HDS Methodologies

For nearly 50 years, advocacy organizations and governmental agencies have used paired testing to detect and measure housing discrimination. In a general paired testing methodology, two individuals are matched on social and economic characteristics but differ on a single focal characteristic that is being tested. The two testers will apply for housing with the same landlord (Evidence Matters 2014). Any difference in how they are treated is interpreted as differential treatment and on average is used to measure discrimination (Evidence Matters 2014). The results of paired testing have been used as evidence in housing discrimination matters because they provide “clear evidence in a form that best suits the manner that trial courts prefer to proceed to decisions; that is, a specific set of facts about well-defined interactions between specific individuals” (Turner and James 2015).  

HUD’s Office of Policy Development and Research (PD&R) has also used paired testing for research purposes to measure the incidence of housing discrimination in the United States. In research studies, paired testing requires a representative sample and consistent protocol that ensure findings from the sample can credibly be generalized to broader patterns in the housing market (Turner and James 2015). For example, PD&R’s 2012 study was designed to produce estimates of Black, Hispanic, or Asian homebuyers experiencing discrimination and estimates of Black and Hispanic persons experiencing rental discrimination for 28 major metropolitan areas in 2012 (Turner et al 2013). Other studies by PD&R have measured discrimination against persons because of disability, familial status, and source of income.  

Following the completion of the 2012 HDS study, articles in the Cityscape journal noted that using paired testing to conduct research of broader patterns of housing discrimination may underreport evidence of discrimination. Research that relies on paired testing to investigate broad patterns in housing discrimination may report a lower bound of the prevalence of discrimination (Turner and James 2015) in certain circumstances, including the following— 

  • Research that constructs samples from advertised housing units – Paired testing may be unable to detect discrimination among all types of housing providers because some housing providers may intentionally avoid or limit public advertising to exclude minorities and other potential residents (Freiberg and Squires 2015; Pitingolo and Ross 2015). 
  • Paired testing research that assesses the initial interaction with housing providers – Some housing providers may change their behavior in subsequent meetings with housing applicants they initially deem favorable, but such changes would not be revealed in the initial interaction (Freiberg and Squires 2015). 
  • Research on segments of the housing market where access is limited such as gated communities, tax credit housing, and nursing homes, among others (Freiberg and Squires 2015). 
  • Research that uses market-wide tests that may have omitted or under sampled smaller geographic areas with high housing discrimination rates (Pitingolo and Ross 2015).

Another recent challenge to measuring housing discrimination is that inquiries about rental units can now be conducted online without the need for tenants to divulge their personal characteristics. New technology permits online housing searches, including searches via smartphones, and home-seekers may also be able to gather more information about available rental and sales units without divulging their personal characteristics. This results in discrimination later in the interaction with housing providers, which is not typically measured in research using paired testing. At the same time, housing providers may be able to detect customer characteristics by screening phone messages, using Caller ID, or even employing real-time, web-based identity verification engines. Testing protocols must adapt to these technological changes, both to avoid detection and to capture differential treatment that might occur before in-person interactions.


Freiberg, Fred and Gregory D. Squires. 2015. “Changing Contexts and New Directions for the Use of Testing.” Cityscape 17 (3): 87‒101.

Pitingolo, Rob and Stephen L. Ross. 2015. “Housing Discrimination Among Available Housing Units in 2012: Do Paired-Testing Studies Understate Housing Discrimination?” Cityscape 17 (3): 61‒86.

Turner, Margery Austin, Rob Santos, Diane K. Levy, Doug Wissoker, Claudia Aranda, and Rob Pitingolo. 2013. Housing Discrimination Against Racial and Ethnic Minorities 2012Washington, DC: U.S. Department of Housing and Urban Development Office of Policy Development and Research.

Turner, Margery Austin, and Judson James. 2015. “Guest Editors’ Introduction: Discrimination as an Object of Measurement.” Cityscape 17 (3): 3‒14