The geography of rental housing discrimination, segregation, and social exclusion: New evidence from Sydney

ABSTRACT

We investigate whether rental housing discrimination directed against 2 predominant ethnic minority groups in Sydney, Australia, is more likely to occur in neighborhoods with a particular mix of ethnicities, socioeconomic profiles, or quality of social goods and whether this geographic pattern reinforces spatial disadvantages of these minorities in a way that abets their social exclusion. We construct measures of differential treatment based on in-person paired testing conducted in 2013, with Anglo, Indian, and Muslim Middle Eastern testers. We summarize 4 dimensions of postcode-level social goods using a principal component analysis reflecting school quality, crime rates, resident employment rates, proximate jobs and job growth, and commuting options. Our ordinary least squares (OLS) regressions show that differential treatment in the Sydney rental market is strongly related to a neighborhood’s ethnic composition and 2 aspects of its social goods involving both desirable and undesirable components but is not related to the socioeconomic characteristics of the neighborhood’s population.

Acknowledgments

The authors gratefully acknowledge the contribution of Rosalie Atie, who helped manage the project and prepare the data. We are also grateful to our colleagues Professor Kevin Dunn (Western Sydney University), Jacqueline Nelson (University of Technology Sydney), and Professor Yin Paradies (Deakin University) for their contributions to the conceptualization of this project.

Funding

This research was supported under the Australian Research Council’s Discovery Projects Funding Scheme, grant number DP110103197.

Notes

1. With one exception: Galster (1990) investigated how steering reduced minority households’ access to better quality schools.

2. A completed test for the purposes of assessing whether differential housing offers were made involved each member of the tester pair completing a phone call to the agent and an inspection of a property; for the purposes of assessing whether differential encouragement to rent occurred, a completed test further required both testers to inspect the dwelling in question.

3. The personal prejudice of an individual rental agent (expressed as an aversion to dealing with minority customers manifested as discriminatory treatment) should not systematically vary according to where the dwelling being offered is located.

4. The customer prejudice hypothesis as applied to the sales housing market suggests that Anglo real estate agents who depend upon commissions for their livelihood are less likely to show minority home seekers options in predominantly Anglo-occupied neighborhoods and/or near the agent’s office for fear of alienating potential Anglo customers who might use them as buyer or seller agents in the future.

5. In this study, which focuses on ethnic minority groups rather than immigrant groups, we use an ancestry-based definition of ethnicity described in note 10, rather than Johnston et al.’s (2001) language spoken at home and birthplace definition.

6. Critics of the audit method of estimating discrimination have argued that the method is flawed by the fundamental difficulty of accurately matching pairs of testers. Most critiques of the method have been raised in studies of employment, rather than housing, discrimination. Heckman (1998) and Heckman and Siegelmann (1993) argue that prospective employers make judgements about an individual’s suitability for a job based on underlying factors that are unlikely to be adequately controlled in the choice of tester pairs. Though controlling for underlying tester characteristics is clearly an important consideration, it is unclear that the problem is equally crucial in housing discrimination studies. Prospective employers are likely to care about a far wider array of personal characteristics of their employees than real estate agents (personality, fit with corporate culture, communication skills, team work, and leadership, among many others). Nevertheless, we designed the tester recruitment, screening, and training process to minimize this problem, as explained in the text.

7. There are 326 postcodes in the Greater Sydney metropolitan area; because we sampled by region, based on which properties were available for rent each week, we tested dwellings in 104 of these postcodes. The highest number of tests in an individual postcode was 17, for a location with a very high proportion of rental housing.

8. We tested for differences in treatment during the phone conversation stage that might be related to the order in which the testers spoke to the agent and found no significant differences. Most tests were conducted with both testers visiting the same 15- or 20-min public inspection—the order in which testers attended these inspections is unlikely to significantly affect our results, and we find no significant differences in whether the Anglo or the minority tester reported they were told that the property had been rented.

9. The disparity reflects the speed at which properties were rented—in all except 10 cases of incomplete tests, inspections were not completed because the property was no longer available. Overall, Anglo–Muslim Middle Eastern test pairs completed 165 phone enquiries, of which 107 inspections were completed; Anglo-Indian test pairs completed 408 phone enquiries, of which 261 inspections were completed.

10. This was a likely outcome; 158 of the total 537 tests did not advance from the phone call to the inspection stage because the property had been rented before the advertised inspection time.

11. In supplemental analyses, we experimented with using the individual factors as dependent variables in our OLS models. These experiments broadly confirmed the conclusions provided in the article but added no additional insights. We therefore report the simpler and more straightforward results provided by our single encouragement to rent index.

12. We did not include housing vacancy rates in the final version of the model. Housing vacancy rates (a measure of dwellings available for rent or sale from a private real estate information service such as the Australian census does not report vacancy rate data) were included in previous versions of PCA analyses of neighborhood attributes. Postcode-level vacancy rates varied from 0.4 to 5.0% across the metropolitan area at the time of the study. Given that we had reservations about the provenance and appropriateness of this data, and because it behaved inconsistently in analyses, we excluded it from the final PCA.

13. We excluded percentage Anglo residents from the final model because it was correlated with percentage recent migrants.

14. The category Muslim Middle Eastern included people reporting their ancestry as Arab, Algerian, Egyptian, Iraqi, Jordanian, Kuwaiti, Lebanese, Libyan, Moroccan, Palestinian, Saudi Arabian, Syrian, Tunisian, Yemeni, Bahraini, Emirati, Omani, Qatari, Iranian, Kurdish, Turkish, or Assyrian. The category Indian included people reporting their ancestry as Anglo Indian, Bengali, Burgher, Gujarati, Indian, Malayali, Nepalese, Pakistani, Punjabi, Sikh, Sinhalese, Maldivian, Bangladeshi, Bhutanese, Fijian Indian, Kashmiri, Parsi, Sindhi, Sri Lankan, Sri Lankan Tamil, Indian Tamil, or Telugu. The Anglo category included people reporting their ancestry as Australian (not Indigenous), New Zealander, British, English, Scottish, Welsh, Channel Islander, Manx, Irish, American, or Canadian.

15. We acknowledge that these usage statistics are ambiguous indicators of the availability of transit, inasmuch as they also will be influenced by auto ownership rates among residents.

16. Because percentage Anglo residents was highly (negatively) correlated with percentage recent immigrants, we exclude percentage Anglo from the analysis in order to limit multicollinearity.

Additional information

Funding

This research was supported under the Australian Research Council’s Discovery Projects Funding Scheme, grant number DP110103197.

Notes on contributors

Heather MacDonald

Heather MacDonald earned her PhD in urban policy and planning from Rutgers University. She is Professor of planning and Head of the School of Built Environment at the University of Technology Sydney. She has published widely on housing policy and finance, spatial analysis, and urban governance.

George Galster

George Galster earned his PhD in economics from the Massachusetts Institute of Technology (MIT) and now serves as Clarence Hilberry Professor of Urban Affairs at Wayne State University. He has published 145 peer-reviewed articles, eight books, and 35 book chapters on a wide range of urban topics. The Urban Affairs Association placed him on their “Service Honor Roll” in 2014 and awarded him the “Contributions to the Field of Urban Affairs” prize in 2016.

Rae Dufty-Jones

Rae Dufty-Jones is a Senior Lecturer in geography and Urban Studies at Western Sydney University. Her research focuses on mobility, housing, and neoliberal governance. Rae is the co-editor of Housing in 21st Century Australia: People, Practices and Policies (Routledge).