Climate Change and Levels of Violence in Socially Disadvantaged Neighborhood Groups

Abstract

The current study examines the link between climate change and neighborhood levels of violence using 20 years of monthly climatic and crime data from St. Louis, MO, USA. St. Louis census tracts are aggregated in neighborhood groups of similar levels of social disadvantage, after which each group is subjected to time series analysis. Findings suggest that neighborhoods with higher levels of social disadvantage are very likely to experience higher levels of violence as a result of anomalously warm temperatures. The 20 % of most disadvantaged neighborhoods in St. Louis, MO, USA are predicted to experience over half of the climate change-related increase in cases of violence. These results provide further evidence that the health impacts of climate change are proportionally higher among populations that are already at high risk and underscore the need to comprehensively address climate change.

Appendices

Appendix 1: Neighborhood Grouping Methodology

In order to create similar neighborhood groups, census data from 2000 (the midpoint in the series) are collected for all 113 census tracts in the city of St. Louis (3 tracts are excluded because their population was below 500) using the Neighborhood Change Database (Neighborhood Change Data Base [computer program]. Washington: The Urban Institute; 2004).

Next, a social disorganization index is developed for each census tract using seven measures of structural disadvantage. These seven measures include the percentage of people below the federal poverty threshold, the unemployment rate, the rate of high school dropouts, the percentage of female headed households, the proportion of young Black males (15–24 years old), the percentage of properties that are rental units, and the percentage of homes that are vacant. Reliability analysis on the elements of the index yields a Cronbach’s alpha of 0.804, suggesting substantial similarity between the individual elements to justify grouping them. Previous neighborhood studies have used similar indexing techniques to measure extreme disadvantage in neighborhoods.7,49,50

The 7 census measures are subsequently standardized and aggregated to create a social disorganization rank score for the remaining 110 census tracts. In order to promote normally distributed dependent variables, a choice is made to group the census tracts into 5 equal groups of 22 to allow for further study. Group 1 is the least disadvantaged group, whereas group 5 is the most disadvantaged group of census tracts. This strategy creates enough monthly counts of violence in the least disadvantaged groups to conduct further analysis.

As Table 1 in the article body indicates, some variability in the total population between the neighborhood groups exists. Considering that violent crime counts are actually higher in the groups of neighborhoods with the lowest population, this should not pose an issue for analysis. The five groups display the expected connection between higher levels of social disadvantage and higher levels of violence. The poverty rate in group 1 (9.191 %), for instance, is well below that of group 5 (43.146 %). What is particularly noteworthy is the large difference in the percentage of young Black males (15–24 years old) in the groups. Group 1 only contains 0.269 % of this high-risk group, whereas group 2 has 10 times as many at-risk youth (2.471 %). This illustrates the continuing racial divide in St. Louis where socially disadvantaged neighborhoods tend to be predominantly African American.

Group 1 (see Table 1) with the lowest levels of disadvantage also has the lowest levels of violent crime. Subsequent groups show increasing crime rates and increasing levels of disadvantage. In fact, group 5 has a violent crime rate more than 10 times that of group 1 (5,059 vs. 494). This indicates that separating distinct neighborhood groups using the disadvantage measure likely captures the essence of socially disorganized neighborhoods and their (in)ability to control crime.4–6

Appendix 2: Additional Analysis Comparison Sites

One of the reviewers brought up an important issue. How do other places stack up to the findings in St. Louis? While it is difficult to locate monthly data by neighborhoods, a quick comparison of four additional cities (Cleveland (NIBRS)¹, New Orleans (UCR), Boston (UCR), and Phoenix (UCR)²), reveals more support for the general idea in this paper (see Table 3). The city in the analysis most comparable to St. Louis is New Orleans, followed by Cleveland. New Orleans also has extremely disadvantaged neighborhoods and an exceptionally high level of violence. Perhaps one of the key differences is that New Orleans has a substantial population of affluent residents in the downtown area, which may explain the slight difference in the “Tempanomaly” variable. What is interesting is that the climate change proxy variables of St. Louis and New Orleans are fairly close (0.739 and 0.651) despite the differences in time period examined. What is of further interest is the fact that New Orleans’ seasonality pattern is smaller than that of all other places. This is not odd because New Orleans has less annual temperature variation than all other places as its climate is subtropical (winters are relatively pleasant). Cleveland also shares many similarities with St. Louis, but unfortunately, only 5 years of data were available at present; this likely underestimates the coefficients for seasonality and climate change. The other two cities (Boston and Phoenix), which were selected here because they house a more affluent population, display smaller coefficients for both seasonality and climate change.

The results of this brief comparison thus fall in line with the results of our neighborhood comparison. The advantage of the neighborhood group approach is that the socioeconomic, cultural, and climatic conditions are kept relatively constant within one city, whereas this is probably not as clear-cut when comparing cities.

Table 3 St. Louis and comparison sites