Housing and Urban Heat: Assessing Risk Disparities

Abstract

Heat is the leading weather-related cause of death in the United States, and housing characteristics affect heat-related mortality. This paper answers two questions. First, how do heat risk measures vary by housing type and location in San José, California? Second, what housing and neighborhood factors are associated with greater heat risk? We first create a parcel dataset with housing, heat risk, and neighborhood characteristics. We then use a combination of descriptive statistics, exploratory mapping, and linear regression models to analyze associations between housing, neighborhoods, and heat risk. The results indicate that households of different housing types face varying degrees of heat risk, and the largest disparities are between detached single-family (lowest heat risk) and multifamily rental (highest heat risk). Air conditioning availability is a major contributing factor: the probability of not having central air conditioning is much lower for detached single-family (44.9%) compared with multifamily rental (73.7%). There are also heat risk disparities for households in neighborhoods with larger proportions of Hispanic and Asian residents. This research demonstrates the need to understand heat risk at the parcel scale and suggests to policymakers the importance of heat mitigation strategies that focus on multifamily rental housing and communities of color.

Keywords:

• Climate adaptation
• extreme heat
• low-income housing
• urban environment
• California

Acknowledgments

We thank HPD’s editor and two anonymous reviewers for their thorough and thoughtful comments. We appreciate feedback from attendees of the UCLA Climate Adaptation Symposium (2021) and Association of Collegiate Schools of Planning conference (2021), including discussant Sara Meerow’s comments at ACSP. We are grateful to Carina Gronlund and Greg Preston for generously sharing R code, and to Peter Friedenbach for his invaluable assistance in using SCU’s high-performance computing lab. All errors are our own.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Notes

1 In terms of attached housing, the assessor differentiates between apartments, which include multiple units on the same parcel, and townhouses/condos, for which each individual unit is assigned an individual parcel.

2 See Appendix Table A2 for a comparative analysis using a much cooler LST scene from September 29, 2021. This scene is 9 °F cooler than the August scene, and represents the most complete dataset available with minimal cloud cover at a similar time of year, to minimize differences from solar incidence. We find consistent results across the two dates. From this we conclude that despite differences in LST due to meteorological conditions, relative spatial distribution and z scores are consistent for the purposes of this analysis.

3 We focus on central AC provision for conceptual and practical reasons. Conceptually, we focus on central AC because the literature shows it to have a greater connection with health outcomes. Practically, our models of window AC had very low sensitivity, meaning they were poor at identifying true negatives.

4 We addressed spatial dependence using OLS models with spatial fixed effects, rather than spatial regression models, for practical reasons. Spatial autocorrelation was indicated by Moran’s I values, calculated on a 25% sample of parcels, ranging from 0.55 to 0.86 depending on the model. This calculation used a k-nearest neighbors spatial weights matrix (k = 5). Spatial regression models, however, proved to be computationally infeasible with the entire parcel dataset, even in a high-performance computing environment. The results of spatial lag models using this 25% random sample of parcels (N = 53,000) are shown in Appendix Table A3. The spatial lag results indicate modest differences from the OLS models, none of which would change the major findings from the analysis.

5 There is worse AC availability in public housing, due to factors described in the Discussion section, relative to other subsidized housing types. Data provided to us by the Santa Clara County Housing Authority indicated that of the agency’s 2,193 units in San José, 32% had central AC, 44% had individual (window) AC, and 24% had no AC.

6 In a separate set of models, shown in Appendix Table A1, we substituted the neighborhood-level independent variables with Gabbe and Pierce’s (2020) adaptive capacity and sensitivity index (ACSI), which includes 19 variables and had previously been used to analyze housing-related heat risk in California. There were only slight differences between the results shown in Table 3 and these models, and the ACSI index was not significantly associated with the heat risk variables across models.

7 This finding was different from much of the literature, and we use scatterplots to illustrate broadly negative tract-level relationships between income and heat risk factors in Figure A1.

Additional information

Notes on contributors

C. J. Gabbe

C. J. Gabbe is an Associate Professor in the Department of Environmental Studies and Sciences at Santa Clara University. His research focuses on land use planning, housing policy, climate adaptation, and environmental justice.

Evan Mallen

Evan Mallen is a Senior Analyst with the Georgia Institute of Technology Urban Climate Lab. His research focuses on urban environmental planning and design, climate adaptation planning, and urban heat risk assessments.

Alexander Varni

Alexander Varni is a graduate of Santa Clara University, with majors in Environmental Studies and Communication. His research interests include climate adaptation, environmental communication, and environmental justice.