Lawrence Sherman: The Need for Research to Keep from Causing Crime with Punishment
Thorsten Sellin was a long-term editor of The Annals of the AAPSS and a criminologist who did the first empirical studies of the death penalty. He concluded from his work that the death penalty had no effect on crime rates. While we still don’t really know the answer to that question, his early work did establish the mold for assessing the effects of public policies on crime. Thus, it is a great honor for me to be elected as a Thorsten Sellin Fellow.
My own work on crime began at the height of the Vietnam War, when my local draft board approved my request to serve as a conscientious objector by working in the New York City Police Department. I didn’t carry a gun or make arrests. But I did spend a lot of time in police cars. The draft board said they approved my work as a civilian analyst because they thought I might get killed.
But I survived to evaluate natural experiments in three cities that had restricted police powers to shoot and kill. My research was cited in the 1985 Supreme Court decision abolishing the Common Law power to kill fleeing suspects. Their decision was based, in part, on my findings that such restrictions caused no increase in either violent crime or attacks on police. In the aftermath of that decision, police killings of U.S. citizens, and particularly African-Americans, were reduced by hundreds of people per year.
I went on to conduct field experiments on a wide range of police practices, including stop-and-frisk tactics against guns. My first experiment in Kansas City showed a 50 percent reduction in gun crimes compared to a similar area where the practice was withheld. This result was replicated six times in other cities, all showing less murder or gunshot wounds. Gun patrols were then widely adopted across the U.S. in the 90s; gun arrests soared, and the national homicide rate dropped by almost half.
Some observers say our research on gun crime may have helped reduce murder. A more important link might be to our discovery that over 60 percent of reported crime occurs at just 3 percent of street addresses. This finding led to some of the largest police departments restructuring their basic patrol operations around the “hot spots” where crime is concentrated rather than around large patrol beats. Police were given further encouragement for that policy by our randomized trial in Minneapolis. That experiment doubled patrol time in 55 high-crime hot spots while leaving it unchanged in 55 others, measuring patrol independently with systematic observation. In the hot spots with more patrol, we found two-thirds less crime than in the hot spots left alone.
Two decades later, the impact of that research remains limited to a few big agencies like New York. Most police departments still waste police patrols in low-crime areas, and put too little patrol into hot spots. Like doctors refusing to wash their hands, police remain resistant to evidence-based practice. So do state legislatures which use research findings that they like but ignore findings that they don’t like—including those from my research on policing domestic violence.
My domestic violence research produced the change in public policy that I regret the most, even though it was the methodological breakthrough I value the most. With the unanimous consent of City Councils in Minneapolis and Milwaukee, I organized police to follow random assignment of arrest or mediation in each case of misdemeanour domestic assault. That Minneapolis experiment became the first randomized trial of arrest in the history of policing. Even better, the experiment had a clear finding of “what worked.” We found that there was half as much repeat violence after arrest as after mediation.
That result was just what a large advocacy community wanted to hear. As an Assistant Attorney General of the U.S. said to me, “it’s a good thing your study got the right result, because otherwise I would have attacked your methodology.” Our initial “arrest works best” finding led to 28 state legislatures, and several other nations, passing laws that mandated arrest in all such cases. This was despite our cautions that the Minneapolis experiment should be replicated before any policy was mandated.
Five years later we were able to repeat the Minneapolis experiment in Milwaukee. There we found the bad news: that while arrest worked best for suspects who had jobs, it backfired terribly for suspects who were job-less. Three independent experiments then confirmed this result: that arrest of unemployed men doubled their level of domestic violence. Yet neither legislatures nor the news media took any notice of this strong evidence against mandatory arrest.
These results finally drove me from doing experiments to building theory. The result was my defiance theory, that criminal sanctions increase crime whenever the people punished are weakly bonded to society and view their punishers as illegitimate. This theory fits the facts of a wide range of criminogenic effects, from arrests of juveniles to the negative effects of British crackdowns on Northern Irish terrorism.
Defiance theory has also been central to our field tests of restorative justice meetings between offenders and their victims, designed to increase the legitimacy of punishment with the emotional impact on offenders of hearing their victims describe the harm offenders caused. In fifteen years on three continents, my colleague Heather Strang and I have completed 12 randomized tests of this police-led practice. Our best results are from 7 controlled trials in the UK, which produced an 8-to-1 return on investment in the costs of crime prevented by restorative justice. Yet even our finding of 27 percent fewer criminal convictions after restorative justice has failed to move the UK government to make use of our results, which we found with their money. At least we can assure the Jerry Lee Foundation and other co-funders of the research that this new approach meets almost any definition of an “evidence-based practice.”
As more of a hedgehog than a fox, I have seen public policy embrace some but not all of the three big ideas reflected in my work: experiments, concentrations, and legitimacy. The first big idea is that to the extent possible, public policies need to have solid randomized experiments as the standard for what is now quite loosely called “evidence-based practice.” The 1997 Report to the U.S. Congress I led on “What Works and What Doesn’t” in preventing crime adumbrated a similar message in President Obama’s inaugural address on all government spending. Since then a small industry has been established to systematically review evidence of crime prevention program effectiveness.
The second big idea is that crime is heavily concentrated in a skewed distribution best known as the “power few.” In the past two decades, policies from policing to parole have slowly invested more in places and people that are high-risk and less in those that are low-risk. But until the evidence of risk is linked to evidence of program outcomes, the use of risk analysis per se may do little good. Untested policies based solely on risk may be no better than those based on tradition or guesswork.
The third idea has had the least effect so far, yet may be the most important. Whatever good it does, criminal justice also causes crime, and a lot of it. One reason for iatrogenic law enforcement may be found in my defiance theory about offenders who view their punishment as illegitimate. But unless we complete a lot more research that is open to that hypothesis, we may continue to cause crime with punishment.
The hedgehog in me says the solution to that problem is idea number 1: better use of better evidence to allocate or withhold spending. So far the Obama administration has encountered stiff opposition to this idea, even in medical practices. Yet the idea is now well-entrenched in the UK’s National Health Service. It is also under discussion at the UK’s national police agency. We may still see a future in which criminology has more impact on policy than at present, especially if it can reach the public to affect politics as well as policy.
I take comfort from the story of James Lind, the doctor whose experiments found a cure for scurvy. Despite the huge cost of this disease to the British navy, it took over 40 years to adopt his recommendation to hand out limes every day to British sailors. As Senator Moynihan once said, politics is not for the short-winded. Neither is research on public policy.
Lawrence W. Sherman is Director of the Jerry Lee Center of Criminology at the University of Pennsylvania and Wolfson Professor of Criminology at the University of Cambridge, where he is a Fellow of Darwin College. These remarks were prepared for his induction as the 2009 Thorsten Sellin Fellow of the American Academy of Political and Social Science.
Press play to hear recording.