President Obama’s talk on gun violence in Chicago two weeks ago continued a transition begun in his State of the Union address. Whereas the current conversation about guns started with a focus on mass shootings, mental health and the killings in Newton, the president’s more recent speeches have centered on communities that suffer violence on a daily basis.
This American Life’s two-part series on Harper High School in Chicago delved deep into one of these communities and helped illustrate a point that most statistics on gun violence miss: So much time and effort is devoted to trying to keep kids safe that could or should be devoted to a more traditional education.
We are well aware by now of how high the costs of gun violence, and of crime more generally, are in these urban communities. But one seemingly straightforward approach to reducing crime and violence — increasing the numbers of police — is largely absent from policy conversations. Sure, conservative commentators extol the benefits of that beautifully euphemistic tactic called “proactive policing.” But would just hiring more police do any good?
A new working paper from the National Bureau of Economic Research tries to answer this question by overcoming some of the classic problems with determining the impact of police on crime. Using a sample of 242 U.S. cities with more than 50,000 people between 1960 and 2010, researchers Aaron Chalfin and Justin McCrary of UC Berkeley draw from several sources of data on police force levels to remedy measurement error. They also include controls to account for the possibility that a more general phenomenon — say, a “tough-on-crime” attitude — affects not only levels of police, but also other law enforcement policies such as harsher sentencing. (This would cause them to pick up on the effect of both the “tough-on-crime” attitude and police.)
The researchers find that police have a significant impact. Across the entire sample of cities, a 10 percent increase in police leads to a 3.4 percent decline in violent crime and a 1.7 percent decline in property crime. In individual crime categories, a 10 percent increase in police leads to a 6.7 percent decline in murder, a 5.6 percent decline in robbery and a 3.4 percent decline in motor vehicle theft.
Chalfin and McCrary go a step further and try to estimate a benefit-cost ratio (BCR) for the provision of an additional police officer, first in the average American city and then for each city in their sample. To do this, they employ a measure called the VSL, or the Value of a Statistical Life, to obtain a rough estimate of the cost of a murder and civil jury awards to estimate the cost of other crimes. For their preferred VSL, Chalfin and McCrary find a BCR of 1.63, “indicating that, in a typical U.S. city, an additional dollar allocated towards policing is predicted to save $1.63 in costs to crime victims.”
In one final group of estimates, Chalfin and McCrary admit they are speculative and might be “heroic” because of the assumptions involved. But these estimates, which outline the benefit-cost ratio of hiring an additional police officer in each of the 242 cities, are revealing.
First, as the researchers note, there is large variability in the benefit of an additional police officer. Some cities — Sunnyvale, Calif.; Stamford, Conn.; Alameda, Calif. — have BCRs well below 1, which indicates that they probably have too many police officers. In contrast, places like Gary, Ind., Baton Rouge, La. and Camden, N.J. have high BCRs that suggest significant benefits to hiring an additional police officer in those cities.
There is a distinct relationship between poverty and city-level BCRs (the correlation is 0.59), but there is no relationship between population and BCR.
The researchers also hint at some of the commonalities among cities with high and low BCRs. Those with high BCRs are both big and small, have both high and low levels of policing, and have high poverty and crime rates. The correlation between poverty and the BCR in the 30 cities with the highest BCRs is 0.46. The correlation between population and the BCR in these cities is -0.2, but is not statistically significant.
The cities with low BCRs are smaller, low-poverty and low-crime cities whose policing levels are “low to moderate.” The correlation between poverty and the BCR in the 30 cities with the lowest BCRs is 0.35, but is not statistically significant. The correlation between population and the BCR in these cities is 0.24 and again not statistically significant.
The paper is particularly timely because it offers the possibility of estimating (very, very roughly) the impact of sequestration on the cost of crime in American cities.
I’ve taken the Center for American Progress’s calculation of state-by-state sequestration impacts and simply scaled them by city population. I then assumed that 5 percent of these roughly estimated city sequestration cuts would be applied to reductions in spending on policing. This type of policing cut is not unreasonable in magnitude and may be quite conservative. It would imply, for instance, a reduction of New York City’s police force by about 140 (nearby Newark, N.J. laid off 160 police officers in 2010) and of only one police officer in Gary. But again, I am only taking a stab in the dark.
The table below shows that sequestration would actually benefit Sunnyvale, with the budgetary savings from the cuts outweighing the increased crime costs. In Chicago and Gary, by contrast, the increased crime costs are much, much higher than the budgetary savings from sequestration (as is implied by a BCR that is greater than one).
The table, rough as it is, is simply meant to illustrate that, to the extent that the sequestration has impacts on the police forces of already cash-strapped cities, crime costs will be affected. And in some cities, particularly those with high poverty, those crime costs will increase significantly.
BCRs, crime costs and VSLs can go some way toward clarifying debates about crime and violence and how to prevent them. In the case of the sequestration, they can highlight that budgetary savings can actually mean substantial real-world costs. But they are also abstractions that cannot substitute the richness of stories like that of Harper High School, which give content to the numbers.