Up-tick in violent crime

As noted earlier this week in Sentencing Law and Policy and documented here and here, Attorney General Gonzales notes that violent crime is up for 2005 and in the first half of 2006, especially in some large urban areas. Here in Washington, DC, for example, murders are not higher in 2006, but a crime emergency was declared this summer to address real and perceived up-ticks in robbery, especially high profile robberies of tourists on the National Mall and in Georgetown.

First, the Attorney General rightly noted that violent crime is still low relative to much of the last 30 years, and he does not offer new federal money to combat this threat, at least until further study. So far makes sense to me.

So USDOJ government is proposing a study:
Specifically looking at trends in gang violence, drug trafficking, and how inmates released from prison may have contributed to the increase. The Usual Suspects, perhaps? Not unreasonable, though I hope this is only some example and not a comprehensive list of topics.

The study will be rolled out in three phases: 1) looking at crime increases in cities, 2) analyzing those results for any trends, and 3) identifying federal programs that can help. Hmmm. Let's check those out a bit.

1. Looking at Cities.
Good idea, from where I sit. Cities have spent a lot of resources on Homeland Security, and traditional law enforcement and prosecution resources are already strained (aren’t they always?), COPS program de-funded, etc. Comparative analysis of urban areas makes sense, if that’s where the up-tick is, and it does seem so.

2. Trends.
It is appropriate for the nation’s law enforcement officer to try and get ahead of the curve. And it is hard to get ahead of a curve, primarily because many curves are “white noise.” And curves continue to curve one way, then another. Whoo, I'm getting dizzy.

Six months or 18 months does not make a trend. The more reasonable time period (a decade or more) and more prominent trend (downwards) is the Crime Drop in America during the 1990’s and into most of this century so far (See Blumstein and Wallman’s (eds.) great book here). The authors thoroughly explore the role of guns, the limited but measurable role of prison expansion (but clear diminishing returns of continued prison growth), changes in drug markets (at first chaotic, then more controlled), policing, economic opportunity, and demographics. A tour-de-force and an important starting place for a prospective study. The popular book Freakonomics also theorizes on the subject here.

The Danger of Microfocus: One problem involves reading too much into a short term spike. This point is rather obvious and is most likely to occur in “moral panics” based on exaggerated claims, media hype, and fear. If It Bleeds It Leads. There will be many turning points in the short run (witness the nonsense of day-to-day interpretation of every blip in the Dow Jones average, as if each daily turn was important to the big picture and in any way intelligible).

The Drive to Find a Cause: Statisticians refer to Regression Toward the Mean. Think of it like gravity, what comes up must some down (or vice versa in this case) – toward some average (natural?) level. Crime was way down, now its up. That might be it, and we don’t need a big study for that.

Bottom line on studying up-tick as a Trend: Hey, if it gets us looking at 3) Policy, then more power to the Justice Dept. Let’s not over-hype the trend yet, though.

3. Policy – specifically federal or state programs/policies.
Well, naturally the federal government is specifically interested in how it spends its (Oops, I mean Our) dollars. But States are the “laboratories” for policy experiments according to Justice Brandeis (New State Ice v. Liebmann, 1932). The federal government needs to fund comparative state studies of policy impact. What works in state CJ policy? What steps are required to do that right? Why?
It’s a big, but not insurmountable task.

Why not just assume more incarceration, or more police for that matter, works? Such political claims have worked before. Well, the medical model of progress suggests that, before prescribing course of treatment, modern medical science requires a strong standard of evidence of effectiveness, usually through random clinical trials before widespread introduction of a treatment. The folks at the Campbell Collaboration are blazing a trail here for CJ policy. Public safety requires our due diligence in determining what works.

Let’s take the example of the role of Incarceration Policy in crime changes as an example. Sentences to a period of incarceration could be studied across both time (1970- 2006, say) and space (geographically) – these usually take a form labeled quasi-experiments (“almost” experiments). Other approaches – true experiments that feature random clinical trials to be used where and when appropriate and feasibleUp-tick in violent crime - are called for. The goal is to separate the random fluctuations from key influences, both policy and environmental causes.

Differences in programs and policies that affect crime could be sifted out from other factors like those mentioned in the Crime Drop, yielding some compelling ideas about what works.

Sure there are obstacles to be overcome:

Sentencing information is not readily available or accessible to independent review, allowing experimental replication etc. And experiments require reporting on the results (as Marc Miller puts it “laboratories are where people conduct experiments and report on their findings.”), thus we have to know what each state did (programs and policies, in detail and at the street level) on a common scale. We have to play nice and share. Not impossible, so let’s get started.

AG Gonzales can leave a lasting legacy in this area. He is making some of the right noises. Let’s support him in this.