The maps below present a classification of Chicago’s 2020 residential census tracts based on multivariate analysis. This approach (sometimes called social area analysis or factorial ecology in geography and sociology) is often used to classify small areas in cities. The ten neighborhood types identified on the maps were derived through a two-step process. First, the TRYSYS program was used to factor 34 important tract-level census variables by the Tryon “key-cluster analysis” method. The data come either from the 2020 census or from the 2018-2022 American Community Survey. Three oblique dimensions were identified. Then each tract was scored on the three dimensions (using a simple sum of standardized scores), and tracts were cluster-analyzed using TRYSYS’s iterative partitioning method. Robert B. Dean did the statistical analysis.
These maps are comparable to those generated for 1990 and 2000 data when I was working at the University of Chicago Library and for 2010 data as reported on this blog. (I’m guilty of repeating some of the language from the 2010 post.) Essentially the same variables were used, and nearly the same geographic area was covered. The three dimensions (or clusters) are very similar to the first three dimensions found in the 1990 and 2000 data (although the order of the first two is different) and the three dimensions identified in 2010. The three dimensions involve [1] measures of wealth and high status; [2] measures of traditional urbanity; and [3] measures of linguistic isolation. These three dimensions account for approximately 95 percent of the communality in the 34 variables. I acknowledge that it could be argued that the use of a set of standard census statistical data pretty much foreordains the identification of three dimensions more or less like these, since most census data fits fairly clearly into one of these categories.
In fact, there are other ways to analyze the statistical data. In both 1990 and 2000, a fourth dimension was identified, associated with family type and age. A similar dimension in the 2010 data was not at all significant, and it’s pretty much completely disappeared in 2020. But in 2020 the initial analysis did once again identify a fourth dimension. It brought together five definers: percent of population African-American, percent of population with a female “householder,” percent of working-age population unemployed, percent of families below the poverty line, and (negatively) percent of population white. This cluster had as great a significance level as cluster 3, which groups together measures of linguistic and cultural isolation. But it had the awkward problem of having a high (although negative) correlation with cluster 1, which groups together measures of wealth and status. The program used here—TRYSYS—does allow the analyst to make some choices at certain points. Subsuming cluster 4 into cluster 1 seemed both statistically plausible and, in fact, sensible, since it has similar explanatory power as identifying four dimensions and simplifies the result. Parsimonious results are prized in statistical and social analysis. But I’d be the first to admit there is something a bit odd about combining measures of wealth with measures of ethnicity and social structure in one dimension. The counterargument would be that this is the analysis to which the statistics point. Per capita income, for example, is significantly correlated (.576) with percent white. Median household income, for example, really does have a high negative correlation (-.646) with percent of households with a female householder. It’s true that the measures of wealth and status have a higher correlation with each other than with the ethnic and socioeconomic variables mentioned above, and the latter have a higher correlation with each other than to the measures of wealth and status.
In the end, whether one keeps dimension 4 or not, cluster analysis of Chicago’s 2020 census data reveals the continued distinctiveness of Chicago’s most disadvantaged African-American neighborhoods. They are not only poorer than other parts of the Chicago region; they have certain social characteristics that separate them from, say, neighborhoods with a large number of recent immigrants that are nearly as poor. Neighborhoods whose population includes a large proportion of ethnic Latin Americans, for example, do not have as large a percentage of female householders or unemployment rates that are quite as high as the poorer African-American neighborhoods..
In general, the broad pattern of Chicago’s social geography appears to have changed only in subtle ways in the first decades of the 21st century. This is not surprising when you consider that Chicago’s population and economy have been pretty stable. A few obvious changes are noted on the page with a description of variables. There definitely has also been a considerable amount of ethnic-specific internal migration in the area. Click here for some maps that demonstrate this.
Note that, because the classifications have changed, the color schemes of the 1990, 2000, 2010, and 2020 neighborhood-type maps, while similar, are not completely comparable.
Note also that, while the program forces every tract to be classified into one neighborhood type or another, most of the neighborhoods identified by the classification algorithm are definitely not homogeneous. Minor changes in the census numbers for many tracts would have moved many of them from one category to another.
No claim can be made that this is a definitive analysis of neighborhood types in Chicago. It is in the nature of this kind of analysis that a change in the variables selected or in the parameters set by the analyst can change the results significantly. The best that can be said is that the maps may provide one useful way of analyzing the differences in Chicago’s residential areas.
I’m inclined to argue that the fact that the analysis does seem to result in maps that show coherent patterns of socioeconomic geography does suggest that there’s at least some validity to the statistical exploration presented here.
One final bit of explanation: The thin black lines on the map are tract boundaries. The heavy black line shows Chicago’s city limits. The heavy dark-grey lines show the area’s freeway network.
Here’s a map showing neighborhood types in Chicago and some of its inner suburbs. Nominal scale is 1:100,000. That’s the scale it would have if printed on a 17-x-22-inch sheet of paper. (Click here for key.)
And here’s a map showing neighborhood types in the Chicago region. Nominal scale is 1:250,000. That’s the scale it would have if printed on a 17-x-22-inch sheet of paper. (Click here for key.)
