In review of:
Chakraborty, J., Grineski, S. E., & Collins, T. W. (2019). Hurricane Harvey and people with disabilities: Disproportionate exposure to flooding in Houston, Texas. Social Science & Medicine, 226, 176–181. doi: 10.1016/j.socscimed.2019.02.039
As the research conducted and presented in this paper is concurrent with the pillar of health geography and epidemiology, the main goal of this paper is to, “[conduct] the first distributive environmental justice study of the relationship between flooding caused by Hurricane Harvey and locations of disabled individual in Harris County, Texas,” the most populated county severally impacted from Harvey-induced inundation (Chakraborty et al., 2019). Having found a gap in the environmental justice framework where vulnerability assessments for natural disasters have primarily examined other social vulnerable groups including, ethnic minorities, sexual minorities, immigrants, and socially and linguistically isolated households, this is the first study to systematically measure the impact of natural disasters on the most physically vulnerable people, despite often being the first victims of a disaster. Thusly, the purpose of this study is to highlight the necessity of creating policy, disaster planning and response regimes to integrate the needs for people with physical and mental disabilities and create a preliminary framework for further case studies.
The study area, Harris County, due to its close proximity to the Gulf of Mexico is reportedly one of the most vulnerable urban areas globally for tropical storms and hurricanes, and during hurricane Harvey nearly 70% of the county was engulfed in over a foot of water, with 12% of residential buildings flooded. Yet, details on the effects of Harvey on people with disabilities is not available though it has been estimated to have have affected over a million people. This shortcoming is part of the logic behind this study.
The data utilized in this study included seven disability variable derived from the 2012-2016 American Community Survey, a digital flood inundation grid derived from high water marks from the U.S. Geologyical Survey, Harris County Flood Control District, federal contractors and FEMA’s Recovery Division, a raster dataset from FEMA’s Hazard and Performance Analysis’s Geospatial Unit contains flood depth values as an attribute of each pixel, and to estimate the flood extent Harvey’s Inundation Footprint, a cartographic product prepared by the FEMA Region 6 Mitigation Division (TX-DR-4322) was used. The flood extent is the dependent variable in this analysis and the set of disability variables in conjunction with other socio-variables like race, age, English proficiency, low income, households with no vehicle, and unemployment, as well as house ownership rate, and vacant housing units as the explanatory variables.
The spatial statistical methodology was to create an OLS, using descriptive statistics to model the relationship between each disability variable on tract level to build a multivariate statistical model, after controlling the other socio-demographic factors. The multivariate model is based on generalized estimating equations (GEE) because, ”they relax several assumptions of traditional regression models and impose no struct distributional assumptions for the variables analyzed while accounting for clustering of neighbourhood tracts,” because they estimate unbiased marginal regression coefficients, and intracluster correlation estimates are not modelled. Next, a intracluster dependency correlation matrix was generated and specified as exchangeable, with normal distribution.
The results showed that all disability variables showed a statistically significant linear trend, meaning a concentration of people of all disability types as the area extent of flooding increases, with cognitive disabilities being having the largest difference between the largest and highest quintile, so are the most impacted.
Overall, I rate this paper a 9/10, it provided a good explanation and defense for why they used the statistical methods they did, and the steps used to estimate the flood extent seemed very thorough. As well, I liked that they accounted for more variables than just different disabilities because there is always intersection between socio-cultural variables. However, I did not like that in their explanation of GEE that they did not provide the different assumptions that that model relaxes, as opposed to other regression models.