“ON DETERRENCE”

James Q. Wilson (1975)

Download Word version

The average citizen hardly needs to be persuaded of the view that crime will be more frequently committed if, other things being equal, crime becomes more profitable compared to other ways of spending one’s time.  Accordingly, the average citizen thinks it obvious that one major reason why crime has gone up is that people have discovered it is easier to get away with it; by the same token, the average citizen thinks a good way to reduce crime is to make the consequences of crime to the would-be offender more costly (by making penalties swifter, more certain, or more severe), to make the value of alternatives to crime more attractive (by increasing the availability and pay of legitimate jobs), or both.  Such opinions spring naturally to minds among persons who notice, as a fact of everyday life, that people take their hands off hot stoves, shop around to find the best buy, smack their children to teach them not to run out into a busy street, and change jobs when the opportunity arises to earn more money for the same amount of effort.

                These citizens may be surprised to learn that social scientists who study crime are deeply divided over the correctness of such views.  To some scholars, especially economists, the popular view is also the scientifically correct one – becoming a criminal can be explained in much the same way we explain being a carpenter or buying a car.  To other scholars, especially sociologists, the popular view is wrong – crime rates do not go up because people discover they can get away with it and will not come down just because society decides to get tough on criminals.

                The debate over the effect on crime rates of changing the costs and benefits of crime is usually referred to as 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 more fearful of committing crime.  But that is something of a misnomer, because the theory of human nature on which is erected the idea of deterrence (the theory that people respond to the penalties associated with crime) is also the theory of human nature that supports the idea 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 reasonably rational and 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 only to the cost side of the equation.  There is no word in common scientific usage to call attention to the benefit side of the equation; perhaps ‘inducement’ might serve.  To a psychologist, deterring persons from committing crimes or inducing persons to engage in non-criminal activities are but special cases of using ‘reinforcements’ (or rewards) to alter behaviour.

                The reason there is a debate amongst scholars about deterrence is that the socially imposed consequences of committing a crime, unlike the market consequences of shopping around for the best price, are characterized by delay, uncertainty, and ignorance.  In addition, some scholars contend that a large fraction of crime is committed by persons who are so impulsive, irrational, or abnormal that even of there were no delay, uncertainty, or ignorance 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.  There is nobody else in view.  If the boy steals the purse, he gets the money immediately.  That is a powerful incentive, and it is available immediately and without doubt.  The costs of taking it are uncertain; the odds are at least fourteen to one that the police will not catch a given robber, end 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 100 result in the imprisonment of the offender. In addition to this uncertainty, whatever penalty may come his way will come only after a long delay; in some jurisdiction, it might take a year or more to complete the court disposition of the offender, assuming he is caught in the first place.  Moreover, this young man may, in his ignorance of how the world works, think the odds in his favour are even greater and that the delay will be even longer.

                Compounding the problems of delay and uncertainty is the fact that society cannot feasibly reduce the uncertainty attached to the chances of being arrested by more than a modest amount and though it can to some degree increase the probability and severity of a prison sentence for those who are caught, it cannot do so drastically by, for example, summarily executing all convicted robbers or even by sending all robbers to 20-year prison terms. Some scholars add 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, is he perhaps ‘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 on crime of differences in the probability and severity of punishment has involved using data about aggregates of people (entire cities, counties, states, and even nations) rather than about individuals.  In a typical study, of which there have been several dozen, 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 socioeconomic features of each state that might affect the supply of robbers (for example, the percentage of persons with low incomes, the unemployment rate, or the population density of the big cities) 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).  Most such studies find, after controlling for socioeconomic differences among the states, that the higher the probability of being imprisoned, the lower the robbery rate.  Isaac Ehrlich, an economist, produced the best known of such analyses using data on crime in the United States in 1940, 1950, and 1960.  To simplify a complex analysis, he found, after controlling for such factors as the income level and age distribution of the population, that the higher the probability of imprisonment for those convicted of robbery, the lower the robbery rate.  Thus, differences in the certainty of punishment seem to make a difference in the level of crime.  At the same time, Ehrlich did not find that the severity of punishment (the average time served in prison for robbery) had, independently of certainty, and effect on robbery rates in two of the three time periods (1940 and 1960).

                But there are some problems associated with studying the effect of sanctions on crime rates using aggregate data of this sort.  One is that 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 on the criminal justice system.

                Moreover, if one omits from the equation some factor that affects the crime rate, then the estimated effect of the factors that are in the equation may be in error because some of the causal power belonging to the omitted factor will be falsely attributed to the included factors.  For example, suppose we want to find out whether differences in the number of policemen on patrol among American cities are associated with differences in the rate at which robberies take place in those cities.  If we fail to include in our equation a measure of the population density of the city, we may wrongly conclude that the more police there are on the streets, the higher the robbery rate, and thus give support to the absurd policy proposition that the way to reduce robberies is to fire police officers.  Since robberies are more likely to occur in larger, densely settled cities (which also tend to have a higher proportion of police), it would be a grave error to omit such measures of population from the equation.  Since we are not certain what causes crime, we always run the risk of inadvertently omitting a key factor from our efforts to see if deterrence works.

                Even if we manage to overcome these problems, a final difficulty lies in wait.  The observed fact (and it has been observed many times) that states in which the probability of going to prison for robbery is low are also states which have high rates of robbery can be interpreted in one of two ways.  It can mean either that the higher robbery rates are the result of the lower imprisonment rates (and thus evidence that deterrence works) or 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 without even 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 report of the Panel on Research on Deterrent and Incapacitative Effects, 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 reaching any strong conclusions about the deterrent effect of existing differences among states or cities in the probability of punishment.  The panel (of which I was a member) noted that ‘the evidence certainly favours a proposition supporting deterrence more than it favours one asserting that deterrence is absent’ but urged ‘scientific caution’ in interpreting this evidence.

                Subsequently, other criticisms of deterrence research, generally along the same lines as those of the panel, were published by Colin Loftin and by Stephen S. Brier and Stephen E. Feinberg.

                Some commentators believe that these criticisms have proved that ‘deterrence doesn’t work’ and thus the decks have now been cleared to 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, a bit premature.

REHABILITATING DETERRENCE

People are governed in their daily lives by rewards and penalties of every sort.  We shop for bargain prices, praise our children for good behaviour and scold them for bad, expect lower interest rates to stimulate home building and fear that higher ones will depress it, and conduct ourselves in public in ways that lead our friends and neighbours to form good opinions of us.  To assert that ‘deterrence doesn’t work’ is tantamount to either denying the plainest facts of everyday life or claiming that would-be criminals are utterly different from the rest of us.  They may well be different to some degree – they most likely have a weaker conscience, worry less about their reputation in polite society, and find it harder to postpone gratifying their urges – but these differences of degree do not make them indifferent to the risks and gains of crime.  If they were truly indifferent, they would scarcely be able to function at all, for their willingness to take risks would be offset by their indifference to loot.  Their lives would consist of little more than the erratic display of animal instincts and fleeting impulses.

                The question before us is whether feasible changes in the deferred and uncertain penalties of crime […] will affect crime rates in ways that can be detected by the data and statistical methods at our disposal.  Though the unreliability of crime data and the limitations of statistical analysis are real enough and are accurately portrayed by the panel of the National Research Council, there are remedies and rejoinders that, on balance, strengthen the case for the claim that not only does deterrence work (the panel never denied that), it probably works in ways that can be measured, even in the aggregate.

                The errors in official statistics about crime rates have been addressed by employing other measures of crime, in particular reports gathered by Census Bureau interviewers from citizens who have been victims of crime.  While these victim surveys have problems of their own (such as the forgetfulness of citizens), they are not the same problems as those that affect police reports of crime.  Thus, if we obtain essentially the same findings about the effect of sanctions on crime from studies that use victim data as we do from studies using police data, our confidence in these findings is strengthened.  Studies of this sort have been done by Itzhak Goldberg at Stanford and by Barbara Boland and myself, and the results are quite consistent with those from research based on police reports.  As sanctions become more likely, crime becomes less common.

[…]

I believe that the weight of the evidence – aggregate statistical analyses, evaluations of experiments and quasi-experiments, and studies of individual behaviour – supports the view that the rate of crime is influenced by its costs.  This influence is greater – or easier to observe – for some crimes and persons than for others.  It is possible to lower the crime rate by increasing the certainty of sanctions, but inducing the criminal justice system to make those changes is difficult, especially if committing the offence confers substantial benefits on the perpetrator, if apprehending and punishing the offender does not provide substantial rewards to members of the criminal justice system, or the crime itself lacks the strong moral condemnation of society.  In theory, the rate of crime should also be sensitive to the benefits of non-crime – for example, the value and availability of jobs – but thus far efforts to show that relationship have led to inconclusive results.  Moreover, the nature of the connection between crime and legitimate opportunities is complex: unemployment (and prosperity!) can cause crime, crime can cause unemployment (but probably not prosperity), and both crime and unemployment may be caused by common third factors.  Economic factors probably have the greatest influence on the behaviour of low-rate, novice offenders and the least on high-rate, experienced ones.  Despite the uncertainty that attaches to the connection between the economy and crime, I believe the wisest course of action for society is to try simultaneously to increase both the benefits of non-crime and the costs of crime, all the while bearing in mind that no feasible changes on either part of the equation are likely to produce big changes in crime rates.

                Some may grant my argument that it makes sense to continue to try to make those marginal gains that are possible by simultaneously changing in desirable directions both the costs of crime and the benefits of non-crime, but they may still feel that it is better to spend more heavily on one side or the other of the cost-benefit equation.  I have attended numerous scholarly gatherings where I have heard learned persons subject to the most searching scrutiny 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 what they mean is that since the evidence on both propositions is equivocal, then it does less harm to believe in – and invest in – the ‘benign’ (that is, job creation) program.  If so, they are 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 all we have done is increase the risks facing persons who commit a crime.  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 innocent persons being victimized.

[…]

justice4victims.org