A few days ago I wrote about a study investigating how public health in different Chicago neighborhoods correlates with the crime rate; not surprisingly, there were high correlations between the crime rate and certain health outcomes, which touch on some of the most important discussions going on about public health in the city and elsewhere.
Then I came across another intriguing study on socioeconomic conditions in the city’s neighborhoods and crime, covering an important period—the increase of the city’s homicide rate to its historic peak in the early 1990s and through its period of greatest decline to 2000 levels—and taking into account the fact that not all neighborhoods benefitted from this decline.
To put this in some context:
1. Crime has declined across the city since the 1990s. The homicide rate has fallen in most neighborhoods since then. But, as Daniel Kay Hertz has illustrated, it’s increased in some neighborhoods, covering roughly a third of the city. There are a lot of likely reasons for this, none definitive, some more likely than others.
2. Sociologists, criminologists, cops, community leaders, policy-makers, and so forth spend lots of time sorting through those reasons in the hopes that they can be applied towards decreasing homicide and violent crime rates. And the biggest and age-oldest debate is economics versus culture. Or, if you prefer, actual economic poverty versus a “culture of poverty.” Which is a generalization, but… a lot of the debate surrounding crime is generalized. (This is a very good piece by Ta-Nehisi Coates on the persistence of the culture-of-poverty argument, which crosses all kinds of political, class, and racial lines.)
And that’s what’s interesting about this study: “Determinants of Chicago Neighborhood Homicide Trends: 1980-2000,” by Florida State criminologist Brian J. Stults. It’s about both. Stults pulls the data, and finds, like Hertz, considerable differences within the change of neighborhood homicide rates over that period, looking at variation across time and space.
Then Stults runs those numbers against data that we have for the causes and influences that are usually associated with the debate over crime: concentrated disadvantage (largely economic), family disruption (divorce, single parents), social disorganization (the built environment and how people live in it, like vacant properties and residential instability), and finally immigrant concentration—since, frequent perception aside, increasing immigrant concentration is generally associated with decreasing crime.
Using those sources, Stults then runs a couple tests. The first is “group-based trajectory analysis": basically, identifying things that are like one another and develop similarly. In Stults’s case, it’s neighborhoods: ones that have consistently low homicide rates, ones that have moderate homicide rates in 1980 and then increase, and so forth.
He starts with two types of neighborhood: both start with low levels of homicide in 1980; one stays low, the other increases. What distinguishes them?
[T]he only characteristics that significantly differentiate between these two trajectory groupings are initial levels and changes in concentrated disadvantage. Specifically, census tracts with higher initial levels of disadvantage, and increasing levels over time, are more likely to be assigned to the trajectory group with increasing levels of homicide.
What about neighborhoods that start with moderate levels of homicide? Stults identifies three groups—one that declines, one that stays the same, and one that increases. Again, “concentrated disadvantage” is the only predictor.
Finally, Stults has three more groups that start at high homicide levels. One gets better fast; one gets better slowly; one stays bad. And, again, concentrated disadvantage is a statistically significant predictor for whether or not homicide rates decline. But it gets a bit more complicated:
Contrary to the previous regressions, however, additional structural characteristics appear as significant predictors in these models. The comparison of group 7 with group 6 shows that tracts with high and increasing levels of family disruption and immigrant concentration are more likely to be assigned to the group with less of a decline in homicide. Likewise, tracts with increasing levels of family disruption are more likely to be assigned to the persistently high trajectory group (group 8) as opposed to the sharply declining group (group 6).
In short, concentrated disadvantage is a really powerful predictor for all sorts of neighborhood change; in neighborhoods with high levels of homicide, other factors play into neighborhood change.
Then Stults takes another run at the data—"hierarchical growth curve modeling,” which, long story short, doesn’t lump the census tracts into groups. (E.g.: “Even the original authors of the GTM method have warned of the risk of reification, explaining that the groups should be used only as approximations of trajectories.")
Using a different method, Stults gets somewhat different findings. Concentrated disadvantage and family disorganization both predict (especially the former) how bad a neighborhood’s homicide problem is to begin with in 1980—but they don’t show where it’s going.
Specifically, concentrated disadvantage and family disruption only affect starting levels of homicide, while social disorganization and immigrant concentration also influence the trend over time… [I]ncreasing disorganization is associated with a more persistently high homicide level, while immigrant concentration tends to accentuate the overall trend by enhancing the initial increase and subsequent decrease.
(Why the pattern for immigrants? Immigrants tend to immigrate into areas with more homicide—perhaps because of availability or affordability—which subsequently declines. This is just one reason why a lot of people think immigration basically saved many of America’s urban cores, and the effect is such that it’s talked about as a possible reason the nation’s crime rate declined so rapidly.)
So: that’s all pretty wonky, but it’s important to note that Stults tried a couple different ways of looking at the data—which, it should be remembered, is a really broad way of looking at all this—and got somewhat different results. Not all social science comes to a clean conclusion.
And the conclusions he does comes to should sound pretty familiar. Concentrated disadvantage is huge; so is social disorganization. Family disorganization pops up as significant, but only around the edges. Unsurprisingly, the sensible conclusion from it is that it requires multiple approaches in concert and in place—improving the economic foundation of the community while keeping it cohesive. But narrowing the task down to what matters most reduces its immensity.
2 days ago