Project Safe Neighborhoods has been among the highest-profile and best-funded national violence prevention initiatives of the past two decades, involving allocations of about $1 billion to U.S. Attorney’s Offices across the country. Evaluations to date have generally been positive, but a new study of the PSN experience in Chicago highlights the challenges of building on early success.

The researchers, Ben Grunwald and Andrew Papachristos, attempted a rigorous, beat-level analysis of the impact of PSN on troubled neighborhoods in the Windy City, which had a distinctive approach to PSN that seemed quite effective at first.

With each of the nation’s U.S. Attorneys authorized to develop locally tailored PSN programs, there was considerable city-to-city variation in implementation. In Chicago, PSN had three primary components: (1) diversion of select gun-related cases from state to federal court, where there were often higher sentences available; (2) creation of a multi-agency “gun team” to investigate gun trafficking cases; and (3) development of in-person offender notification forums involving parolees and probationers who had been convicted of gun crimes. Forum attendees were both warned of the punishment they would face if they committed new gun crimes and offered social services to support better choices in the future.

PSN was rolled out in two phases in Chicago. The first, beginning in 2002, focused on the high-crime, high-poverty 11th and 15th police districts on the city’s west side. An initial evaluation covering the initiative’s first three years found a 37% decrease in homicides in the targeted communities, with the largest reductions associated with higher participation levels in the offender notification forums.

Then, in 2005 and 2006, PSN was expanded to cover four additional police districts, thereby reaching about one-quarter of the city and almost all of the highest-crime districts. However, funding and personnel were not increased.

In order to study the impact of PSN, Grunwald and Papachristos focused on beats, a subdistrict level of police organization encompassing a few thousand residents. They studied twenty-four beats receiving PSN treatment in the initial (2002) rollout, and thirty beats that were part of the second phase. They further identified a matching set of non-PSN beats that were similar in important respects, such as pre-PSN homicide rates.

Grunwald and Papachristos first sought to determine what happened in the initial-rollout beats through 2010. (The earlier Chicago study included only 2002-2004 data.) Dishearteningly, after the big homicide drop in the early PSN years, they found that homicide rates rebounded between 2006 and 2010, although they continued to remain below pre-PSN levels. Overall, the net drop in homicide was about 10% over the PSN years. The matched control (i.e., non-PSN) beats followed a similar pattern of declining homicide before 2006 and rising afterward, suggesting that at least some of the net drop in the PSN beats was probably due to other factors besides just the PSN program.

Consistent with the earlier Chicago study, the new paper found a statistically significant correlation between offender notification forums and homicide decline. However, when the researchers excluded the first three years from analysis, the association became much weaker and lost statistical significance, suggesting that the forums became less effective over time.

Next, Grunwald and Papachristos analyzed the thirty second-phase beats to determine whether PSN was as successful there as it had been in the initial-phase beats. They found that PSN had less of a splash in the new areas. Homicide declined modestly in the first year, but then tended to increase or remain stable. Overall, they found no statistically significant correlation between PSN operation and either gun homicide or homicide generally. Likewise, there was no statistically significant correlation between forum participation and either measure of homicide.

In short, the data through 2010 point to declining PSN effectiveness in the initial-rollout areas and little or no impact at all in the second-phase areas.

These are disappointing results — and, of course, do not even take into account the recent horrific spike in killings in Chicago.

What went wrong? Grunwald and Papachristos’s data do not permit any definitive answer to this question, but they suggest two possibilities. First, “the effects of large-scale social interventions can dissipate over time. It is possible, for example, that focused deterrence strategies like PSN exert a short-term shock on homicide rates that dissipates over time as residents become accustomed to the new and more punitive criminal justice ‘regime’” (155). Second, the expansion of PSN without any corresponding increase in PSN resources “may have diluted the treatment dosage” to the point that PSN lost much of its effectiveness.

Both explanations seem quite plausible to me. There may be another dynamic at work, as well. I have an admittedly pessimistic view of efforts to achieve big improvements in the effectiveness of established organizations. (My ideas here, I might note, are not limited to criminal-justice or even governmental agencies.) Here is a recurring pattern. First, a new initiative within the organization is established. The exciting new program tends to draw some of the most capable individuals in the organization (possibly to the detriment of established programs). When talented people focus their attention on an objective and are given resources and institutional support, they are almost always able to produce some demonstrable gains in the short term. However, it can be very difficult to maintain success as the program expands. If no personnel are added, then the original cadre will be spread increasingly thinly and thereby lose effectiveness. If new personnel are added, they will tend to be less capable and dedicated — both because the program is no longer the exciting Next Big Thing and because new people added to a larger, existing program will have less of a sense of ownership of the program than the people who initially launched it. Meanwhile, those initial leaders will drift away over time, pulled into higher-level management positions or drawn to some even newer program offering the gratification of quick success through the reaping of low-hanging fruit somewhere else. Note, too, that quick successes are sometimes achieved through short-term strategies that may create bigger problems over the long run. (It is possible, for instance, that more aggressive police stop-and-frisk practices fit this model.)

I am not familiar enough with the Chicago story to know whether or to what extent it reflects these particular organizational pathologies. (Grunwald and Papachristos did note, however, that “the PSN coordinators and core support team had to conduct several ‘reboot’ trainings to promote buy-in among new personnel and ensure continued fidelity to the offender notification forums” (142).) However, having observed other promising crime-reduction initiatives fall short of expectations over the years, my sense is that some of these organizational dynamics are not infrequently part of the problem.

The new paper is Ben Grunwald & Andrew V. Papachristos, Project Safe Neighborhoods in Chicago: Looking Back a Decade Later, 107 J. Crim. L. & Criminology 131 (2017).

Cross posted at Life Sentences.