Local Geographic Variation of Public Services Inequality: Does the Neighborhood Scale Matter?

Abstract

This study aims to explore the effect of the neighborhood scale when estimating public services inequality based on the aggregation of social, environmental, and health-related indicators. Inequality analyses were carried out at three neighborhood scales: the original census blocks and two aggregated neighborhood units generated by the spatial "k"luster analysis by the tree edge removal (SKATER) algorithm and the self-organizing map (SOM) algorithm. Then, we combined a set of health-related public services indicators with the geographically weighted principal components analyses (GWPCA) and the principal components analyses (PCA) to measure the public services inequality across all multi-scale neighborhood units. Finally, a statistical test was applied to evaluate the scale effects in inequality measurements by combining all available field survey data. We chose Quito as the case study area. All of the aggregated neighborhood units performed better than the original census blocks in terms of the social indicators extracted from a field survey. The SKATER and SOM algorithms can help to define the neighborhoods in inequality analyses. Moreover, GWPCA performs better than PCA in multivariate spatial inequality estimation. Understanding the scale effects is essential to sustain a social neighborhood organization, which, in turn, positively affects social determinants of public health and public quality of life.

Keywords: GWPCA; PCA; health; inequality; neighborhood; public services; scale.

Figure 1

The study area—Quito, Ecuador.

Figure 2

Overall workflow. White boxes represent input, white boxes with dashed borders represent specific methods, and grey boxes represent results. PCA = principal components analyses, GWPCA = geographically weighted principal components analyses.

Figure 3

The first principal component of the GWPCA in three zoning systems.

Figure 4

Different public services inequality results based on three different scales of neighborhoods units: (a) original census blocks; (b) self-organizing map (SOM)_based zoning system; and (c) spatial the “k”luster analysis by tree edge removal (SKATER)_based zoning system. The x-axes represent the number of neighborhood units in each zoning system, y-axes represent the PCA_ and GWPCA_based measures of public services inequality based on seven accessibility variables (NonDri: no access to drinking water; NonSew: no access to the sewerage system; NonCol: no access to a garbage collection service; NonEle: no access to the public electricity grid; Dist_H: limited access to health care services; Dist_E: limited access to educational services; Green: limited access to green areas).

Figure 5

Spatial visualization of the GWPCA-based public services inequality in three different zoning systems: (a) SOM-based zoning system; (b) SKATER-based zoning system; and (c) original census block.

Figure 6

Local outliner detection for the public services accessibility indicators for the original census blocks zoning system: (a) for the area of the new city center; and (b) for the area of the old city center; x-axes represent the seven accessibility variables (NonDri, NonSew, NonCol, NonEle, Dist_H, Dist_E, Green); and y-axes represent the re-scaled spatial structures of seven accessibility variables.

Figure 7

Local outliner detection for the public services accessibility indicators: (a) SKATER_based zoning system for the new city center area; (b) SKATER-based zoning system for the area of the old city center; (c) SOM_based zoning system for the new city center area; and (d) SOM_based zoning system for the area of the old city center; x-axes represent the seven accessibility variables (NonDri, NonSew, NonCol, NonEle, Dist_H, Dist_E, Green); and y-axes represent the re-scaled spatial structures of seven accessibility variables.