Associations Between Street Connectivity and Poverty

Abstract

Patterns of street connectivity and poverty in the US and Mexico are investigated by means of spatial analysis and statistical techniques. The evaluation is conducted at three levels of spatial aggregation: 389 (74) metropolitan areas, 3,106 (2,432) counties (municipalities), and 58,953 (27,413) census tracts of the US (Mexico). The article explores whether the physical configuration of the street network may affect geographical concentration of poverty. To quantitatively measure differences in network patterns, we consider six metrics: street density, intersection density, regularity, betweenness, closeness and information centrality. The results reveal that relationships between connectivity and poverty have opposite signs in the two countries. For example, whereas intersection density and betweenness centrality positively influence the spatial agglomeration of low-income households in the US, the reverse pattern of association is observable in Mexico. Furthermore, differences in street layouts between low-income and high-income neighborhoods are more pronounced among Mexican cities.

Appendix

In Table 2 we study regression models results using PM2.5 emissions, traffic fatalities and crimes as dependent variables. For ease of exposition, we examine only ordinary least squares (OLS) regressions including street network indicators and controls for socioeconomic characteristics. Our estimations yielded four findings. First, consistent with past research (Lu and Liu 2016; Sobstyl et al. 2018), one would expect to see a strong association between the network structure and air pollutant emissions. Estimated coefficients support this pattern for both countries. Second, aside from regularity, all network features identify channels through which the built-up environment may affect road traffic deaths in the US. This confirms the early evidence of Ewing et al. (2003), Houston et al. (2004) or Moeinaddini et al. (2014). Conversely, we did not find signs that the same holds true for Mexico. Third, in both countries under review, the composition of the street space structure influence criminal behavior, however, important variables differ among counties and municipalities. Fourth, the Pearson’s $r$ correlations between poverty and the three secondary dependent variables show opposite directions in the two countries. For example, in the US, there is a presence of positive relationship ($r=0.169$, p-value $<0.01$) between the percentage of the population living below the national poverty and reported offences per 1,000 people. At higher poverty levels, the reverse occurs in Mexico ($r=-0.334$, p-value $<0.01$). This last finding partially explains the observed contrary effects of connectivity variables on the response when looking at the two nations.

Table 2 Estimated coefficients for PM2.5 emissions, traffic fatalities and crimes at the county/municipality–level