American legal response centered on the Racketeer Influenced and Corrupt Organizations Act of 1970, which allowed prosecutors to charge an enterprise's pattern of activity rather than isolated offenses, reaching leaders who never touched a crime scene personally. RICO substantially damaged traditional American mafia families.

Contemporary scholarship emphasizes that most organized crime is less hierarchical than the classic image. Much of it is better described as fluid networks of loosely connected entrepreneurs assembling for particular ventures, which is harder to decapitate because there is often no head. Transnational activity, in drugs, smuggling, trafficking, counterfeiting, and fraud, follows trade routes and regulatory gaps rather than ethnic organization charts.

Key idea: Organized crime is best understood as a business supplying protection and contract enforcement where the state cannot or will not, and modern forms are more often fluid networks than hierarchies, which limits decapitation strategies.

Cybercrime and the fraud blind spot

Cybercrime divides usefully into two kinds. Cyber-dependent offenses could not exist without computers: hacking, ransomware, denial of service attacks, malware. Cyber-enabled offenses are old crimes at new scale: fraud, extortion, harassment, theft.

The scale is large and growing. The FBI's Internet Crime Complaint Center recorded roughly 880,000 complaints with reported losses over 12 billion dollars in 2023, and reported losses in excess of 16 billion dollars in 2024. Those figures count only complaints filed with one American agency, so they are a floor rather than a total.

Several features distinguish cybercrime analytically. Offending scales without proportional effort, since one actor can attempt millions of frauds. Distance collapses, so offender and victim are frequently in different jurisdictions, which makes enforcement enormously harder. Attribution is difficult. And business victims often decline to report, fearing reputational and regulatory consequences.

The theoretical challenge is real. Majid Yar argued that routine activity theory transfers imperfectly to cyberspace, because the theory's core requirement, convergence in time and space, does not hold when an offender and target need never occupy the same place, and when the target may be data that can be copied without being removed. The concepts of suitable target and capable guardian survive the translation; physical convergence does not.

Now the measurement point, which matters for everything in this course. Traditional American crime statistics were built for a world of physical property offenses and handle fraud badly. Fraud has historically been poorly captured in police-based counts, and much of it is never reported to police at all. When you read that property crime has fallen dramatically since the 1990s, the accurate statement is that recorded conventional property crime has fallen dramatically while a large and growing category of acquisitive victimization sits substantially outside the measurement system. Both parts of that sentence are needed.

Key idea: Cybercrime scales without proportional effort and crosses jurisdictions, and because fraud is poorly captured by traditional crime statistics, claims about the property crime decline must be qualified by what the measures cannot see.

Common misconceptions

• Burglars are skilled professionals who case targets for days. Most decide quickly under pressure for cash and work close to home.
• Better locks are the main defense against burglary. Occupancy cues, visibility, and escape routes weigh more heavily in target selection.
• Organized crime is defined by rigid ethnic hierarchies. Much of it operates as fluid networks assembled for particular ventures.
• Mafias exist because of culture. The protection theory explains them as suppliers of enforcement where the state does not provide it.
• Property crime simply fell. Recorded conventional property crime fell sharply while fraud and cybercrime grew largely outside the measures.

Recap

• Active offender interviews show burglary is quick, cash-driven, local, and governed by occupancy and visibility cues.
• Stolen goods markets are part of the causal system, and commodity theft tracks resale prices.
• Repeat and near-repeat victimization concentrate risk in the weeks after an offense and in adjacent properties.
• Vehicle immobilizers drove the international collapse in car theft, and their absence drove a recent American resurgence.
• Organized crime sells protection where the state does not, and cybercrime and fraud are largely invisible in traditional crime measures.

Sources

1. Federal Bureau of Investigation. (n.d.). Internet Crime Complaint Center (IC3). ic3.gov
2. Farrell, G., Tilley, N., & Tseloni, A. (2014). Why the crime drop? Crime and Justice, 43(1), 421-490. doi.org
3. Britannica. (n.d.). Organized crime. Encyclopaedia Britannica. britannica.com
4. Bureau of Justice Statistics. (n.d.). Property crime. Office of Justice Programs. bjs.ojp.gov
5. Wikipedia contributors. (n.d.). Cybercrime. Wikipedia. en.wikipedia.org

The big picture

This lesson pairs two subjects that are rarely taught together, and the pairing is the point. Drug offending is punished heavily and produces relatively contained direct harm to others. Corporate offending is punished lightly and produces enormous harm, including deaths. Put side by side, they raise the question that critical criminology asked in Module 4, now with numbers attached: what governs how hard a society punishes a harm?

The answer is not harm. It is a combination of who commits the offense, how visible it is, how easy it is to prove, and how the conduct is culturally coded. Understanding that mechanism is more useful than being outraged by it.

Both halves also have unusually good policy evidence, and in both cases the evidence points somewhere other than where policy has gone.

Key idea: How hard a society punishes a harm depends less on the harm's size than on the offender's status, the offense's visibility, and the difficulty of proof, and drug and corporate offending illustrate the two extremes.

The drug-crime relationship

Drug use and crime correlate strongly. Arrestees test positive at high rates, and offenders report far higher substance use than the general population. What that correlation means is the interesting part, and there are three candidate models.

The first is that drugs cause crime. The second is that crime causes drug use, since offending funds and normalizes it. The third is that both flow from common causes such as impulsivity, disadvantage, and peer environment.

The evidence supports a mixture, with an important refinement. For many offenders, delinquency precedes heavy drug use rather than following it, which undercuts the simple first model. But studies that follow individuals across time consistently find that offending frequency rises during periods of heavy use and falls during periods of abstinence or treatment. The best summary is that drugs do not typically create offenders, but they substantially intensify offending among people already at risk.

Paul Goldstein's tripartite framework then separates three distinct pathways from drugs to violence.

Psychopharmacological violence results from the drug's direct effects on the user, through intoxication, irritability, or withdrawal. Economic-compulsive violence occurs when users commit acquisitive crime to fund a habit and violence follows from the robbery or theft. Systemic violence is built into the drug market itself: territorial disputes, enforcement of debts, punishment of informants, robbery of dealers who cannot call the police.

When Goldstein and colleagues classified New York City homicides in the late 1980s, the largest share of drug-related killings by a wide margin were systemic. Very few were psychopharmacological, and alcohol, not illegal drugs, was the substance most often directly implicated in violence.

That result reorganizes the policy question. If most drug-related violence comes from the market rather than the substance, then it is a consequence of the market's illegality, since illegal markets have no courts. This is the same protection problem from the previous lesson, now driving homicide.

Key idea: Drugs intensify offending among people already at risk rather than creating offenders, and Goldstein's framework shows that most drug-related violence is systemic market violence rather than a pharmacological effect.

Enforcement versus treatment

Here the evidence is unusually pointed.

On enforcement, a systematic review by Dan Werb and colleagues examined studies of the effect of drug law enforcement on violence and found that the large majority reported enforcement associated with increases rather than decreases in drug market violence. The proposed mechanism is straightforward: removing a supplier destabilizes a market, and destabilized illegal markets are settled by force. This does not mean enforcement is never justified, but it is evidence against the assumption that more of it reliably reduces violence.

On treatment, the evidence is considerably better. A meta-analysis of drug courts by Ojmarrh Mitchell and colleagues found that adult drug court participants had substantially lower recidivism than comparison groups, with a difference of roughly ten to twelve percentage points in the studies reviewed. Effects for juvenile drug courts were smaller.

Stronger still is the evidence on medication for opioid use disorder. A systematic review and meta-analysis by Luis Sordo and colleagues found that all-cause mortality during methadone or buprenorphine treatment was roughly half or less of mortality out of treatment. Since offending falls during treatment periods, this is simultaneously a health intervention and a crime intervention, and it has among the strongest evidence bases in this entire course.

On decriminalization, Portugal is the case most often cited. Its 2001 reform moved possession of small quantities from criminal to administrative handling while expanding treatment. Drug-related HIV infections and drug-induced deaths fell substantially, treatment uptake rose, and there was no explosion in use. The honest caveat is that the reform bundled decriminalization with a large treatment expansion, so attributing the outcomes to the legal change alone is not warranted by the design.

Key idea: Reviews find drug law enforcement associated with increased market violence, while drug courts show modest recidivism reductions and medication for opioid use disorder roughly halves mortality, making treatment the better-evidenced approach.

White-collar crime

Edwin Sutherland coined white-collar crime in 1939 and defined it as crime committed by a person of respectability and high social status in the course of his occupation. The definition was a deliberate provocation, because it forced criminology to notice that its theories had been built entirely on the poor.

Modern scholarship splits it. Occupational crime is committed by individuals against their employer or clients: embezzlement, expense fraud, theft of trade secrets. Corporate or organizational crime is committed by the organization, or by people acting for it, against employees, customers, competitors, or the public: price fixing, securities fraud, environmental violations, safety violations, consumer fraud.

The distinction matters because the second type has no obvious individual beneficiary, is often the product of routine organizational pressure rather than deviant intent, and is precisely the type least likely to be prosecuted.

On scale, dollar estimates should be handled with care because they are soft. Even so, standard estimates of annual losses from white-collar and corporate offending in the United States exceed total reported losses from all conventional property crime by a large multiple, with health care fraud alone estimated in the tens of billions of dollars per year.

The physical harm is easier to underestimate and harder to dismiss. Workplace fatalities in the United States run in the thousands annually, a meaningful share involving violations of safety rules. Add deaths and injuries from unsafe products, from pollution, and from pharmaceutical and marketing misconduct, and the death toll attributable to corporate conduct is not obviously smaller than that from homicide. It is simply distributed differently: dispersed, delayed, statistical, and hard to attribute to a single actor.

Key idea: White-collar crime divides into occupational offending against an employer and corporate offending for the organization, and its financial and physical harms plausibly exceed those of conventional property and violent crime.

Why corporate crime is under-punished

The gap between harm and punishment has identifiable mechanisms, not just a conspiracy.

Proof is genuinely hard. Corporate offenses are documentary, technical, and slow. A homicide investigation has a body and a scene. A price-fixing case has millions of emails and a dispute about what a meeting meant.

Responsibility diffuses. Organizations distribute decisions across people and time, so no single actor may have both the guilty knowledge and the guilty act. Criminal law was built for individuals and handles distributed agency badly.

Much of it is routed to regulators. Conduct that could be charged criminally is frequently handled as a civil or administrative matter, producing fines that companies can treat as operating costs.

Resources are asymmetric. A prosecutor with a caseload faces a firm that can fund years of expert litigation. Prosecutors respond rationally by settling.

Institutional caution followed a shock. After the accounting firm Arthur Andersen collapsed in 2002 following indictment, taking tens of thousands of innocent jobs with it, prosecutors grew wary of charging firms directly. Deferred and non-prosecution agreements, in which a company pays and promises reform without conviction, became a standard tool. Judge Jed Rakoff argued influentially that the aftermath of the 2008 financial crisis showed the cost of this approach, since the crisis produced very few prosecutions of senior individuals.

Cultural coding. The conduct does not look like crime. There is no confrontation, the offender resembles the people who make and enforce the law, and neutralizations are readily available: everyone does it, no one was really hurt, the rules are arbitrary. Those are Sykes and Matza's techniques of neutralization, and they operate far more smoothly in a boardroom than on a street.

Key idea: Corporate crime is under-punished because proof is hard, responsibility diffuses across organizations, conduct is routed to regulators, resources are asymmetric, prosecutors turned to deferred agreements after Arthur Andersen, and the conduct does not read culturally as crime.

The deterrence asymmetry

Now a finding that cuts against the usual reflexes, and it is one of the most useful in this course.

Lesson 4 showed that severity deters weakly for street crime, because offenders discount the future, act under emotion or intoxication, and do not know the penalties. Every one of those conditions is reversed for corporate offenders.

They plan. They have lawyers who tell them exactly what the penalties are. They have enormous stakes in their futures, including careers, licenses, reputations, and liberty they expect to keep. They act deliberately over months. They are, in short, close to the rational calculator that classical criminology imagined and that street offenders are not.

The implication is that the population where deterrence should work best is the one the system deters least, and the population where deterrence works worst receives the most severe sanctions. That is not a slogan; it follows directly from the deterrence evidence applied consistently. It also suggests where prevention effort has the best expected return: certainty of detection through auditing and inspection, personal liability for identifiable decision makers rather than corporate fines alone, and penalties large enough that violation is not a rational operating cost.

Be careful not to overclaim. The direct evaluation literature on corporate crime deterrence is thin, precisely because these cases are rare and hard to study, and firms respond to enforcement in complex ways including relabeling and offshoring. The argument above is an inference from general deterrence findings, not a demonstration, and it should be stated as such.

Key idea: Corporate offenders match the rational calculator model far better than street offenders do, so deterrence should work best precisely where it is applied least, though the direct evaluation evidence for this inference remains thin.

Common misconceptions

• Drugs cause crime directly. Most drug-related violence is systemic market violence, and offending often precedes heavy use.
• More drug enforcement reduces drug violence. A systematic review found most studies reported the opposite association.
• Portugal proves decriminalization works by itself. The reform bundled legal change with a large treatment expansion, so the design cannot isolate either.
• White-collar crime is only financial. Safety, product, and environmental offenses produce deaths and injuries at meaningful scale.
• Corporate crime goes unpunished purely because of corruption. Proof difficulty, diffused responsibility, regulatory routing, and resource asymmetry do much of the work.

Recap

• Drugs intensify offending among those already at risk more than they create offenders.
• Goldstein's framework separates psychopharmacological, economic-compulsive, and systemic violence, with systemic dominant.
• Enforcement is associated with more market violence, while drug courts and medication treatment have better evidence.
• White-collar and corporate crime cause financial and physical harm plausibly exceeding conventional crime.
• Corporate offenders fit the rational calculator model, so deterrence should work best where it is applied least.

Sources

1. Werb, D., Rowell, G., Guyatt, G., Kerr, T., Montaner, J., & Wood, E. (2011). Effect of drug law enforcement on drug market violence: A systematic review. International Journal of Drug Policy, 22(2), 87-94. doi.org
2. Mitchell, O., Wilson, D. B., Eggers, A., & MacKenzie, D. L. (2012). Assessing the effectiveness of drug courts on recidivism. Journal of Criminal Justice, 40(1), 60-71. doi.org
3. Sordo, L., Barrio, G., Bravo, M. J., Indave, B. I., Degenhardt, L., Wiessing, L., Ferri, M., & Pastor-Barriuso, R. (2017). Mortality risk during and after opioid substitution treatment. BMJ, 357, j1550. doi.org
4. Britannica. (n.d.). White-collar crime. Encyclopaedia Britannica. britannica.com
5. National Institute on Drug Abuse. (n.d.). Treatment and recovery. National Institutes of Health. nida.nih.gov

The big picture

Plot arrests against age for almost any population, in almost any country, in almost any decade, and you get the same shape. Offending climbs steeply through adolescence, peaks somewhere in the late teens or early twenties, and then falls away, slowly at first and then decisively through the twenties and thirties.

This is the age-crime curve, and it is the most robust empirical regularity in criminology. It is also, in a way, the field's most awkward fact, because no theory in this course predicted it and every theory has had to accommodate it.

Two enormous questions follow. Why does nearly everyone stop? And why does a small group not? Answering them turns out to matter for sentencing policy in a very direct way, because a system that imprisons people through the years they would have stopped anyway is buying much less public safety than it thinks.

Key idea: The age-crime curve, in which offending peaks in late adolescence and then declines sharply, is criminology's most robust regularity, and explaining both the near-universal decline and the persistent minority is the central task of life-course research.

The curve and the debate about it

The curve's details vary in ways worth knowing. Property offending typically peaks earlier, around ages sixteen to eighteen, and violent offending somewhat later. The curve is much higher for males than females but has broadly the same shape. And its peak has shifted modestly across historical periods.

In 1983 Travis Hirschi and Michael Gottfredson made a provocative argument: the age effect is invariant across social and cultural conditions, cannot be explained by any known variable, and is therefore not something criminology can explain. Their conclusion was that the field should abandon longitudinal studies of criminal careers as a waste of resources, since the age distribution would swamp everything.

Alfred Blumstein, Jacqueline Cohen, and David Farrington replied with a decomposition that has held up. The aggregate curve mixes two very different quantities. Participation, sometimes called prevalence, is the proportion of a population actively offending at a given age. Frequency, often written as lambda, is how often an active offender offends.

When those are separated, the aggregate decline after the peak turns out to be driven overwhelmingly by falling participation. Fewer and fewer people remain active. Among those who stay active, individual offending frequency declines much more gently. The aggregate curve is therefore mostly a story about people stopping, not about active offenders slowing down.

That distinction is not academic. If the decline were about slowing down, incapacitating known offenders for longer would keep paying off. If it is about people stopping, then long sentences increasingly hold people who had already exited the population of offenders.

Key idea: Decomposing the age-crime curve shows the post-peak decline reflects falling participation rather than active offenders slowing down, which directly weakens the incapacitation case for very long sentences.

Moffitt's developmental taxonomy

Terrie Moffitt proposed in 1993 that the curve conceals two distinct populations with different causes.

Adolescence-limited offenders are the large majority. Their offending is normative, social, minor, and temporary. Moffitt explained it through the maturity gap: modern societies grant biological adulthood years before social adulthood, leaving adolescents with adult capacities and juvenile status. Delinquency becomes a way of claiming autonomy, and it is learned partly by mimicking the small group of persistent offenders, who appear enviably independent. When adult roles become available, the behavior loses its purpose and stops.

Life-course-persistent offenders are a small group, in Moffitt's estimate something like five percent of males, who begin very early and continue across the life span. Their pathway starts with subtle neuropsychological difficulties, in temperament, attention, language, or executive function, which interact with high-risk family and neighborhood environments. Difficult children evoke harsh responses, fail early at school, are rejected by conventional peers, and accumulate deficits, a process of cumulative continuity that narrows options at each stage.

The Dunedin longitudinal study in New Zealand, which Moffitt worked with, broadly supported the distinction: an early-onset group showed neuropsychological and family risk profiles absent in the adolescence-limited group and persisted into adulthood.

Later work complicated it. Trajectory analyses typically find more than two groups, including low-level chronic offenders and adult-onset offenders whom the theory did not predict. Follow-ups also found that many adolescence-limited offenders were not entirely fine in adulthood, showing elevated substance and financial problems. The taxonomy is best treated as a productive simplification rather than a literal typology.

Key idea: Moffitt distinguished a large adolescence-limited group whose offending reflects the maturity gap from a small life-course-persistent group with early neuropsychological and environmental risk, though later research finds more groups than two.

Sampson and Laub and the age-graded theory

Robert Sampson and John Laub took a different route. They recovered the raw data from Sheldon and Eleanor Glueck's study of 500 delinquent and 500 non-delinquent Boston boys from the 1940s, reanalyzed it, and then traced the surviving men and interviewed them into their sixties and seventies.

Their age-graded theory of informal social control holds that what restrains offending is social bonds appropriate to each life stage, and that these bonds can form or dissolve at any age. Childhood does not determine adulthood.

They emphasize turning points: marriage or a stable partnership, steady employment, military service, and residential change. Sampson, Laub, and Wimer analyzed the marriage effect using within-individual methods that compare a man to himself in married and unmarried periods, which removes every stable difference between men. They estimated marriage associated with roughly a thirty-five percent reduction in the odds of offending.

Four mechanisms are proposed. A turning point restructures routine activities, changing who you spend time with and where. It supplies direct supervision by a partner or employer. It creates an investment you would lose, which is Hirschi's commitment in a new key. And it opens the possibility of a changed identity.

Their most pointed finding was negative. Following the men to age seventy, they concluded that childhood risk factors could not predict which men would persist and which would desist. Prediction failed at the individual level even with exceptionally rich childhood data. They argued for human agency as an irreducible part of the story: men made choices, and those choices were not determined by their measured backgrounds.

Key idea: Sampson and Laub found that bonds forming in adulthood, especially marriage and steady work, reduce offending substantially, and that childhood risk factors could not predict individual trajectories even with outstanding data.

Desistance as a process

Desistance research asks what stopping actually looks like from the inside, and its findings are consistent across studies.

Desistance is a process, not an event. It typically involves ambivalence, reduction in frequency and seriousness, relapse, and gradual settling, rather than a clean stop. This means any study measuring desistance as a single reoffense event is measuring something crude.

Shadd Maruna interviewed persisting and desisting offenders and found a striking difference in how they narrated their lives. Desisters told what he called redemption scripts, in which the person's real self was always decent, the offending was caused by circumstances now overcome, and the difficult past has been given purpose, often through helping others. Persisters told fatalistic accounts of being trapped. Maruna's careful claim is that identity narrative is part of the process of maintaining change, not simply a report about it.

Peggy Giordano's cognitive transformation model describes four steps: a general openness to change, exposure to a hook for change such as a relationship, a job, or a programme, the ability to envision a replacement self, and finally a reappraisal in which the old behavior comes to seem incompatible with who one now is. The hooks matter, but only for someone ready to grab them, which explains why identical opportunities produce different results.

Ray Paternoster and Shadd Bushway added the feared self: change is often initiated by a vivid, frightening picture of who one will become if nothing changes, which arrives before any external opportunity does.

Key idea: Desistance is a gradual and reversible process in which identity change, through redemption narratives, replacement selves, and feared future selves, works together with external hooks such as relationships and work.

What this means for policy

Three implications follow directly, and they are among the most practically important conclusions in this course.

First, age must enter risk assessment. Since participation falls steeply with age, a fifty-five-year-old with a long record is far less likely to reoffend than his record alone suggests. Instruments that weight criminal history heavily without weighting age properly will systematically overestimate the risk of older people.

Second, very long sentences buy diminishing public safety. Someone serving decades is being incapacitated largely through years in which he would have desisted anyway. This is an argument about efficiency, not about desert; retribution can still justify a long sentence, but incapacitation cannot justify it as well as intuition suggests.

Third, since desistance runs through employment, relationships, and identity, policies that permanently damage employability, housing, and social ties work directly against the process that ends offending. That includes long-term collateral consequences of conviction. Lesson 16 examines what interventions actually support desistance.

One honest caution. Because desistance research is mostly observational, the causal status of turning points is not fully settled. Within-individual designs help enormously, and the marriage estimates use them, but selection into marriage and employment cannot be entirely ruled out. The finding is strong; it is not a randomized trial.

Key idea: Because participation in offending falls steeply with age, risk assessment must weight age, very long sentences buy little incapacitation benefit, and policies that damage employment and social ties work against the desistance process.

Common misconceptions

• The age-crime curve means older offenders commit crimes less often. The decline is driven mainly by fewer people offending at all.
• Moffitt's taxonomy describes exactly two kinds of offender. Later trajectory research consistently finds more groups, including adult-onset offenders.
• Childhood risk factors determine adult outcomes. Sampson and Laub found rich childhood data could not predict individual trajectories to age seventy.
• Desistance is a single decision to stop. It is a gradual, ambivalent, relapse-prone process involving identity change.
• Long sentences maximize incapacitation benefits. They increasingly hold people through years in which they would have desisted anyway.

Recap

• The age-crime curve peaks in late adolescence and declines through the twenties and thirties, driven mainly by falling participation.
• Moffitt distinguished adolescence-limited offending, explained by the maturity gap, from life-course-persistent offending with early neurodevelopmental and environmental roots.
• Sampson and Laub showed that adult bonds such as marriage and steady work reduce offending, with marriage associated with roughly a thirty-five percent reduction.
• Desistance is a gradual process involving redemption narratives, replacement selves, and external hooks for change.
• Age belongs in risk assessment, long sentences buy diminishing incapacitation, and collateral consequences work against desistance.

Sources

1. Moffitt, T. E. (1993). Adolescence-limited and life-course-persistent antisocial behavior: A developmental taxonomy. Psychological Review, 100(4), 674-701. doi.org
2. Sampson, R. J., Laub, J. H., & Wimer, C. (2006). Does marriage reduce crime? A counterfactual approach to within-individual causal effects. Criminology, 44(3), 465-508. doi.org
3. Laub, J. H., & Sampson, R. J. (2001). Understanding desistance from crime. Crime and Justice, 28, 1-69. doi.org
4. National Institute of Justice. (n.d.). Desistance from crime. Office of Justice Programs. nij.ojp.gov
5. Wikipedia contributors. (n.d.). Life course approach. Wikipedia. en.wikipedia.org

The big picture

The United States built the largest incarceration system in the world, roughly quadrupling its imprisonment rate between the mid-1970s and 2008, and it did so with an explicit crime-control justification. This lesson asks what was bought.

It then turns to policing, where the evidence is in much better shape. Some policing strategies have been evaluated repeatedly with credible designs and work. Others, including some of the most famous and most politically defended, do not survive testing. The gap between the reputation of a strategy and the evidence for it is enormous in both directions.

This is the most policy-relevant lesson in the course, and the one where you should be most alert to your own priors. Some findings here will please people who want less punishment; others will please people who want more police. Both are in the evidence.

Key idea: Incarceration and policing are the state's two main crime-control tools, and the evidence supports some specific policing strategies far more strongly than it supports expanding imprisonment.

Three ways prison could reduce crime

Prison can only reduce crime through three mechanisms, and they must be evaluated separately because the evidence differs for each.

General deterrence is the effect on people who are not imprisoned but who observe the threat. Lesson 4 covered this: sentence severity produces small effects at best, while certainty of apprehension does the work. Prison expansion operated almost entirely on the severity lever.

Incapacitation is crime prevented because the person is physically unable to offend against the public. This effect is genuinely real. The debate is entirely about magnitude, and the magnitude depends on who the marginal prisoner is.

After-effects are what happens to the person's offending after release, compared with what would have happened under a non-custodial sanction. This is where the evidence has moved furthest from intuition.

Key idea: Prison can reduce crime only through general deterrence, incapacitation, or post-release effects, and evaluating imprisonment means assessing these three separately rather than as one lump.

Incapacitation and diminishing returns

Incapacitation seems simple: a person in prison cannot burgle a house. Two complications make the arithmetic much harder.

The first is the identity of the marginal prisoner. When a system imprisons few people, it imprisons the highest-rate offenders. As it expands, it reaches progressively lower-rate offenders. The hundred-thousandth prisoner prevents far less crime than the first thousand.

Second is replacement. In organized illegal markets, removing a seller may create a vacancy someone else fills, so the market's offending continues with a different person in the role. This applies most strongly to drug distribution.

The empirical work follows this logic. Research by Raymond Liedka, Anne Piehl, and Bert Useem examined how the crime-reducing effect of imprisonment varies with scale and found the elasticity declining as prison populations grow, approaching zero and possibly reversing at high levels. The National Research Council's 2014 consensus report concluded that the incremental crime-prevention effect of the very high incarceration rates reached in the United States is modest, and that the growth of incarceration was not the main driver of the crime decline.

Notice what this does and does not say. It does not say prison prevents no crime. It says the United States is far out on a curve where each additional prisoner buys very little, which is an argument about where to operate rather than about whether prison works at all.

Key idea: Incapacitation is real but subject to steep diminishing returns because expansion reaches lower-rate offenders and because some roles are simply refilled, and major reviews conclude the marginal effect at American scale is modest.

After-effects: does prison make people worse?

The most surprising body of evidence concerns what prison does to the people who go through it.

Reviews comparing custodial with non-custodial sanctions for similar offenders, including work by Daniel Nagin, Francis Cullen, and Cheryl Lero Jonson, generally find that imprisonment does not reduce reoffending relative to lesser sanctions and may slightly increase it. David Roodman's systematic review for Open Philanthropy, which reanalyzed the strongest studies rather than merely counting them, reached a similar conclusion: after-effects are on average around zero and possibly criminogenic.

The mechanisms are plausible and align with earlier lessons. Prison severs employment and relationships, which Lesson 14 identified as the engines of desistance. It concentrates people with delinquent definitions, which Lesson 9 predicts will matter. It labels, which Lesson 10 showed has measurable effects through blocked opportunity. And it imposes lasting collateral consequences on housing, licensing, and employment.

An important counterpoint keeps this from becoming a slogan. A well-identified Norwegian study by Manudeep Bhuller and colleagues, using random assignment of cases to judges of differing severity, found that imprisonment there reduced reoffending and increased employment, apparently through rehabilitation and training programs delivered inside. The comparison is instructive: the effect of incarceration is not a fixed property of confinement, but a function of what the confinement contains and what the alternative was.

There is also a neighborhood-level argument. Todd Clear's coercive mobility hypothesis holds that when a neighborhood loses a very high proportion of its young men to prison and receives them back continuously, the churn itself destabilizes families and informal social control enough to raise crime. The evidence is suggestive rather than settled, but it points at a cost that individual-level studies cannot see.

Key idea: Reviews find that imprisonment does not reduce reoffending relative to non-custodial sanctions and may slightly increase it, while a Norwegian study with strong identification found the opposite, indicating that the effect depends on what prisons actually do.

Police numbers

Now the other side, and here the evidence turns more favorable to state action.

Estimating the effect of police on crime is hard because cities hire more police when crime rises, which biases naive estimates toward zero or the wrong sign. Credible studies therefore use variation that is not driven by local crime: federal hiring grants, terror alert deployments, mass layoffs after budget crises.

Aaron Chalfin and Justin McCrary's careful analysis, correcting for measurement error in police employment counts, estimated an elasticity of violent crime with respect to police of roughly minus 0.5 to minus 0.7. In plain terms, a ten percent increase in officers is associated with something like a five to seven percent reduction in violent crime. That is a large effect by criminological standards, and it is consistent with the deterrence finding that certainty matters, since more officers raise the perceived probability of apprehension.

Later work by Chalfin and colleagues added an important complication. Each additional officer was estimated to prevent a meaningful fraction of a homicide per year, with larger benefits in predominantly Black communities that experience the most violence, while the same officers also generated additional arrests for low-level offenses, with costs falling disproportionately on those same communities. The honest summary is that police reduce serious violence and, as currently deployed, also increase low-level enforcement, and that these are separate dials rather than a single package.

Key idea: Well-identified studies estimate that adding police substantially reduces violent crime, with an elasticity around minus 0.5 to minus 0.7, while the same expansions also increase low-level arrests, making the benefits and costs separable questions.

Hot spots policing and focused deterrence

Two specific strategies have the strongest evidence in policing, and both follow directly from findings earlier in this course.

Hot spots policing concentrates resources on the small number of micro-places that generate most crime, which Lesson 7 established. Anthony Braga and colleagues have maintained a systematic review of hot spots evaluations, now covering dozens of studies including randomized trials. The pooled result is a statistically significant crime reduction of modest to moderate size. Crucially, the reviews find no evidence of net displacement to surrounding areas, and often find diffusion of benefits instead. Within hot spots work, problem-oriented approaches that diagnose and fix the specific conditions of a place tend to outperform simply adding patrol time.

Focused deterrence, also called pulling levers, began with the Boston Gun Project in the 1990s. It identifies the small networks responsible for most serious violence, communicates directly and explicitly to them that violence will bring concentrated enforcement attention, and simultaneously offers services and a way out. A Campbell Collaboration review by Braga, Weisburd, and Turchan pooled two dozen evaluations and found a moderate overall crime-reduction effect, strongest for programs targeting gang and group-involved violence.

Both deserve a methodological caveat. Most focused deterrence evaluations are quasi-experimental rather than randomized, several were conducted by teams involved in designing the programs, and results have been more variable in replication than in original sites. The strategies are among the best-supported in policing and are not magic.

Key idea: Hot spots policing and focused deterrence have the strongest evidence in policing, producing modest to moderate crime reductions without net displacement, though most focused deterrence evaluations are quasi-experimental and replication has been uneven.

Broken windows and stop-and-frisk

Now the strategies whose reputations exceed their evidence.

Broken windows, proposed by James Q. Wilson and George Kelling in 1982, argues that visible disorder signals that nobody is in control, which invites more serious crime, so addressing minor disorder prevents major offending. It became the intellectual justification for aggressive misdemeanor enforcement in the 1990s.

The evidence splits the theory in half, and the split is the most useful thing to remember. A meta-analysis by Braga, Brandon Welsh, and Cory Schnell examined disorder policing evaluations and found that community and problem-solving approaches to disorder, cleaning up places, fixing conditions, and working with residents, produced significant crime reductions. Aggressive order-maintenance strategies built on misdemeanor arrests did not produce significant reductions. Bernard Harcourt and Jens Ludwig's reanalysis of the New York City evidence, using the Moving to Opportunity data, found no support for the broken windows mechanism there and argued the observed decline was better explained by regression to the mean in previously spiking precincts.

Stop-and-frisk as practiced in New York City provides an unusually clean test. Recorded stops rose to roughly 685,000 in 2011, overwhelmingly of young Black and Hispanic men, with very low rates of weapon recovery. In Floyd v. City of New York in 2013, a federal court found the practice unconstitutional as applied. Stops then fell by more than ninety percent. Crime did not rise; it continued to fall. Whatever the tactic contributed at its peak, the city removed almost all of it without the predicted consequences. Research on more targeted stops at high-crime micro-places finds modest deterrent effects at best, which suggests the useful version of the tactic was a small fraction of what was being done.

Key idea: Meta-analysis finds that problem-solving approaches to disorder reduce crime while aggressive misdemeanor enforcement does not, and New York's ninety percent reduction in stops without any crime increase is strong evidence against mass stop-and-frisk.

Legitimacy, and a lesson in humility

A different tradition, associated with Tom Tyler, argues that most compliance with law comes not from fear of sanctions but from legitimacy, the belief that authorities have the right to be obeyed. Legitimacy is built through procedural justice: treating people with dignity, giving them voice, explaining decisions, and behaving neutrally. The correlational evidence linking perceived procedural justice to cooperation and compliance is substantial and consistent.

Whether training officers in procedural justice changes behavior is a separate question, and the answer is genuinely unsettled. A widely publicized evaluation of a large procedural justice training programme reported substantial reductions in complaints and use of force. The authors subsequently discovered a coding error and retracted the paper. That episode is worth remembering for two reasons: it is a model of how science is supposed to correct itself, and it is a warning against building policy on a single striking study before replication.

Body-worn cameras illustrate the same caution from a different direction. Early studies were enthusiastic, but the largest randomized trial, conducted in Washington DC with over two thousand officers, found no statistically significant effects on use of force or civilian complaints. Cameras may be justified for accountability and evidentiary reasons; the behavioral effects that were promised did not appear at scale.

Key idea: Perceived legitimacy and procedural justice are strongly associated with cooperation, but evidence that training changes officer behavior is unsettled, and both a retracted training study and the null body-camera trial warn against building policy on single early studies.

Common misconceptions

• Prison obviously prevents crime, so more is better. Incapacitation faces steep diminishing returns, and after-effects average around zero or slightly criminogenic.
• Research shows prison is always criminogenic. A well-identified Norwegian study found the opposite, indicating the effect depends on what prisons do.
• Hot spots policing just moves crime around the corner. Systematic reviews find no net displacement and often a diffusion of benefits.
• Broken windows policing is supported by evidence. Problem-solving approaches to disorder work; aggressive misdemeanor enforcement does not.
• Ending stop-and-frisk would have caused a crime wave. New York cut stops by over ninety percent and crime continued to fall.

Recap

• Prison works only through deterrence, incapacitation, or after-effects, and these must be assessed separately.
• Incapacitation is real but faces steep diminishing returns, and major reviews find the marginal American effect modest.
• Custodial sanctions do not reduce reoffending relative to non-custodial ones and may slightly increase it, though the Norwegian case shows this is not inevitable.
• Adding police substantially reduces violent crime while also increasing low-level arrests.
• Hot spots policing and focused deterrence have the best evidence, while aggressive misdemeanor enforcement and mass stop-and-frisk do not.

Sources

1. National Research Council. (2014). The growth of incarceration in the United States. National Academies Press. nap.nationalacademies.org
2. Roodman, D. (2017). The impacts of incarceration on crime. Open Philanthropy. openphilanthropy.org
3. Braga, A. A., Weisburd, D., & Turchan, B. (2018). Focused deterrence strategies and crime control. Criminology & Public Policy, 17(1), 205-250. doi.org
4. Braga, A. A., Welsh, B. C., & Schnell, C. (2015). Can policing disorder reduce crime? A systematic review and meta-analysis. Journal of Research in Crime and Delinquency, 52(4), 567-588. doi.org
5. National Institute of Justice. (n.d.). CrimeSolutions: Policing. Office of Justice Programs. crimesolutions.ojp.gov

The big picture

This final lesson answers the question the whole course has been building toward: what actually reduces offending?

The answer is more encouraging than the field believed fifty years ago and much narrower than program brochures suggest. Some interventions work reliably. A smaller number make things worse, and several of those remain popular. The difference is not warmth, cost, or intuitive appeal, but whether the program targets the things that actually predict reoffending, using methods that actually change behavior. The lesson closes with the skill that outlasts every fact in this course: how to read a criminology study without being fooled, by anyone, including yourself.

Key idea: A well-defined set of interventions reliably reduces reoffending, a smaller set reliably increases it, and the difference lies in what they target and how, not in how appealing they sound.

From nothing works to what works

In 1974 Robert Martinson published a review of correctional treatment evaluations that was widely summarized in three words: nothing works. The actual paper was more careful, and Martinson himself substantially qualified his position in 1979. But the slogan escaped and did enormous damage, helping justify a turn away from rehabilitation toward deterrence and incapacitation.

What rescued the field was meta-analysis. Instead of counting studies as successes or failures, researchers pooled effect sizes across hundreds of evaluations and asked which program characteristics predicted success. Treatment effects turned out to be real but highly variable, and the variation was systematic. Mark Lipsey's synthesis of hundreds of juvenile program evaluations found the clearest pattern. Therapeutic approaches, meaning counseling, skill building, mentoring, and restorative programs, produced meaningful recidivism reductions, typically on the order of ten percentage points or more for well-implemented programs. Control-oriented approaches built on surveillance, discipline, and deterrence produced negligible effects and sometimes negative ones. The distinction was not severity but whether the program taught something.

Key idea: The nothing works slogan misrepresented a more careful review, and meta-analysis later showed that therapeutic programs reduce recidivism meaningfully while surveillance and discipline approaches do not.

Risk, need, and responsivity

The most useful framework in correctional practice comes from Donald Andrews and James Bonta. Risk-need-responsivity has three principles, each counterintuitive in a way that matters.

The risk principle says match intervention intensity to assessed risk. This is not a moral claim but an empirical one: intensive programming delivered to low-risk individuals frequently increases their recidivism, apparently by disrupting the conventional ties already protecting them and by placing them alongside higher-risk peers. Doing more for someone who needs little is not neutral.

The need principle says target criminogenic needs, meaning changeable factors that actually predict reoffending. Andrews and Bonta's central eight are antisocial attitudes, antisocial peers, an antisocial personality pattern including impulsivity, family and marital problems, substance abuse, poor school or work performance, unstructured leisure, and criminal history, which is static rather than changeable. Notice what is missing. Low self-esteem, anxiety, and vague personal distress are real human needs but are not criminogenic, and programs targeting them do not reduce reoffending. This single point explains a large share of failed programming.

The responsivity principle says use methods that work, principally cognitive-behavioral and social learning approaches, and adapt delivery to the person's abilities, motivation, and circumstances.

Programs adhering to all three principles show the largest recidivism reductions, with community-based programs meeting all three reporting reductions of roughly a quarter, while programs adhering to none show effects near zero or in the wrong direction. That gradient, from more principles to bigger effects, is what makes the framework credible.

Key idea: Risk-need-responsivity holds that intensity should match risk, that programs must target criminogenic needs rather than general well-being, and that cognitive-behavioral methods should be used, with adherence to all three principles producing the largest reductions.

Cognitive-behavioral programs

The single best-supported intervention type is cognitive-behavioral therapy adapted for offending populations. It works on the thinking patterns that precede offending: hostile attribution of others' motives, cognitive distortions that justify harm, poor consequence anticipation, and automatic responses to provocation. Participants practice recognizing these patterns and rehearsing alternatives.

A meta-analysis by Nana Landenberger and Mark Lipsey found an average recidivism reduction of roughly twenty-five percent, with the best-implemented programs achieving considerably more. Effects were larger with better implementation and higher-risk participants, exactly as the risk principle predicts.

Two randomized Chicago trials tested this outside corrections. Sara Heller and colleagues evaluated Becoming a Man, a school-based programme using cognitive-behavioral methods with adolescent boys, and found violent-crime arrests down by roughly a third to nearly a half across study years, with better school engagement. Heller's separate evaluation of a summer jobs programme found violent-crime arrests roughly forty percent lower over the following year, despite the jobs lasting only eight weeks. In both, effects appeared on violent crime rather than all offending, a clue that the intervention changes responses in charged moments rather than economic circumstances.

Key idea: Cognitive-behavioral programs reduce recidivism by roughly a quarter on average, with larger effects for higher-risk participants and well-implemented programs, and randomized trials outside corrections show substantial reductions in violent arrests.

What does not work, and what backfires

An equally important body of evidence identifies interventions that fail or cause harm.

Scared Straight programs, which bring at-risk juveniles into prisons for confrontation with inmates, are the field's clearest case of harm. A Campbell review of randomized trials found participants more likely to offend than controls, with pooled odds roughly sixty to seventy percent higher. The programs are intuitive, cheap, emotionally satisfying, and criminogenic.

Boot camps emphasizing military discipline show no consistent recidivism reductions relative to comparison sanctions, though participants often report feeling changed, and DARE in its original form showed no meaningful effect on later drug use across many evaluations despite enormous reach.

Intensive supervision without treatment does not reduce reoffending and reliably increases technical violations, because more supervision detects more rule breaking. A program can therefore appear to increase failure while changing no behavior, a measurement trap worth remembering.

Why do intuitive programs fail so consistently? They target non-criminogenic needs, they aggregate high-risk peers in what researchers call deviancy training, and they rely on frightening or shaming people rather than building capability. Deterrence, as Lesson 4 showed, is weak precisely where these programs assume it is strong.

Key idea: Scared Straight increases offending, boot camps and DARE show no effect, and intensive supervision without treatment mainly increases detected violations, because such programs target the wrong needs, group high-risk peers, and rely on fear rather than skill.

Restorative justice

Restorative justice brings victims, offenders, and affected community members together, typically in a facilitated conference, to address what happened, who was harmed, and what would repair it.

A Campbell Collaboration review by Heather Strang, Lawrence Sherman, and colleagues pooled randomized trials of face-to-face conferencing. Repeat offending was modestly reduced, with effects at least as strong for violent as for property offenses, which surprised many observers. Victim outcomes were the larger story: participating victims reported substantially higher satisfaction, lower desire for revenge, and reduced post-traumatic stress symptoms than those who went through conventional processing.

Two honest caveats. The number of high-quality trials remains modest and many involve less serious cases with consenting participants, which limits generalization. And restorative processes raise real concerns where the parties are unequal in power, which is why their use in intimate partner violence is contested among researchers and advocates alike. Even so, it is the only intervention in this lesson that produces a large measured benefit for victims, who are otherwise nearly invisible in crime policy evaluation.

Key idea: Randomized evidence shows restorative conferencing modestly reduces repeat offending and substantially improves victim satisfaction and psychological outcomes, with caveats about case selection and power imbalances.

Prevention before the system

The interventions with the longest reach begin before anyone has offended.

High-quality early childhood programs have produced long-run arrest reductions in several well-known trials, including the Perry Preschool project, which followed participants for decades. The Nurse-Family Partnership, which sends nurses to first-time low-income mothers from pregnancy through a child's second birthday, reported substantial reductions in child arrests at fifteen-year follow-up in its original trial.

Then comes the caution this course has repeated throughout: replications have been mixed. Later Nurse-Family Partnership trials found benefits on some outcomes and not others, a large British trial found no effect on its primary outcomes, and Multisystemic Therapy has produced weaker or null results in some independent evaluations abroad.

This is the developer effect, one of the most important patterns in the evaluation literature: programs evaluated by the people who created them show larger effects than the same programs evaluated independently. It is not usually fraud. It reflects implementation quality, favorable site selection, and the many small decisions that go one way when you believe in something. Prefer clearinghouses that grade evidence and require independent replication, such as Blueprints for Healthy Youth Development and CrimeSolutions, over any single striking study.

Key idea: Early childhood and home visiting programs have produced long-run crime reductions in landmark trials, but replication has been uneven and the developer effect means independent evaluations should carry more weight than original ones.

Victims

Criminology neglected victims for most of its history, and victimology corrects that. Victimization is not randomly distributed. Lifestyle-exposure theory, developed by Michael Hindelang, Michael Gottfredson, and James Garofalo, holds that risk follows from routine activities that increase contact with likely offenders in unguarded settings. National Crime Victimization Survey data show violent victimization concentrated among the young and elevated among people with low household income, while the elderly, who fear crime most, are victimized least.

The most striking finding is the victim-offender overlap. Across many studies, offending is among the strongest predictors of being victimized, and victimization predicts later offending, so the two categories are far less separate than law and public discussion assume. Repeat victimization applies to people as well as properties, which makes a recent victim an efficient target for prevention effort.

Institutionally, the Victims of Crime Act of 1984 created a federal fund, supported by criminal fines rather than taxes, that finances state compensation and victim assistance. Research on secondary victimization documents that the justice process itself frequently harms victims through delay, insensitive questioning, and loss of control, which is part of why restorative approaches score so well on victim outcomes.

Key idea: Victimization concentrates by lifestyle and age rather than by fear, offending is among the strongest predictors of victimization, and the justice process itself often harms victims.

Studying and working in criminology

Research and analysis roles include crime analysts in police departments, research analysts in state statistical and corrections agencies, policy analysts at nonprofits and legislative offices, and academic researchers; these reward statistical skill far more than most students expect. Practice roles include probation and parole officers, victim advocates, and program staff, while law and public policy are common graduate destinations. Forensic science is a separate field requiring a natural science background. The Bureau of Labor Statistics Occupational Outlook Handbook is a better guide to qualifications, pay, and demand than any program's marketing.

Key idea: Criminology leads to research and analysis roles that reward quantitative skill, to practice roles in supervision and advocacy, and to law and policy, but not to forensic science.

How to read a criminology study

Here is the checklist. Applying it to the next crime claim you encounter is the most valuable thing you can take from this course.

Question	Why it matters
What is the counterfactual?	Compared with what? A program with no comparison group tells you nothing.
What is the design?	Randomized trial, natural experiment, matched comparison, or simple correlation. This determines what can be concluded.
What is the outcome measure?	Arrest measures policing as well as behavior; self-report and conviction measure different things.
How large is the effect?	Statistically significant is not the same as meaningful. Ask for percentage points or a standardized effect size.
How long is the follow-up?	Effects that appear at six months often shrink by two years.
Who evaluated it?	Developer-led evaluations report larger effects than independent ones.
Has it replicated?	A single striking result is a hypothesis, not a finding.
Is there a systematic review?	If Campbell, CrimeSolutions, or a comparable body has reviewed the question, start there.

Two extra habits. Watch for attrition: if a program reports only on people who completed it, the dropouts have been silently removed, and dropouts are usually the highest risk. Intention-to-treat analysis, counting everyone assigned regardless of completion, avoids this. And remember regression to the mean, since interventions are often launched where crime just spiked and unusually high numbers fall on their own.

Key idea: Ask what the counterfactual was, what design produced the estimate, how the outcome was measured, how big and durable the effect is, who ran the evaluation, and whether it has replicated, and prefer systematic reviews to single studies.

Common misconceptions

• Nothing works in rehabilitation. Meta-analysis reversed that conclusion decades ago; therapeutic programs work while control approaches do not.
• More treatment is always better. Intensive programming delivered to low-risk people often increases their reoffending.
• Building self-esteem reduces crime. Self-esteem is not a criminogenic need, and targeting it does not reduce recidivism.
• Scared Straight deters young people. Randomized evidence shows it increases subsequent offending.
• Victims and offenders are separate populations. Offending is among the strongest predictors of victimization.

Recap

• Meta-analysis overturned the nothing works era, showing therapeutic programs reduce recidivism while control approaches do not.
• Risk-need-responsivity matches intensity to risk, targets criminogenic needs, and uses cognitive-behavioral methods.
• Cognitive-behavioral programs cut recidivism by roughly a quarter, while Scared Straight, boot camps, and DARE fail or backfire.
• Restorative conferencing modestly reduces reoffending and substantially improves victim outcomes.
• Read every study for its counterfactual, design, outcome measure, effect size, evaluator, and replication record.

Sources

1. Landenberger, N. A., & Lipsey, M. W. (2005). The positive effects of cognitive-behavioral programs for offenders. Journal of Experimental Criminology, 1(4), 451-476. doi.org
2. Petrosino, A., Turpin-Petrosino, C., Hollis-Peel, M. E., & Lavenberg, J. G. (2013). Scared Straight and other juvenile awareness programs for preventing juvenile delinquency. Campbell Systematic Reviews, 9(1), 1-55. doi.org
3. Strang, H., Sherman, L. W., Mayo-Wilson, E., Woods, D., & Ariel, B. (2013). Restorative justice conferencing using face-to-face meetings of offenders and victims. Campbell Systematic Reviews, 9(1), 1-59. doi.org
4. Heller, S. B., Shah, A. K., Guryan, J., Ludwig, J., Mullainathan, S., & Pollack, H. A. (2017). Thinking, fast and slow? Some field experiments to reduce crime and dropout in Chicago. Quarterly Journal of Economics, 132(1), 1-54. doi.org
5. Bureau of Labor Statistics. (n.d.). Occupational Outlook Handbook. U.S. Department of Labor. bls.gov