Thinking About Crime
by James Q. Wilson
THE average citizen hardly needs to be persuaded that crimes will be committed more frequently if, other things being equal, crime becomes more profitable than other ways of spending one's time. Accordingly, the average citizen thinks it obvious that one major reason why crime has increased is that people have discovered they can get away with it. By the same token, a good way to reduce crime is to make its consequences to the would-be offender more costly (by making penalties swifter, more certain, or more severe), or to make alternatives to crime more attractive (by increasing the availability and pay of legitimate jobs), or both.
These citizens may be surprised to learn that social scientists who study crime are deeply divided over the correctness of such views. While some scholars, especially economists, believe that the decision to become a criminal can be explained in much the same way as we explain the decision to become a carpenter or to buy a car, other scholars, especially sociologists, contend that the popular view is wrong--crime rates do not go up because would-be criminals have little fear of arrest, and will not come down just because society decides to get tough on criminals.
This debate over the way the costs and benefits of crime affect crime rates is usually called a debate over deterrence--a debate, that is, over the efficacy (and perhaps even the propriety) of trying to prevent crime by making would-be offenders fearful of punishment. But the theory of human nature that supports the idea of deterrence--the theory that people respond to the penalties associated with crime--also assumes that people will take jobs in preference to crime if the jobs are more attractive. In both cases, we are saying that would-be offenders are rational and that they respond to their perception of the costs and benefits attached to alternative courses of action. When we use the word "deterrence," we are calling attention to only the cost side of the equation. No word in common scientific usage calls attention to the benefit side of the equation, though perhaps "inducement" might serve.
The reason scholars disagree about deterrence is that the consequences of committing a crime, unlike the consequences of shopping around for the best price on a given automobile, are complicated by delay, uncertainty, and ignorance. In addition, some scholars contend that many crimes are committed by persons who are so impulsive, irrational, or abnormal that even if delay, uncertainty, or ignorance were not attached to the consequences of criminality, we would still have a lot of crime.
Imagine a young man walking down the street at night with nothing on his mind but a desire for good times and high living. Suddenly he sees a little old lady standing alone on a dark corner, stuffing the proceeds of her recently cashed Social Security check into her purse. Nobody else is in view. If the young man steals the purse, he gets the money immediately. The costs of taking it are uncertain--the odds are at least ten to one that the police will not catch a robber, and even if he is caught, the odds are very good that he will not go to prison, unless he has a long record. On the average, no more than three felonies out of a hundred result in the imprisonment of the offender. In addition, whatever penalty may come his way will come only after a long delay--in some jurisdictions, a year or more might be needed to complete the court disposition of the offender, assuming he is caught in the first place. Moreover, this young man might, in his ignorance of how the world works, think the odds against being caught are even greater than they are, or that delays in the court proceedings might result in a reduction or avoidance of punishment.
Compounding the problem of delay and uncertainty is the fact that society cannot feasibly increase by more than a modest amount the likelihood of arrest, and though it can to some degree increase the probability and severity of prison sentences for those who are caught, it cannot do so drastically, by, for example, summarily executing all convicted robbers, or even by sentencing all robbers to twenty-year prison terms. Some scholars note a further complication: the young man may be incapable of assessing the risks of crime. How, they ask, is he to know his chances of being caught and punished? And even if he does know, perhaps he is driven by uncontrollable impulses to snatch purses whatever the risks.
As if all this were not bad enough, the principal method by which scholars have attempted to measure the effect of deterrence on crime has involved using data about aggregates of people (entire cities, counties states, and even nations) rather than about individuals. In a typical study, the rate at which, say, robbery is committed in each state is "explained" by means of a statistical procedure in which the analyst takes into account both the socio-economic features of each state that might affect the supply of robbers (for example, the percentage of persons with low incomes, the unemployment rate, the population density of the big cities, the proportion of the population made up of young males) and the operation of the criminal-justice system of each state as it attempts to cope with robbery (for example, the probability of being caught and imprisoned for a given robbery, and the length of the average prison term for robbery).
Isaac Ehrlich, an economist, produced the best-known of such analyses, using data on crime in the United States in 1940, 1950, and 1960. He found, after controlling for such things as the income level and the age distribution of the population, that the higher the probability of imprisonment for those convicted of robbery, the lower the robbery rate. At the same time, he did not find that the severity of punishment (the average time served in prison for robbery) had, independent of its certainty, an effect on robbery rates in two of the three time periods (1940 and 1960).
But some grave problems are associated with using aggregate data of this sort in studies of the effect of sanctions on crime rates. One is that many of the most important factors are not known with any accuracy. For example, we are dependent on police reports for our measure of the robbery rate, and these undoubtedly vary in accuracy from place to place. If all police departments were inaccurate to the same degree, this would not be important; unfortunately, some departments are probably much less accurate than others, and this variable error can introduce a serious bias into the statistical estimates of the effect of the criminal-justice system.
Even if we manage to overcome this problem, a further difficulty lies in wait. States in which the probability of going to prison for robbery is low are also states that have high rates of robbery (other things being equal). This fact can be interpreted in at least two ways. It can mean that the higher robbery rates are the results of the lower imprisonment rates (and would thus be evidence that deterrence works), but it might also mean that the lower imprisonment rates are caused by the higher robbery rates. To see how the latter might be true, imagine a state that is experiencing, for some reason, a rapidly rising robbery rate. It arrests, convicts, and imprisons more and more robbers as more and more robberies are committed, but it cannot quite keep up. The robberies are increasing so fast that they swamp the criminal-justice system; prosecutors and judges respond by letting more robbers off without a prison sentence, or perhaps even without a trial, in order to keep the system from becoming hopelessly clogged. As a result, the proportion of arrested robbers who go to prison goes down while the robbery rate goes up. In this case, we ought to conclude not that prison deters robbers but that high robbery rates "deter" prosecutors and judges.
The best analysis of these problems in statistical studies of deterrence is to be found in a 1978 report of the Panel on Research on Deterrent and Incapacitative Effects, which was set up by the National Research Council (an arm of the National Academy of Sciences). That panel, chaired by Alfred Blumstein, of Carnegie-Mellon University, concluded that the available statistical evidence (as of 1978) did not warrant any strong conclusions about the extent to which differences among states or cities in the probability of punishment might alter deterrent effect. The panel (of which I was a member) noted that "the evidence certainly favors a proposition supporting deterrence more than it favors one asserting that deterrence is absent," but urged "scientific caution" in interpreting this evidence.
Other criticisms of deterrence research, generally along the same lines as those of the panel, have led some commentators to declare that "deterrence doesn't work," and that we may now get on with the task of investing in those programs, such as job-creation and income maintenance, that will have an effect on crime. Such a conclusion is, to put it mildly, premature.
It is possible, as some critics of deterrence say, that rising crime rates swamp the criminal-justice system, so that a negative statistical association between, say, rates of theft and the chances of going to prison for theft may mean not that a decline in imprisonment is causing theft to increase but rather that a rise in theft is causing imprisonment to become less likely. This might occur particularly with respect to less serious crimes, such as shoplifting or petty larceny; indeed, the proportion of prisoners who are shoplifters or petty thieves has gone down over the past two decades. But it is hard to imagine that the criminal-justice system would respond to an increase in murder or armed robbery by letting some murderers or armed robbers off with no punishment. Convicted murderers are as likely to go to prison today as they were twenty years ago. Moreover, the deterrent effect of prison on serious crimes like murder and robbery was apparently as great in 1940 or 1950, when these crimes were much less common, as it is today, suggesting that swamping has not occurred.
Still more support for the proposition that variations in sanctions affect crime can be found in the very best studies of deterrence--those that manage to avoid the statistical errors described above. In 1977, Alfred Blumstein and Daniel Nagin published a study of the relationship between draft evasion and the penalties imposed for draft evasion in each of the states. After controlling for the socio-economic characteristics of the states, they found that the higher the probability of conviction for draft evasion, the lower the evasion rates. This is an especially strong finding, because the study is largely immune to the problems associated with other analyses of deterrence. Draft evasion is more accurately measured than street crimes, and draft-evasion cases could not have swamped the federal courts in which they were tried, in part because such cases made up only a small fraction (about 7 percent) of the workload of these courts, and in part because federal authorities had instructed the prosecutors to give high priority to these cases. For all these reasons, Blumstein and Nagin felt they could safely conclude that draft evasion is deterrable.
White-collar crime can also be deterred. In the late 1970s, Michael Block, Fred Nold, and J. G. Sidak, then at Stanford, investigated the effect of enforcing the antitrust laws on the price of bread in the bakery business. When the government filed a price-fixing complaint against colluding bakery firms, and when those firms also faced the possibility of private suits claiming treble damages for this price-fixing, the collusion ended and the price of bread fell.
Another way of testing whether deterrence works is to look not at differences among states or firms at one point in time but at changes in the nation as a whole over a long period of time. Historical data on the criminal-justice system in America are so spotty that such research is difficult to do here, but it is not at all difficult in England, where the data are excellent. Kenneth I. Wolpin, of Yale, analyzed changes in crime rates and in various parts of the criminal-justice system (the chances of being arrested, convicted and punished) for the period 1894 to 1967, and concluded that changes in the probability of being punished seemed to cause changes in the crime rate. He offered reasons for believing that this causal connection could not be explained away by the argument that the criminal-justice system was being swamped.
Given what we are trying to measure--changes in the behavior of a small number of hard-to-observe persons who are responding to delayed and uncertain penalties--we will never be entirely sure that our statistical manipulations have proved that deterrence works. What is impressive is that so many (but not all) studies using such different methods come to similar conclusions. More such evidence can be found in studies of the death penalty. Though the evidence as to whether capital punishment deters crime is quite ambiguous, most of the studies find that the chances of being imprisoned for murder do seem to affect the murder rate. Even after wading through all this, the skeptical reader may remain unconvinced. Considering the difficulties of any aggregate statistical analysis, that is understandable. But, as we shall shortly see, the evidence from certain social experiments reinforces the statistical studies.
Ideally, we would like to know how the probability or severity of a possible punishment will affect the behavior of persons who might commit a serious crime. Such persons probably constitute only a small fraction of the total population, but they are the important fraction. Most of us would not commit a serious crime because of the bite of conscience, reinforced by the fear of embarrassment should our misconduct be detected. A few of us may commit serious crimes with only small regard to the risks, unless those risks can be made great and immediate. What we would like to know is how changes in the prospective costs of crime, and in the prospective benefits of pursuing legitimate alternatives to crime, affect the behavior of those individuals who are "at risk"--that is, persons who lack strong, internalized inhibitions against misconduct, who value highly the excitement of breaking the law, who have a weak stake in conformity, who are willing to take greater chances than the rest of us, and who greatly value quick access to ready cash. Such persons tend, disproportionately, to be young males. As Philip J. Cook, at Duke University, has argued, would-be offenders don't have to be entirely rational or fully informed for the criminal-justice system (or the legitimate labor market) to have an effect on them. They need only attach some value to the consequences of their actions. We already know they value one consequence--the money or pleasure produced by crime. There is no reason to believe they are indifferent to all other consequences, including the risks of being caught.
Most of us are probably not very well informed about the true costs of crime: being law-abiding, we probably imagine that the chances of being caught are higher than in fact they are, and that the severity of the sentence (measured in years in prison) is greater than it really is. But most of us depend for our information on newspaper stories, on detective programs on television, and on our own deep fear of being exposed as a disreputable person. Persons at risk--young men hanging around on street corners, and thieves who associate with other thieves--tend to depend for their information on the accounts of other young men or other thieves who have had a run-in with the police or the courts and who therefore can supply a crudely accurate estimate of the current risks of arrest, prosecution, and sentencing.
The behavior of these persons, thus informed, is what we wish to observe. Only a few careful efforts have been made to measure the deterrent effect of the sanctions of the criminal-justice system on individuals, as opposed to cities or states. One such effort was carried out in Cook County, Illinois, and aimed at the young offender. Charles A. Murray and Louis A. Cox, Jr., in their book, Beyond Probation, measured the number of arrests per month of 317 Chicago boys who had been incarcerated for the first time by the Illinois Department of Corrections. Though young (their average age was sixteen), they were scarcely novices at crime; they had been arrested an average of thirteen times each before receiving their first prison sentences. Nor were their offenses trivial: as a group, they had been charged with fourteen homicides, twenty-three rapes, more than 300 assaults and a like number of auto thefts, nearly 200 armed robberies, and more than 700 burglaries. The patience of the court exhausted, these youthful offenders were eventually sent off to a correctional institution, where they served an average sentence of ten months. Murray and Cox followed them for (on the average) seventeen months after their release. During this period, the frequency with which they were arrested (i.e., arrests per month per 100 boys) declined by about two thirds. To be exact, the members of this group of hard-core delinquents were arrested an average of 6.3 times each during the year before being sent away but only 2.9 times each during the seventeen months on the street after release.
The Murray and Cox study, one of the few of its kind, adds some support to the deterrence theory. But it, like others, focuses on persons who have already committed crimes; we remain uncertain about the effect of changes in the criminal-justice system on those who have not yet committed any crime. To study behavior that does not occur is all but impossible, though we may ask, as some scholars have, various persons- often students--whether they would commit or have committed a crime when they perceived the penalties to be of a given severity and a given probability. One such study was done among students at an eastern college, another among high school students in Arizona, and a third among adult Americans generally. The authors of all three studies found that the persons who believed they were likely to be punished for a particular criminal act were less likely to report (anonymously) having committed the act than were persons who thought they probably wouldn't be punished. The studies were broadly consistent with the view that deterrence works, but all are difficult to interpret. No one can be confident that the number of offenses the persons reported bears any relationship to the number they actually committed. More important, the studies raise the possibility that what actually deters these persons (very few of whom commit any serious acts with any frequency) is not what they guess to be the chances of being caught but the moral opprobrium with which such acts are viewed. For most people in most circumstances, the moral quality of their actions, and the internalized inhibitions against misconduct arising out of that moral code, are probably the major deterrents to crime. Interviewing people may highlight that fact, but it cannot tell us what happens at the margin, when society alters the certainty or severity of punishment for a given offense. And for purposes of public policy, that is exactly what we want to know. The best way to find out is to conduct an experiment.
Perhaps the most dramatic evidence of the operation of deterrence- dramatic because it involved a true experiment on individuals engaging in what some believe is a wholly emotional crime--comes from an effort in Minneapolis to find out how the police can best handle incidents of spouse assault. The conventional wisdom had been that if one or both parties to such an assault were handled by the officer informally--by mediation or referral to a social-work agency--the parties would be better off than if the assaulter were arrested. And the police themselves often preferred not to make an arrest, because it took time and effort and often led to no prosecution when the victim refused to press charges. With the advice of the Police Foundation, a group of Minneapolis officers began handling their misdemeanor spouse-assault cases by randomly assigning the assaulter to one of three dispositions: arresting him, counseling him, or sending him out of the house to cool off. Over 200 cases were treated in this experimental fashion and followed up for six months. The assaulters who were arrested were less likely to be reported to the police for a subsequent assault than were those advised and much less likely than those sent out of the house. And this was true even though, in the vast majority of cases, the arrested person spent no more than a week in jail.
A tougher and, for policy purposes, more useful test of deterrence would be to alter the sentence a person gets without altering police conduct. We have surprisingly few careful studies of the results of doing that, even though it is regularly done. Many states have passed mandatory minimum sentences for certain offenses, and some have tried to eliminate plea bargaining, or at least to ensure that serious offenders cannot have the charges against them reduced simply in order to induce a guilty plea. Unfortunately, most of these changes were made under circumstances that rendered any serious evaluation of their effect difficult, if not impossible.
The explanation for this disappointing experience, in the opinion of the study group, was that difficulties in administering the law weakened its deterrent power, with the result that most offenders and would-be offenders did not experience any significantly higher risk of apprehension and punishment. There was no increase in the number of arrests, and a slight decline in the proportion of arrests leading to indictments as well as in the proportion of indictments resulting in conviction. Offsetting this was a higher probability that a person convicted would go to prison. The net effect was that the probability of imprisonment for arrested drug dealers did not change as a result of the law--it was about one imprisonment per nine arrests both before and after passage of the law. On the other hand, the sentences received by those who did go to prison were more severe. Before the law was passed, only 3 percent of persons imprisoned for selling an ounce or more of heroin received a sentence of three years or more. After the law went into effect, 22 percent received such sentences. Perhaps because sentences became more severe, more accused persons demanded trials instead of pleading guilty; as a result the time needed to dispose of the average drug case nearly doubled.
Does the experience under the Rockefeller law disprove the claim that deterrence works? The answer is no, but that is chiefly because deterrence theory wasn't satisfactorily tested. If "deterrence" means changing behavior by increasing either the certainty or the swiftness of punishment, then the Rockefeller law, as it was administered, could not have deterred behavior because it made no change in the certainty of punishment and actually reduced its swiftness. If on the other hand, "deterrence" means changing behavior by increasing the severity of punishment, then deterrence did not work in this case. What we mainly want to know, however is whether heroin trafficking could have been reduced if the penalties associated with it had been imposed more quickly and in a higher proportion of cases.
Severity may prove to be the enemy of certainty and speed. As penalties get tougher, defendants and their lawyers have a greater incentive to slow down the process, and those judges who, for private reasons, resist heavy sentences for drug dealing may use their discretionary powers to decline indictment, accept plea bargains, grant continuances, and modify penalties in ways that reduce the certainty and the celerity of punishment. The group that evaluated the Rockefeller law suggested that reducing severity in favor of certainty might create the only real possibility for testing the deterrent effect of changes in sentences.
The Bartley-Fox gun law in Massachusetts was administered and evaluated in ways that avoided some of the problems of the Rockefeller drug laws. In 1974, the Massachusetts legislature amended the law that had long required a license for a person carrying a handgun, by stipulating that a violation of this law would entail a mandatory penalty of one year in prison, which sentence could not be reduced by probation or parole or by judicial finagling. When the law went into effect, in April of 1975, various efforts were made to evaluate both the compliance of the criminal-justice system with it and the law's impact on crimes involving handguns. James A. Beha, II, then at the Harvard Law School, traced the application of the law for eighteen months, and concluded that, despite widespread predictions to the contrary, the police, prosecutors, and judges were not evading the law. As in New York, more persons asked for trials, and delays in disposition apparently increased, but in Massachusetts, by contrast with the experience in New York, the probability of punishment increased for those arrested. Beha estimated in 1977 (at a time when not all the early arrests had yet worked their way through the system) that prison sentences were being imposed five times more frequently on persons arrested for illegally carrying firearms than had been true before the law was passed. Owing to some combination of the heavy publicity given to the Bartley-Fox law and the real increase in the risk of imprisonment facing persons arrested while carrying a firearm without a license, the casual carrying of firearms seems to have decreased. This was the view expressed to interviewers by participants in the system, including persons being held in jail, and it was buttressed by a sharp drop in the proportion of drug dealers arrested by the Boston police who, at the time of their arrest, were found to be carrying firearms.
Three studies were made of the impact of the Bartley-Fox law on serious crime. The authors of those studies used slightly different methods, but came in general to the same conclusion--namely, that a measurable decline had occurred in some crimes that involve the casual use of firearms. Moreover, the proportion of crimes in which guns were used did not go down in other large cities during this time. In sum, the Bartley-Fox law seems, at least during the years in which its effect was studied, to have increased the risk associated with carrying a gun, to have reduced the frequency with which guns were casually carried, and thereby to have reduced the rate at which certain gun-related crimes were committed.
An effort to achieve the same results in Michigan did not work out as well, in large measure because the judges there--and in particular the judges in Wayne County, which includes Detroit--did not apply the law as expected. The Michigan Felony Firearm Statute, which went into effect in 1977, required the imposition of a two-year prison sentence for possessing a firearm while committing a felony, and the two-year firearm sentence was to be added on to whatever sentence was imposed for the other felony. But no real change occurred in either the certainty or the severity of sentences issued to gun-carrying felons. Many judges would reduce the sentence given for the original felony (say, assault or robbery) in order to compensate for the add-on. In other cases, the judge would dismiss the gun count. Given this evasion, it is not surprising that the law had little effect on the rate at which gun-related crimes were committed.
Several states have recently altered the legal minimum drinking age; because of the effect of teenage drinking on highway fatalities, these changes have been closely studied. Between 1970 and 1973, twenty-four states lowered their legal drinking ages. Shortly thereafter, Allan F. Williams and his associates at the Insurance Institute for Highway Safety examined the effect of these reductions on highway accidents, and concluded that the changes in the laws had contributed to an increase in fatal motor-vehicle accidents. Reacting to the implications of such findings at least fourteen states, beginning in 1976, raised their minimum drinking ages from eighteen or nineteen to twenty or twenty-one. Williams and his colleagues studied these changes in nine states and concluded that when young persons could not legally buy alcoholic beverages, fewer fatal auto accidents occurred at night (when most drink related accidents take place). Alexander C. Wagenaar, at the University of Michigan, looked closely at the Michigan experience, and came to the same conclusion. When the legal drinking age there was lowered to eighteen, the number of persons aged eighteen to twenty who were involved in accidents and reportedly had been drinking began to rise; when the drinking age was raised to twenty-one, the number of such persons in crashes began to fall. Comparable conclusions were reached in a study of the consequences of altering the legal drinking age in Maine.
In sum, the evidence from these experiences is that changes in the probability of being punished can lead to changes in behavior, though this may not happen when the new laws exist on paper and not in practice, or when the benefits to be had from violating the law are so great as to make would-be perpetrators indifferent to the slight alteration in the risks facing them. For example, when the prospective gains from heroin trafficking or obtaining (and supplying) illegal abortions are very large, these gains can nullify the effect of modest changes in the costs of these actions, especially when (as with the New York drug law and the Michigan firearms law) the criminal-justice system does not in practice impose greater risks. When the system in fact makes the behavior much more costly, as it did with the 317 juveniles in Chicago, we observe a reduction in crime. And when the prospective benefits from violating the law are small (as with teenage drinking, or perhaps with carrying an unlicensed gun), even small changes in the risks can have significant effects on behavior.
All this means that it is difficult but not impossible to achieve increased deterrent effects through changes in the law. To obtain these effects, society must walk a narrow line--the penalties must be sufficiently great to offset, at the margin, the benefits of the illegal act, but not so great as to generate in the criminal-justice system resistance to their prompt imposition. If social experiments show that under some circumstances crime rates go down when penalties go up, can other experiments show that crime rates go down when jobs are easier to find?
Let us return to our original example. The young man is still yearning for the money necessary to enjoy some high living. Let us assume that he considers finding a job. He knows he will have to look for one; this will take time. Assuming he gets one, he will have to wait even longer to obtain his first paycheck. But he knows that young men have difficulty finding their first jobs, especially in inner-city neighborhoods such as his. Moreover, he cannot be certain that the job he might get would provide benefits that exceed the costs. Working forty hours a week as a messenger, a dishwasher, or a busboy might not seem worth the sacrifice in time, effort, and reputation on the street corner that it entails. The young man may be wrong about all this, but if he is ignorant of the true risks of crime, he is probably just as ignorant of the true benefits of alternatives to crime.
Compounding the problems of delay, uncertainty, and ignorance is the fact that society cannot feasibly make more than modest changes in the employment prospects of young men. Job-creation takes a long time, when it can be done at all, and many of the jobs created will go to the "wrong" (i.e., not criminally inclined) persons; thus, unemployment rates among the young will not vary greatly among states and will change only slowly over time. And if we wish to see differences in unemployment rates (or income levels) affect crime, we must estimate those effects by exactly the same statistical techniques we use to estimate the effect of criminal-justice sanctions.
The problem of measurement error arises because we do not know with much accuracy the unemployment rate among youths by city or state. Much depends on who is looking for work and how hard, how we count students who are looking only for part-time jobs, and whether we can distinguish between people out of work for a long period and those who happen to be between jobs at the moment. Again, since inaccuracies in these data vary from place to place, we will obtain biased results.
The problem of omitted factors is also real, as is evident in a frequently cited study done in 1976 by Harvey Brenner, of Johns Hopkins University. He suggested that between 1940 and 1973, increases in the unemployment rate led to increases in the homicide rate. But he omitted from his analysis any measure of changes in the certainty or the severity of sentences for murder, factors that other scholars have found to have a strong effect on homicide.
Finally, the relationship between crime and unemployment (or poverty) is probably complex, not simple. For example, in a statistical study that manages to overcome the problems already mentioned, we might discover that as unemployment rates go up, crime rates go up. One's natural instinct would be to interpret this as meaning that rising unemployment causes rising crime. But rising crime might as easily cause rising unemployment. If young men examining the world about them conclude that crime pays more than work--that, for instance, stealing cars is more profitable than washing them--they may then leave their jobs in favor of crime. Some young men find dealing in drugs more attractive than nine-to five jobs, but, technically, they are "unemployed."
Perhaps both crime and unemployment are the results of some common underlying cause. In 1964, the unemployment rate for black men aged twenty to twenty-four was 12.6 percent; by 1978, it was 20 percent. During the same period, crime rates, in particular those involving young black men, went up. Among the several possible explanations are the changes that have occurred where so many young blacks live, in the inner parts of large cities. One such change is the movement out of the inner cities of both jobs and the social infrastructure that is manned by adult members of the middle class. The departure of jobs led to increased unemployment; the departure of the middle class led to lessened social control and hence to more crime. If we knew more than we now know, we would probably discover that all three relationships are working simultaneously: for some persons, unemployment leads to crime; for others, crime leads to unemployment; and for still others, social disintegration or personal inadequacy leads to both crime and unemployment.
That several of these relationships are in fact at work is suggested by Brenner's study, which noted that the murder rate went up with increases in per capita income and inflation as well as with a rise in joblessness. But if the stress of joblessness leads to more murders, what is it about increases in average income (or in inflation) that also leads to more murders? And if society attempts to reduce the murder rate by reducing unemployment, how can it do this without at the same time increasing the murder rate, because, as a result of increased employment, it has managed to increase per capita incomes or stimulate inflation?
I do not say this to explain away the studies purporting to show that unemployment or poverty causes crime, for in fact--contrary to what many people assert--very little research shows a relationship between economic factors and crime. Robert W. Gillespie, of the University of Illinois, reviewed studies available as of 1975 and found that out of ten looking for evidence of the existence of a significant relationship between unemployment and crime, only three were successful. In 1981, Thomas Orsagh and Ann Dryden Witte, both of the University of North Carolina, reviewed the studies that had appeared since 1975 and found very little statistically strong or consistent evidence to support the existence of such a connection. The evidence linking income (or poverty) and crime is similarly inconclusive, and probably for the same reasons: grave methodological problems confront anyone trying to find the relationship, and the relationship, to the extent that it exists is probably quite complex (some people may turn to crime because they are poor, some people may be poor because they have turned to crime but are not very good at it, and still others may have been made both poor and criminal because of some common underlying factor). To quote Orsagh and Witte: "Research using aggregate data provides only weak support for the simple proposition that unemployment causes crime....[and] does not provide convincing tests of the relationship between low income and crime."
Though preventing crime and delinquency through job programs of the sort developed by the Great Society seemed a lost hope, somewhat more success was reported from efforts to reduce crime among ex-offenders. Philip Cook followed 325 men who had been released from Massachusetts prisons in 1959 and found that parolees who were able to find "satisfactory" jobs (not just any job) were less likely than other parolees to have their parole revoked because they committed a new crime during an eighteen-month follow-up period. This was true even after controlling for the personal attributes of the parolees, such as race, intelligence, marital status, education, prior occupation, and military service.
Findings such as Cook's may have reinforced the belief of policy-makers that if only we would re-integrate the ex-offender into the labor market, we could cut crime and at the same time save money through reduced prison populations. By the early 1970s, forty-two states had adopted some variety of "work-release" program for prisoners, in which convicts nearing the end of their prison terms were released into the community in order to work at various jobs during the day, returning to prison at night or on weekends. Gordon P. Waldo and Theodore G. Chiricos, of Florida State University, evaluated the results of work-release in Florida, and did so on the basis of a particularly sophisticated research design. Eligible inmates were randomly assigned to either a work-release or a non-release group. And many different measures of recidivism were calculated--not just whether the offender was later arrested but also the rate of arrests per month. Waldo and Chiricos found no differences whatsoever in the re arrest rate (or in any measure of recidivism) between persons in and out of work-release. An equally unpromising result was found by Ann Dryden Witte in North Carolina, though there work-release may have led offenders to commit somewhat less serious crimes.
If work-release seems not to reduce crime rates, perhaps it is because it focuses on work rather than on wealth. Perhaps if ex-offenders had more money, especially during the crucial few months after their release, they would not need to steal in order to support themselves. Some preliminary evidence gave credence to this view. In Baltimore, about 400 ex-convicts were randomly assigned to one of four groups--those receiving nothing, those receiving employment assistance, those receiving financial aid, and those receiving both job-placement services and financial aid. After two years, it was clear that getting employment counseling did not affect the chances of being re-arrested, but getting financial aid ($60 a week for thirteen weeks) did make a small difference (about 8 percent). This finding, however, brought with it a host of problems. For one thing, recidivism was defined as re-arrest, not the rate of re-arrest (thus possibly obscuring changes in the frequency with which persons committed crimes). Moreover, the study excluded first offenders, alcoholics, heroin users, and persons who had not committed property offenses.
A fuller test of the combined effects of employment and wealth on criminal behavior was made in Georgia and Texas. Called TARP (Transitional Aid Research Project), it involved randomly assigning about 9,000 ex-convicts in each state to groups that, on release from prison, received financial aid, job-placement services, or nothing. This experiment was not only much larger than the one in Baltimore but also included all available categories of offenders, and it used the number of arrests of an individual (and not simply the fact of arrest) as the measure of the outcome. It also arranged that the financial aid that ex-convicts received would be reduced by any income they received from jobs--a more realistic arrangement than that used in Baltimore, where the ex-convicts could keep their financial aid whether or not they worked.
The ex-convicts who received financial aid and/or employment counseling had about the same arrest rate after release as the group that received no aid or counseling. Moreover, individuals who received TARP financial aid worked less than those who did not, so the money could be said to have discouraged, rather than encouraged, employment. The authors of the evaluation, however, were not discouraged by these findings. A complex statistical analysis led them to suggest that if the financial assistance could have been administered so as not to cause unemployment, then the payments might have prevented some crime. That speculation has been challenged by critics. What is not in dispute is that, as administered, the TARP payments did not reduce crime.
Welfare recipients and ex-addicts benefited from supported work, but ex convicts and youthful school dropouts did not. Over a twenty-seven-month observation period, the school dropouts in the project were arrested as frequently as the school dropouts in the control group, and the ex offenders in the project were arrested more frequently (seventeen more arrests per 100 persons) than ex-offenders in the control group.
The clear implication, I think, of the supported-work project--and of all the studies to which I have referred--is that unemployment and other economic factors may well be connected with criminality, but the connection is not a simple one. If, as some people assume, "unemployment causes crime," then simply providing jobs to would-be criminals or to convicted criminals would reduce their propensity to commit crimes. We have very little evidence that this is true, at least for the kinds of persons helped by MDRC. Whether crime rates would go down if dropouts and ex-convicts held on to their jobs we cannot say, because, as the supported-work project clearly showed, within a year and a half after entering the program, the dropouts and ex-convicts were no more likely to be employed than those who had never entered the program at all--despite the great and compassionate efforts made on their behalf. Help, training, and jobs may make a difference for some persons--the young and criminally inexperienced dropout; the older, "burned-out" ex-addict; the more mature (over age thirty-five) ex-convict. But ex-addicts, middle aged ex-cons, and inexperienced youths do not commit most of the crimes that worry us. These are committed by the young chronic offender.
Marvin E. Wolfgang and his colleagues at the University of Pennsylvania, following the criminal careers of about 10,000 boys born in 1945 who lived in Philadelphia from their tenth to their eighteenth birthdays, found that about one third of the boys were arrested. For about half of these, their criminal "careers" stopped there. However, once a juvenile had been arrested three times, the chances that he would be arrested again were greater than 70 percent. These findings are consistent with the view that for novice offenders (to say nothing of non-offenders), some combination of informal social control, the deterrent effect of punishment, and the desire for normal entry into the world of work serves to restrain the growth of criminality. This is the group to whom we should look for evidence of the effects of changes in the probability and severity of punishment and of changes in job availability. At the other end of the scale, 6 percent of the Philadelphia boys committed five or more crimes before they were eighteen, accounting for more than half of all the recorded delinquencies of the 10,000 boys and most of the violent crimes committed by the entire cohort. The evidence from MDRC is consistent with the view that job programs are not likely to be effective with these offenders. Since we have only a few studies of the effect of deterrence on individuals (as opposed to large aggregates of people), we cannot be confident that increasing the certainty or severity of punishment would affect this group of hard-core, high-rate offenders, but evidence in the Murray and Cox study suggests that it might.
Some may agree with me but still feel that we should spend more heavily on one side or the other of the cost-benefit equation. At countless scholarly gatherings, I have heard learned persons examine closely any evidence purporting to show the deterrent effect of sanctions, but accept with scarcely a blink the theory that crime is caused by a "lack of opportunities." Perhaps they feel that since the evidence on both propositions is equivocal, it does less harm to believe in--and invest in- the "benign" (i.e., job-creation) program. That is surely wrong. If we try to make the penalties for crime swifter and more certain, and it should turn out that deterrence does not work, then we have merely increased the risks facing persons who are guilty of crimes in any event. If we fail to increase the certainty and swiftness of penalties, and it should turn out that deterrence does work, then we have needlessly increased the risk of being victimized for many innocent persons.
We can alter the crime rate in ways other than by manipulating the incentives confronting a would-be offender. Even if no criminal paid any heed to the risks he ran--an unlikely state of affairs--we could still reduce the crime rate by separating offenders (in prisons or on desert islands) from the rest of us. We call this "incapacitation." Almost no one doubts that incapacitation works: a man in prison cannot harm persons outside of prison, though scholars do disagree as to how large a reduction in crime we will obtain for a given cost (in money for prisons and in the freedom lost by the inmates).
Many people doubt that manipulating incentives works. Settling that issue is important, because, in theory, changing the penalties and opportunities operating in society would be a more powerful crime-control technique than simply incapacitating known criminals. To the extent that behavior is affected by rewards, would-be criminals can be induced not to become criminals at all and existing criminals can be persuaded to reduce the rate at which they offend. Were sanctions sufficiently swift and certain (and were alternatives to crime sufficiently attractive), the long prison terms required by the incapacitation strategy might not be necessary.
The behavior of most of us is affected by even small (and possibly illusory) changes in the costs attached to it. We are easily deterred by a crackdown on drunk driving, especially if it is highly publicized, and our willingness to take chances when filling out our tax returns is influenced by how likely we think an audit may be. Why, then, should we not see big changes in the crime rates when we make our laws tougher?
The answer is not that, unlike the rest of us, burglars, muggers, and assaulters are irrational. I am struck by the account given in Sally Engle Merry's book, Urban Danger of her extended interviews with youthful offenders in a big-city neighborhood she observed for a year and a half. She found that these young men had a sophisticated, pragmatic view of their criminal enterprises, even though they were neither "white-collar" criminals nor highly professional burglars. They distinguished carefully between affluent and less-affluent targets, spoke knowledgeably about the chances of being caught in one part of the district as opposed to another, understood that some citizens were less likely to call the police than others, knew which offenses were most and which were least likely to lead to arrest and prosecution, and had formed a judgment about what kinds of stories the judges would or would not believe. Though many committed crimes opportunistically, or in retaliation for what they took to be the hostile attitudes of certain neighbors, they were neither so impulsive nor so emotional as to be unaware of, or indifferent to, the consequences of their actions.
The chronic offenders who cause us all so much misery are probably especially aware of the risks they run. When Jan and Marcia Chaiken, of the Rand Corporation, analyzed the data gathered from more than 2,000 inmates of jails and prisons in California, Michigan, and Texas, they found that the chances of going to prison for a crime were higher in Texas than in California, and that the inmates sensed it. The California inmates were twice as likely as the ones in Texas to say that they thought they could commit the same crime again without getting caught. The Chaikens found, in short, what appeared to be some connection among what criminals think will happen, what in fact does happen, and how criminals behave.
Why, then, do crime rates seem so hard to change? The answer is not that criminals are irrational. It is rather that they evaluate their risks and opportunities by standards somewhat different from those of the rest of us, and society cannot easily change those risks and opportunities in ways that will have a noticeable effect on criminal behavior.
Chronic offenders may attach little or no importance to the loss of reputation that comes from being arrested; in certain circles, they may feel that an arrest has enhanced their reputation. They may attach a low value to the alleged benefits of a legitimate job, because it requires punctuality, deferential behavior, and a forty-hour week, all in exchange for no more money than they can earn in three or four burglaries carried out at their leisure. These values are not acquired merely by trying crime and comparing its benefits with those of non-criminal behavior; if that were all that was involved, far more of us would be criminals. These preferences are shaped by personal temperament, early familial experiences, and contacts with street-corner peers--by character forming processes that occur in intimate settings, produce lasting effects, and are hard to change.
Whereas the drinking driver, the casual tax cheat, or the would-be draft evader, having conventional preferences, responds quickly to small changes in socially determined risks, the chronic offender seems to respond only to changes in risks that are sufficiently great to offset the large benefits he associates with crime and the low value he assigns to having a decent reputation. Changing risks to that degree is not impossible, but changing those risks permanently and for large numbers of persons is neither easy nor inexpensive, especially since (as we saw in Wayne County, with the felony firearm statute, and in New York, with the Rockefeller drug law) some members of the criminal-justice system resist programs of this kind.
One third of all robberies committed in the United States are committed in the six largest cities, even though they contain only 8 percent of the nation's population. The conditions of the criminal-justice system in those cities range from poor to disastrous. The New York Times recently described one day in New York City's criminal courts. Nearly 4,000 cases came up on that day; each received, on the average, a three-minute hearing from one of seventy overworked judges. Fewer than one case in two hundred resulted in a trial. Three quarters of the summonses issued in the city are ignored; 3.7 million unanswered summonses now fill the courts' files. It is possible that some measure of rough justice results from all this--that the most serious offenders are dealt with adequately, and that the trivial or nonexistent penalties (other than inconvenience) imposed upon minor offenders do not contribute to the production of more chronic offenders. In short, these chaotic courts may not, as the Times described them, constitute a "system in collapse." But could such a system reduce the production of chronic offenders by increasing the swiftness, certainty, or severity of penalties for minor offenders? Could it take more seriously spouse assaults where the victim is reluctant to testify? Or monitor more closely the behavior of persons placed on probation on the condition that they perform community service or make restitution to their victims? Or weigh more carefully the sentences given to serious offenders, so as to maximize the crime-reduction potential of those sentences? It seems most unlikely. And yet, doing some or all of these things is exactly what is required by any plan to reduce crime by improving deterrence. For reasons best known to state legislators who talk tough about crime but appropriate too little money for a big-city court system to cope properly with lawbreakers, the struggle against street crime that has supposedly been going on for the last decade or so is in large measure a symbolic crusade.
By contrast, a steadily growing body of evidence suggests that the family affects criminality and that its effect, at least for serious offenders, is lasting. Beginning with the research of Sheldon and Eleanor Glueck in Boston during the 1930s and 1940s, and continuing with the work of Lee Robins, William and Joan McCord, and Travis Hirschi in this country, Donald West and David Farrington in England, Lea Pulkinnen in Finland, Dan Olweus in Norway, and many others, we now have available an impressive number of studies that, taken together, support the following view: Some combination of constitutional traits and early family experiences accounts for more of the variation among young persons in their serious criminality than any other factors, and serious misconduct that appears relatively early in life tends to persist into adulthood. What happens on the street corner, in the school, or in the job market can still make a difference, but it will not be as influential as what has gone before.
If criminals are rational persons with values different from those of the rest of us, then it stands to reason that temperament and family experiences, which most shape values, will have the greatest effect on crime, and that perceived costs and benefits will have a lesser impact. For example, in a society where people cannot be under continuous official surveillance, the pleasure I take in hitting people is likely to have a greater effect on my behavior than the occasional intervention of some person in a blue uniform who objects to my hitting others and sets in motion a lengthy and uncertain process that may or may not result in my being punished for doing the hitting.
In a sense, the radical critics of America are correct. If you wish to make a big difference in crime rates, you must make a fundamental change in society. But what they propose to put in place of existing institutions, to the extent that they propose anything at all except angry rhetoric, would leave us yearning for the good old days when our crime rate may have been higher but our freedom was intact.
There are, of course, ways of re-organizing a society other than along the authoritarian lines of radical Marxism. One can imagine living in a society in which the shared values of the people, reinforced by the operation of religious, educational, and communal organizations concerned with character formation, would produce a citizenry less criminal than ours is now without diminishing to any significant degree the political liberties we cherish. Indeed, we can do more than imagine it; we can recall it. During the latter half of the nineteenth century, we managed in this country to keep our crime rate lower than it might have been in the face of extensive urbanization, rapid industrialization, large-scale immigration, and the widening of class differences. We did this, as I have argued elsewhere ("Crime and American Culture," The Public Interest, Winter, 1982), by investing heavily in various systems of impulse control through revival movements, temperance societies, uplift organizations, and moral education--investments that were based on and gave effect to a widespread view that self-restraint was a fundamental element of character.
These efforts were designed to protect (and, where necessary, to replace) the family, by institutionalizing familial virtues in society at large. The efforts weakened as the moral consensus on which they were based decayed: self-expression began to rival self-control as a core human value, at first among young, well-educated persons, and eventually among persons of every station. Child-rearing methods, school curricula, social fashions, and intellectual tendencies began to exalt rights over duties, spontaneity over loyalty, tolerance over conformity, and authenticity over convention.
The criminal-justice system of the nineteenth century was probably no swifter or more certain in its operations than the system of today, at least in the large cities, and the economy was even more subject to booms and busts than anything we have known since the 1930s. The police were primitively organized and slow to respond, plea bargaining was then, as now, rife in the criminal courts, prisons were overcrowded and nontherapeutic, and protection against the vicissitudes of the labor market was haphazard or nonexistent. Yet these larger social processes may have had a greater effect on crime rates then than they do today, because then, unlike now, they were working in concert with social sentiments: society condemned those whom the police arrested, the judge convicted, or the labor market ignored. Shame magnified the effect of punishment, and perhaps was its most important part.
Today, we are forced to act as if the degree of crime control that was once obtained by the joint effect of intimate social processes and larger social institutions can be achieved by the latter alone. It is as if we hope to find in some combination of swift and certain penalties and abundant economic opportunities a substitute for discordant homes, secularized churches, intimidated schools, and an ethos of individual self-expression. We are not likely to succeed.
Nor are we likely to reproduce, by plan, an older ethos or its accompanying array of voluntary associations and social movements. And, since we should not abandon essential political liberties, our crime-control efforts for the most part will have to proceed on the assumption--shaky as it is- that the things we can change, at least marginally, will make a significant difference. We must act as if swifter and more certain sanctions and better opportunities will improve matters. Up to a point, I think, they will, but in reaching for that point we must be prepared for modest gains uncertainly measured and expensively priced.
Brighter prospects may lie ahead. By 1990, about half a million fewer eighteen-year-old males will be living in this country than were living here in 1979. As everyone knows, young males commit proportionately more crimes than older ones. Since it is the case in general that about 6 percent of young males become chronic offenders, we will in 1990 have 30,000 fewer eighteen-year-old chronic offenders; if each chronic offender commits ten offenses (a conservative estimate) per year, we will have a third of a million fewer crimes from this age group alone. But other things may happen as well as the change in numbers. A lasting drop in the birthrate will mean that the number of children per family will remain low, easing the parental problem of supervision. A less youthful society may be less likely to celebrate a "youth culture," with its attendant emphasis on unfettered self-expression. A society less attuned to youth may find it can more easily re-assert traditional values and may be more influenced by the otherwise marginal effects of improvements in the efficiency of the criminal-justice system and the operation of the labor market. Natural and powerful demographic forces, rather than the deliberate re-establishment of an older culture, may increase the values of those few policy tools with which a free society can protect itself. In the meantime, justice requires that we use those tools, because penalizing wrong conduct and rewarding good conduct are right policies in themselves, whatever effect they may have.
Copyright 1983 by James Q. Wilson. All rights reserved.
The Atlantic Monthly; September 1983; Thinking About Crime; Volume 252, Number 3; pages 72-88.