Last week, the U.S. Bureau of Justice Statistics issued a new report compiling nearly two decades of data on gun crime, Firearm Violence, 1993-2011. No doubt, many readers will pore over the report’s abundant tables and graphs looking for support for their views on gun control. However, I was most struck by a breakdown of firearm violence based on population size (table 5). Among the six size-based categories, the most dangerous places were cities of 500,000-999,999 — the category containing Milwaukee (pop. 597,867). These mid-large cities not only have rates of gun crime that are about four times higher than cities of less than 100,000, but they are also forty-four percent higher than cities of one million or more.
More specifically, according to the National Crime Victimization Survey, there were 4.6 nonfatal firearm victimizations per 1,000 persons age twelve or older in the mid-large cities in 2010 and 2011. (Nationally, homicides constitute only two percent of all gun-related crimes, so the NCVS numbers would not change much if fatalities were included, too.) The second-highest rate was 3.9, for cities with 250,000-499,999.
The numbers look very different today than they did in 1996-1997, when the Milwaukee-sized cities were tied for second place with 7.3 victimizations per 1,000, and the medium-sized cities (250,000-499,999) led with 10.3.
I have two reactions to the data. First, the relationship of community size to gun violence is in some respects predictable, and in others quite puzzling.
It is predictable at the lower end. One would stereotypically think of smaller communities as places in which there is less crime because people know one another, which may lead to a greater sense of social responsibility, more effective informal social controls, and more effective policing. And, indeed, almost invariably since 1996-1997, the lowest rates of gun violence have been found in unincorporated communities and other places of less than 100,000. (In the BJS chart, the former communities are referred to, oddly, as “not a place.”) Moreover, the mid-small cities (100,000-249,999) also normally do quite well. Indeed, there is often a wide gap between the mid-small cities and the medium-sized cities. For instance, the mid-small rate in 2010-2011 was 1.3, as compared to the medium rate of 3.9. We might hypothesize that, as cities grow, there is some tipping point around the quarter-million mark at which the things that work in smaller communities stop working.
But, above 250,000, things get more puzzling: there is no consistent pattern in how the medium, mid-large, and large cities perform. If there is a tipping point at about the quarter-million mark, there seems little incremental damage done with further growth. Indeed, cities of more than one million often have lower rates of gun violence than one or more of the lower-level categories. Most strikingly, in 2006-2007 and 2007-2008, the largest cities even had lower rates of gun violence than the places of less than 100,000. By contrast, in every two-year period since 2005-2006, the Milwaukee-sized cities have been the most dangerous. I don’t have a good explanation for this, but let me suggest one possibility: economies of scale in policing. The larger cities with larger police departments may be able to realize certain economies of scale that tend to make for more efficient policing than is typically found in medium and mid-large cities.
Second, the data put into context some of the claims made by police leaders regarding the effectiveness of new policing technologies and strategies. Consider, for instance, a hypothetical new police chief taking office in an average mid-large city in 1996. If that police chief were still around today, he or she would undoubtedly claim credit for an impressive-sounding drop in gun violence from 7.3 per 100,000 in 1996-1997 to 4.6 in 2010-2011 — a drop of more than one-third. The chief could doubtlessly point to new policies or practices adopted on his or her watch that plausibly contributed to the drop.
There are, however, two important problems highlighted by the data. First, the one-third drop is not actually all that impressive when compared with even larger drops elsewhere. In fact, the one-third decline in gun violence in mid-large cities was the smallest rate of decline for any of the six categories covered by the BJS report between 1996-1997 and 2010-2011. Second, the fact that comparable or even larger-percentage drops were recorded in the smallest communities over the same time period casts doubt on how much policing innovations contributed to the improved gun-violence numbers in the bigger cities. It seems doubtful that cities of less than 100,000, much less unincorporated communities with no active government, have been wringing large gains from, say, compstat or hot-spots policing. Rather, the decline of gun violence in small places may point to the existence of broader social or cultural changes that have played an important role on a national scale. (See also figures 9 and 10 in the BJS report, which show large declines in gun violence since the mid-1990s across all major geographic regions of the country and across the urban-suburban-rural divide.) So, our hypothetical police chief may merely be benefitting from a rising tide raising all boats (or more accurately, I suppose, a sinking tide lowering all boats).
This is not to say that policing is irrelevant. For instance, one imagines that, among the mid-large cities, several did better than the average drop of one-third and several did worse; policing may play an important role in explaining some of these differences. Rather, the point is that it can be difficult to disentangle police performance as a cause of changes in crime rates from broader social forces over which police departments have little control. We should be wary of heaping too much praise on police departments for declining crime rates, just as we should be wary about heaping blame on police departments for increasing crime rates.
Cross posted at Life Sentences.
You can follow any responses to this entry through the RSS 2.0 feed. You can skip to the end and leave a response. Pinging is currently not allowed.