Rental Housing Discrimination on the Basis of Mental Disabilities: Results of Pilot Testing

Rental Housing Discrimination on

the Basis of Mental Disabilities:

Results of Pilot Testing

Study of Rental Housing

Discrimination on the Basis

of Mental Disabilities:

Final Report

U.S. Department of Housing and Urban Development | Ofﬁce of Policy Development and Research

RENTAL HOUSING

DISCRIMINATION ON THE BASIS

OF MENTAL DISABILITIES:

RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING

DISCRIMINATION ON THE BASIS

OF MENTAL DISABILITIES:

FINAL REPORT

Prepared by

Joy Hammel

Janet Smith

University of Illinois at Chicago

Susan Scovill

Consultant

Ron Campbell

M. Davis and Company, Inc.

Rui Duan

HDR, Inc.

Prepared for

U.S. Department of Housing and Urban Development

Ofﬁce of Policy Development and Research

Submitted by

M. Davis and Company, Inc.

August 2017

Acknowledgments

This project could not have been completed successfully

without the commitment, dedication, and hard work

of many. We particularly thank the research team and

administrative support staff at M. Davis and Company, Inc.

Thanks are also due to Madeline Hoffman for assisting with

coordination, quality control, and guidance throughout

the testing process and to May Raad of HDR, Inc., for their

analysis contributions. Finally, this study would not have

been possible without the work of our local testing organi-

zations, Access Living in Chicago (Jamie Wichman, Kelly

Chen) and The Equal Rights Center in Washington, D.C.

(Kristen Barry and Valentine Khaminwa), and our outstand-

ing group of testers—many of whom are members of the

very disability community that is the focus of this study.

A panel of expert advisers made invaluable contributions

to the authors’ understanding of the housing experiences

and challenges faced by people with mental disability,

reasonable accommodations, and protocol design and

data analysis. The panel members (and their organiza-

tional afﬁliations at the time of the study) were Michael

Allen, Relman, Dane & Colfax, PLLC; Anne Cameron

Caldwell, Research and Innovations Ofﬁcer and Director

of the Autism NOW Resource Center, The ARC; Arlene

Kanter, Co-Director of the Syracuse University Center on

Human Policy, Law, and Disability Studies; Marti Knisely,

Community Support Initiative, the Technical Assistance

Collaborative; Katherine McDonald, Associate Professor,

Syracuse University; Tia Nelis, Self-Advocacy Specialist at

the Rehabilitation Research and Training Center, Institute

on Disability and Human Development, University of

Illinois at Chicago; Jan Peters, Executive Vice President

and Chief Operating Ofﬁcer, Eden Housing Inc.; Susan

Ann Silverstein, Senior Attorney, Foundation Litigation,

AARP; Karen Tamley, Commissioner of the Mayor’s Ofﬁce

for People with Disabilities, City of Chicago; Victoria

Tedder, Housing Counseling and Advocacy Program, San

Francisco Independent Living Resource Center; and Sam

Tsemberis, Pathways to Housing. Many thanks are due to

Kim Savage for her assistance in managing the expert panel.

Anne Fletcher from the U.S. Department of Housing and

Urban Development, Ofﬁce of Policy Development and

Research, provided excellent guidance and oversight

throughout this research effort.

Finally, special thanks and remembrance to Bonnie Milstein,

a lifelong advocate for disability rights and social justice,

who served on the initial project team and coordinated the

expert panel.

Disclaimer

The contents of this report are the views of the contractor and do not necessarily reﬂect the views or policies of the U.S. Department

of Housing and Urban Development or the U.S. government.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

iii

Contents

Executive Summary ................................................................ vi

Introduction ....................................................................vi

Summary of Findings ..................................................... vii

Recommendations ......................................................... viii

Study Overview ...................................................................... 1

Purpose and Goals of the Study ...................................... 1

Background and Previous Research on Housing

Discrimination Against People With MD .................. 1

Overview of Testing Focused on Individuals With MD .... 2

Paired Testing Protocols and Field Management ..................... 4

Overview of Testing for Housing Discrimination ............. 4

Strengths and Limitations of Paired Testing in Different

Testing Methods....................................................... 4

Overview of the Paired Testing Design and Process ......... 5

Pretesting ........................................................................ 5

E-mail Testing Protocols .................................................. 6

Telephone Testing Protocols ............................................ 9

In-Person Testing Protocols ........................................... 12

Overall Data Collection Oversight, Management, and

Quality Control...................................................... 16

Sampling Methods ................................................................ 17

Sampling Frame ............................................................ 17

Allocation of Tests Across Market Size ........................... 17

Sample of Metropolitan Areas........................................ 18

Selection of Advertised Units ........................................ 20

Risk of Detection ........................................................... 21

Sampling of Units ......................................................... 21

Analysis of Testing Findings ................................................. 23

Summary of Tests Conducted ........................................ 23

Measuring Differential Treatment .................................. 24

Data Analysis ................................................................ 27

Measuring Discrimination ............................................. 27

Findings of Incidence of Discrimination ............................... 29

Data Reliability and Limitations .................................... 29

E-mail Testing Findings ................................................. 29

Reasonable Accommodation Requests via E-mail ............ 31

Telephone Testing Findings ........................................... 32

Reasonable Accommodation Requests via Telephone ...... 34

In-Person Testing Findings ............................................ 36

Findings From the Evaluation of Different Testing Modes ..... 38

People With Disabilities As Testers: Is It Feasible? ......... 38

Tester Proﬁle: How Do You Isolate and Reveal Disability? .. 38

E-mail Testing: Strengths and Weaknesses ..................... 40

Telephone Testing: Strengths and Weaknesses ............... 41

In-Person Testing: Strengths and Weaknesses As

Evaluated by Testers With MD and Companion

Testers ................................................................... 42

Summary, Conclusions, and Recommendations .................... 47

Project Leadership and Design ...................................... 49

Recruitment and Selection of Testers ............................. 49

Training and Data Collection ......................................... 49

Appendix A: Qualitative Analyses and Findings on Overt

and Subtle Discrimination Against People With Mental

Disabilities by Different Testing Modes ................................. 50

Qualitative Analysis Methodology ................................. 50

Data Analysis ................................................................ 51

Results .......................................................................... 53

Discussion and Conclusions .......................................... 61

Appendix B: References ........................................................ 63

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

List of Tables

Table 1. Testing Timeline .................................................. 5

Table 2. Subject Lines Assigned Randomly for All

E-mail Tests ......................................................... 6

Table 3. E-mail Tester Scripts ............................................ 7

Table 4. Reasonable Accommodation Request by Type

of Disability ......................................................... 8

Table 5. Telephone Tester Scripts .................................... 10

Table 6. Examples of How MD May Affect Functional

Aspects of Life and Maintenance of Housing ..... 14

Table 7. Sampling Frame for People With MD, by Test

Type .................................................................. 17

Table 8. Rental Market Sampling Frame .......................... 17

Table 9. Allocation of Tests Across Market Size ............... 17

Table 10. Distribution of Population With Disabilities

and Proportion With Cognitive Disabilities,

2009–2011 ........................................................ 18

Table 11. Pool of Possible Markets (MSAs) Based on

Statistical Criteria for Inclusion ......................... 19

Table 12. Selected Markets by Size and Olmstead Status .... 20

Table 13. Sample Allocation Plan Across Communities

and Test Types by Market Size ........................... 20

Table 14. Criteria for Inclusion in/Exclusion From Unit

Sample .............................................................. 21

Table 15. Total Tests Conducted by Test Mode .................. 23

Table 16. E-mail Sample Allocation by Market Size .......... 23

Table 17. Telephone Testing Allocation by Market Size ..... 24

Table 18. In-Person Sample Allocation by Location .......... 24

Table 19. Differential Treatment Indicator Matrix .............. 25

Table 20. Possible Outcomes Matrix ................................. 27

Table 21. Differential Treatment in Response Received

for Renters With MD (I/DD+MI) Found in

E-mail Testing ................................................... 29

Table 22. Summary of Differential Treatment of Renters

With MD (I/DD+MI) Found in E-mail Testing ... 30

Table 23. Differential Treatment in Response Received

for Renters With MI Found in E-mail Testing .... 31

Table 24. Summary of Differential Treatment of Renters

With MI Found in E-mail Testing ...................... 31

Table 25. Summary of Reasonable Accommodation

Requests by Testers With MD Found in E-mail

Testing ............................................................... 32

Table 26. Summary of Differential Treatment of Renters

With MD Found in Telephone Testing ............... 32

Table 27. Summary of Differential Treatment of Renters

With MI Found in Telephone Testing ................ 33

Table 28. Summary of Differential Treatment of Renters

With I/DD Found in Telephone Testing ............. 33

Table 29. Summary of Reasonable Accommodation

Requests by Testers With MD Found in

Telephone Testing .............................................. 34

Table 30. Summary of Reasonable Accommodation

Request Responses by Speciﬁc Markets ............. 35

Table 31. Summary of Differential Treatment of Renters

With MD Found in In-Person Testing ................ 36

Table 32. Summary of Differential Treatment of Renters

With MI Found in In-Person Testing ................. 36

Table 33. Summary of Differential Treatment of Renters

With I/DD Found in In-Person Testing .............. 37

Table 34. Pros and Cons of E-mail Testing ........................ 41

Table 35. Pros and Cons of Telephone Testing .................. 42

Table 36. Pros and Cons of In-Person Testing ................... 46

Table A-1. Samples of Telephone and In-Person Testing ..... 51

Table A-2. Subtle Discrimination Indicator Keyword

Groups .............................................................. 52

Table A-3. Response Difference in E-mail Content .............. 53

Table A-4. Subtle Discrimination Responses in Multiple

Categories ......................................................... 53

Table A-5. Subtle Discrimination in E-mail Testing ............. 55

Table A-6. Housing Provider Responses to RA Requests in

E-mail Testing ................................................... 55

Table A-7. Open Door/Closed Door in E-mail Testing ........ 55

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Table A-8. Subtle Discrimination in Telephone Testing ....... 57

Table A-9. Overt Discrimination in Telephone Testing ........ 58

Table A-10. Housing Provider Responses to RA Requests

in Telephone Testing .......................................... 58

Table A-11. Open Door/Closed Door in Telephone Testing... 58

Table A-12. Subtle Discrimination in In-Person Testing ........ 60

Table A-13. Overt Discrimination in In-Person Testing ......... 60

Table A-14. Open Door/Closed Door in In-Person Testing ... 61

List of Figures

Figure 1. Flow of Sample Selection and Testing Process ... 22

Figure 2. Response Rate of E-mail Paired Tests ................. 23

Figure 3. Response Rates of E-mail Paired Tests for

Pro tected Testers Identifying as a Person With

MI .................................................................... 30

Figure A-1. E-mail Testing Responses ................................... 50

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Executive Summary

Introduction

More than 15 million people in the United States have some

type of mental disability (MD). Many of these individuals seek

community-based housing in the rental market. As a result of the

U.S. Supreme Court’s decision in Olmstead v. L.C., 527 U.S. 581

(1999), an increasing number of individuals with disabilities

are moving from nursing homes and other institutional and

segregated settings into community-based settings. As a result,

equality of access to the rental housing market has become

even more of a critical policy issue.

This study represents the ﬁrst comprehensive examination of

discrimination in the rental housing market against people

with MD, speciﬁcally focusing on people with mental illness

(MI) and those with intellectual or developmental disability (I/

DD). The goal of the study is to increase the understanding of

the forms and prevalence of housing discrimination against this

community as its members seek market-rate rental housing.

The overall study design included a stepwise approach to

testing the feasibility of involving individuals with MI and I/DD

in discrimination testing and to obtaining data regarding their

treatment by rental housing providers. The study was initiated

with a comprehensive review of relevant literature and the

convening of multiple focus groups that included individuals

with MI and I/DD, disability and fair housing organizations,

and policymakers. As very little discrimination research to date

has actually included these target populations as participants in

the process, the focus groups were particularly informative in

identifying speciﬁc issues faced by individuals with MI and I/

DD throughout the process of seeking rental housing. Barriers

identiﬁed by the focus group process included—

• Negative and stigmatizing reactions and attitudes by

housing providers and their agents.

• Improper requests to disclose personal and disability-relat-

ed information.

• A lack of understanding about what types of reasonable

accommodations may be needed by individuals with MD.

• The denial of proper requests for reasonable accommodations.

• Steering toward speciﬁc housing, buildings, or units based

on an individual’s disability.

From focus group input, the authors learned that individuals

with MI and I/DD experience potential discrimination at

different points of time in the rental process: when initially

attempting to gain information about available units, during the

application process, and even after moving in and during their

tenure in their home.

After the literature review and focus group phases were con-

ducted, protocols for conducting discrimination testing were

designed. A matched-pair testing design was created to examine

differences in treatment by housing providers between indi-

viduals with MI and I/DD (protected testers) and individuals

without such disabilities (control testers). Paired testers were

matched on key characteristics other than disability, such as

income, race, gender, and age. The testing sought to examine

differences in testers’ treatment across three modes of contact

with housing providers: e-mail, telephone, and in person. These

modes generally represent the primary ways people, including

individuals with MI and I/DD, might seek rental housing. In

two of the modes, e-mail and telephone testing, a request for

a reasonable accommodation was also made by the protected

tester to allow for the examination of housing providers’

responses.

An initial set of exploratory tests was administered to examine

and reﬁne testing protocols before broad-scale testing. After the

exploratory testing, broad-scale pilot testing was conducted in

multiple urban rental markets across the United States. Testing

in each mode was divided equally between the two subcatego-

ries of MD (MI and I/DD), and more than 1,000 matched-pair

tests were administered—359 by e-mail in nine rental markets,

668 by telephone in nine rental markets, and 101 in person in

two large rental markets.

This study represents the ﬁrst time that individuals with MI

and I/DD have been enlisted to serve as in-person testers in

a comprehensive, multicity housing discrimination study.

As a result, this study also contributes signiﬁcantly to our

knowledge about research methodologies that are accessible,

inclusive, and participatory to individuals with MI and I/DD.

Based on feedback from individuals with MI and I/DD, a testing

methodology was developed for this study in which individuals

without MI and I/DD were trained and accompanied protected

testers as companions during in-person tests. These companion

testers served three roles: ﬁrst, to act as a friend and natural

support to the protected tester during the testing; second, to

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

vii

further disclose the MD of the protected tester to the housing

provider; and third, to assist in preparing detailed ﬁeld notes

to document their testing experiences.

Summary of Findings

The testing conducted in this study documented signiﬁcant

levels of adverse differential treatment toward individuals with

MI and I/DD when compared with individuals who did not

have MD. Pilot testing revealed that individuals with MI and I/

DD were—

• Less likely to receive a response to their inquiry (17.55 percent

of people without disabilities received a response com-

pared with 9.19 percent of people with MI and I/DD in

e-mail testing).

• Less likely to be told an advertised unit was available (5.94

percent of people without disabilities were told that the

advertised unit was available compared with 0.99 percent

of people with MI and I/DD in in-person testing).

• Less likely to be invited to contact the housing provider (7.69

percent of people without disabilities were invited to

contact the housing provider to see the unit compared

with 0.00 percent of people with MI and I/DD in e-mail

testing).

• Less likely to be invited to inspect the available unit (21.26

percent of people without disabilities were invited to

inspect the unit compared with 16.47 percent of people

with MI and I/DD in telephone testing).

• More likely to be encouraged to look at a different unit than the

one advertised in telephone testing, a potential indicator

of steering people with MI and I/DD toward speciﬁc

buildings or areas within rental complexes, resulting in

segregated living patterns.

• Treated adversely at disparate rates depending on disability

type, with higher rates of adverse treatment found for

individuals with MI than for individuals with I/DD. This

ﬁnding may indicate that individuals with MI face more

negative stereotypes and stigma from rental housing

providers.

In both e-mail and telephone testing, a signiﬁcant percentage

of individuals with MI and I/DD also experienced adverse treat-

ment with respect to a request for a reasonable accommodation.

The reasonable accommodations requested were identiﬁed as

speciﬁc to the protected tester’s functional needs, such as the

use of an assistance animal to manage emotional and cognitive

issues or the need for a verbal reminder from the landlord to

pay the rent on time.

• The willingness of a housing provider to grant a request for an

accommodation varied by mode of testing, with the rate of

granting a request for a reasonable accommodation being

signiﬁcantly higher when the request was made by telephone

(59.14 percent willing to accommodate; 40.86 percent not

willing to accommodate) than by e-mail (15.38 percent

willing to accommodate; 84.62 percent not willing to

accommodate).

• Regardless of testing mode, a signiﬁcant percentage of people

with MI and I/DD were given a negative response to their

reasonable accommodation request, ranging from outright

denials to more subtle barriers, such as an indication that

the individual with MD would be responsible for actively

seeking out and appealing the denial of his or her request

with higher-level managers whose name and contact

information were then not provided.

The research also found that many housing providers simply

did not respond to housing inquiries but that the probability

of a response increased based on the interpersonal nature of

the interaction. For example, in-person inquiries received the

highest rates of response, followed by telephone inquiries, with

e-mail inquiries the most likely to be ignored.

Each of the three testing modes used (e-mail, telephone, and

in person) yielded a different form of testing data that, when

viewed together, not only produced evidence of discrimination

but also showed promise for bringing testing to a national scale

and provided insights on opportunities for future testing.

Finally, this study yielded signiﬁcant ﬁndings on how to con-

duct testing with testers with MD. In-person testing speciﬁcally

involving a person with MD demonstrated that people with MI

and I/DD can serve as effective testers.

The evidence from the

In-person testing protocols did not include a request by the protected tester for a reasonable accommodation.

In addition, testing results varied by metropolitan statistical area (MSA), with the rate of granting a reasonable accommodation request being lowest in the Chicago-

Joliet-Naperville, IL-IN-WI MSA (38.6 percent), and highest in the Albuquerque, NM MSA (75.5 percent).

Contributing to these outcomes were training and support provided to the testers, the use of a companion tester, and the overall coordination of the testing by

community-based disability rights organizations.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

viii

in-person testing suggests strongly that both populations could

also do telephone testing if the same approach to training and

debrieﬁng is used.

Recommendations

These ﬁndings led the authors to make the following recom-

mendations.

• A broad-based education initiative should be created to

educate housing providers, including owners and their

agents, about fair housing rights and obligations, includ-

ing appropriate policies and practices when dealing with

individuals with MI, I/DD, and other MD.

• Public and private housing, disability, and civil rights

organizations should redouble efforts to engage and educate

the community of people with MI and I/DD about their rights

under federal, state, and local antidiscrimination laws, how

to recognize potential discrimination, and what actions to

take when faced with discrimination.

• Additional research is necessary to better understand the

scope and severity of the discriminatory barriers encountered

by individuals with MI and I/DD as they seek, obtain, and

retain accessible rental housing. Findings from this pilot

study effort suggest that a national testing effort would be

feasible.

• Individuals with MI and I/DD can and should be an integral

part of future housing research and testing, housing dis-

crimination education initiatives, and efforts to strengthen

housing policy to promote equal housing opportunity.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Study Overview

Purpose and Goals of the Study

The existing research related to housing discrimination and its

effect on individuals with mental disability (MD)

is particular-

ly limited; therefore, HUD launched this study with the goal of

increasing the understanding of the experiences of individuals

with MD as they seek rental housing.

The following research objectives identiﬁed by HUD guided

this study.

• Develop and implement different testing methods and

approaches, evaluating their strengths and weaknesses, to

establish a series of protocols for conducting paired testing

to determine and measure the degree to which people

with MD experience discrimination in the search for rental

housing.

• Conduct pilot testing using different testing procedures to

develop protocols for generating reliable estimates of any

difference in the treatment of people with MD during the

process of searching for rental housing.

• Produce a qualitative and quantitative analysis of Title VIII

Automated Paperless Ofﬁce Tracking System (TEAPOTS)

data.

• Produce ﬁve short papers that expand the understanding

in the ﬁeld of housing discrimination on the basis of MD.

The purpose of this study was not to generate national

estimates of housing discrimination on the basis of MD, but

rather to pilot methods for conducting paired testing with this

protected class with the goal of demonstrating a potentially

feasible approach to future nationwide testing.

Additional short papers have been developed to inform

researchers, practitioners, and consumers about the shape

and scope of housing discrimination against people with MD.

These papers focus on the following related topics.

• Short Paper 1: Systematic Literature Review of Research

on Discrimination in Rental Housing on the Basis of Mental

Disabilities.

• Short Paper 2: Getting Into and Maintaining Housing in the

Private Rental Market: Experiences of People With Mental

Disabilities.

• Short Paper 3: The Potential for Discrimination in Rental

Housing for People With Mental Disabilities Moving Out of

Institutions to Community Living Under Olmstead.

• Short Paper 4: Accessible & Participatory Methods for

Involving People With Mental Disabilities in Housing Discrimi-

nation Testing.

• Short Paper 5: Rental Housing Access & Discrimination

Experienced by People With Multiple Disabilities.

Background and Previous Research

on Housing Discrimination Against

People With MD

The federal Fair Housing Act (FHA)

prohibits discrimination

in the sale, rental, and ﬁnancing of housing and in other

housing-related transactions, on the basis of seven protected

classes: race, color, religion, national origin, sex, familial status,

and disability. Housing providers are prohibited from consid-

ering these protected characteristics as the basis for rejecting

or refusing to negotiate with individuals seeking housing or

housing-related services and from misrepresenting or limiting

housing opportunities based on any protected characteristic.

In addition to having the basic prohibitions of the FHA against

housing discrimination for all protected classes, people with

disabilities have three additional protections under the FHA:

(1) multifamily housing with four or more units, built for

ﬁrst occupancy after March 13, 1991, must meet speciﬁc, if

Mental disability is deﬁned as “(1) having a mental or psychological disorder or condition that limits a major life activity, including working; (2) any other mental

or psychological disorder or condition that requires special education or related services; (3) having a record or history of a mental or psychological disorder or

condition which is known to the employer or other entity covered by this part; or (4) being regarded or treated by the employer as having, or having had, any mental

condition that makes achievement of a major life activity difﬁcult” (Foster v. City of Oakland, 2009 U.S. Dist. LEXIS 70094).

TEAPOTS provides data records regarding ﬁled discrimination complaints. The analysis of TEAPOTS data on housing complaints based on MD is documented in a

separate report.

These papers have been published by HUD separately.

Title VIII of the Civil Rights Act of 1968 as amended in 1988, 24 U.S.C. §§ 3601, et seq.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

relatively modest, accessibility requirements that allow for a

person with disabilities to access and use both the housing

units and associated public use and common use areas; (2)

housing providers must make reasonable accommodations

to their rules, policies, practices, and services necessary for

people with disabilities to equally enjoy the property; and

(3) housing providers must allow residents with disabilities,

at the residents’ expense, to make reasonable modiﬁcations

to physical structures necessary in order for them to use and

enjoy the property.

In addition, in 1999, the United States Supreme Court ruled in

Olmstead v. L.C., 527 U.S. 581, that unjustiﬁed segregation of

people with disabilities constitutes discrimination in violation

of Title II of the Americans with Disabilities Act.

Although the

beneﬁciaries of this ruling are all people with disabilities, the

decision has had a particularly signiﬁcant effect on individuals

with MD living in restrictive and segregated settings, including

nursing homes and institutions. As states seek to comply with

the Olmstead ruling, people with MD are moving out of insti-

tutional settings in larger numbers and increasingly seeking

housing in the private rental market.

Individuals with MD often face multiple challenges when they

seek housing in the rental housing market. Challenges may

include both economic barriers and stigma or suspicion on

the part of housing providers that limits their access to diverse

housing choices. Complaints based on disability make up the

largest number of housing discrimination complaints ﬁled with

federal, state, and local fair housing agencies and with private

fair housing groups (HUD, 2014a). In 2014, disability com-

plaints made up 51.84 percent, or 14,271 of the total number

of 27,528 fair housing complaints ﬁled with HUD, its partner

state and local agencies, and private fair housing enforcement

organizations (NFHA, 2015).

A signiﬁcant portion of these

complaints involves people with MD. For example, HUD (n.d.,

Table 1) noted that, in ﬁscal year 2010, fully 40 percent of the

disability complaints involved MD, which includes people with

psychiatric disability or mental illness (PD/MI) and intellectual

or developmental disability (I/DD).

In the United States, the term “mental disability” includes

a wide range of conditions. According to the U.S. Census

Bureau’s Survey of Income and Program Participation, about

6.3 percent of the U.S. population in 2010, approximately 15.2

million people age 15 or older, had difﬁculty with cognitive,

mental, or emotional functioning (Brault, 2012). This number

includes people with both PD and I/DD. Using a broader deﬁni-

tion and different methodology, however, the Substance Abuse

and Mental Health Services Administration (SAMHSA) estimates

a much larger number of people who have experienced some

form of PD in a year: 44.5 million adults age 18 or older.

This

total comprises nearly 20 percent of the adult U.S. population.

Of this total, about 10.4 million adults (4.6 percent) of U.S.

population have severe MI (APA, 2000). Braddock et al. (2011)

used a ﬁgure of 4.8 million for the population with I/DD.

Little is known about the forms of discrimination people with

MD experience when searching for housing. According to

NCD (2001), people with any disability are likely to encounter

discrimination in three ways: (1) the housing is inaccessible, (2)

the landlord assumes that renters with disabilities are unable to

live independently, and/or (3) the housing staff are unwilling to

modify rules and policies that exclude people with disabilities.

For people with MD, access to housing also may be affected by

histories of homelessness or incarceration, both of which can be

potential sources of additional discrimination.

Overview of Testing Focused on

Individuals With MD

Although fair housing testing related to the barriers faced by the

disability community has been conducted in some metropolitan

areas, research on housing discrimination against people

with MD is particularly limited. Before this study, the most

signiﬁcant examination of the level and nature of this type of

disability discrimination was published in a 2005 HUD Disabil-

ity Discrimination Study report (DDS 2005; HUD, 2005).

DDS 2005 explored the development of methodologies in

testing for housing discrimination on the basis of MD using a

paired testing methodology previously used in testing for hous-

ing discrimination on the basis of race and ethnicity. DDS 2005

employed a handful of in-person tests by individuals with MD

and telephone tests by testers without disabilities purportedly

calling on behalf of individuals with MD.

The Americans with Disabilities Act of 1990, 42 U.S.C. §§ 12101, et seq.

This percentage is an increase from 49.51 percent of all complaints submitted in 2013.

SAMHSA deﬁnes mental illness as the presence of mental, behavioral, or emotional disorder based on the diagnostic criteria in APA (2000).

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Key ﬁndings of DDS 2005 exploratory testing included—

• It is feasible to conduct in-person testing, and people with

MD can effectively serve as testers.

• It is challenging to recruit people with MD who can

handle the role of a tester, and some individuals may need

extra support and assistance.

• Tests in which the person with MD was accompanied by

a “friend” without MD proved to be credible and effective

but signiﬁcantly raised the cost of testing.

• Testers with MD may be more effective if they conduct

tests relatively infrequently, and it would be preferable to

extend the testing over a relatively long period of time,

unless a large pool of testers with MD could be recruited.

The pilot testing documented in this report sought to build on

the key ﬁndings observed in DDS 2005.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Paired Testing Protocols and Field Management

Overview of Testing for Housing

Discrimination

Testing is an investigative technique that serves as a powerful

tool for directly observing differences in treatment in the

practices of housing providers and their agents. Testers pose

as individuals seeking housing, contacting housing providers

in a variety of ways to inquire about housing opportunities.

The origins of paired testing as a method of studying housing

discrimination and identifying differential treatment lie in fair

housing enforcement, in which testing is often used to identify

individual housing providers who are violating housing

discrimination laws.

When paired testing is applied to large,

representative samples and implemented with rigorous con-

trols using a quasi-experimental research design, it can provide

reliable estimates of the differences in treatment among

different populations and of the prevalence of discriminatory

behavior toward different protected classes. Paired testing also

can be used to identify discrimination by a speciﬁc provider

that violates the FHA.

HUD has used testing as an investigative and research tool

for more than 40 years to identify and document discrimi-

natory housing practices.

HUD has funded a wide variety

of testing-based initiatives, including four national studies

using the methodology of matched-pair testing to measure the

levels of housing discrimination based on race and ethnicity,

and HUD has also funded national studies on discrimination

against same-sex couples, families with children, people who

are deaf, and people who use wheelchairs.

For people with

disabilities, testing can be used to identify instances of direct

discrimination, to obtain evidence of unequal treatment, and

to determine whether reasonable modiﬁcations or accommoda-

tions are permitted to enable a person with a disability to fully

enjoy living in the housing.

Strengths and Limitations of Paired

Testing in Different Testing Methods

Testing can be conducted in a variety of conﬁgurations,

depending on the issues being examined, but the most common

and effective approach to observing differences in treatment

based on a particular characteristic is matched-pair testing.

In this type of testing, two testers are assigned proﬁles that

make them similarly situated and qualiﬁed for the housing

being tested, differing only in their membership in a protected

class. The testers contact a housing provider and inquire

about available housing. Matched-pair testing allows for the

observation of housing providers’ actions and statements and

measures whether and how people are treated differently

during the housing search process. The tester exhibiting the

characteristic being tested (for example, a disability) is referred

to as the protected tester and the tester without the characteristic

in the pair is the control tester. The key to testing is to ensure

pairs are well matched on all variables except the one variable

or characteristic to be tested—this procedure ensures that

test results are unambiguous. The matched characteristics of

the tester pairs in this study included race, gender, age range,

education, employment, household size, and income. The

single variable that differed between matched testers was the

existence or absence of a particular MD.

To minimize the risk

of detection by housing providers during testing, some of the

characteristics assigned to a test pair were slightly enhanced to

favor the protected tester.

HUD has noted that: “Fair housing advocates have turned to testing as the most effective tool to investigate violations of fair housing law” (HUD, 2014b).

Testing cannot capture all forms of housing discrimination. It is most effective at the early stages of the search for housing, including contacting providers to

inquire about housing opportunities and visiting properties to view units and obtain information about the rental (or sales, lending, and insurance) process and

associated costs. Testers generally have ﬁctitious proﬁles, and testing rarely proceeds through the actual application stage. Testing also does not effectively capture

issues found later in the housing process, such as decisions made based on submitted applications or how residents in a protected class are treated after they move

in, such as maintenance or lease termination.

See HUD (2013a) and HUD (2015).

Because of the need to isolate the single variable of MD (either MI or I/DD) and the need to ensure that the proﬁles made the testers ﬁnancially qualiﬁed for the

housing, the testing protocols could not reﬂect the real life experience of many people with MD searching for rental housing, whose lives and characteristics are

much more complex than the rigors of research testing would allow.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Overview of the Paired Testing Design

and Process

This study included testing components in nine selected rental

markets using matched-pair testing via e-mail, telephone, and

in-person visits.

The overall goals were to (1) enhance the

understanding of the barriers faced by individuals with MD

in their search for rental housing, (2) determine the type and

nature of discrimination that might be encountered, (3) eval-

uate the utility of different approaches to paired testing when

conducting research on housing discrimination on the basis

of MD, and (4) assess the feasibility of employing individuals

with MD as testers. The testing in this study was not designed

to produce a national estimate of rental discrimination, but

rather was designed to develop and evaluate the utility of

different testing approaches and techniques for measuring

housing discrimination among this particular protected class.

As shown in Table 1, the testing for this study took place over

a 1 1/2-year period in 2013 and 2014. The timeframes for

each mode of testing were staggered, with each new testing

mode beginning with pretesting to conﬁrm the effectiveness

of that mode and to assess the need for any methodological

adjustments. The testing protocols in this study built on the

protocols established in previous studies; the e-mail testing

protocols mirrored those used in a national e-mail testing

study of same-sex couples, and the telephone testing and

in-person testing protocols mirrored those established in the

DDS 2005 study, which conducted a small number of explor-

atory telephone and in-person tests based on MD employing

some individuals with MD as protected testers.

The e-mail and telephone testing in this study was conducted

from a central location in Philadelphia, Pennsylvania. The

in-person testing was performed by local testing organizations

in two markets: Access Living in the Chicago-Joliet-Naperville,

IL-IN-WI Metropolitan Statistical Area (MSA; hereafter, Chi-

cago) and The Equal Rights Center in the Washington-Arling-

ton-Alexandria, DC-VA-MD-WV MSA (hereafter, Washington,

D.C.). These testing organizations recruited, trained, and

managed the in-person testing in the ﬁeld with support and

supervision from the study team.

Pretesting

Before beginning the pilot testing, pretests were conducted for

each testing mode (e-mail, telephone, and in-person testing).

The purpose of the pretesting was to ensure that (1) data

collection protocols and strategies were designed correctly;

(2) testers were trained to execute the tests appropriately; (3)

administrative procedures were effective; (4) the data entry,

coding, and interpretation process accurately captured the

experiences of testers; and (5) feedback was produced quickly,

without detection, so as to permit adjustments in the testing

protocols.

Table 1. Testing Timeline

2013 2014

3rd Q 4th Q 1st Q 2nd Q 3rd Q 4th Q

Testing

E-mail

Pretest

Pilot testing

Washington, D.C.

/Chicago

Remaining markets/RA request

Telephone

Pretest

Pilot testing Washington, D.C.

/Chicago

Pilot testing remaining markets

In person

Pretest

Pilot testing

RA = reasonable accommodation.

Washington-Arlington-Alexandria, DC-VA-MD-WV Metropolitan Statistical Area.

Chicago-Joliet-Naperville, IL-IN-WI Metropolitan Statistical Area.

The rental markets tested were in the following metropolitan statistical areas (MSAs): the Albuquerque, NM; Chicago-Joliet-Naperville, IL-IN-WI; Cincinnati-

Middletown, OH-KY-IN; Fresno, CA; Harrisburg-Carlisle, PA; Nashville-Davidson--Murfreesboro--Franklin, TN; New Haven-Milford, CT; Syracuse, NY; and

Washington-Arlington-Alexandria, DC-VA-MD-WV MSAs.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Focus group input and a review of relevant literature provided

information about how people with MD engage in the housing

search process and identiﬁed unique factors affecting their

housing search. Information from national and local disability

groups in Washington, D.C., and Chicago helped to identify

the different sources used to locate housing units available for

rent. These organizations also assisted in providing testers with

consistent, realistic, and believable identities and backgrounds.

In‐person testers with MD portrayed their own disability and

additional characteristics consistent with the tester proﬁle

provided by the study team.

Testing protocols were designed so as to create identities that

clearly disclosed the protected tester’s MD by incorporating

supporting information such as previous hospitalization or

institutionalization and afﬁliation with relevant community

support agencies to ensure that the presence of MD was

unambiguously communicated to the housing provider. This

background information was also used to provide testers with

ready answers to the typical types of questions asked by rental

agents, such as, “Why are you looking at this property?” and

“Where do you work?”

E-mail Testing Protocols

The e-mail testing did not use individuals serving as testers

in the traditional sense. Replicating the protocols of HUD

(2013a), the de facto “tester” was the text of each paired e-mail

and not the individual sending the e-mails. Test administrators

sent out all e-mails inquiring about advertised rental units and

then reviewed and coded the responses. The text of the paired

e-mails contained similar language asking about the housing

unit for rent. The only difference between the two e-mails was

that one e-mail stated the person making the inquiry had MD

(either MI or I/DD). A total of 359 paired e-mail tests were ad-

ministered, including 78 paired e-mail tests that incorporated

a request for a reasonable accommodation in the e-mail of the

protected tester.

E-mail Testing Script

A beneﬁt of using e-mail for testing is that the message is

completely controlled and consistent for the protected tester

and the control tester in the pair. Moreover, an e-mail response

from a housing provider gives verbatim statements to code and

analyze in relation to the initial e-mail inquiry. For each e-mail

test, the subject line (see Table 2) and the message were short

and relatively neutral in tone, with standard information being

requested.

Table 2. Subject Lines Assigned Randomly for All

E-mail Tests

1. One bedroom apartment

2. Apartment for rent

3. Unit for rent

4. Craigslist - 1br apartment

5. Craigslist Rental

6. Unit availability

7. Apartment posted on craigslist

8. Craigslist apartment posting

9. Inquiry about your apartment

10. Saw your Craigslist Apartment ad

11. In response to your Craigslist listing

A variety of e-mail “scripts” were developed to serve as the

body of the message (see Table 3). In the body of the text, the

protected tester discloses disability early in the text in one of

the following direct manners.

MI—“I have a mental health issue.”

I/DD—“I have a developmental disability.”

Control—none.

The e-mail text also contained a secondary reveal by the

protected tester, with a parallel addition by control testers,

referring to their previous residence.

MI—psychiatric hospital (“I have been living in a psychiatric

center for the past year”).

I/DD—group home (“I have been living in a group home for

the past year”).

Control—dorm, parents, divorce (“I have been living with my

parents, or just ﬁnished school and was living in a dorm, or just

got divorced”).

The “dual disclosure” text samples in table 3 include reveal of

the disability plus the alternate living scenarios, which provided

a sufﬁcient number of options and ensured that the text sam-

ples were comparable.

Including both a statement about having a disability and a

type of previous living arrangement was designed to provide

clear and unambiguous evidence to a housing provider that

the prospective tenant had MD. Disclosing the current housing

situation of the protected tester is important because it not

only signals that the person has a disability (a requirement for

that living situation), but it also signals that the applicant does

not have experience living in the private market, which can

make the individual appear to be more of a risk to the housing

provider. This same justiﬁcation applies to the disclosure of the

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Table 3. E-mail Tester Scripts

Protected Tester Control Tester

Hello. I am interested in your 1 br. Apt. on <specify address>. Is it

available? Is there a time I could come by and see it? I have a <disability

identiﬁer> and have been living in <alternate housing identiﬁer> and am

ready to get my own apartment. Thank you. <name>

Hi, my name is <name>, and I am very interested in your 1br apt

located at <specify address>. I have been living in <alternate housing

identiﬁer>. Could you let me know if it is still available and when a

convenient time to view it would be? Thank you. <name>

Hello. Your apartment (<specify location>) seems to be what I had in

mind. Is it still available? I’d like to come by and view the apartment;

could you contact me with an available time to do so? I have a <disability

identiﬁer> and I’ve been living in <alternate housing identiﬁer> but I want

to get my own apartment now. Thanks, <name>

Hello. I saw your 1 br apt ad on CL at <specify address>. My name

is <name> and I am looking for a place that matches this description.

I’ve been living in <alternate housing identiﬁer> and I’m ready to get

my own apartment. Is it available and, if so, can I see the unit? Thanks,

<name>

Hello, I am writing in response to your listing for the 1 bedroom

apartment located at <specify address>. Is it available? May I come and

see it? I’ve been living in <alternate housing identiﬁer> because I have a

<disability identiﬁer> and I’m ready to get my own apartment. Thank you

for your time, <name>.

I saw your ad for a 1Br apartment on Craigslist located at <specify

location>. Is this apartment still available? I would like to set a time to

see it. Can you tell me what hours you would be available so we can

schedule a visit? I’m looking forward to getting my own apt. I have

been living in <alternate housing identiﬁer>. I can be reached at this

return email. Thanks, <name>

I saw your ad for a 1Br apartment on Craigslist located at <specify loca-

tion>. Is this apartment still available? I would like to set a time to see it.

Can you tell me what hours you would be available so we can schedule

a visit? I’m looking forward to getting my own apt. I have been living in

<alternate housing identiﬁer> due to my <disability identiﬁer>. I can be

reached at this return email. Thanks, <name>

Hello, I am writing in response to your listing for the 1 bedroom

apartment located at <specify address>. Is it available? May I come

and see it? Let me know when would be a good time to come by. I’ve

been living in <alternate housing identiﬁer> and I’m ready to get my

own apartment. Thank you for your time, <name>.

Hello. I saw your 1 br apt ad on CL at <specify address>. My name is

<name> and I have a <disability identiﬁer>. I am looking for a place that

matches this description. I’ve been living in <alternate housing identiﬁer>

because I have a <disability identiﬁer> and I’m ready to get my own

apartment. Is it available and, if so, can I see the unit? Thanks, <name>

Hello. Your apartment (<specify location>) seems to be what I had in

mind. Is it still available? I’d like to come by and view the apartment;

could you contact me with an available time to do so? I have been

living in <alternate housing identiﬁer> but I want to get my own apart-

ment now. Thanks, <name>

Hello, I would like set up a time to see the place you advertised on

<specify the address>. I was wondering, is this apartment still available?

Also, is there any particular time that works best for a showing? I have

a <disability identiﬁer> and have been living in <alternate housing iden-

tiﬁer>. Please contact me for your most convenient time to view it. Best

regards, <name>

Hi, I am writing in response to your listing for the 1 bedroom apartment

located at <specify address>. Is it available? May I come and see it?

I’ve been living in <alternate housing identiﬁer> and I’m ready to get

my own apartment. Thank you for your time, <name>.

current housing situation of the control tester, which is neces-

sary for parity. The inclusion of phrases such as “I’m ready to

get my own apartment” and “I’m excited about my own place”

provides counterbalancing information that mirrors the proﬁle

of the protected tester and raises a similar ﬁrst apartment

concern for a housing provider.

In 78 of the paired e-mail tests, the protected tester included a

request for a reasonable accommodation. Table 4 summarizes

the different reasonable accommodation requests used in those

additional tests. A request to have an assistance animal live

at the test property was included because it is an increasingly

common request by people with MD, particularly for people

with some forms of MD, and is also becoming a common

source of fair housing complaints ﬁled.

The use of a request for a reminder of rent being due was an

accommodation recommended by the study’s expert panel. It

reﬂects a common housing-related concern for people with MD,

especially because it may help a tenant keep his or her apart-

ment. Before this study, little was known about how housing

providers would actually respond to this type of request, which

requires the housing provider to be actively engaged with the

tenant on a regular basis.

HUD uses the term “assistance animals,” which includes emotional support animals. According to HUD’s notice on Service Animals and Assistance Animals for

People with Disabilities in Housing and HUD-Funded Programs, an assistance animal “is an animal that works, provides assistance, or performs tasks for the beneﬁt

of a person with a disability, or provides emotional support that alleviates one or more identiﬁed symptoms or effects of a person’s disability” (HUD 2013b: 2).

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Table 4. Reasonable Accommodation Request by Type of Disability

MI I/DD

1. I have a support animal, a cat, which helps me cope with my chronic

pain. Will that be okay?

I do have some trouble remembering things and I’d like to know if you

can call me with a reminder that my rent is due a couple of days before

it’s due. Could you do that?

2. I have a companion dog named Bob. He makes it possible for me

to go outside. He will need a grassy area by the apartment. Is there

someplace he can use?

Could you call me with a reminder that my rent is due a few days

before? Sometimes I have trouble remembering things and staying

organized.

3. I have a therapeutic cat to help me deal with my illness. I want to

make sure that’s ok with you.

I can’t always remember things. Could you come by or call me to

remind me that my rent is due?

4. I have some memory problems. Could you call me with a reminder

that my rent is due a couple of days before it’s due?

My dog is my support animal and I want to make sure she can stay

with me in this apartment.

5. I will need a reminder that my rent is due a couple of days before.

Could you call or come by to remind me?

My dog provides emotional support for me and it’s important that I can

keep her in this apartment. Will that be okay?

6. I sometimes have trouble staying organized and remembering. Could

you call or come by to remind me that the rent is due in a couple of

days?

I have a cat as an emotional support and companion animal. I want to

be sure it’s ok with you for me to keep her.

I/DD = intellectual or developmental disability. MI = mental illness.

Training for E-mail Testing

In January 2014, project training was conducted for e-mail test

administrators. All e-mail test administrators were rigorously

trained to ensure a common understanding of the project’s

overall purpose, quality assurance responsibilities, and terms

and deﬁnitions. Training topics included an introduction to

the project, an overview of fair housing and discrimination

testing, e-mail testing procedures, and a review of testing

forms. Special emphasis was placed on the fact that test

administrators were to be objective collectors of information.

The conﬁdential nature of all study and testing information

was also emphasized with all study participants. This training

conformed to standard survey research guidelines and replicat-

ed those from HUD (2013a).

E-mail test administrators also received technical training

for e-mail generation and coding of e-mail responses. This

training was critical to ensure team members’ understanding

of speciﬁc coding guidelines and requirements and ensure

consistency and efﬁciency during the testing process. For the

coding component of the e-mail testing, the training approach

involved learning a process that could be effectively and

consistently replicated across multiple test administrators. The

same e-mail response was initially coded by the whole team.

Any differences were discussed and resolved. As this coding

was critical to the success of the project, this process continued

for several days until test administrators were uniform in their

coding of test e-mail responses.

Conducting the E-mail Testing and Quality

Control

Sampling of the advertised rental units within markets used

both automated and manual methods. An automated data-min-

ing tool (the “scraper tool”) was used to select and collect data

from advertisements on craigslist in each study market. In

addition to automated sampling, local newspapers and other

materials featuring rental housing advertisements in the selected

markets were sampled manually, with every seventh post select-

ed. All ads were ﬁltered using the preselected criteria including

size of the unit, rental amount, and e-mail address. Ads that did

not provide the necessary information, and those describing

units set aside for low-income residents, seniors, or people with

disabilities, were excluded from the sample. A more detailed

description of the process for selecting advertised units for the

overall study is provided in the Sampling Methods section.

After rental advertisements from each market were collected,

postings from each rental market were placed in a database,

and matched-pair e-mail testing commenced. During the data

collection period (April through August 2014), rental advertise-

ments were randomly assigned to matched pairs of e-mails from

a person with MD and a person without MD.

Quality control (QC) was intrinsic to every level of the e-mail

testing process. The procedures described in this section were

used in identifying housing units and providers to be tested,

in executing the tests, and in collecting test data to minimize

the risk of detection by housing providers who might become

involved in more than one mode of testing.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

The scraper tool provided samples from internet-advertised

rental units for the e-mail tests. QC mechanisms were em-

bedded in the scraper tool to remove duplicates, and the test

administrator was responsible for providing a ﬁnal QC review

of the pooled rental advertisements generated by the scraper.

The test coordinator reviewed randomly selected entries and

noted any exceptions to the protocols, which were logged in

an exception ﬁle so the same decision rules could be applied to

subsequent exceptions. For example, if the scraper tool picked

up an apartment listing that was actually part of a single-family

home, this listing was removed from the sample and logged so

that the same protocol would be followed for all future instances.

For each e-mail test, the e-mail text, subject lines, and tester

names were designed to appear as natural housing inquiries.

The tests were also matched at the semantic level, so that

level of interest, length of e-mail text, and word choice were

all similar. Thus, in all aspects, except for the disability of the

hypothetical individual who sent the housing inquiry e-mail,

the paired e-mail tests were matched.

The order of which e-mail was sent ﬁrst, either the control or

protected tester’s e-mail, was randomized to address the poten-

tial effect on the test outcome of the order in which the inquiry

was received by the housing provider. The counterbalance

device used was a Microsoft Access table, which was created

to randomly select the order in which the control or protected

e-mails were sent. Microsoft Access forms were used to schedule

and generate the outgoing e-mails, ensuring consistency in

timing of outgoing e-mails. Both parts of each pair of e-mails

were sent to the selected housing provider within a 2-hour

window. All outgoing e-mails were saved to a computer server

and linked to the Access database so that accuracy could be

conﬁrmed.

Test administrators used a sample generation form to compose

e-mails. The scraper tool ﬁlled in most of the required ﬁelds

of the form. Required ﬁelds were checked for accuracy by the

test administrator, and any ﬁelds found to be missing were

completed based on the original advertisement. All the samples

for e-mail testing were created within 48 hours of the posting

of the advertisement. After e-mails were generated based on

the sample generation form, test administrators conﬁrmed that

the e-mail addresses were correct, the e-mail subject lines and

text were different between the e-mails, the e-mail addresses and

sender names matched, and the “when to send” ﬁelds were logical.

The ordering of paired tests was randomly assigned, and both

e-mails within a given pair were sent within 2 hours by an

automatic scheduler. The test coordinator tracked outgoing

inquiries from, and incoming responses to, the test e-mail

accounts daily, and the tracking lists were compared against—

• The number of outgoing and incoming e-mails saved to the

server.

• The number of e-mails coded for each entry in the

tester-level table.

• Access-generated e-mail contents for randomly selected

entries.

Only one inquiry was sent to each advertised property, and

there was no followup if a housing provider did not respond to

an e-mail. Tests remained open throughout the entire testing

period, although most responses (90.1 percent) occurred within

2 days of the original e-mail. If a response from the advertised

property was received, testers did not send a return e-mail. For

cases in which a test e-mail received multiple responses from

the same advertised property, all were coded in the Access database.

The Microsoft Access program used for coding input was

custom designed to ensure complete, accurate, and high-quality

data. For example, acceptable ranges for responses were pre-

programmed to help prevent data entry error. In addition, the

possibility of incomplete or missing data was prevented by not

allowing a coder to proceed to the next ﬁeld or screen until all

ﬁelds were complete. During the actual testing phase, data were

examined daily and validation was performed on 20 percent

of each test administrator’s work using double-entry coding to

respondents’ e-mails to verify responses before inclusion in the

ﬁnal datasheet and before being signed off by the test coordi-

nator. All tests underwent rigorous review by the study team to

help ensure high quality, consistency, completeness, accuracy,

clarity, and usability.

Telephone Testing Protocols

As in the MD exploratory testing of DDS 2005, the telephone

tests in this study were conducted by people without MD pur-

portedly calling on behalf of someone with MD. Given the large

number of telephone tests to be conducted over a relatively

short period of time and the challenges of recruiting, training,

and supporting testers with MD noted in DDS 2005, practical

considerations and concern for the burden on the testers with

MD led to the decision to use proxy testers who did not have

MD for the telephone testing in this study. Therefore, the

telephone testing process involved the recruitment and training

of people without disabilities to serve either as the control tester

or as the protected tester in the role of a proxy calling on behalf

of a relative or friend with MI or I/DD.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

After the pretesting, it was determined that the telephone test-

ing would also include a request for a reasonable accommoda-

tion related to the stated disability on behalf of the person with

MD. The two reasonable accommodation requests used were

either (1) a request that the housing provider give the tenant

a reminder to pay rent or (2) a request that the tenant be

allowed to have an assistance animal in the unit.

One of the

two accommodation requests was made for each MI or I/DD

telephone test. One-half of the testers for each group requested

one of the two accommodations. Reasonable accommodation

requests were included in all telephone test proﬁles and tester

instructions. Reasonable accommodation requests were to be

made after the information regarding availability and the rental

unit was obtained, and thus requests were actually made in

629 of the 668 telephone tests conducted, because in 39 of the

tests, the conversation with the provider ended before reaching

the accommodation request.

Procedures and Scripts for Tester

Telephone Calls

The order of telephone calls to the housing provider by the

protected tester and the control tester was randomized, and the

call placements occurred no less than 1 hour and no more than

4 hours apart. The timing strategy was designed to increase the

likelihood of speaking with the same housing provider or agent

and to minimize the possibility of an incomplete test because

the unit was rented between the paired calls.

Proxy Relationships

The telephone testing protocol originally included the use of

proxies calling on behalf of both the control and protected

testers. During the telephone exploratory testing, however, it

was determined that use of a proxy for both testers within the

short timeframe created an unacceptable risk of detection by

housing providers, and thus the proxy caller for control testers

was dropped and the control testers called on their own behalf.

The tester calling as the proxy for the person with MD posed as

a friend or family member (and not a mental health professional

or case manager) of the potential renter. Protected proxies

expressly stated that they were calling for a person with MD

(“My friend/sibling has a mental health disability or illness” for

MI or “has a developmental disability” for I/DD).

As with e-mail tests, the disclosure of the protected tester’s dis-

ability occurred early in the communication and was designed

to be unambiguous. Table 5 includes the text of the telephone

testing scripts used by the proxy tester to disclose disability.

The proﬁles for both the protected proxy and the control tester

included that neither potential renter was currently living in his

or her own rental unit. The proxy caller for the individual with

MD was instructed to state that the person with MD had been

living in a psychiatric center for the past year (MI) or had been

living in a group home in the community for the past year (I/

DD). The control tester was instructed to state that he or she

was returning from school or relocating to the area.

Reveal of Disability

Although the proxy for the protected tester disclosed the MD to

the housing provider very clearly and early in the conversation,

the proxy was instructed not to provide a speciﬁc diagnosis (for

example, schizophrenia). Instead, any elaboration if requested

would be provided in the form of information on functional

aspects of the disability using common descriptions provided

by the focus groups and identiﬁed in the literature review

(for example, has emotional issues, difﬁculty remembering or

organizing, or needs help with taking care of self or managing

payments).

Table 5. Telephone Tester Scripts

Control Tester Protected Tester

My name is <ﬁrst name> and I’m calling about the apartment. I’ve been

in school and I’m looking for my ﬁrst place.

I’m <ﬁrst name>. My <brother/sister> has a <disability identiﬁer>. He/

she has been living in <alternate housing identiﬁer> and I’m helping

him/her look for an apartment.

I’m <ﬁrst name>. I just moved to town and I’m looking for my ﬁrst

apartment.

My name is <ﬁrst name>. My friend has a <disability identiﬁer> and has

been living in <alternate housing identiﬁer> and I’m helping him/her

look for an apartment.

In the pretesting, one of the reasonable requests used was for an additional key for a caregiver, but this request was determined to be too “soft;” that is, no housing

provider objected to providing an additional key, and their consideration of the key request seemed unrelated to the issue of MD—anyone who wanted an additional

key could get one. This accommodation request was then changed for the pilot testing to a request that the housing provider give the tenant a reminder to pay rent.

The decision to have the proxy be a friend or family member, as opposed to a mental health professional or a case manager, was based on previous research that

identiﬁed a set of concerns or assumptions that might be triggered by a call from a mental health professional or case manager (for example, severity of disability,

need for active treatment, or anticipation of need for signiﬁcant accommodations).

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Examples of functional aspects that were evaluated in the pilot

test included—

• Goes to “X” for mental health treatment/services/support

(MI); or receives I/DD services in the community from

“X.” Control proxies said nothing about this issue.

• Requires assistance to cognitively or emotionally manage

or experiences difﬁculty with remembering or emotionally

managing; has someone who comes in 2 days a week to

help take care of him or her and the apartment (MI and I/

DD); only if asked by the landlord, additional details may

include that the person helps with paying bills, managing

money, cooking, and laundry. Control proxies said

nothing about this issue.

Statements about the functional aspects of the disability did

not request or require any action on the part of the housing

providers and, more importantly, were designed not to be confused

or conﬂated with a request for a reasonable accommodation.

Reasonable Accommodation Requests

Any request for a reasonable accommodation was made at

the end of the telephone call after all other information for

the test had been collected. The intent of the design was to

provide a second level of analysis related to the reasonable

accommodation request. As with the e-mail testing, reasonable

accommodation requests were designed to reﬂect the types of

accommodations commonly requested by people with MD as

part of the search for rental housing. Reasonable accommoda-

tion requests speciﬁcally avoided issues of “source of income”

or poverty that could introduce other variables into the test.

The accommodations sought in this test were—

• Need for an assistance animal and a waiver of any

associated fees.

• Request for a reminder from the housing provider by

telephone or in person to the person with the MD that

rent was due, as they have difﬁculty remembering.

Telephone Tester Recruitment

Recruitment of telephone tester candidates began in late

November 2013. Local fair housing advocacy organizations

and social work departments at universities in the tristate area

of Delaware, New Jersey, and Pennsylvania were contacted,

and recommendations for tester candidates were solicited.

The recruitment goal was to have both male and female

African-American and Caucasian matched pairs of testers

participate in training and testing.

All candidates were administered a telephone screening by the

test coordinator. This screening was followed by an in-person

interview during which the candidates’ voices were recorded.

The purpose of the recordings was to identify and screen out

candidates with discernible regional accents or other vocal

idiosyncrasies that might contribute to bias.

A review of testing background and experience level of each

candidate informed the preliminary decision on tester suitabil-

ity. Demographic information, which was used to determine

tester pairings, was provided by every candidate. The ﬁnal

decisions on tester pairings—that is, which of the two testers in

a pair would assume the control and the protected-class roles

and who would be paired with whom—were based on the race,

age, gender, and experience levels of each tester and were made

by the project’s subject matter experts (SMEs) and the test coor-

dinator. As a result of these screening efforts, the initial training

class consisted entirely of experienced fair housing testers, all of

whom had previously conducted telephone testing.

Training for Telephone Testers

Classroom and practical training were conducted in late

January 2014, over a 2-day period in Philadelphia. The ﬁrst day

of training consisted of a classroom orientation that covered a

variety of topics, including—

• A history of housing discrimination testing.

• Types of discriminatory housing practices.

• The process of fair housing testing.

• The strategic goals of the project.

• The telephone testing data collection process.

• Tester instructions and guidelines.

• Review of tester forms.

• Completion of testing assignments.

• Narrative writing.

• QC practices.

Practical training was conducted on the second day of training

in the call center where the telephone testing would be imple-

mented. Testers were given hands-on instruction on the use of

tester assignment and test report forms and were instructed on

how to use the data repository computer application to store

test result data. The training also included instruction on call

center telephone use and inbound voicemail access procedures.

Each trainee made practice test calls to SMEs on site and at

remote locations using carefully constructed tester proﬁles.

Trainees received immediate feedback from the test coordina-

tor, SMEs, or both after each call.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Supplemental and Review Training

At times during the 9-month exploratory and pilot testing pe-

riod, additional telephone testers were trained to maintain the

demographic consistency of available testing pairs. Additional

Caucasian male and female paired teams were added after data

collection had begun. Supplemental trainings were led by the

test coordinator and included practice telephone tests to offsite

SMEs and hands-on telephone and computer training.

At the midpoint of the data collection period, SMEs conducted

a second classroom training session for all testers, during

which critical testing protocols and QC procedures were re-

viewed and emphasized.

This 1-day mandatory training was

conducted at the training facility for all telephone testers.

Conducting Telephone Pilot Testing and

Quality Control Procedures

Telephone tests were not conducted by individuals with MD,

but instead relied on a proxy design. Proxies posed as a friend

or family member seeking rental housing on behalf of an indi-

vidual with MD. Control testers contacted the housing provid-

er on their own behalf. Proxies calling on behalf of an individ-

ual with MD were provided detailed background information

on the proﬁle of the person on whose behalf they were calling.

Control testers were provided similar background information

for testing purposes. As in all matched-pair testing, the proﬁles

of the individuals whose treatment is to be compared (here,

the people on whose behalf the proxies were calling) were

matched to the control tester on all relevant characteristics

except the variable being tested—the presence of MD. Thus,

paired potential renters’ proﬁles were of the same gender and

race and assigned personal and ﬁnancial characteristics that

were as similar as possible, such as approximate move-in

date, desired rent range, credit score, current employment,

and income level. To avoid detection by housing providers as

a result of two testers representing individuals with identical

characteristics, the proﬁles of potential renters with disabilities

showed them to be slightly more qualiﬁed for the housing—for

example, having a higher income level or credit score—than

the paired potential renter without a disability. In addition,

to avoid bias caused by the proxies, proxies were of the same

gender, race, and approximate age as control testers.

Telephone tests were designed so that the proxy was to reveal

the protected potential renter’s MD to the housing provider at

the beginning of the conversation using a direct reveal of the

disability and a description of functional needs without provid-

ing an actual diagnosis.

The type of functional issues identiﬁed

matched the designated disability group (MI or I/DD).

All telephone testers were briefed on the nature of the QC

pro cedures instituted for this study. Some QC processes were

embedded in the selection of rental housing assignments and

test order. For example, the order of the calls within each tele-

phone test (control call made ﬁrst or protected call made ﬁrst) was

randomized to address any potential inﬂuence on test outcomes.

Rental housing assignments in the various markets were auto-

matically screened via data-mining tools using criteria such

as rent range, market location, and property owner or agent

uniqueness. The test assignment forms were automatically

populated with information from this data-mining process, then

printed and provided to each pair of proxies and testers. Testers

reviewed the test assignment forms and made telephone calls

based on a randomized order. The second telephone call in

each test was made within 1 to 4 hours after the initial test call

to reduce the chance of detection but to increase the likelihood

of speaking to the same property owner or agent.

QC procedures also were in place at the tester level. Proxies and

testers were trained that, after the completion of each telephone

engagement with a property owner or agent and before the

completion of a test narrative, the tester would be debriefed by

the onsite test coordinator and test details would be discussed.

Only after this debrieﬁng would the proxy or tester complete

the test narrative and enter the results from the telephone test

form into the database.

Finally, all test forms and narratives for each paired test were

sent to SMEs for review and approval. Only after this approval

would a paired test be considered complete.

In-Person Testing Protocols

People with MD served exclusively as protected testers for the

in-person testing in this study. For in-person tests, which had

a smaller sample size and a longer execution period than other

modes of testing, the testers with MD (both MI and I/DD) were

Although a variety of issues were covered in this training, the importance of the timing of the protected tester’s identiﬁcation of disability and the completion of the

matched control test were particularly stressed.

To compare the results of each paired test, it was essential that the housing provider unambiguously know from the beginning of the call that the protected

potential renter had MD. The proxies clearly and directly stated to the housing provider at the start of the conversation that they were calling on behalf of someone

with MD (MI or I/DD).

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

accompanied by someone without MD posing as a friend.

This

protocol of a companion accompanying the protected tester

replicates what was adopted in the DDS 2005 for in-person

testing. The protected tester-and-companion team approach

provided support and assistance for the testers with disabilities

to ensure that test protocols were followed and that reporting

was complete and accurate.

Control testers were people without MD and were not accom-

panied. The tests were designed to reveal whether rental hous-

ing providers treated testers differently on the basis of MD. In

each test, the protected tester’s MD was revealed to the housing

provider by the protected tester at the beginning of the visit

to the property. In making this disclosure, the protected tester

used a direct reveal and description of functional needs with-

out providing an actual diagnosis. The disability identiﬁed was

the same for all testers in each disability group (MI or I/DD).

As in all matched-pair testing, the paired testers were matched

as to all relevant characteristics except for the variable being

tested—the MD. Paired testers were of the same gender and

race and were assigned personal and ﬁnancial characteristics

designed to be as similar as possible, including, as described in

more detail subsequently, unit size needed, potential move-in

date, desired rent range, credit, current employment, and in-

come level. To avoid detection by housing providers as a result

of two testers having identical characteristics, protected tester

proﬁles were slightly more qualiﬁed for the housing being

sought; for example, having a slightly higher income level or

length of employment.

Procedures and Scripts for In-Person Tests

Advance calls were made by test coordinators before each

in-person visit to conﬁrm that the housing unit remained

available, the location, and ofﬁce hours and to determine

whether an appointment would be needed to see the unit. The

need for appointments differed in the two cities, with nearly

all Chicago properties requiring appointments but most of

the Washington, D.C. properties not (allowing for the testers

to “walk in” to visit the properties). Appointment contacts, if

necessary, were made primarily by telephone calls within 3

days of the publication of the advertisement by the companion

testers (for MI and I/DD testers) and by the control testers.

Testers’ companions did not reveal that they were looking

for an apartment for someone else, and no disclosure of any

disability was made during the call.

Test coordinators created test assignments based on the infor-

mation in the rental advertisements and developed proﬁles for

test assignments that matched the paired control and protected

testers on their relevant personal and ﬁnancial characteristics—

for example, gender, race, approximate age, income, credit,

ﬁnancial resources, employment, household composition, and

housing preferences such as the unit size, rent, and date need-

ed—and included an assigned e-mail address and telephone

number for each tester (which were monitored by the test

coordinators). Assignment forms for protected testers included

instructions on revealing their MD. All testers were seeking

their ﬁrst apartment and were ﬁnancially qualiﬁed for the units

they sought. The protected testers were either living in the

community (in a group home or with family) or coming out of

an institutional setting. Control testers had been living at home

with family or away at school. To avoid identical characteristics

that might lead to detection, protected testers’ proﬁles made

them slightly more qualiﬁed for the housing being sought; for

example, having a slightly higher income level or length of

employment. Test coordinators reviewed the assignment forms

with testers before each test.

Protected testers were accompanied on property visits by

companions who provided support to testers during data col-

lection. For example, when testers forgot to ask a question of a

property owner or agent, “got off track,” or “froze,” companion

testers were able to naturally prompt the tester or provide a

response. Companion testers also reassured testers and helped

them feel less nervous when in a new area of their community.

Companion testers also assisted by taking notes, organizing

and remembering information, helping out when unanticipated

situations arose, and in some cases providing transportation.

Both testers and companions approved of this model, reporting

that a companion’s presence helped some testers feel less

anxious because immediate support was available, that testers

and companion testers complemented one another during data

collection by offering different ideas, and that working as a pair

made the experience more enjoyable.

Testers and companion

testers also noted that the presence of a companion may also

have added credibility to the disclosure of a disability by testers.

Paired testers attempted to visit an assigned property on the

same day, but always visited within 24 hours of each other.

The order for whether the control or protected tester visited

the property ﬁrst was randomized, as previously described.

The study team thought it was important to ask testers in what situations they believed they needed assistance (for example, writing, reading, or having a quiet

room in which to complete forms) and to be ﬂexible in responding to testers’ emergent needs. The need for support decreased over time as testers increased their

skills and comfort. This support improved project efﬁciencies by better enabling a smaller group of testers to comfortably conduct multiple tests rather than having a

larger pool of testers who might conduct only a single test.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Testers used protocols designed to allow for an assessment of

differential treatment; for example, protected testers revealed

their MD very early in the conversation, before information on

the unit was obtained. During site visits, testers inquired about

the advertised property and gathered information on the avail-

ability of the advertised unit or other similar units, addresses

and unit numbers, rental amounts, the application process,

fees, utilities, other terms and conditions, and amenities.

Testers asked to see the unit advertised and were willing to see

a model unit or other similar units. Testers were encouraged to

take notes during their visits.

Immediately after each test, the testers notiﬁed the test

coordinator that the test was completed, ﬁlled out the test

report form, and wrote a narrative. Protected testers and their

companions completed test reports and narratives together.

Testers turned their reports into the test coordinator by mid-

night of the day of the test and the narrative within 24 hours.

Test coordinators reviewed the completed test reports and

narratives and debriefed each tester. Test results were entered

into the testing database by the study team, and a weekly

report of completed tests was produced to monitor progress.

All documentation from in-person tests was reviewed and

approved by a designated member of the testing QC team.

Tester Proﬁle and Reveal Strategy

It was critical to the test design that protected testers reveal

their disability at the beginning of the interaction so that the

presence of their MD was known to the housing providers or

agents before information about the unit was discussed. To

accomplish this reveal, protected testers were instructed to

greet a housing provider or agent, identify themselves by ﬁrst

name, and state either “I have a developmental disability” or “I

have a mental health disability.” Their companions were then

introduced as someone who was helping them to ﬁnd rental

housing and speciﬁcally was tasked with remembering what

questions to ask about the unit and helping to take notes about

the apartment.

The protected testers were not to name their speciﬁc disability

by providing a diagnosis or condition but were to reveal only

the broad category of “developmental disability” or “mental

health disability.” Test assignment forms provided each tester

with information about how the disability affected the tester’s

daily life (for example, difﬁculty remembering things or

difﬁculty staying organized). This information was mentioned

during the course of the test. See Table 6 for examples.

Test assignment forms also provided information about

protected testers’ current housing situation (for example, living

with family or in an institution, such as a nursing home or psy-

chiatric facility), which was also revealed at some point during

the test. Test assignment forms for control testers also included

information about their current housing situation. Control

testers were instructed to tell a housing provider or agent that

this was their ﬁrst apartment search because they had been

living in the housing assigned in their proﬁle (for example, with

family or away at school).

Table 6. Examples of How MD May Affect Functional Aspects of Life and Maintenance of Housing

Functional Aspects of MD

MI I/DD

Afﬁliation with a

community service

provider

Goes to “X” for mental health treatment/services/support Receives I/DD services in the community from “X”

Needs supports

with housing-related

management activities

(1) Requires assistance to help with organizing and

keeping track of things.

(1) Requires assistance to help with organizing and

keeping track of things.

(2) Has difﬁculty with remembering. (2) Has difﬁculty with remembering.

(3) Has someone who comes in 2 days a week to help

take care of him or her and the apartment (both).

(3) Has someone who comes in 2 days a week to help

take care of him or her and the apartment (both).

Only if asked by the landlord, additional details may

include that the person helps with paying bills, managing

money, cooking, and laundry.

Only if asked by the landlord, additional details may

include that the person helps with paying bills, managing

money, cooking, and laundry.

I/DD = intellectual or developmental disability. MD = mental disability. MI = mental illness.

Pretesting was used to evaluate the manner in which the MD was conveyed to the housing provider and to highlight any detection issues that might arise either

when setting up an appointment to visit a property or during the actual visit.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

In-Person Tester Recruitment and

Training

Test coordinators from community-based fair housing testing

organizations in the two metropolitan areas selected for

in-person testing, Access Living in Chicago and The Equal

Rights Center in Washington, D.C., recruited potential testers

from their existing tester pools (particularly for control testers

with no disability) and from a variety of community groups

and organizations in which people with MI and I/DD actively

participated or from whom they received housing or commu-

nity living services. These organizations’ extensive contacts and

continual networking with community groups and people with

disabilities were essential to this effort and would be difﬁcult

for organizations not familiar with, and embedded within,

disability communities to replicate. The use of these communi-

ty-based testing organizations was a critical component of the

project and a key to success in actively supporting people with

I/DD and PD during the in-person testing.

To establish a pool of testers, test coordinators needed to allo-

cate signiﬁcant time to the tester recruitment process, initially

recruiting about three times the number of testers needed

and continuing to recruit and train additional testers during

the actual testing period. For example, in some cases, most

notably among potential testers with I/DD, availability and

reliability were issues (for example, scheduling training, not

showing up for scheduled training, the availability of reliable

transportation, and the need to navigate new transportation

routes without support). The lack of availability and reliability

was particularly problematic given the need to coordinate the

schedules of protected testers with companion testers for initial

training, practice testing, and conducting each in-person test.

Because of employment disparities among the populations,

people with MD were found to be more available to work

as testers, but the ability to recruit control testers was more

challenging because of their higher employment rate.

Test coordinators communicated extensively with potential

testers to determine whether the tester’s abilities matched the

testing and reporting responsibilities involved in the testing

process, to assess whether candidates could complete the

essential tasks required of a tester. In these interactions, test

coordinators provided potential testers a candid and thorough

description of the project and their expected role (including an

emphasis on the responsibilities of being a tester versus other

roles within the project), assessing how the potential tester

understood his or her role, and the potential tester’s comfort

with pretending to be searching for an apartment and disclosing

his or her disability during the housing search process.

Test coordinators also noted the importance of identifying can-

didates who would be good companion testers. Strong rapport

between a tester with disabilities and his or her companion was

key to the tester’s comfort and to the quality and efﬁciency of

testing. This rapport was accomplished by creating opportu-

nities throughout the project—from initial training, to during

practice tests, and through the completion of all test assign-

ments—for each protected tester and the assigned companion

to get to know and become comfortable with each other. Test

coordinators regularly monitored how these relationships

progressed. If one member of a pair was not comfortable, pairs

were rematched.

Separate training sessions were held for each of the three

groups of testers—MI, I/DD (together with their companions),

and control testers. Based on initial insights from multiple focus

groups and discussions with the expert panel at the start of the

project, training approaches were modiﬁed for each group to

maximize their learning. For example, testers with I/DD had

project information broken into smaller units, followed imme-

diately by opportunities to practice a small skill, leading up to

practicing a full test under controlled circumstances, and ﬁnally

moving on to an actual practice test in the ﬁeld. Testers with

MI received initial information in one sitting and then practiced

different scenarios.

The study team maintained ﬂexibility in the tester training

protocols, simplifying and limiting the quantity of background

information and lecture for the MI and I/DD groups and

focusing on role-play, practice exercises, and open discussions

to enable the testers to gain a level of comfort with both the

testing process and working with a companion.

For the MI

and I/DD groups, training materials deemphasized complicated

text, reduced the number of ideas presented at one time, and

incorporated graphics, photos, and images to illustrate more

abstract points such as civil rights.

Frequent breaks were provided during all training sessions to

allow for testers to assimilate new information and experiences.

The training sessions also helped test coordinators in making

ﬁnal determinations about a potential tester’s involvement in

the actual in-person tests.

Trainings included opportunities ﬁrst for testers to observe testing skits performed by test coordinators and project staff, and then for testing pairs to actively

engage in several different testing scenarios with role-play (moving from working in pairs with an assigned project staff member to role-play in front of the whole

group and getting feedback afterward).

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Conducting In-Person Pilot Testing and

Quality Control Procedures

The in-person pilot testing was conducted in Chicago and

Washington, D.C., from June to November 2014. Samples of

available rental units were gathered from rental advertisement

contact data extracted from online rental advertisements by the

automated scraper tool and weekly random selections of ap-

proximately 20 to 25 rental property advertisements provided

by the study team to the test coordinators in each city.

Overall Data Collection Oversight,

Management, and Quality Control

For each mode of testing, testers received training and

retraining as needed during the data collection period. During

initial training, testers engaged in role-playing, using the actual

testing protocols. This process facilitated learning and was the

basis for the evaluation of tester capabilities and an assessment

of whether additional training was needed. This process of

evaluation continued throughout training, exploratory testing,

and actual pilot testing.

E-mail Testing

The research team used a proprietary Microsoft Access program

that was customized for this particular paired testing data

collection. Logic and other QC checks are imbedded in the program

to facilitate a high level of accuracy when recording data. In

addition, e-mail testers were assigned speciﬁc markets to facili-

tate familiarity with local address formats and to reduce errors.

Telephone Testing

Telephone testers were paired into teams, and the team conﬁg-

uration remained the same over the course of the testing, absent

illness or other uncontrollable events. After the completion of

each test, results were reviewed at least two times, once by the

QC supervisor and a second time by senior testing experts. The

reviews validated the tests, ensured protocols were followed,

and conﬁrmed that the information recorded was complete,

clear, and unambiguous.

In-Person Testing

In-person tests were reviewed by the test coordinator within

24 hours of test completion. In addition, each paired test was

reviewed by senior testing experts to ensure the complex testing

protocols were followed and to conﬁrm that the information

recorded was complete, clear, and unambiguous.

In all testing modes, overall, QC was implemented with stringent

reviews of protocol and processes throughout the data collec-

tion period.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Sampling Methods

Sampling Frame

The sampling frame for the paired testing in this project was

designed with the goal of completing 1,000 matched-pair tests:

500 matched-pair tests to assess discrimination against people

with MI and 500 matched-pair tests to assess discrimination

against people with I/DD (Table 7). An additional 100 paired

tests were to be conducted by e-mail speciﬁcally to assess the

willingness of housing providers to make reasonable accom-

modations for people with the two identiﬁed categories of MD.

Table 8 summarizes the sampling frame for the project and

how the 1,000 matched-pair tests were allocated among the

three selected rental market strata—medium-sized markets

(population of 500,000 to 1,250,000), large-sized markets

(population of 1,250,000 to 2,500,000), and very large-sized

markets (population of 2,500,000 or more). As noted, tests

in all markets were equally divided between those focused on

individuals with MI and those with I/DD.

To avoid issues of detection, small markets with populations

less than 500,000 were excluded from all pilot testing. Tele-

phone and e-mail testing were conducted in all three market

size types, whereas in-person testing was conducted only in the

very large markets.

Allocation of Tests Across Market Size

Test allocation was guided by market and population char-

acteristics. As shown in Table 9, the number of test markets

selected within each stratum for the pilot testing mirrors the

distribution of the population across metropolitan statistical

Table 7. Sampling Frame for People With MD, by Test Type

Matched-Pair Test Type Total MI I/DD

E-mail 200 100 100

E-mail + reasonable accommodation 100 50 50

Telephone 600 300 300

In person 100 50 50

Total 1,000 500 500

I/DD = intellectual or developmental disability. MD = mental disability. MI = mental illness.

Table 8. Rental Market Sampling Frame

Matched-Pair Test Type Total

Market Population Size

Medium

(500,000 to 1,250,000)

Large

(1,250,000 to 2,500,000)

Very Large

(2,500,000 or more)

n = 4 n = 2 n = 2

E-mail 300 20 40 240

Telephone 600 200 200 200

In person 100 0 0 100

Total 1,000 220 240 540

Table 9. Allocation of Tests Across Market Size

Market Population Size

Total

Medium

(500,000 to 1,250,000)

Large

(1,250,000 to 2,500,000)

Very Large

(2,500,000 or more)

U.S. distribution

Percent of population 21 21 58 100

Percent of all markets 56 23 21 100

Pilot testing

Number of markets 4 2 2 8

Percent of markets 50 25 25 100

Percent of paired tests 20 20 60 100

Although the plan was to test in four medium-sized markets, possible detection before the completion of data collection in one of the markets,

New Haven-Milford, CT, caused the addition of Syracuse, NY. Thus, nine total markets were included in the ﬁnal analysis.

Source of U.S. population and market distribution: 2010 American Community Survey, SF1, Table P1

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

areas (MSA) and micropolitan statistical areas based on 2010

American Community Survey (ACS) data. The percentage

distribution of paired tests across strata similarly mirrors the

population percentage distribution based on 2010 census data.

For example, because approximately 20 percent of the U.S.

population resides in markets with populations ranging from

0.5 to 1.25 million, 20 percent of the total number of paired

tests was implemented across markets of this size.

Sample of Metropolitan Areas

Based on the sampling design and sample frame, the selection

of markets and allocation of tests across markets were con-

ducted in two stages: (1) identify and characterize the pool of

potential markets for inclusion and (2) select the markets for

testing.

Stage 1: Identifying and Characterizing

Potential Markets

Using MSA population data, 374 MSAs were divided into four

categories based on size—270 small-sized markets (population

less than 500,000), 58 medium-sized markets (population of

500,000 to 1,250,000), 25 large-sized markets (population

of 1,250,000 to 2,500,000), and 22 very large-sized markets

(population of 2,500,000 or more).

To further differentiate these markets, the proportion of people

with any disability and proportion of people with cognitive

disabilities, based on the 2009–2011 ACS and the 2010

census, were considered.

These data were included to ensure

that each selected market contained a typical (within a range)

proportion of people with MD and a sufﬁcient population from

which testers for in-person testing could be drawn. These data

are summarized in Table 10.

Markets were also differentiated by whether the state had an

Olmstead Plan, any active Olmstead-related lawsuits, or both

and by if the state received Centers for Medicare and Medicaid

Services, or CMS, funds to transition people with MD out of

institutions to the community via a Money Follows the Person

grant. If the state met either of these conditions, markets were

further differentiated by how many people had transitioned and

whether the state had met its transition goals.

The presence

and strength of state-level efforts to move individuals with

disabilities from institutional settings to community-based

settings was presumed to have an effect on rental markets by

increasing awareness of the rights of people with disabilities

under the Fair Housing Act and the Americans with Disabilities

Act among both housing providers and prospective tenants,

thus potentially affecting testing.

Stage 2: Selecting the Study Markets

This study used a purposive sampling approach to identify

speciﬁc markets for testing based on population size and certain

characteristics, as described previously. A purposive approach

was used for several reasons.

• The selected markets had to reﬂect a proportion of people

with MD that was typical (in other words, an unusually

high or low percentage of people with disabilities should

not reside in the market).

Table 10. Distribution of Population With Disabilities and Proportion With Cognitive Disabilities, 2009–2011

MSA

Size

Number

of MSAs

Proportion of Population With Disabilities

(%)

Proportion of Population With Cognitive Disabilities

(%)

Median Average Minimum Maximum Median Average Minimum Maximum

Very large 22 9.6 10.2 7.8 13.6 30.5 31.1 20.8 48.2

Large 25 11.5 11.9 7.7 19.8 22.6 24.9 9.3 53.6

Medium 58 12.2 12.2 7.2 16.4 22.9 23.9 15.0 40.0

MSA = metropolitan statistical area.

Sources: 2009–11 American Community Survey; 2010 census

To avoid issues of detection, the 270 small-sized markets were excluded from all pilot testing.

The ACS and census count an individual as having a cognitive disability if he or she reported that, due to a physical, mental, or emotional condition, they had

“serious difﬁculty concentrating, remembering, or making decisions.” The term “cognitive disabilities” is formally used within census data to be inclusive of the same

categories as the term “mental disabilities” as used by HUD. It includes MI and I/DD. As the source of the data in Table 9 is from census data, the term cognitive

disabilities is used. For all other purposes of this study, the term “mental disabilities” is used.

This grouping includes any plan instituted as a result of a judgment or settlement, or any pending litigation, arising out of the U.S. Supreme Court’s ruling in

Olmstead v. L.C., 527 U.S. 581 (1999) (unjustiﬁed segregation of people with disabilities constitutes discrimination in violation of Title II of the Americans with

Disabilities Act).

Data for these measures were attained from Ng, Wong, and Harrington (2013), on Olmstead Plan and active lawsuits, and from Denny-Brown et al. (2011), on

People transitioned as of June 2011 and 2010 Transition Goal achieved.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

• The selected markets for in-person testing had to contain

a sufﬁcient population of people with MD from which

testers could be drawn.

• The selected markets for in-person testing had to include

local fair housing and enforcement agencies that had

expertise in working with people with MD.

• To the extent possible, the research team sought to measure

any effects of Olmstead on differential treatment of people

with MD; thus, at least one market in a state that is imple-

menting an Olmstead Plan was intentionally selected.

After applying these criteria to produce a pool of potential

mar kets, the list was narrowed further by identifying those

markets that represent a typical mix of people with disabilities

relative to the rest of the population and speciﬁcally a typical

mix of people with cognitive disabilities (as deﬁned by ACS

and the census) as a proportion of all people with disabilities.

Using these data and the previously discussed criteria, the ﬁnal

market pool contained 17 medium and 7 large MSAs (see Table

11) based on those rates that fell approximately between the

25th and 75th percentiles for disability and cognitive disability.

These markets preliminarily selected were then discussed with

the expert panel regarding any speciﬁc factors such as Olmstead

implementation activities and disability-related fair housing

litigation that might bias sample selection and results. The

panel’s feedback was taken into consideration in making the

ﬁnal market selections. As a ﬁnal step, an analysis of secondary

data on key characteristics relevant to the testing process of

each market was conducted, including an assessment of the

percentage of total housing that was available for rent, rental

vacancy rates, and craigslist coverage of housing rentals.

Table 11. Pool of Possible Markets (MSAs) Based on Statistical Criteria for Inclusion

MSA

State-Level Data

Has an

Olmstead Plan

Active Disability

Lawsuits

People Transitioned

as of June 2011

2010 Transition

Goal Achieved

Large markets

Cincinnati-Middletown, OH-KY-IN Yes Yes 1,139 Yes

Memphis, TN-MS-AR No No NA NA

Milwaukee-Waukesha-West Allis, WI Yes Yes 145 No

Nashville-Davidson--Murfreesboro--Franklin, TN No No NA NA

Providence-New Bedford-Fall River, RI-MA No No NA NA

Sacramento--Arden-Arcade--Roseville, CA Yes Yes 537 No

Virginia Beach-Norfolk-Newport News, VA-NC No No 269 Yes

Medium markets

Albany-Schenectady-Troy, NY Yes Yes 370 Yes

Albuquerque, NM No No NA NA

Allentown-Bethlehem-Easton, PA-NJ No Yes 719 Yes

Bakersﬁeld-Delano, CA Yes Yes 537 No

Baton Rouge, LA No No 164 No

Buffalo-Niagara Falls, NY Yes Yes 370 Yes

Cape Coral-Fort Myers, FL No Yes NA NA

Columbia, SC No No NA NA

Fresno, CA Yes Yes 537 No

Greenville-Mauldin-Easley, SC No No NA NA

Harrisburg-Carlisle, PA No Yes 719 Yes

Jackson, MS Yes Yes NA NA

New Haven-Milford, CT Yes Yes 561 Yes

New Orleans-Metairie-Kenner, LA No No 164 No

North Port-Bradenton-Sarasota, FL No Yes NA NA

Rochester, NY Yes Yes 370 Yes

Syracuse, NY Yes Yes 370 Yes

MSA = metropolitan statistical area. NA = data not reported or available for that MSA.

Source: Final Research Design & Data Collection Plan, approved September 13, 2013

As shown in Table 10, the range of people with disabilities varies across the MSAs (the range is the difference between minimum and maximum values) and is

much narrower than ranges observed in the distribution of people with cognitive disabilities within the group with disabilities. The data on cognitive disabilities from

the ACS are imputed. The observed higher variability in the percentage of people with cognitive disabilities within the population with disabilities per market may be

a result of how the imputation was done, or it may be representative of the true distribution.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

The nine MSA markets selected for this project are identiﬁed in

Table 12.

Table 13 shows the planned sample allocation of paired

tests across the selected markets within each size category.

Approximately 60 percent of the allocated matched-pair tests

fall within the very large markets and 20 percent fall in each of

the large- and medium-sized markets.

Selection of Advertised Units

Analysis of 2011 American Housing Survey (AHS) data reveals

that most people who moved into a rental housing unit in the

previous year reported they found their current apartment via

word of mouth (34 percent), followed by a sign outside of a

building (11 percent), craigslist (11 percent), and the large

category of “other.”

For the purpose of this study, neither

word of mouth nor signs on buildings were effective means for

identifying the large numbers of available units necessary for

paired testing. Housing providers also use multiple outlets to

advertise their available units, however—especially in online

markets, such as craigslist and other sources.

Table 12. Selected Markets by Size and Olmstead Status

Olmstead Status

Market (MSA) Size

Very Large Large Medium

Olmstead Plan or active Olmstead-

related settlement

Chicago-Joliet-Naperville, IL-IN-WI Cincinnati-Middletown, OH-KY-IN Fresno, CA

New Haven-Milford, CT

Syracuse, NY

No Olmstead Plan

Washington-Arlington-Alexandria,

DC-VA-MD-WV

Nashville-Davidson--Murfreesboro--

Franklin, TN

Albuquerque, NM

Harrisburg-Carlisle, PA

MSA = metropolitan statistical area.

Table 13. Sample Allocation Plan Across Communities and Test Types by Market Size

Matched-

Pair Test

Type

Very Large Markets

(2,500,000 or more)

Large Markets

(1,250,000 to 2,500,000)

Medium Markets

(500,000 to 1,250,000)

Total

Chicago-

Joliet-

Naperville,

IL-IN-WI

Washington-

Arlington-

Alexandria,

DC-VA-MD-

Nashville-

Davidson--

Murfreesboro-

-Franklin, TN

Cincinnati-

Middletown,

OH-KY-IN

Albuquerque,

Fresno, CA

Harrisburg-

Carlisle, PA

New Haven-

Milford, CT/

Syracuse, NY

MI I/DD MI I/DD MI I/DD MI I/DD MI I/DD MI I/DD MI I/DD MI I/DD

Matched-pair tests with reasonable accommodation request

E-mail 10 10 10 10 10 10 10 10 3 3 3 3 2 2 2 2 100

Telephone 50 50 50 50 50 50 50 50 25 25 25 25 25 25 25 25 600

Total 60 60 60 60 60 60 60 60 28 28 28 28 27 27 27 27 700

Matched-pair tests without reasonable accommodation request

E-mail 50 50 50 50 — — — — — — — — — — — — 200

In person 25 25 25 25 — — — — — — — — — — — — 100

Total 75 75 75 75 — — — — — — — — — — — — 300

Total matched-pair tests

E-mail 60 60 60 60 10 10 10 10 3 3 3 3 2 2 2 2 300

Telephone 50 50 50 50 50 50 50 50 25 25 25 25 25 25 25 25 600

In person 25 25 25 25 — — — — — — — — — — — — 100

Total 135 135 135 135 60 60 60 60 28 28 28 28 28 27 27 27 1,000

I/DD = intellectual or developmental disability. MI = mental illness.

Possible detection in one of the markets, the New Haven-Milford, CT Metropolitan Statistical Area (MSA), before the completion of data collection

caused the addition of the Syracuse, NY MSA. Thus, nine total markets were included in the ﬁnal analysis.

The MSA selection process placed priority on including markets consistent with the expert panel’s criteria for sufﬁcient Craigslist coverage, including markets

with and without Olmstead provisions, and on accounting for the number of rental units and vacancy rates. The MSA selection process also considered geographic

representation. Given the total number of markets included in the sampling frame and the need to delineate by Olmstead status for markets with sufﬁcient Craigslist

coverage, however, those factors took precedence over a perfect regional distribution of study markets.

Note that, given the challenges of determining in the AHS if a person has a disability, these data were not further manipulated to produce cross-tabulations by

disability. Also, the data include all people with disabilities, which means they include people with both physical disabilities and MD.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Therefore, the sample of units for possible testing was drawn

primarily from craigslist. Print sources were also included as

possible sources of rental advertising, particularly in markets

where craigslist is not sufﬁcient and local newspapers did

not have an online presence for apartment rental listings to

generate a sufﬁcient sample.

Basing their observations on the literature review and focus

group feedback, the study team noted that people with MD

use a variety of outlets to search for available rental housing,

including mental health and social service agencies. The study

team intentionally sampled from sources used by all renters,

excluding any outlets that speciﬁcally help people with MD get

into subsidized housing, that have access to a pool of landlords

with whom the agency has an existing relationship, or both.

These sources were excluded to obtain a neutral market sample

of units that had not been prescreened to accept people with

disabilities and that would therefore reﬂect the rental search

experience in the general market.

Risk of Detection

A key concern when selecting properties to test was detection

by housing providers that testing was under way in their mar-

ket. For example, testing for discrimination against two sub-

groups (MI and I/DD) within the same market could increase

the risk of detection if the same housing provider is contacted

for each subgroup. Craigslist and local online newspaper apart-

ment rental ads can likewise contain multiple advertisements

from the same housing provider. For example, in housing

markets where realty agencies list available rental units, such

agencies can post multiple advertisements by multiple agents

within the same company. If those agents received inquiries

for different available rental units from testers using the same

or similar proﬁles, they might have detected the testing, which

could have, in turn, affected the results of the study.

One strategy to minimize the likelihood of detection by a housing

provider during testing would have been to sample housing

providers only once within each MSA. This strategy, however,

would have resulted in a dramatic reduction in the pool of

housing providers and testable properties, given that housing

providers may have multiple sites within a market, use a

man agement company or leasing agent that represents many

properties, or both. Instead of adopting this strategy, other

precautions to avoid detection were taken. These precautions

included—

• Using software to track and ﬁlter out duplicate rental

advertisements within a speciﬁed timeframe (for example,

3 weeks).

• Ensuring that no housing provider was tested more than

once for the same type of MD.

• Lengthening the timeframe between tests if an advertised

unit was selected for both MI and I/DD testing.

• Staggering the tests to ensure that the different modes of

testing (e-mail, telephone, and in person) were conducted

sequentially—e-mail ﬁrst, then telephone, then in-person—

with the e-mail and telephone testing in the Chicago and

Washington, D.C. markets administered ﬁrst, before the

in-person testing in those markets began.

Sampling of Units

Sampling of the advertised rental units within MSAs was

accomplished using two methods—one automated and the

other manual.

Under the automated method, to streamline the sampling of

advertised housing units, a data scraper tool was used to select

and collect data from advertisements on craigslist in each study

MSA. The tool ﬁltered the advertisements using the preselected

criteria in table 14, excluding advertisements that did not

Table 14. Criteria for Inclusion in/Exclusion From Unit

Sample

Criteria for inclusion in unit sample

1. Proper unit size (studio, one bedroom, or two bedrooms).

2. Rent within speciﬁed range.

3. Adequate provider contact information available, including—

• Unitaddress(forin-persontesting).

• Telephonenumber(fortelephoneandin-persontesting).

• E-mailaddress(fore-mailtesting).

Criteria for exclusion from unit sample

1. Housing restricted solely to seniors and/or people with disabilities

(for example, single-site housing).

2. Rents outside speciﬁed range (highest and lowest 10% were

thrown out, to exclude misplaced or misleading listings).

3. For in-person testing sample, recent (within 30 days) selection for

e-mail or telephone testing.

Two-bedroom units, a unit size still credibly sought by an individual,

were included because of a limited number of available one-bedroom

units, especially for in-person tests.

A multipronged approach to marketing units is encouraged by various resources for rental housing providers. See, for example, Wagner (2008).

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

provide the rent or other pertinent information for each testing

mode. Table 14 summarizes the criteria for inclusion and

exclusion used in this sampling.

Under the manual sampling method, local newspapers and

other materials featuring rental housing advertisements in the

selected markets were sampled, with every 7seventh post se-

lected. Filters analogous to those identiﬁed for the data scraper

tool were also applied in the manual sampling process.

The automatically and manually mined advertisements were

loaded daily into a Microsoft Access database. Data entry

personnel then reviewed and removed any new advertisement

that met one of the following three conditions: (1) included

data for a housing provider that was already in the database

(phone number, address, agent/owner name, e-mail addresses,

and so on), (2) referenced units that had already been included

in the study sample, or (3) included contact information in the

advertisement that matched any of the contact information for

an advertisement that had been included in the study within

the previous 3 weeks. Information on properties ultimately

selected for testing were then provided weekly to test coordina-

tion teams for each MSA for assignment to testers.

Figure 1 illustrates the data collection process, from the identiﬁ-

cation of sample rental unit advertisements, through the testing,

to the analysis of results.

Figure 1. Flow of Sample Selection and Testing Process

Sample/Testing Process Flow

E-mail testing

Telephone testing

In-person testing

Automated and

manual aggregation

of craigslist local

listing sample

Conduct 200

paired tests

(0.5–1.25M, 1.25–2.5M,

2.5M+ markets)

Conduct 100 paired

tests w/RA request*

(0.5–1.25M, 1.25–2.5M,

2.5M+ markets)

*Protected tester only

Analysis database

Testing data, tester forms,

original housing posts, etc.

Sampling database

Manual aggregation of

other (e.g. newspaper)

local listing sample

Conduct 600 paired

tests

(0.5–1.25M, 1.25–2.5M,

2.5M+ markets)

Conduct 100 paired

tests

(2.5M+ markets only)

RA = reasonable accommodation.

This method of avoiding detection was successful in M. Davis and Company, Inc.’s most recent housing discrimination study (HUD, 2013a).

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Analysis of Testing Findings

Summary of Tests Conducted

During the pilot testing period, 1,128 paired tests were ad-

ministered. Paired testing was conducted using three different

modes of testing protocols: e-mail, telephone, and in-person

testing. Some subset of both telephone tests and e-mail tests

included a request for a reasonable accommodation by the

protected tester. Table 15 summarizes the distribution of tests

by test mode.

Table 15. Total Tests Conducted by Test Mode

Test Mode Tests

E-mail 359

(Subset of e-mail tests with a reasonable accommodation

request)

(78)

Telephone 668

(Subset of telephone tests with a reasonable accommodation

request)

(629)

In person 101

Total tests 1,128

E-mail Testing

Table 16 summarizes the number and distribution of e-mail

tests administered for people with MI and people with I/DD

across three distinct market sizes. Via e-mail, 359 matched-pair

tests were administered, with approximately 73.54 percent of

the tests conducted within the very large markets, 17.5 percent

within the large markets, and the remaining 8.9 percent within

the medium markets.

Figure 2 represents the response rates of the 359 paired e-mail

tests that were conducted during the pilot testing.

Based on the existence, nature, and extent of response from

housing providers, paired testing using an e-mail protocol

enables researchers to observe two different levels of differential

treatment. The ﬁrst level of differential treatment that can be

observed is the disparity between the number of tests for which

the control tester received a response from the housing provider

and the protected tester did not. The lack of a response to the

protected tester when the control tester received a response is

considered in measuring discrimination.

Figure 2. Response Rate of E-mail Paired Tests

Control

tester

received a

response

n = 154

Only control

tester

received a

response

n = 63

Both

testers

received a

response

n = 91

Protected

tester

received a

response

n = 124

No response from provider

n = 172

Some response from provider

n = 187

Total e-mail tests conducted

n = 359

Only

protected

tester

received a

response

n = 33

Table 16. E-mail Sample Allocation by Market Size

Matched-Pair Test Type

Market Size

Total

Very Large

(2,500,000 or More)

Large

(1,250,000 to 2,500,000)

Medium

(500,000 to 1,250,000)

Matched-pair tests

MI 120 15 8 143

I/DD 121 10 7 138

Total 241 25 15 281

Matched-pair tests with request for reasonable accommodation

MI 13 19 8 40

I/DD 10 19 9 38

Total 23 38 17 78

Sum total 264 63 32 359

I/DD = intellectual or developmental disability. MI = mental illness.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

The second level of differential treatment that can be observed

is the difference in how the housing providers interact with the

testers by tracking the incidence of positive treatment, such as

stating whether the unit was available or inviting the potential

renter to contact the housing provider. To be included in the

second-level analysis for differential treatment, both members

of the pair needed to receive an e-mail response within a given

time period. In this study, in 91 paired e-mail tests, both

testers in the pair received a response from the housing pro-

vider within the requisite timeframe. The outcomes observed

in these pairs were used to measure different dimensions of

discrimination against people with MD.

Telephone Testing

Table 17 summarizes the number and distribution of tele-

phone tests conducted for people with MI and I/DD across

three market sizes. Data are differentiated into three sets of

tests: (1) the total number of all matched-pair tests with and

without an accompanying reasonable accommodation request,

(2) the number of matched-pair tests with an accompanying

reasonable accommodation request, and (3) the number of

matched-pair test without a reasonable accommodation request.

Of the 668 telephone tests conducted, the protected tester made

a reasonable accommodation request in 629 (94.2 percent).

Approximately 30.4 percent of the tests were conducted in the

very large markets, 39.4 percent in the large markets, and the

rest in the medium markets.

In-Person Testing

Table 18 summarizes the number and distribution of in-person

matched-pair tests, which were conducted only in the two very

large markets. This mode of testing did not include a reasonable

accommodation request.

Table 18. In-Person Sample Allocation by Location

Matched-Pair

Test Type

Market

Total

Chicago Washington, D.C.

MI 26 24 50

I/DD 26 25 51

Total 52 49 101

I/DD = intellectual or developmental disability. MI = mental illness.

The 101 tests completed exceeded the target goal. The distribu-

tion across disability types was virtually equal, with one extra

test in the I/DD sample. The pairs were matched so that both

testers within a pair had the same race and gender; however,

the total tester pool was not evenly split across race or gender.

Overall, the sample had more women than men for the MI tests

and more men than women for the I/DD tests. MI testing pairs

were also more likely to be African-American, and I/DD testing

pairs were more likely to be Caucasian.

Measuring Differential Treatment

Results from the e-mail, telephone, and in-person tests were

analyzed to determine whether the protected tester with MD

was treated unfavorably relative to the control tester. This

determination was based on information received by the testers

during their interactions with the housing provider, which

were characterized by the research team as either favorable or

unfavorable. Previous studies of housing discrimination have

Table 17. Telephone Testing Allocation by Market Size

Matched-Pair Test Type

Market Size

Total

Very Large

(2,500,000 or More)

Large

(1,250,000 to 2,500,000)

Medium

(500,000 to 1,250,000)

Total matched-pair tests

MI 110 142 105 357

I/DD 92 121 98 311

Total 202 263 203 668

Matched-pair tests with reasonable accommodation requests

MI 106 128 105 339

I/DD 88 108 94 290

Total 194 236 199 629

Matched-pair tests without reasonable accommodation requests

MI 4 15 0 19

I/DD 4 12 4 20

Total 8 27 4 39

I/DD = intellectual or developmental disability. MI = mental illness.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

shown that housing providers can respond in various favorable

or unfavorable manners to an applicant. In this study, the

following indicators of favorable or unfavorable responses that

testers could receive included—

• Whether testers received a response to their inquiry, for

the ﬁrst level of unfavorable treatment testing (e-mail

testing mode).

• If testers were told the advertised unit or similar unit in

the same building or development was available.

• How many available units testers were told about.

• If testers were invited to inspect the unit.

• If testers were advised to call the housing provider (e-mail

only).

• Whether testers were asked to provide additional informa-

tion regarding their qualiﬁcations as an applicant (that is,

their credit score/income/employment).

• Whether testers were reminded about qualiﬁcations they

must possess to rent the unit.

• Whether the unit was available and, if not, whether testers

were given a reason for the unit not being available.

• If testers were given the same information on rents,

security deposits, application fees, or other fees.

• If testers were given the same information on specials or

incentives to rent.

• If testers were given the same information on amenities

and services available.

• Whether testers were encouraged to look at a different unit

owned by the same landlord (that is, “This unit actually isn’t

available, but I have another unit in the same building you

might be interested in”).

• Whether testers were referred to a different complex altogether.

• Whether the housing provider was willing to grant a

reasonable accommodation (only when applicable).

Table 19 summarizes a treatment indicator matrix, showing the

behavior observed in the testing that was assessed to determine

how the outcomes data from all tests were analyzed by testing

mode—e-mail, telephone, and in person. Although some indi-

cators are shared by all three modes, each also has indicators

unique to that mode. The data were broken out into four categories:

(1) contact with housing providers, (2) availability, (3) tester

engagement, and (4) reasonable accommodation, each of which

are described in detail in Table 19.

Table 19. Differential Treatment Indicator Matrix (1 of 2)

Outcome

Category

Outcome

Subcategory

Test Modality

E-mail Telephone In Person

Tester

engagement

Unit selection

method

Method(s) for selection of unit(s) to view

Agent or tester selects unit(s) to view

Details of

discussed unit(s)

Address Address Address

Apt. number Apt. number Apt. number

Type Type Type

Number of bedrooms Number of bedrooms Number of bedrooms

Rent Rent Rent

Security deposit Security deposit Security deposit

Date of availability Date of availability Date of availability

Fees Fees Fees

Apartment shown

Apartment condition ready for move in

(assessed only for vacant units)

Total number of shown apartments that meet

tester needs

Information

disclosed by

agent/housing

provider

Application mentioned Application mentioned Application mentioned

Application offered Application offered Application offered

Duration of application

process

Duration of application

process

Duration of application process

Co-signer mentioned? Co-signer mentioned? Co-signer mentioned?

Security deposit? Security deposit? Security deposit?

Other fees Other fees Other fees

Amenities Amenities Amenities

Incentives Incentives Incentives

Encouraging/discouraging

remarks

Encouraging/discouraging

remarks

Encouraging/discouraging remarks

Disability comments Disability comments Disability comments

Referred to alternative

residence type

Referred to alternative

residence type

Referred to alternative residence type

Comments/questions

on race, color, national

origin, religion, sex, disability,

children, age, marital status,

source of income, or sexual

orientation

Comments/questions

on race, color, national

origin, religion, sex, disability,

children, age, marital status,

source of income, or sexual

orientation

Comments/questions on race, color, national

origin, religion, sex, disability, children, age,

marital status, source of income, or sexual

orientation

Future contact Future contact Future contact

Unit location comments Unit location comments Unit location comments

Other

qualiﬁcations

Escorted to apts

Proof of citizenship prior to showing

Require personal ID Require personal ID Require personal ID

Housing provider/agent

indicated applicant is “not

qualiﬁed”

Housing provider/agent

indicated applicant is “not

qualiﬁed”

Housing provider/agent indicated applicant is

“not qualiﬁed”

References References References

Credit check (as part of

application)

Credit check (as part of

application)

Credit check (as part of application)

Criminal background check Criminal background check Criminal background check

Reasonable

accommodation

Housing provider/agent

responded to request?

Housing provider/agent

responded to request?

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Data Analysis

Analysis of the responses of housing providers observed during

the testing followed the approach used in the 2000 Housing

Discrimination Study (HUD, 2002) and DDS 2005 studies,

which assessed the extent to which one tester is consistently

favored over the other in the treatment received from housing

providers based on their inquiries. For example, tests are clas-

siﬁed as “control tester favored” if the control tester received

favorable treatment on at least one indicator of interest and the

protected tester received no favorable treatment on the same

indicator(s).Tests are conversely classiﬁed as “protected tester

favored” if the protected tester received favorable treatment on

at least one indicator of interest and the control tester received

no favorable treatment on the same indicator(s).

Typical treatment indicators that have been tracked in past

housing discrimination studies, including HUD (2013a), a

paired e-mail testing study, include whether each tester (1)

received a response, (2) in a pair received a response, (3) was

told the unit was available, (4) was told to contact the provid-

er, and (5) was invited to inspect the unit. In addition to track-

ing these ﬁve indicators, this study explored other indicators

related to quality of applicant, costs, and encouragement using

a validated qualitative approach to analyze e-mail data (see

Appendix A, Qualitative Analyses and Findings on Overt and

Subtle Discrimination Against People With Mental Disabilities

by Different Testing Modes).

The testing conducted in this study estimates the incidence

of discrimination by studying favorable and unfavorable out-

comes for each test dimension. A favorable outcome represents

a response in which the tester received an afﬁrmative value on

the test dimension, and an unfavorable outcome represents a

response in which the tester received a negative value on the

test dimension. Each test had four possible combinations of

outcomes: (1) both the control and protected testers received

favorable outcomes, (2) both testers received unfavorable

outcomes, (3) only the control tester received a favorable

outcome, or (4) only the protected tester received a favorable

outcome. The relationships among the four combinations are

summarized in Table 20.

Table 20. Possible Outcomes Matrix

Possible Outcomes

Protected

Favorable Unfavorable

Control Favorable

(1) Both (3) Control favored

Unfavorable

(4) MI or I/DD favored (2) Both

I/DD = intellectual or developmental disability. MI = mental illness.

Discrimination is assumed when the proportion of tests that

favors control testers is greater than the proportion that favors

protected testers. The difference in the percentages between

outcomes 3 and 4 represents the net difference in outcomes.

The value of the net difference is what is assessed in the

hypothesis test. The hypothesis is declared formally in the

following equations.

: %

= %

or %

– %

= 0.

: %

> %

or %

– %

> 0.

Measuring Discrimination

In this study, discrimination was estimated separately by test

mode (e-mail, telephone, and in person) for each disability

group. The measure of discrimination is expressed as—

D = p

– p

In this expression, p

is the proportion of pairs in which

the control applicant receives a favorable outcome and the

applicant with disabilities receives an unfavorable outcome,

and p

is the proportion of pairs in which the control applicant

receives a negative outcome and the applicant with disabilities

receives a positive outcome. The standard error of D is—

+ n

– p

SE (D)

+ n

– n

)

( |n

– n

| – 1)

– p

where n is the total number of pairs. Given that D is approx-

imately normally distributed, we may compute conﬁdence

intervals using—

+ n

– p

SE (D) =

+ n

– n

)

( |n

– n

| – 1)

– p

where Z

is the appropriate critical value of the standard normal

distribution.

The test statistic for McNemar’s Test is—

+ n

– p

SE (D) =

+ n

– n

)

( |n

– n

| – 1)

– p

In this expression, n

is the number of pairs in which the con-

trol tester receives a favorable outcome and the protected tester

receives an unfavorable outcome, and n

is the number of pairs

in which the control tester receives an unfavorable outcome and

the protected tester receives a favorable outcome. Under the null

hypothesis of no discrimination, with a sufﬁciently large number

of discordant samples (n

and n

), the test statistic has an

asymptotic chi-squared distribution with one degree of freedom.

If either n

or n

is small (n

10 +

< 25), then an exact binomial

test can be used where n

is compared to the binomial distribution

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

with size parameter n = n

+ n

and probability 0.5.

The

binomial exact p-value can be approximated with the following

continuity-corrected version of McNemar’s Test.

+ n

– p

SE (D) =

+ n

– n

)

( |n

– n

| – 1)

– p

The test statistic has an asymptotic chi-squared distribution

with one degree of freedom. In this study, a one-tailed version

of this test was done by dividing the two-tailed p-value from the

chi-squared or exact binomial test by 2. The estimates of inci-

dences of discrimination are provided at 99-, 95-, and 90-percent

levels of signiﬁcance by protected subclass (MI or I/DD).

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Findings of Incidence of Discrimination

This section presents the data generated by the three testing

modes (e-mail, telephone, and in person) and summarizes

ﬁndings from the tests using the dimensions described in the

previous section to determine the incidence of discrimination

identiﬁed in each mode.

Data Reliability and Limitations

Each mode of testing is examined by different dimensions of

discrimination. In cases for which the matched-pair sample

size was sufﬁcient, data were further cross-tabulated by (1)

market size, (2) Olmstead status (that is, whether the pair was

in a state with an Olmstead Plan or Olmstead-related litigation),

(3) type of disability (MI or I/DD), (4) gender, and (5) race.

E-mail Testing Findings

The analysis of the e-mail tests focused on the differences in

the responses of housing providers to people with MD (MI and

I/DD samples combined = protected class) compared with peo-

ple without disabilities (control) across the following indicators

on two different levels.

1. Level 1 (applicable to administered tests in which both

testers sent an e-mail in inquiry; n = 359).

• Did each tester receive a response to his or her

inquiry (favorable)?

2. Level 2 (applicable only to completed tests in which both

testers received a response; n = 91).

• Were both testers told the advertised unit is available

(favorable)?

• Were both testers invited to inspect the unit (favorable)?

• Was the unit available and, if not, were both testers

given a reason for the unit not being available (favor-

able)?

• Were both testers advised to call the housing provider

(favorable)?

• Were protected testers asked to provide additional

information regarding their qualiﬁcations as an appli-

cant (not favorable)?

• Were protected testers reminded about qualiﬁcations

they must possess to rent the unit (not favorable)?

• Were protected testers encouraged to look at a differ-

ent unit owned by the same landlord instead of the

unit advertised (not favorable)?

The ﬁrst step in the statistical analysis was to determine whether

testers from the control group were more likely to receive a

response to the initial inquiry than their matched protected

testers. Figure 2 summarizes the number and distribution of

e-mail responses versus nonresponses to the initial rental inqui-

ries from testers both with and without MD. Of the 359 tests

administered, only 91 tests (25 percent) resulted in responses

to both testers in the pair, and, in 172 tests (48 percent), neither

tester received a response. Thus, in 73 percent of the e-mail

tests, the control and protected group received equal treatment.

Table 21 presents the results of this ﬁrst level of analysis. As

shown, people with MD (MI and I/DD combined) were signiﬁ-

cantly less likely to receive any response back to their rental

inquiry via e-mail (17.55 percent of tests favored the control

class compared with 9.19 percent that favored the protected

class). The difference in these two percentages captures the net

measure of preferential treatment given to the control testers

Table 21. Differential Treatment in Response Received for Renters With MD (I/DD+MI) Found in E-mail Testing

Test Dimension

Both

Testers

(%)

Neither

Tester

(%)

Control

Tester

Only

(%)

Protected

Tester

Only

(%)

Net

Measure

p-Value

Sample

Size

(n)

Which tester(s) received a response to their e-mail? 25.35 47.91 17.55 9.19 8.36*** 0.0011 359

I/DD = intellectual or developmental disability. MD = mental disability. MI = mental illness.

*** Signiﬁcant at the .01 level.

Throughout the Findings of Incidence of Discrimination section, in the event the sample size or matched-pair data available were not sufﬁcient to yield conclusive

statistical results, this limitation is documented and indicates the need for additional testing with larger sample sizes in the future.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

(8.36 percent).

This positive net measure (8.36 percent)

suggests that people with MD (MI and I/DD combined) were

signiﬁcantly less likely to receive a response back to their rental

inquiry and that a greater percentage of people in the control

group were favored. Moreover, the resulting p-value indicates

that the difference in treatment between control and protected

classes is signiﬁcant.

A second level of analysis considers several dimensions of the

response to the different testers, and thus only those tests in

which both testers within a pair received a response (n = 91,

or 25 percent of the administered tests) can be included in the

analysis. Table 22 demonstrates that of the 91 tests in which

both the control tester and the paired protected tester received

a reply, protected testers identifying as having MD were less

likely to be forwarded to a housing provider to make an ap-

pointment or to see the unit (7.69 percent favored the control

class compared with 0.00 percent that favored the protected class).

As described in the Sampling Methods section, the protected

tester identiﬁed himself or herself as an individual with either

MI or I/DD. Figure 3 represents the response rates of the 183

paired e-mail tests that were conducted by protected testers

identifying as a person with MI.

Table 23 documents the results of the ﬁrst level analysis to

determine whether testers from the control group were more

likely to receive a response to the initial inquiry than their

matched protected tester. Of the 183 paired tests completed in

which the protected testers revealed that they were people with

MI (as opposed to I/DD), only 42 (22.95 percent) of the tests

resulted in responses to both testers in the pair, and 93 (50.82

percent) of the tests resulted in no response to either tester.

Figure 3. Response Rates of E-mail Paired Tests for

Protected Testers Identifying as a Person With MI

No response from provider

n = 93

Some response from provider

n = 90

Total MI tests conducted

n = 183

Control

tester

received a

response

n = 79

Only control

tester

received a

response

n = 37

Both

tester

received a

response

n = 42

Only

protected

tester

received a

response

n = 11

Protected

tester

received a

response

n = 53

MI = mental illness.

Thus, in 74 percent of the e-mail tests for which the tester

revealed MI, the control and protected groups received similar

treatment. If the responses were different within test pairs,

however, control testers were favored at a rate of 20.22 percent

compared with only 6.01 percent for the protected testers. The

resulting net measure of 14.21 percent and p-value of 0.0001

demonstrate that these differences are statistically signiﬁcant.

Table 22. Summary of Differential Treatment of Renters With MD (I/DD+MI) Found in E-mail Testing

Test Dimension

Both

Testers

(%)

Neither

Tester

(%)

Control

Tester

Only

(%)

Protected

Tester

Only

(%)

Net

Measure

p-Value

Sample

Size

(n)

Unit available 28.57 56.04 8.79 6.59 2.20 0.3953 91

Invited to inspect 72.94 11.76 7.06 8.24 – 1.18 0.5000 85

Given reason for unit not available 28.57 53.85 8.79 8.79 0.00 0.5000 91

Contact housing provider 41.76 50.55 7.69 0.00 7.69*** 0.0078 91

Asked for additional qualiﬁcation information 96.70 2.20 1.10 0.00 1.10 0.5000 91

Reminded about qualiﬁcations to rent 95.60 3.30 1.10 0.00 1.10 0.5000 91

Encouraged to look at different unit than one advertised 6.59 93.41 0.00 0.00 0.00 NA 91

I/DD = intellectual or developmental disability. MD = mental disability. MI = mental illness. NA = test statistics cannot be produced due to no

observations in both discordant pairs.

*** Signiﬁcant at the .01 level.

The sample size of 85 is due to the lack of response data for one of the pairs involved for this particular test dimension. The set of pairs without

comparable responses cannot be used in the analysis.

A positive net difference (control – protected net difference) means a greater percentage of people in the control class were favored, whereas a negative net differ-

ence would have meant a greater percentage of people in the protected class were favored. The p-value shows whether differences noted were signiﬁcant and, if so, at

what level.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Table 23. Differential Treatment in Response Received for Renters With MI Found in E-mail Testing

Test Dimension

Both

Testers

(%)

Neither

Tester

(%)

Control

Tester

Only

(%)

Protected

(MI) Tester

Only

(%)

Net

Measure

p-Value

Sample

Size

(n)

Which tester(s) received a response to their e-mail? 22.95 50.82 20.22 6.01 14.21*** 0.0001 183

MI = mental illness.

*** Signiﬁcant at the .01 level.

Because only 42 completed tests received responses from the

housing provider to both the control and the protected (MI)

testers, subsequent indicators of favorable or unfavorable treat-

ment could be analyzed using only these 42 tests. The results

summarized in Table 24 of the second level analysis show that

people with MI, when compared with the control tester, were

less likely to be referred to the housing provider to actually see

the unit or to make an appointment to do so.

When these same analyses were conducted on the set of com-

pleted tests in which protected testers identiﬁed themselves as

people with I/DD rather than people with MI, no signiﬁcant

differences were observed between the treatment of the

protected and control testers. This ﬁnding suggests potentially

different treatment of people by diagnosis or type of MD.

Differences in treatment were also noted based on other

factors. Despite small sample sizes, notable trends were shown

in the following areas.

• Size of market—In the very large markets, people with

MD were less likely to receive a reply to their inquiries

and less likely to be forwarded to a housing provider to set

up an appointment to see the advertised unit than in other

size markets. This lower level of response in Chicago and

Washington, D.C., was noted particularly with respect to

people with MI.

• Gender—Both men and women with MI were less likely

to receive a reply to their inquiries relative to paired

control testers, but unfavorable ﬁndings were higher in

signiﬁcance for men (signiﬁcant at the .001 level for men

compared with the .017 level for women). To the contrary,

treatment differences based on gender were not found for

testers identifying as having an I/DD.

• Current living situation (whether the person was living

in a community or institutional setting before the

rental request)—Overall, protected testers with MD were

less likely to receive a reply to their e-mail inquiries than

were control testers. Between the two different scenarios

presented by the protected testers—either moving from

an institutional setting (Olmstead) or moving from a family

home or other community setting (non-Olmstead)—protected

testers who indicated they were moving from an institutional

setting were even less likely to receive a response to their

request (signiﬁcant at the .004 level for Olmstead versus the

.090 level for non-Olmstead).

Reasonable Accommodation Requests

via E-mail

Within the total pool of 359 attempted e-mail tests, 78 paired

tests included a speciﬁc request from the protected tester for a

reasonable accommodation (the control tester in the pair did

not make a reasonable accommodation request). These rea-

sonable accommodation requests, such as use of an assistance

animal or the need for verbal reminders to pay rent at a certain

time, were designed to relate directly to people with MD. Of

the 78 housing providers who received e-mail inquiries that

Table 24. Summary of Differential Treatment of Renters With MI Found in E-mail Testing

Test Dimension

Both

Testers

(%)

Neither

Tester

(%)

Control

Tester

Only

(%)

Protected

(MI) Tester

Only

(%)

Net

Measure

p-Value

Sample

Size

(n)

Unit available 21.43 61.90 9.52 7.14 2.38 0.5000 42

Invited to inspect 70.00 5.00 12.50 12.50 0.00 0.5000 40

Given reason for unit not available 21.43 59.52 9.52 9.52 0.00 0.5000 42

Contact housing provider 45.24 45.24 9.52 0.00 9.52* 0.0625 42

Asked for additional qualiﬁcation information 100.00 0.00 0.00 0.00 0.00 NA 42

Reminded about qualiﬁcations to rent 95.24 4.76 0.00 0.00 0.00 NA 42

Encouraged to look at different unit than one advertised 4.76 95.24 0.00 0.00 0.00 NA 42

MI = mental illness. NA = test statistics cannot be produced due to no observations in both discordant pairs.

* Signiﬁcant at the .10 level.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

included a reasonable accommodation request, only 39 (50

percent) sent e-mail responses back to the protected testers. Of

those housing providers who did respond, roughly 15 percent

were willing to grant the requested accommodation.

Table 25 summarizes the results of the tests requesting a

reasonable accommodation. Only 6 of the 39 housing pro-

viders who responded (15.384 percent) responded favorably

to the reasonable accommodation request.

An additional 3

housing providers (7.69 percent) responded to the request for

a reasonable accommodation but in an unfavorable manner,

either by stating that they were not willing to provide the

requested accommodation or by giving a more ambiguous

reply that left the impression that the request for a reasonable

accommodation was not likely to be approved. The remaining

30 responding housing providers simply did not address the

request for a reasonable accommodation. These housing pro-

viders’ avoidance of addressing the reasonable accommodation

request was considered to be unfavorable treatment.

Table 25. Summary of Reasonable Accommodation

Requests by Testers With MD Found in E-mail Testing

Type of

Disability

Reasonable Accommodation Requests

Response

Received

Willing To

Provide (n)

Willing To

Provide (%)

MI 21 3 14.29

I/DD 18 3 16.67

Total 39 6 15.384

I/DD = intellectual or developmental disability. MD = mental disability.

MI = mental illness.

Overall, the vast majority of people with MD making an e-mail

request for a reasonable accommodation (84.6 percent) were

denied their request outright or were left, on their own, to

determine how to pursue their request or to appeal a housing

provider’s denial.

Additional qualitative analyses of e-mail replies to accommoda-

tion requests are provided in Appendix A, Qualitative Analyses

and Findings on Overt and Subtle Discrimination Against

People With Mental Disabilities by Different Testing Modes.

Telephone Testing Findings

This section summarizes ﬁndings from the telephone testing to

measure discrimination against people with MD as they seek

rental housing. Responses to tester inquiries were analyzed by

assessing the following responses to testers.

• Were both testers told the advertised unit is available

(favorable)?

• Were both testers invited to inspect the unit (favorable)?

• Was the unit available and, if not, were both testers given a

reason for the unit not being available (favorable)?

• Were protected testers asked to provide additional

information regarding their qualiﬁcations as an applicant

(not favorable)?

• Were protected testers reminded about qualiﬁcations they

must possess to rent the unit (not favorable)?

• Were protected testers encouraged to look at a different

unit owned by the same landlord than the unit advertised

(not favorable)?

Analyses focus on differences in treatment between people with

MD (MI and I/DD) and control testers without disabilities. As

shown in Table 26, no signiﬁcant difference was identiﬁed with

respect to whether potential renters with or without MD are told

whether an advertised unit is available, although the trend sug-

gests that people with MD are less likely to be told an advertised

Table 26. Summary of Differential Treatment of Renters With MD Found in Telephone Testing

Test Dimension

Control

Tester Only

(%)

Protected

Tester Only

(%)

Net

Measure

p-Value

Sample Size

(n)

Unit available 6.59 4.94 1.65 0.1050 668

Invited to inspect 21.26 16.47 4.79** 0.0219 668

Given reason for unit not available 5.39 5.39 0.00 0.5000 668

NOT asked for additional qualiﬁcation information 19.71 20.99 – 1.28 0.33085 629

NOT reminded about qualiﬁcations to rent 16.18 24.76 – 8.58 0.0004

618

NOT encouraged to look at different unit than one advertised 7.78 4.79 2.99** 0.0188 668

MD = mental disability.

** Signiﬁcant at the .05 level.

The null hypothesis is not rejected as test statistics indicate signiﬁcance in the other direction.

The testing demonstrated slightly higher levels of willingness to provide accommodations for people with I/DD (16.7 percent) versus people with MI (14.3 percent).

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

unit is available than people without disabilities (p = .105).

People with MD (both I/DD and MI), however, were treated

signiﬁcantly less favorably in two ways.

1. Protected testers were less likely to be invited to inspect an

advertised unit.

2. Protected testers were more likely to be steered to a

different unit than the one advertised, even when the

advertised unit was available.

Additional instances of signiﬁcant differences in treatment

were found during telephone testing when the tests were ana-

lyzed based on the type of disability revealed by the protected

tester. As shown in Table 27, when speciﬁcally looking at the

sample of testers with MI (n = 357), testers with MI were—

• Signiﬁcantly less likely to be told an advertised unit was

available than the control tester.

• Signiﬁcantly less likely to be invited to inspect an advertised

unit and signiﬁcantly more likely to be forwarded to a

different unit than the one advertised, even when the

advertised unit was said to be available.

• Signiﬁcantly more likely to be asked to provide additional

information about their qualiﬁcations.

• Signiﬁcantly more likely to be sent ambiguous signs of

rentals not being available.

Analysis of the tests conducted for people with I/DD (n = 311),

revealed a signiﬁcant unfavorable difference in that people with

I/DD were more likely to be forwarded to a different unit than

the one advertised, even when the advertised unit was said to

be available (Table 28). This ﬁnding suggests that people with

I/DD are being steered toward a speciﬁc building, wing, or area

within a property.

Table 27. Summary of Differential Treatment of Renters With MI Found in Telephone Testing

Test Dimension

Control

Tester Only

(%)

Protected (MI)

Tester Only

(%)

Net

Measure

p-Value

Sample Size

(n)

Unit available 7.56 4.48 3.08** 0.0467 357

Invited to inspect 19.61 14.01 5.60** 0.0339 357

Given reason for unit not available 5.60 5.04 0.56 0.3728 357

NOT asked for additional qualiﬁcation information 22.87 17.68 5.19* 0.0702 328

NOT reminded about qualiﬁcations to rent 15.63 25.00 – 9.37 0.0043

320

Sent ambiguous sign of availability 7.56 5.04 2.52* 0.0899 357

NOT encouraged to look at different unit than one advertised 7.28 4.20 3.08** 0.0429 357

MI = mental illness.

* Signiﬁcant at the .10 level. ** Signiﬁcant at the .05 level.

The null hypothesis is not rejected as test statistics indicate signiﬁcance in the other direction.

Table 28. Summary of Differential Treatment of Renters With I/DD Found in Telephone Testing

Test Dimension

Control

Tester Only

(%)

Protected (I/DD)

Tester Only

(%)

Net

Measure

p-Value

Sample Size

(n)

Unit available 5.47 5.47 0.00 0.5000 311

Invited to inspect 23.15 19.29 3.86 0.1481 311

Given reason for unit not available 5.14 5.79 – 0.65 0.3658 311

NOT asked for additional qualiﬁcation information 16.28 24.58 – 8.30 0.0121 301

NOT reminded about qualiﬁcations to rent 16.78 24.50 – 7.72 0.0190 298

Sent ambiguous sign of availability 5.47 5.47 0.00 0.5000 311

NOT encouraged to look at different unit than one advertised 8.36 5.47 2.89* 0.0850 311

I/DD = intellectual or developmental disability.

* Signiﬁcant at the .10 level.

This ﬁnding suggests that people with MI are being steered toward a different building or a speciﬁc wing or area within a building.

See Appendix A: Qualitative Analyses and Findings on Overt and Subtle Discrimination Against People With Mental Disabilities by Different Testing Modes, for

examples of this subtle discrimination.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Telephone testing revealed that people with MI and I/DD

were less likely to be reminded about qualiﬁcations or asked to

provide additional information about their qualiﬁcations than

people without disabilities.

Further testing and qualitative

debrieﬁng of housing providers would be needed to determine

whether this indicator is favorable or unfavorable.

The following differences by size of market, gender, and previous

living status were also observed as trends in telephone testing.

• Size of market—People with both MI and I/DD were

treated less favorably in large- and medium-sized markets,

as compared with very large markets, in several areas,

including (1) they were less likely to be told an advertised

unit is available, (2) they were less likely to be invited to

inspect that unit, and (3) they were encouraged more to

look at a different unit than the one advertised.

• Gender—By contrast with e-mail ﬁndings, in telephone

testing, women with MI were (1) less likely to be invited in

to view the unit (22.70 percent of tests favored control-group

women compared with 6.38 percent that favored women

with MI; p = .0002), (2) less likely to be asked to provide

additional information regarding their qualiﬁcations as an

applicant (32.14 percent favored control-group women

compared with 11.43 percent that favored women with

MI; p = .0001), and (3) more likely to be encouraged

to look at a different unit owned by the same landlord

(14.18 percent favored control-group women compared

with 7.09 percent that favored in women with MI; p = .034).

Men with MI were less likely to be asked about qualiﬁca-

tions or any additional information than women with MI.

Signiﬁcant gender differences were not found for people

with I/DD.

• Current living situation—Protected testers who indicat-

ed to the housing provider that they had been living in an

institutional setting before renting (Olmstead) were (1) less

likely to receive a reply to their inquiries, (2) less likely

to be invited to inspect a unit, and (3) more encouraged

to look at a different unit than the one advertised, when

compared with the protected testers who reported to

providers that they were living in a family home or

community-based settings before renting (non-Olmstead).

Reasonable Accommodation Requests

via Telephone

During telephone testing, if protected testers were invited to see

a unit, a request for a reasonable accommodation was made. A

total of 629 telephone tests included a request for a reasonable

accommodation from the protected tester. Table 29 summarizes

the results of the responses to these requests.

Table 29. Summary of Reasonable Accommodation

Requests by Testers With MD Found in Telephone

Testing

Type of

Disability

Reasonable Accommodation Requests

Response

Received

Willing To

Provide (n)

Willing To

Provide (%)

MI 339 187 55.16

I/DD 290 185 63.79

Total 629 372 59.14

I/DD = intellectual or developmental disability. MD = mental disability.

MI = mental illness.

By contrast with ﬁndings from the e-mail tests (in which only

15.384 percent of housing providers expressed willingness

to provide a reasonable accommodation when asked), 59.14

percent of the housing providers who were asked to provide

reasonable accommodations during telephone testing stated

that they would be willing to make the accommodation or

would view the request favorably given disclosure of the

disability. This ﬁnding may indicate that housing providers

are more likely to engage in a conversation and openly discuss

accommodations on the telephone than in an e-mail. Telephone

testing still showed a signiﬁcant percentage of accommodation

requests that were not accepted (40.86 percent), however. The

remaining housing providers either stated outright that they

were not willing to provide the accommodation or gave more

ambiguous replies that more subtly pointed toward a negative

decision.

Thus, a large percentage of people with MD (40.86

percent) were denied their request for a reasonable accommo-

dation or left to ﬁgure out how to pursue that request on their

own given negative input in response to their request.

In addition, reasonable accommodation response rates differed

by type of disability (see Table 29), revealing that a higher

percentage of housing providers were willing to provide

This ﬁnding may suggest a favorable ﬁnding that housing providers know not to ask these types of questions, or it may suggest unfavorable treatment in that

housing providers may be less likely to ask for any detailed rental information from applicants with MI and I/DD after having already shown less chance of inviting

them to see or inspect a unit.

For example, many housing providers or their agents stated that they did not think the owner would allow the requested accommodation or that it was unlikely

the request would be considered, and they said that the applicant would need to ﬁnd and call a representative at the housing provider to verify this information but

did not give any contact information to do so.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

accommodations to people with I/DD (63.79 percent) than

to people with MI (55.16 percent). Requests for a reasonable

accommodation were also more likely to be considered and

granted in large markets (65.7 percent) and medium markets

(66.7 percent) than in very large markets (47.4 percent) or

small markets (41.9 percent). This analysis again produced

a higher acceptance rate for people with I/DD than with MI,

regardless of market size.

Table 30 summarizes the varying levels of afﬁrmative

responses to requests for reasonable accommodations within

different- sized markets tested. Given small sample sizes tested,

particularly in smaller city markets, results are limited to show-

ing only potential trends. This limit demonstrates the need for

larger-scale testing to determine signiﬁcant differences between

metropolitan statistical areas based on market size.

Additional qualitative analyses of telephone responses to

requests for reasonable accommodations are provided in

Appendix A, Qualitative Analyses and Findings on Overt and

Subtle Discrimination Against People With Mental Disabilities

by Different Testing Modes.

Table 30. Summary of Reasonable Accommodation Request Responses by Speciﬁc Markets

Type of

Disability

Counts

Willing To

Accommodate

(%)

99% C.I. 95% C.I. 90% C.I.

Total

Requested

Willing To

Accommodate

Lower

(%)

Upper

(%)

Lower

(%)

Upper

(%)

Lower

(%)

Upper

(%)

Chicago-Joliet-Naperville, IL-IN-WI

MI 55 18 32.7 16.4 49.0 20.3 45.1 22.3 43.1

I/DD 46 21 45.7 26.7 64.6 31.3 60.0 33.6 57.7

MI+ I/DD 101 39 38.6 26.1 51.1 29.1 48.1 30.6 46.6

Washington-Arlington-Alexandria, DC-VA-MD-WV

MI 51 24 47.1 29.1 65.1 33.4 60.8 35.6 58.6

I/DD 42 29 69.0 50.7 87.4 55.1 83.0 57.3 80.8

MI+ I/DD 93 53 57.0 43.8 70.2 46.9 67.1 48.5 65.4

Nashville-Davidson--Murfreesboro--Franklin, TN

MI 53 35 66.0 49.3 82.8 53.3 78.8 55.3 76.7

I/DD 53 37 69.8 53.6 86.1 57.5 82.2 59.4 80.2

MI+ I/DD 106 72 67.9 56.2 79.6 59.0 76.8 60.5 75.4

Cincinnati-Middletown, OH-KY-IN

MI 75 48 64.0 49.7 78.3 53.1 74.9 54.9 73.1

I/DD 55 35 63.6 46.9 80.3 50.9 76.3 53.0 74.3

MI+ I/DD 130 83 63.8 53.0 74.7 55.6 72.1 56.9 70.8

Albuquerque, NM

MI 26 19 73.1 50.7 95.5 56.0 90.1 58.8 87.4

I/DD 23 18 78.3 56.1 100.0 61.4 95.1 64.1 92.4

MI+ I/DD 49 37 75.5 59.7 91.3 63.5 87.6 65.4 85.6

Fresno, CA

MI 27 17 63.0 39.0 86.9 44.7 81.2 47.7 78.2

I/DD 24 21 87.5 61.3 98.5 67.6 97.3 70.8 96.5

MI+ I/DD 51 38 74.5 58.8 90.2 62.5 86.5 64.5 84.5

Harrisburg-Carlisle, PA

MI 28 14 50.0 25.7 74.3 31.5 68.5 34.5 65.5

I/DD 25 14 56.0 30.4 81.6 36.5 75.5 39.7 72.3

MI+ I/DD 53 28 52.8 35.2 70.5 39.4 66.3 41.6 64.1

New Haven-Milford, CT

MI 10 6 60.0 19.1 92.3 26.2 87.8 30.4 85.0

I/DD 5 3 60.0 8.3 97.7 14.7 94.7 18.9 92.4

MI+ I/DD 15 9 60.0 27.4 92.6 35.2 84.8 39.2 80.8

Syracuse, NY

MI 17 7 41.2 10.4 71.9 17.8 64.6 21.5 60.8

I/DD 14 6 42.9 8.8 76.9 16.9 68.8 21.1 64.6

MI+ I/DD 31 13 41.9 19.1 64.8 24.6 59.3 27.4 56.5

C.I. = conﬁdence interval. I/DD = intellectual or developmental disability. MI = mental illness.

Added after a potential detection issue was discovered in New Haven-Milford, CT.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

In-Person Testing Findings

This section summarizes ﬁndings from in-person testing to

measure discrimination against people with and without MD

as they seek rental housing. Key questions that form the basis

for analysis include—

• Were both testers told the advertised unit is available

(favorable)?

• Was the unit available and, if not, were both testers given

a reason for the unit not being available (favorable)?

• Were protected testers asked to provide additional

information regarding their qualiﬁcations as an applicant

(not favorable)?

• Were protected testers reminded about qualiﬁcations they

must possess to rent the unit (not favorable)?

• Were protected testers encouraged to look at a different

unit owned by the same landlord than the unit advertised

(not favorable)?

The ﬁrst analysis focuses on differences between testers with

MD (MI and I/DD samples combined) and testers without

disabilities. As shown in Table 31, people with MD were sig-

niﬁcantly less likely than people without disabilities to be told

that the advertised unit was available (5.94 percent favoring

the control tester compared with 0.99 percent favoring the

protected tester) and were less likely to be given a valid reason

for why the unit was not available (3.96 percent favoring

the control tester compared with 0.00 percent favoring the

protected tester). No other signiﬁcant differences in treatment

were identiﬁed during the in-person testing.

The sample size varies if data were missing or not available for

one of the matched pairs; only balanced matched-pair data

were analyzed.

As in the other modes of testing, when additional characteristics

of testers were considered in the analysis, evidence of differ-

ential treatment based on disability type, current living status,

region, and gender were observed.

As shown in Table 32, people with MI were less likely to be

told that an advertised unit is available (8 percent of tests

favored the control group compared with 0 percent that favored

the MI group) during in-person testing.

No signiﬁcant differences in treatment were found between tes-

ters with I/DD and control testers without MD (see Table 33).

The following additional differences by gender, race, and living

situation were also found to be signiﬁcant in in-person testing.

• Gender—Women with MI were less likely to be told

an advertised unit was available (11.43 percent of tests

favored women without disabilities compared with 0

percent that favored women with MI; p = .06) and more

likely to be encouraged to look at a different unit than

the one available (20 percent favored women without

disabilities compared with 5.71 percent that favored in

women with MI; p = .09) during the in-person testing.

Table 31. Summary of Differential Treatment of Renters With MD Found in In-Person Testing

Test Dimension

Control

Tester Only

(%)

Protected

Tester Only

(%)

Net

Measure

p-Value

Sample Size

(n)

Unit available 5.94 0.99 4.95* 0.0625 101

Given reason for unit not available 3.96 0.00 3.96* 0.0625 101

NOT asked for additional qualiﬁcation information 16.49 21.65 – 5.16 0.2055 97

NOT reminded about qualiﬁcations to rent 18.75 19.79 – 1.04 0.4347 96

NOT encouraged to look at different unit than one advertised 12.87 12.87 0.00 0.5000 101

MD = mental disability.

* Signiﬁcant at the .10 level.

Table 32. Summary of Differential Treatment of Renters With MI Found in In-Person Testing

Test Dimension

Control

Tester Only

(%)

Protected (MI)

Tester Only

(%)

Net

Measure

p-Value

Sample Size

(n)

Unit available 8.00 0.00 8.00* 0.0625* 50

Given reason for unit not available 4.00 0.00 4.00 0.2500 50

NOT asked for additional qualiﬁcation information 17.02 25.53 – 8.51 0.2517 47

NOT reminded about qualiﬁcations to rent 18.75 20.83 – 2.08 0.5000 48

NOT encouraged to look at different unit than one advertised 16.00 6.00 10.00 0.1133 50

MI = mental illness.

* Signiﬁcant at the .10 level.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Table 33. Summary of Differential Treatment of Renters With I/DD Found in In-Person Testing

Test Dimension

% Tester Favored

Control

Tester Only

(%)

Protected (MI)

Tester Only

(%)

Net

Measure

p-Value

Sample Size

(n)

Unit available 3.92 1.96 1.96 0.5000 51

Given reason for unit not available 3.92 0.00 3.92 NA 51

NOT asked for additional qualiﬁcation information 16.00 18.00 – 2.00 0.5000 50

NOT reminded about qualiﬁcations to rent 18.75 18.75 0.00 0.5000 48

NOT encouraged to look at different unit than one advertised 9.80 19.61 – 9.81 0.1509 51

I/DD = intellectual or developmental disability. NA = test statistics cannot be produced due to no observations in control, protected, or both groups.

Such observations were not made for men (for additional

qualitative analyses of the content of these responses, see

appendix A).

• Race—Although sample sizes and matched-pair data were

too small to be conclusive to analyze most differences,

African-Americans with MI were less likely to be told

an advertised unit was available (9.76 percent favored

African-Americans without disabilities compared with

0.00 percent that favored African-Americans with MI;

p = .06) than African-American control testers.

• Current living situation—Testers with MI who identiﬁed

as living in a nursing home or institution before renting

(Olmstead group) were less likely to be told an advertised

unit was available (13.33 percent favored people without

disabilities compared with 0.00 percent that favored peo-

ple with MI living in nursing homes; p = .06) and trending

toward more likely to be encouraged to look at a different

unit (23.33 percent favored people without disabilities

compared with 6.67 percent that favored people with MI

living in nursing homes; p = .09). No signiﬁcant ﬁndings

were observed in test results for applicants with I/DD who

identiﬁed as living in large group homes versus with family

before renting.

In addition to the analysis of these quantitative ﬁndings, all

qualitative ﬁeld notes and examples from in-person testing

were analyzed for additional overt and subtle indications of

discrimination. An innovative mixed-methods approach was

used for analyzing these qualitative data, which yielded rich

descriptive ﬁndings on additional areas of potential discrimina-

tion that were not measured in the quantitative analyses. These

qualitative methods and ﬁndings are summarized separately in

appendix A, Qualitative Analyses and Findings on Overt and

Subtle Discrimination Against People With Mental Disabilities

by Different Testing Modes.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Findings From the Evaluation of Different Testing Modes

Two primary objectives of this study were to (1) develop

and implement different testing methods and approaches for

measuring housing discrimination on the basis of MD and

(2) evaluate the strengths and weaknesses of each approach,

to establish a set of protocols for conducting paired testing

to measure the degree to which people with MD experience

discrimination in the search for rental housing. This section

of the report evaluates the three testing methods implemented

over the course of this research effort, focusing on the unique

aspects of implementing paired testing with, and on behalf of,

people with MD.

DDS 2005 explored the development of methodologies to test

for housing discrimination on the basis of MD and served as

a starting point for developing the research approach for this

study. The ﬁnal report from DDS 2005 provided the study team

with a set of considerations when employing people with MD

as testers and also provided some insights on how to reveal

disability and how to construct parallel stories for paired testing.

This information was supplemented with a review of current

research on testing, insights from focus groups that included

people with MD, and guidance from the study’s expert panel to

develop the testing protocols implemented in this project.

People With Disabilities As Testers: Is

It Feasible?

HUD, the study team, the expert panel, and focus groups

strongly supported the participation of people with MD

throughout the project, both as advisors and as testers.

Although the recommendations of DDS 2005 indicated such

an approach was feasible for in-person testing, the report also

noted the potential difﬁculty of recruiting a sufﬁcient number

of testers with MD and the concern that the effectiveness of the

testers might decline if too many tests were conducted within

a given time period. With these concerns in mind, people with

MD were engaged early in the planning process to help design

the testing protocols and to participate in the testing process.

The study team sought to balance the equally important

practical need to produce credible results in a reasonably short

period of time with the need to contain costs. This balance

imposed practical limitations on methodologies, but it also

supported the consideration of different approaches, both for

the beneﬁt of the testers with MD and for the efﬁcient execution

of the study.

The ﬁnal research design included pilot testing that employed

people with and without MD as testers. People with MD

served exclusively as protected testers for the in-person testing,

completing each test accompanied by a companion without

MD, who was another trained tester posing as friend helping

in the search process. This testing design was based on the

recommendation of the expert panel, was conﬁrmed by focus

groups that included people with MD as a reasonable strategy,

and replicated the approach of the 2005 DDS.

The conclusion reached by the study team, as discussed

subsequently, was that producing reliable fair housing test

results while working with individuals with MD—both MI and

I/DD—is feasible.

Tester Proﬁle: How Do You Isolate

and Reveal Disability?

A key challenge to conducting matched-pair testing for housing

discrimination on the basis of MD is ensuring that the tester

can convincingly and unambiguously reveal the presence of

MD to a housing provider. Based on a review of the literature,

ﬁndings from focus groups, and discussions with the expert

panel, the following recommendations were made regarding (1)

the severity of a tester’s disability, (2) the disclosure-and-reveal

sequence, and (3) the proﬁle of the tester.

Severity of Disability

After reviewing existing research on testing with this popula-

tion, and keeping in the spirit of the goals of the Fair Housing

Act and the Americans with Disabilities Act, the study team

determined not to restrict the type or severity of MD for testers,

other than being conﬁdent that the person with MD was able

to complete the assigned test. To make this determination, a

Focus groups noted that people with MD who brought along another person for support when searching for housing reported feeling safer and more comfortable

in the housing search process. Focus group respondents also noted that even when a companion was present, people with MD still experienced potential

discrimination by landlords. The expert panel, which also included people with MD, agreed that this approach would likely yield good data by having two people

to observe and record each test. The concern that the presence of a companion might affect or bias a test was discussed with the expert panel, and it was ultimately

determined that the effect of a companion accompanying the protected tester would likely be minimal.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

veriﬁcation process was used that was equivalent to seeking

informed consent when enrolling a research subject. For

this project, tester candidates were required to be their own

guardian and be capable of giving the equivalent of informed

consent, recognizing that this restriction would likely reduce

the pool of potential MI and I/DD testers. The research team

then ensured that the candidate understood what was expected

of him or her and could repeat the requirements back to

research team member. The ultimate goal was to ensure that

all testers were able to fully execute the test, including the

reporting requirements after the test, not only at the time of

recruitment but also on the day of each scheduled test. To

this end, the recruitment, interview, and training process was

critical for successful testing involving people with MD.

Revealing Disability

Several challenges were identiﬁed as being inherent in testing

for housing discrimination on the basis of MD.

• Unlike other housing discrimination research studies,

which have focused on externally recognizable character-

istics, such as race and ethnicity, most MD is not visible or

immediately recognizable (Corrigan and Penn, 1999) and

therefore must be revealed during testing.

• Although many people with MD in the real world may

be reluctant or unwilling to reveal their disability, for the

purpose of these tests, the MD must be clearly, credibly,

and consistently disclosed to the housing provider or agent.

• Many people with MD have multiple or co-occurring

mental or physical disabilities; however, the testing design

can focus on only the single issue of the presence of MD.

• Although people with MD may disproportionately have

low incomes, proﬁles for both control and protected

testers had to depict them as income qualiﬁed for the

units they were seeking in order to avoid introducing

another variable into the tests.

Recognizing that the necessarily narrow design of the testing

implemented for the purposes of this research effort does not

represent the multiple and complex factors that affect the

typical experience of many people with MD during the housing

search process, different strategies were developed across the

three test modes to isolate the variable of interest in this project.

Because data are limited on testing for housing discrimination

against people with MD, the study team turned to recent

research on discrimination during the search for employment

for additional guidance for its disclosure strategies. Revealing

disability

during a job interview has parallels to the housing

search process; that is, some individuals may reveal the pres-

ence of a disability right at the start of the housing search pro-

cess, and others may hide their disability and continue to do so

after moving into a unit. The timing of disclosure of disability

during matched-pair testing was studied by Dalgin and Bellini

(2008)using a tester proﬁle that contained a 9-month gap in

employment. In that study, the control testers with no disability

offered that they were taking care of their mother who had a

serious illness during that time. The testers with disabilities

disclosed that they had been ill and needed to take some time

off to deal with their own health concerns. Disclosure of the

speciﬁc nature of the health concern (that is, disability) then

occurred later when voluntarily shared by the candidate either

in a short or long response (that is, more or less detail). This

study determined that extensive information on disability does

not change the outcome, because no difference was evident in

response to the shorter version of disclosure. This approach

assured that the employer knew the applicant had MD early in

the interview process and then later received some additional

information about that disability.

The DDS 2005 exploratory testing grappled with how the

protected tester could reveal the presence of a disability to a

housing provider. For this project, it was generally agreed that,

although most people in this group would not typically disclose

their disability outright to a housing provider, it would be

credible to do so if the each tester disclosed while explaining

why they had no recent rental history. As a result, in all tests,

protected testers disclosed that their lack of rental history was

due to being in a mental health treatment facility (for people

with MI) or in a group home (for people with I/DD). Paired

control testers disclosed that they had either been a student

living in a dorm, had lived with family, or had lived abroad.

MacDonald-Wilson, et al. (2011:192) deﬁned disclosure as “revealing information about one’s diagnostic label, mental health condition, or psychiatric disability”

and identify seven different types or reasons for disclosure.

1. Full disclosure—no one excluded from knowing.

2. Selective disclosure—to people whom the person trusts.

3. Strategically timed—after building trust over time; may or may not be full disclosure.

4. Targeted—condition of employment.

5. Nondisclosure—hide it from all.

6. Inadvertent—reveal instead of being “found out.”

7. Forced—something came up that required it (e.g., hospitalization).

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Another approach to revealing disability in testing is with a

request for a reasonable accommodation. Common accommoda-

tions requested can include assistance animals, variance of the

rent due date to accommodate the date of receipt of disability-

related income, request for a live-in aid, waiver of minimum

income or credit requirements, acceptance of rent from a

third- party payer, or agreement to communicate through a

case worker or family member.

Kaye (2012: 16) suggested

that “disclosure of a PD [psychiatric disability] through a

written accommodation request accompanying an application

can cause a housing applicant to be negatively stereotyped as

less capable than an applicant without a PD, despite similar

qualiﬁcations” and can increase likelihood of discrimination.

Establishing a Suitable Proﬁle

A key concern in this project was to make the testing situation

as realistic as possible. This goal meant that the tester proﬁles

were required to align with the reality experienced by people

with MD but also to not trigger suspicion by the housing

provider or generate bias. The following was recommended.

• Type of disability—The speciﬁc type of disability was not

stated during a test. Instead, protected testers gave the

broad category of having either MI or I/DD and then revealed

more about functional aspects of the disability during the

interview (telephone or in-person) or in the e-mail.

• Visible characteristics—Race, gender, and age of test

pairs were driven by actual tester demographics but were

matched across pairs for all test types.

• Income—The income of each tester was established by

the study team and was in the range needed to afford the

unit being sought. Furthermore, testers did not claim SSDI or

any other disability-related income, nor did they indicate

that they were using a Housing Choice Voucher or any other

form of housing assistance. This condition was to eliminate

any potential bias based on source of income, which is a

protected class in some jurisdictions but not others.

• Employment—It was agreed that testers should be

working to eliminate concerns that unemployment might

be a source of bias for some housing providers. Being

employed, however, does not ensure sufﬁcient income to

afford many apartments (using the “30 percent of income”

rule), therefore, control and protected testers’ proﬁles were

designed to include employment that realistically paid

enough to cover the rent at the property being tested.

Using the testing protocols described in previous sections, pro-

tected class testers were trained to reveal their disability in such

a way that it was unambiguous and could not be discredited

even if the protected tester did not “manifest symptoms.” The

approach taken in this study was, thus, full disclosure of MD in

all test modes by the individual acknowledging the presence of

a disability (1) directly (for example, “I have a mental disability”)

and (2) indirectly (for example, rental history). In some cases,

disability was also revealed by asking for a reasonable accom-

modation; however, such a request was made only after the ﬁrst

two disclosure methods were completed and all information

required for the test to be considered complete was attained.

In all cases, the functional aspects of disability, rather than the

type of disability, were provided when disclosing any details.

Tester proﬁles were matched for consistency between the

protected tester and control tester. Similar backgrounds were

used to establish a similarity in rental history. The guidance for

testers with disabilities during in-person testing and proxies

during the telephone testing, and the exact language used in

e-mail testing, are presented subsequently.

E-mail Testing: Strengths and Weak-

nesses

E-mail testing is important given the growing use of this medi-

um for communication in the rental housing search process. As

a testing mode, it has several strengths when focusing on indi-

viduals with disabilities. It also has some limitations, including

the inability to interpret a nonresponse from a housing provider

and the effect such a nonresponse has on the overall set of

paired test results. One strategy for addressing this concern is

to enlarge the sample, or to oversample. This approach, too, can

also have drawbacks, however, especially in smaller markets

where increasing the sample size can also increase the risk of

test detection.

A key strength of e-mail testing is the elimination of potential

fatigue among testers, which was a particular concern raised in

the discussion of working with people with MD as testers. With

the e-mail as the “tester,” this mode also reduces human error

in reporting because it provides verbatim statements on both

sides of the transaction—each member of the tester pair and

the housing provider. This design allows for the message and

initial contact of the tester to be controlled, including when and

how the disability is revealed, thus ensuring consistency across

Kaye (2012: 31) suggested an additional reasonable accommodation request to reveal a PD by having “two applicants requesting that they can break lease with no

penalty, but one is asking in case of psychiatric hospitalization and the other is asking in case of military deployment.”

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

all tests, which in turn can produce reliable and unambiguous

quantitative measures to determine if clear patterns of discrim-

ination exist against people with MD.

E-mail testing also provides data that allow for a more systematic

and speciﬁc analysis to ﬁnd evidence of subtle discrimination.

Since passage of the FHA, researchers involved in paired

testing studies have observed that it is rare to ﬁnd a “smoking

gun” of proof (Hanson, Hawley, and Taylor, 2011). E-mail

testing provides written data directly from a housing provider

that can then be used to more accurately assess the subtle nature

and extent of discrimination (Hanson, Hawley, and Taylor,

2011; HDLP, 2012).

Subtle discrimination may be used to discourage someone in

a protected class from pursuing a particular housing option.

It can also include more favorable treatment of the protected

class. For example, recent testing in Vermont involving in-person

testing of people with disabilities, including MD, found that

27 percent of the control testers were positively favored, but

neither protected nor control testers experienced outright dis-

crimination.

The e-mail data produced for this project were

coded and analyzed using qualitative methods, which found

evidence of (1) subtle discrimination against the protected

tester and (2) more overt differences in treatment that could

be categorized as discrimination (Hanson, Hawley, and Taylor

(2011). See Appendix A, Qualitative Analyses and Findings on

Overt and Subtle Discrimination Against People With Mental

Disabilities by Different Testing Modes for a more extensive

discussion of this topic.

This analysis also highlighted issues relating to the timing of

e-mails in testing, and particularly when the initial housing

inquiry e-mail was received and responded to. In this testing, it

could only be assumed from the e-mail time stamp that the order in

which an e-mail response was received by the testing team was

also the order in which it was read by the housing provider.

Although a new frontier in testing, this project conﬁrms that

e-mail testing research can generate reliable results that, based

on statistical analysis of quantitative data, can be used to deter-

mine discrimination. These same data can be coded for both

quantitative and qualitative analysis to look systematically for

evidence of subtle forms of discrimination, which may be of use

in future testing efforts. Table 34 summarizes the pros and cons

of the e-mail method of testing.

Table 34. Pros and Cons of E-mail Testing

Pros Cons

Cost effective for a large number of tests. Captures only ﬁrst point-of-contact step of housing search process.

Automated system generates predetermined script. No live interaction with housing provider.

Logistically straightforward. Many larger housing providers use canned, scripted replies regardless

of initial request—unclear whether provider is responding to the

disability component.

Test administrators, but no testers—eliminates tester fatigue and error. Potential for low response rate requires larger sample size.

Less time consuming than other modes (e-mail is the data so no report

forms or debrieﬁng required).

Data are limited to the e-mail returned.

Provides exact language used by housing provider for analysis. Difﬁcult to interpret nonresponses.

Telephone Testing: Strengths and

Weaknesses

The number of telephone tests that were completed,

including those tests that included a request for a reasonable

accommodation, exceeded the goals for this test mode. This

result demonstrates that telephone testing can produce a high

volume of tests for analysis in a relatively short period of time

and can be used successfully to produce useful data.

The telephone mode of testing also has its limitations, howev-

er, including—

• Testers have no face-to-face interaction with housing

providers, so interpreting responses to questions is based

solely on what people say over a telephone, which may be

different than if the conversation occurred in person.

• Compared with e-mail testing, telephone testing produces

detailed ﬁeld notes to use in subsequent analysis. Although

a positive, it is also a negative because it requires additional

time for testers to produce their reports.

• The particular test design here, which used proxies, is not

the equivalent of actually speaking with a person with MI

or I/DD. The results, although valid and useful, may not be

as socially valid as using a person with MD to do the testing.

Housing Discrimination Law Project of Vermont Legal Aid (2012).

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Based on the successful use of testers with MD for in-person

testing as part of this project, the study team believes that it

would be feasible to conduct telephone testing that employs

people with MD to serve as protected testers, making calls and

requesting information about the apartment for themselves,

assuming that a strong commitment to training and oversight

is in place. As noted previously, the social validity of this testing

is likely to increase if people with disabilities are used in the

testing process to make telephone calls.

Table 35 summarizes the observed pros and cons of telephone

testing. As with e-mail testing, most of the issues are not unique

to testing for discrimination against people with MD.

Table 35. Pros and Cons of Telephone Testing

Pros Cons

Live interaction w/housing provider. No face-to-face interaction.

Use of proxy without a disability as friend or family member to conduct

test allows for more ﬂexibility in recruiting, training, and scheduling

testers.

Not the equivalent of actually speaking to people with MI or I/DD.

Can be conducted from anywhere in the United States (one centralized

location or group of remote locations).

Requires being in a location with sufﬁcient MD community to recruit

from if using population to test.

Testers have script in front of them during test and record results in real

time.

Although detailed ﬁeld notes a positive, they require time for testers

to produce, unlike e-mail, for which returned e-mail response alone is

used.

I/DD = intellectual or developmental disability. MD = mental disability. MI = mental illness.

In-Person Testing: Strengths and

Weaknesses As Evaluated by Testers

With MD and Companion Testers

In-person testing is a labor-intensive process for any protected

class, given the recruitment and retention strategies needed to

ensure that the required number of paired tests is executed. In

this study, one signiﬁcant concern was that people with MD

would require more time between tests to reduce potential for

fatigue, and that therefore completing the planned number

of tests within a reasonable timeframe would necessitate the

recruitment of a larger number of testers, which can have sig-

niﬁcant cost implications. In addition, the use of a companion

tester raised concerns regarding both increased costs and a

possible delay in testing as a result of having to coordinate

multiple schedules for each test.

The initial literature review and focus groups, input from the

expert panel, and a small set of in-person pretests served as

preparatory steps to working with in-person testers with MI

and I/DD. These efforts were designed to provide key insight

into housing discrimination experienced by this population

and ideas for how to involve community members as in-person

fair housing testers, including the recruitment, selection, and

training of testers and the data collection process.

The in-person testing had a smaller sample size and a longer

execution time, and it was conducted in only two, very large

housing markets. To evaluate the inclusion and active partic-

ipation of people with MD as testers, four focus groups with

key participants and stakeholders were conducted, including

(1) project leaders, (2) testers with I/DD, (3) testers with MI,

and (4) companion testers.

Additional earlier focus groups, primarily consisting of testers,

were held in the weeks immediately after the completion of

in-person testing and data collection so participants could reﬂect

on their testing experiences. The detailed approach and results

of the focus groups, which includes quotes from participants,

can be found in one of the short papers associated with this

project, Accessible & Participatory Methods for Involving People

with Mental Disabilities In Housing Discrimination Testing.

Strategies for Tester Recruitment, Selection,

Retention, and Matching

Test coordinators from the two participating fair housing testing

organizations broadly recruited potential testers by contacting

a variety of community groups and organizations with which

people with MI and I/DD had active contact, such as programs

that provide housing or community living services. Because

of widespread employment disparities among this population,

some people with MI and I/DD were found to be more available

to work as testers; however, the availability of testers who were

currently unemployed may also have restricted the recruitment

of testers with full-time jobs or busy schedules.

Test coordinators also implemented extended interactions with

potential testers to allow for the evaluation of the “ﬁt” between

the tester and the testing responsibilities involved (that is,

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

discerning if the candidate could complete the essential job

duties). During these interactions, test coordinators provided

potential testers a candid and thorough description of the

project and their expected role, including an emphasis on

the responsibilities of being a tester versus other roles in the

project, assessing how the potential tester understood his or

her role, and the potential tester’s comfort with pretending to

be searching for an apartment and disclosing their disability

during the housing search process. All these issues were

proactively addressed during the initial tester training. For

some, initial training helped test coordinators make a ﬁnal

determination about a potential tester’s involvement in the

actual in-person tests.

In addition, the study team learned how best to deliver train-

ing to potential testers with MD, by simplifying and limiting

background information and lecture and instead focusing on

short practice exercises, open discussion of any questions, and

other modiﬁcations that enable testers to gain comfort carrying

out a test and working with a companion tester.

Test coordinators also noted that they needed to allocate

signiﬁcant time to the recruitment process, needing initially

to recruit about three times the number of testers needed. For

example, in some cases, most notably among potential testers

with I/DD, availability and reliability were issues (for example,

getting training scheduled, not showing up for scheduled

training, the availability of reliable transportation, and the

need to navigate a new transportation route without support).

Finding available testers was especially difﬁcult given the need

to coordinate I/DD testers with companion testers for initial

training, practice testing, and each in-person test.

Despite these challenges, test coordinators from the two

participating testing organizations were able to recruit the

required number of testers due to their extensive contacts and

continual networking with community groups and people with

disabilities. These established connections with community

groups and people with disabilities were essential to this effort

and would be difﬁcult for organizations not familiar with and

embedded within disability communities to replicate. The use

of these community-based testing organizations was a critical

component of the project and a key to success in actively

supporting people with MD during the in-person testing effort.

Test coordinators also noted that it was important to identify

people who would be good companion testers, and they noted

that ensuring a strong rapport between a tester with disabilities

and his or her companion tester was key to the protected

testers’ comfort and to the quality and efﬁciency of testing. Test

coordinators also noted that it was especially important that

they consult with both the protected tester and the companion

independently, and early in the partnership, to ensure that the

pair was a good match. This assessment of the match began

during the initial training, when the team purposively built in

many opportunities for pairs to get to know each other and

become comfortable with each other during practice tests. Test

coordinators and testers agreed that a tester’s choice of his or

her companion was very important to the comfort and ﬁdelity

of the tester and data collection and agreed that pairs should be

rematched if one of the members was not comfortable.

Preparing and Supporting Testers: Initial

Training

Project leaders worked to design tester training to effectively

engage testers with both MI and I/DD and to actively support

their learning. For both groups, training materials deempha-

sized complicated text, reduced the number of ideas presented

at one time, and incorporated graphics, photos, and images

to illustrate more abstract points such as civil rights. The

trainings also included opportunities to observe testing skits

performed by test coordinators and project staff ﬁrst, and then

for testing pairs to actively engage in several different testing

scenarios by role-plays (moving from working in pairs with an

assigned project staff to role-playing in front of the whole group

and getting feedback thereafter), and they included frequent

breaks to take in information learned. Basing their approach

on initial insights from stakeholder focus groups and expert

panel meetings conducted at the start of the project, project

leaders delivered the training differently for each group to

maximize their learning. For example, testers with I/DD had the

information broken into smaller units followed immediately by

opportunities to practice a small skill ﬁrst and then build up

to practicing a full test and then onto an actual practice test in

the ﬁeld, and testers with MI received initial information in one

sitting and then practiced different scenarios.

Testers reported that these experiences—particularly the oppor-

tunity to practice and receive individualized feedback—helped

them learn their responsibilities and feel prepared for their role.

Some testers noted that when the time between training and testing

in the ﬁeld was signiﬁcant, it was beneﬁcial to have opportuni-

ties to brush up on their skills and review the testing protocol.

Project leaders observed that ensuring that the training had a

good deal of time devoted to hands on learning was essential.

Preparing and Supporting Testers:

Ongoing Support During In-Person Testing

The in-person paired testing was also designed to support

testers during data collection in ways that testers and compan-

ion testers found helpful. Project leaders provided scripts and

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

checklists to help testers remember the personal information

they would need to provide during testing. They also provided

reminders via e-mail, telephone, or both (cell phone communi-

cations were used frequently before, during, and after testing)

and were readily available to respond to questions and assist

with unanticipated problems.

Many project leaders and testers agreed that being able to debrief

with test coordinators immediately after each test (either by

telephone or in person) was very helpful. The testing protocols

required debrieﬁng within 48 hours of the actual test and, in

many cases, testers completed debrieﬁng immediately after the

test or within 24 hours. For some testers, being able to complete

the forms at the testing coordination agency allowed them

access to onsite computers and to test coordinators in case any

questions arose. These factors were reported to have helped

the testers submit completed forms quickly and accurately.

Whether for training or data collection, project leaders thought

it was important to ask testers what they needed assistance

with (for example, writing, reading, or a quiet room to

complete forms in) and to be ﬂexible in responding to testers’

emergent needs. Many noted that the level of support needed

decreased over time as testers increased their skills and level

of comfort. Although some testers initially felt nervous, they

reported becoming increasingly comfortable as they gained

experience. Thus, it was important to support a key group of

testers so they could do many tests and feel comfortable in

doing so, rather than trying to have a large pool of testers who

completed only one or two tests each.

Preparing and Supporting Testers: Use of

Companion Testers As Ongoing Supports

Companion testers also provided support to protected testers

during data collection. For example, when testers forgot to ask

a question of a landlord, “got off track,” or “froze,” companion

testers were able to naturally prompt the tester or provide a

response. Companion testers also reassured testers and helped

them feel less nervous during data collection or when in a new

area of their community. Companion testers were also able to

provide assistance by taking notes, organizing and remember-

ing information, helping out when unanticipated situations

arose, and in some cases with transportation.

Many testers with MD and companion testers thought that

the presence of the companion testers may have helped the

protected tester feel less stressed, because immediate support

was available. It was also noted that protected testers and

their companion testers complemented one another during

data collection by offering different ideas and that working as

a pair made the experience more enjoyable. Protected testers

and companion testers noted that the support provided by

companion testers during data collection may have also added

credibility to the disclosure of a disability by testers. On the

other hand, some companion testers wondered whether their

presence tempered potential opportunities for differential

treatment by landlords.

Most thought that companion testers were most effective when

they saw their relationship with the tester as a partnership;

that is, when companions understood that they were there to

support testers, not to be testers. To achieve this understanding,

companion testers needed to be ﬂexible to their paired tester’s

needs and preferences. The dynamic of successful relationships

was evident when testers and companion testers had discussed

and agreed on each individual’s role and responsibility, had

good communication before and after data collection, and both

contributed to the completion and approval of reporting forms.

Mentoring, training, and role-playing may be required to help

companion testers excel in these supporting roles. The need

for training on how to be a good companion tester may have

been particularly relevant for companion testers who had been

testers in other fair housing testing situations. Although their

previous experience provided great knowledge and insight, they

also had to transition to playing a supporting role, rather than

the primary role, during a test.

When directly asked whether companion testers were needed,

the study team agreed that most testers with I/DD would need

a companion tester to complete data collection. The study team

also believes, however, that most testers with MI, and perhaps

some testers with I/DD, could initially beneﬁt from a compan-

ion but could potentially conduct testing on their own as they

gained experience.

Documenting Differential Treatment and

the Role of the Companion Tester

Some protected testers reported that they thought they

experienced negative treatment by housing providers or agents

and that the companion tester helped to document and handle

that treatment. For example, one tester was asked repeatedly

whether he had a criminal background, what his income was,

and whether he could afford the unit. Another thought the

housing provider was suggesting she may not be intelligent

enough to live in the building, and another was told that the

landlord would “pray for her.” Others reported landlords acting

distracted (talking on the telephone or looking at their watch),

being abrupt in interactions, refusing to provide an application

(likely because an application had been received earlier in the

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

day), or simply “giving attitude.” Companion testers had the

sense that testers appreciated having someone with them to

witness and validate this treatment.

Recommendations for Expanded Testing

in the Future

Although they understood the scientiﬁc value of the current

testing approach and its focus on one speciﬁc disability group,

many testers stated that they wished the project could be

broadened to better reﬂect the actual experiences of individuals

with MD, especially because many people with MD experience

more than one functional limitation. For example, some

people with I/DD also were anticipating some future physical

accessibility issues and wanted to be able to document those as

well. They also thought it would be illuminating to study the

qualities of accessible housing, particularly what that means

and looks like for people with MD.

In considering the expansion of matched-pair testing to

study housing discrimination on the basis of MD to a larger

geographic scale, project leaders, testers, and companions had

many recommendations. For example, project leaders—

• Thought that having an experienced, centralized oversight

team would be important to ensuring that the data

collection protocols are being administered similarly

across sites (including for training), to facilitate ongoing

communications among sites, and to provide ongoing

support and accommodations, as needed.

• Conﬁrmed the importance of some strategies that were

used in this project, such as the need for housing organiza-

tions to partner closely with organizations run by people

with disabilities within testing regions to succeed in this

work.

• Stressed the importance of allowing adequate time for

the recruitment, training, and ongoing support of testers

with MD, with some suggesting the initial recruitment of

approximately three times as many testers as originally

needed.

• Believed that holding training in places that are familiar

to potential testers, such as at disability and community

organizations, could help attract testers with MD because

they are more likely to trust these sites.

• Emphasized that, in preparing testers and companion

testers for data collection, the research team and com-

panion testers should not make assumptions about what

testers can and cannot do.

• Agreed that experienced testers with I/DD and MI, and

companion testers, might be well-equipped to train future

testers, and their role as peer mentors and trainers should

be considered in future studies.

• Emphasized the value of building in time for mentoring

and one-on-one training.

Testers with MI and I/DD and their companion testers also

made valuable suggestions, including—

• Agreeing on the beneﬁts of making sure that everyone has

an opportunity to role-play and complete a practice test

with feedback.

• Noting that having all testers engage in more than one

practice test might be advantageous.

• Suggesting that, although each person is unique, trainings

should include some general information about MI and I/DD

at the start of tester interaction that, in a respectful way,

acknowledges the considerable heterogeneity of people

with MD.

• Recommending that they receive a checklist detailing

everything they should do before a test (including practical

information, such as use the restroom).

• Expressing a desire that they do tests in areas where their

personal characteristics match those of the local communi-

ty and in areas that “feel safe.”

• Noting that it would be helpful to attempt to make sure

that apartment-viewing appointments were with leasing

agents and not with maintenance personnel.

• Suggesting that they should also be involved in making

calls to arrange apartment-viewing appointments, because

some proﬁle information is requested at that point of contact.

• Raising the issue of how to ensure that testers receive

support for any emotional reactions they may have during

data collection.

• Encouraging testers and companion testers to return to the

testing coordination agency after each test to debrief and

complete reporting forms, and after testing would also be

a potential time to support them emotionally, give them

additional resources if needed, or both.

• Urging that test assignment and report forms be as clear as

possible so that testers and companions would have fewer

questions for test coordinators.

• Raising the possibility of alternative payment options

so that those who receive disability beneﬁts could more

fully participate without having their subsidized disability

income being at risk.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Finally, several testers suggested that all testers should have the

chance to reﬂect and debrief at the end of the project, similar

to what was done in debrieﬁng focus groups with a small

sample of testers and staff in this study. Testers expressed

that the focus groups enabled them to validate their feelings,

revisit their experiences and make sense of them, reconnect

with other testers and companion testers, and validate their

contributions and role in the overall project.

The pros and cons of in-person testing are summarized in Table 36.

In-person testing yields richer data than the other modes, but it

costs more and takes more time to complete the tests. Although

that is true for any protected class, the pros and cons are ampli-

ﬁed when using people with disabilities as testers. Depending

on the objective, the added time and costs associated need to be

weighed against the data produced.

Table 36. Pros and Cons of In-Person Testing

Pros Cons

Face-to-face interaction with a housing provider captures richest, most

realistic interaction information, including verbal and nonverbal responses.

Logistically complex; scheduling of two testers and companion for

protected tester to conduct tests and debrief.

Test reaches deeper into rental transaction. Tester recruitment, training, support, and retention are challenging.

Involves people with MI and I/DD in conducting research, which increases

social validity of ﬁndings.

Potential for tester fatigue and emotional reactions need to be consid-

ered in training and debrieﬁng support.

Use of companion without disability provides additional attention to

detailed observations and ﬁeld notes for improved rigor of analysis of

ﬁndings.

Requires longer time in the ﬁeld for testing to coordinate tests.

Upfront screening and training reduces attrition, and people with MD

were found to be effective testers.

Requires more preparation, time, and resources, especially collaboration

and involvement of disability and community organizations that under-

stand how to provide accommodations to testers with MI and I/DD.

I/DD = intellectual or developmental disability. MD = mental disability. MI = mental illness.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Summary, Conclusions, and Recommendations

The pilot testing effort documented in this report represents

the largest and most comprehensive effort to date to validate

methods of testing for housing discrimination against people

with mental disabilities. The pilot testing also represents the

ﬁrst time that people with MI and I/DD have been enlisted to

serve as in-person testers in a housing discrimination study of

this size and therefore contributes signiﬁcantly to the knowl-

edge about disability-related housing discrimination and the

design of research methodologies that are accessible, inclusive,

and participatory to people with MI and I/DD.

The testing conducted in this study documented signiﬁcant

levels of adverse differential treatment toward individuals with

MD when compared with the treatment of a control group of

individuals who did not have MD. Such adverse differential

treatment included observations that individuals with MI or I/

DD were—

• Less likely to receive a response to their inquiry (17.55 percent

of tests favored the control group of people without dis-

abilities compared with 9.19 percent that favored people

with MI and I/DD in e-mail testing).

• Less likely to be told an advertised unit was available (5.94

percent favored people without disabilities compared with

0.99 percent that favored people with MI and I/DD in

in-person testing).

• Less likely to be invited to contact the housing provider to see

the unit (7.69 percent favored people without disabilities

compared with 0.00 percent that favored people with MI

and I/DD in e-mail testing).

• Less likely to be invited to inspect the unit (21.26 percent

favored people without disabilities compared with 16.47

percent that favored people with MI and I/DD in telephone

testing).

• More likely to be encouraged to look at a different unit than the

one advertised, a potential indicator of steering people with

MI and I/DD toward speciﬁc buildings or areas within

rental complexes, resulting in segregated living patterns.

Rates of adverse treatment also varied between disability types,

with higher rates of adverse treatment noted for individuals

with MI than for individuals with I/DD. This ﬁnding may indicate

that individuals with MI face more negative stereotypes and

stigma from housing providers and their agents.

This study,

however, was not large enough to yield conclusive evidence

about what motivation might be underlying the observed dif-

ferential treatment between people with MI as compared with

people with I/DD, nor was the study designed to examine dif-

ferential discrimination based on type of MD. Future larger-scale

research should explore these questions and might include

additional focus groups with housing providers to discuss study

ﬁndings and to seek housing providers’ reactions and responses

to the results of fair housing testing.

In both e-mail and telephone testing, a signiﬁcant percentage

of individuals with MI and I/DD also experienced adverse treat-

ment with respect to requests for a reasonable accommodation.

The reasonable accommodations requested were identiﬁed as

speciﬁc to the protected tester’s functional needs. The willing-

ness of a housing provider to grant a request for a reasonable

accommodation varied by mode of testing, with requests for

a reasonable accommodation made by telephone receiving

signiﬁcantly more favorable responses than requests made by

e-mail.

This ﬁnding may reﬂect the fact that housing providers

were able to have a more personal or indepth conversation with

the tester with MD via telephone, resulting in more favorable

responses to these requests. On the contrary, in e-mail testing,

replies were often more formal and very limited in scope. These

experiences may advise the use of telephone over e-mail testing

in future large-scale discrimination testing.

Negative responses to testers’ requests for a reasonable accom-

modation ranged from outright denials to more subtle barriers,

such as indications that the individual with MD would be

responsible for actively seeking out and appealing the denial

of their request with higher-level managers whose name and

contact information were then not provided.

These ﬁndings

on responses to requests for reasonable accommodations

further add to potential discrimination against people with

This observation was supported in the qualitative data obtained from focus groups of people with MI that were conducted both before and after the testing to

reﬂect on their experiences.

In-person testing protocols did not include a request by the protected tester for a reasonable accommodation.

In addition, testing results varied by rental market, with the rate of granting a reasonable accommodation request lowest in Chicago-Joliet-Naperville, IL-IN-WI

(38.6 percent) and highest in Albuquerque, NM (75.5 percent).

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

MD, and it is recommended that reasonable accommodation

requests be included in in-person testing in the future, using

a test design similar to that used in telephone testing in this

project, wherein requests for a reasonable accommodation are

not made unless the tester is invited to inspect the unit.

This study also yielded signiﬁcant ﬁndings on how to imple-

ment fully inclusive fair housing testing employing people with

MD. All three modes of testing that were carried out (e-mail,

telephone, and in person) showed promise for bringing testing

for housing discrimination on the basis of MD to a national

scale. Each produced a different form of testing data that not

only yielded evidence of discrimination but also provided

insights on opportunities for future testing.

Because online housing searches and e-mail communications

with housing providers during these searches is anticipated

to increase, efforts to use e-mail testing in this project should

inform future e-mail testing. Observations include—

• E-mail testing is a relatively cost-effective method of

tracking changes over time in discrimination in initial

contacts people with MD have with housing providers

(that is, trying to get in the door).

• Longitudinal data showing a change in the rates of differ-

ential treatment of people with MD could be efﬁciently

collected by subsequent e-mail testing of the same sample

of housing providers.

• E-mail could also be useful for monitoring how housing

providers are responding to reasonable accommodation

requests, if a followup e-mail is added to further interact

on the reasonable accommodation request. This technique

could also ensure that the housing provider is responding

directly to the request as opposed to sending a scripted or

form reply.

The in-person testing in this study that speciﬁcally involved

testers with MD demonstrates that people with MI and I/DD

can serve as effective testers when the proper training and

supports are provided.

As the ﬁrst large-scale in-person testing that had people with

MD as testers, this project demonstrated the importance of

disability-focused and directed training. The entire process of

engagement with testers, from recruitment to training, testing,

and debrieﬁng, was grounded in constant feedback from

the people with MD and members of the expert panel. The

evidence from the in-person testing in this project suggests

strongly that both populations can also conduct telephone

testing if the same approach to training and debrieﬁng is used.

This project demonstrates that the training and use of disability

rights organizations to coordinate and debrief testers is a highly

effective design for testing focused on disability and was a

critical component of the successful completion of in-person

testing in the given timeframe in this study.

The use of companion testers, although adding cost and

coordinating time, resulted in several important beneﬁts to the

project, including—

• Providing valuable cognitive and emotional assistance to

testers with MD before, during, and after each test.

• Providing a natural way to reveal the MD via a friend in

conversation while viewing a unit.

• Providing a means to conﬁrm what had been heard and to

clarify what was observed during each test, thus providing

richer data and ﬁeld note observations.

• For testers who used public transit, companions often

(although they were not expected to) provided transit

assistance or went along on public transportation with the

tester with MD, ensuring that both testers would get to

appointments in a timely fashion.

Findings from this project, although important in documenting

how housing discrimination may occur against people with

MI and I/DD as they seek rental housing, are limited given the

small sample sizes, particularly when attempting analyses on

variations of treatment between subcategories of paired tests,

such as gender, race, type of disability, and geography. Future

research, done on a larger scale, could examine other important

issues, for example—

• Are people with different types of MD, or different requests

for other types of reasonable accommodations, met with

less favorable responses or more discrimination?

• How do factors such as age, gender, race, and previous

living status contribute to or interact with MD to affect

rental housing access?

• How does income or economic status interact with MD to

affect housing access and choice?

• How does the use of rent subsidies (for example, Housing

Choice Vouchers) and other government-supported

sources of income for housing affect housing access and

choice for people with MD?

Finally, efforts to ﬁnd protocols to study and test for housing

discrimination after initial contact with a housing provider (for

example, application results, before move in, after move in,

changes in response to requests for reasonable accommodations

over time, and criteria for lease termination) should continue.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Based on the results of the study, the authors make the follow-

ing recommendations.

• A broad-based education initiative should be created to ed-

ucate housing providers, including owners and landlords

and their agents, about fair housing rights and obligations

and about appropriate policies and practices when dealing

with individuals with MI, I/DD, and other MD.

• Public and private housing, disability, and civil rights

organizations should redouble efforts to engage and educate

the community of people with MI, I/DD, and other MD about

their rights under federal, state, and local antidiscrimina-

tion laws, how to recognize potential discrimination, and

what actions to take when faced with discrimination.

• Additional research is necessary to better understand the

scope and severity of the discriminatory barriers en-

countered by individuals with MI and I/DD as they seek,

obtain, and retain accessible rental housing.

• Individuals with MI and I/DD can and should be an integral

part of future housing research and testing, housing dis-

crimination education initiatives, and efforts to strengthen

housing policy to promote equal housing opportunity.

In addition, protected and companion testers involved in the

in-person testing for this project offered the following summary

recommendations for including people with MD in future

housing discrimination research.

Project Leadership and Design

• Project leadership is deeply important to people with

MD. Having test coordinator sites embedded within local

disability and housing rights organizations was especially

valuable in supporting testers with disabilities.

• Testing sites and all project members should actively foster

commitment to, and belief in, the inclusion of people with

MD at all stages of research projects.

• An inclusive project should foster coleadership, in which

the research team and the protected community colead

and collaborate to design and conduct discrimination

research with a goal of bringing housing discrimination

experts together with people with personal experience

with disability to fully and accurately understand the

actual experiences people have when seeking housing.

• Future testing should use experienced testers with MD

(such as those from this project) and their experiences in

protocol development and training as peer trainers and

mentors in a train-the-trainer model.

Recruitment and Selection of Testers

• Reinforce responsibilities and expectations of participation

with potential testers to encourage increased reliability and

validity of ﬁndings.

• Provide sufﬁcient time, resources, and expertise to recruit,

select, and support team members with MD.

• Have a sustained selection process for testers and compan-

ion testers that allows for time to build relationships and

develop a ﬁt between test pairs.

• Involve experienced testers with MD in tester recruitment

and selection as they also have key contacts and networks

throughout protected communities.

Training and Data Collection

• Be proactive and ﬂexible in accommodating team members

with MD in training and data collection, including speciﬁc

attention to strategies to make participation more accessi-

ble and inclusive to this community.

• Have a choice and range of supports and accommodations

available to testers, and match emotional support to

individual needs, adjusting the supports as needed as

testers gain experience.

• Provide testers with choice throughout the process,

including a choice of whether to stay with a companion

tester or to choose another that is a better ﬁt.

• Ensure that all testers have time to role-play, practice, and

receive individualized feedback during trainings.

• Train companion testers to provide positive, effective support

to testers with MD in a respectful, person-directed way.

• Assess the value of companion testers—between the added

credibility and validity added to the capture of testing data

versus the potentially unanticipated effect that their pres-

ence may actually decrease or buffer differential treatment.

• Design all materials, illustrations, role-plays, and videos

in all training and data collection instructions and forms

using clear, concise, and plain language.

• Use technology to enable people with MD to take on

increasingly independent roles in testing, such as the use

of voice-operated smartphones or tablets during data

collection and reporting.

• Discuss in initial trainings the attitudes that may be

encountered during tests and how to capture attitudes and

treatment in more detail in reporting.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Appendix A: Qualitative Analyses and Findings on Overt

and Subtle Discrimination Against People With Mental

Disabilities by Different Testing Modes

The purpose of this additional indepth qualitative study was

to provide detailed qualitative analyses of the overt and subtle

discrimination experienced by people with mental disability

(MD)—that is, people with mental illness (MI) and people with

intellectual or developmental disability (I/DD)—as gathered

from multiple sources including e-mail, telephone, and in-person

testing, ﬁeld notes, and tester debrieﬁng reports. The following

research questions were used to guide the qualitative analyses.

• Is there evidence of subtle discrimination on the basis of

mental disability in the e-mail, telephone, and in-person

tests conducted during this study, and what does it look like?

• Is there evidence of overt discrimination on the basis of

mental disability in the telephone and in-person tests

conducted during this study, and what does it look like?

Qualitative Analysis Methodology

The structure and process for the analyses was guided by

Ahmed and Hammarstedt (2008), who examined subtle

discrimination on the basis of sexual orientation in the rental

housing market using e-mail testing, and by Hanson, Hawley,

and Taylor (2011), who proposed a methodology for qualita-

tively analyzing e-mail contents for subtle discrimination. As

in these previous studies, this study used matched-pair tests in

that each pair was matched in key demographics (for example,

age, gender, and race or ethnicity) except for MD status, with

one tester having an identiﬁed MD and one tester not having MD.

The tester with MD (protected tester) speciﬁcally disclosed

having either MI or I/DD. This reveal occurred in multiple

ways: (1) disclosing having lived in a psychiatric facility or

nursing home for the past year, (2) directly identifying himself

or herself as a person with I/DD or MI, and (3) requesting a rea-

sonable accommodation (that is, using an assistance animal or

requesting a verbal rental payment reminder). The other tester

in the pair (control tester) did not identify as having a disability

and instead identiﬁed as someone who recently graduated from

college or who had been living with his or her parents to show

a matched previous rental history. Both testers in a pair were

also given matching proﬁles that included similar income and

job histories, at a level adequate to pay for a one-bedroom rent-

al unit in that market, and other similar personal demographics

other than disability status.

E-mail Testing

As reﬂected in the larger discrimination study, 359 e-mail pairs

were sent out, including 176 pairs with testers with I/DD and

183 pairs with testers with MI. Out of those 359, 187 pairs

received at least one response from an agent or landlord. To

qualitatively compare the e-mails in this particular phase of

the study, we selected only e-mail pairs that both received

responses (n = 91 pairs). Of those 91 pairs, 9 were not codable

for the following reasons: (1) different agent responded, (2) agent

noticed the testing and test was not used, or (3) lack of quali-

tative data to code. Therefore, 82 pairs were codable, including

45 pairs with a tester with I/DD and 37 pairs with a tester with

MI. Out of the 82 e-mail pairs, 25 e-mail tests also included a

reasonable accommodation request that could be analyzed (see

Figure A-1).

Figure A-1. E-mail Testing Responses

E-mail pairs sent out

(n = 359)

E-mail pairs with

at least

one response

(n = 187)

E-mail pairs with

responses to both

(n = 91)

E-mail pairs

not codable

(n = 9)

E-mail pairs

coded

(n = 82)

I/DD testers

(n = 45)

MI testers

(n = 37)

With RA request

(n = 10)

With RA request

(n = 15)

I/DD = intellectual or developmental disability. MI = mental illness. RA = reasonable accommodation.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Telephone and In-Person Testing

A total of 668 pairs participated in telephone tests of ran-

domly selected available apartment ads across nine different

metropolitan statistical areas (MSAs). In-person testing

involved a visit to view the apartment by 101 pairs in the

Chicago-Joliet-Naperville, IL-IN-WI, and the Washington-Ar-

lington-Alexandria, DC-VA-MD-WV MSAs, 83 of which had

valid qualitative data to additionally code. See Table A-1.

Table A-1. Samples of Telephone and In-Person

Testing

Type Sample

Type of Disability

Total

I/DD MI

Telephone Valid pairs 300 352 652

Missing

11 5 16

Total 311 357 668

In person Valid pairs 48 35 83

Missing

3 15 18

Total 51 50 101

I/DD = intellectual or developmental disability. MI = mental illness.

Qualitative data were not available to code.

Data Analysis

A mixed-methods approach was used to analyze data, beginning

with qualitative analyses, followed by quantitative frequency

counts and analyses of summary themes, to answer the re-

search questions and fully understand discrimination practiced

by housing providers. Each type of testing used different data

analysis methods, as described in the following section.

E-mail Testing Only: Quantitative

Coding of Keywords Indicating Subtle

Discrimination

The housing discrimination study by Hanson, Hawley, and Tay-

lor (2011) provided signiﬁcant guidance for the interpretation

of the results of e-mail testing, especially because e-mail testing

yields an explicit and deﬁned text-based inquiry and response

that can be analyzed verbatim.

Hanson, Hawley, and Taylor describe several dimensions of

coding favorable and unfavorable treatment in landlord-tenant

correspondence. They describe favorable treatment as including

examples “if landlords describe the unit in a positive way, if

they mentioned other available units, if they invited further

contact, and if they used generally friendly language” (Hanson,

Hawley, and Taylor, 2011: 281) and unfavorable treatment as

asking the tenant for more information about their employment

or background or also as using more discouraging language,

such as emphasizing additional fees and building rules. In

addition to the keyword list Hanson, Hawley, and Taylor cre-

ated, the analysis here added more unfavorable and favorable

treatment dimensions speciﬁc to MD (for example, asking for

further disability-related information or disclosure) that were

identiﬁed after a round of pilot coding.

Using the subtle discrimination keywords listed in Table A-2, two

members of the study team coded the frequency of positive and

negative languages used in e-mails. These counts were com-

pared between the protected and control groups and divided

into four categories: (1) present in neither, (2) present in both,

(3) control group only, and (4) protected group only. Evidence

of subtle discrimination was coded if positive language was

shown more frequently in the control tester only or if negative

language was more frequently present in the protected tester only.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Table A-2. Subtle Discrimination Indicator Keyword Groups

Category Keywords

Positive language categories

Positive descriptive New Clean Quiet Nice Good Beautiful

Other

(for example,

wonderful)

View unit Tour

Show

Stop

Come

Visit

Meet

See

View

Look

Appointment

Other

Tour time offered Speciﬁc time Range of time

Prompts Quick

Hurry

Fast

Today

Tomorrow

Call Application Look forward to

Other (for example, Let me

know, questions)

Greetings Hi

Hello

Hey

Dear Good morning/

afternoon

Others

Polite Thank you

Thanks

Sincerely Regards

Best

Please Love to

Happy to

Other

Other units Second Several Other

Negative language categories

Fees Application fee Deposit fee $ Fee starting Other

(for example,., pet policy, move-in fee)

Employment Employment Employer Income Paystub Credit Other

(for example, source of rent)

Background Criminal Background Veriﬁcation

Verify

Other

Discouraging Already rented No longer Not available Best wishes

Good luck

Other

History Rental history Eviction Other

Negative descriptive Old Needs work Other

Disability

Handicap

Diagnosis

Medication

Hospitalization

Treatment

Treated

Accommodation Service dog Other (for example, mentally

competent, caregiver)

Indicates keywords that were added in this study.

Qualitative Interpretation and Grounded

Theory Analysis: E-mail, Telephone, and

In-Person Testing

Because additional qualitative data were available and new,

given the focus on testers with MD in this study, a grounded

theory qualitative approach was used to generate a framework

to analyze and code both overt discrimination (as deﬁned

within the Fair Housing Act) and more subtle, potential

discrimination. After coding the keywords, the two coders

also qualitatively described their overall impression of each

pair of e-mails, concerning subtle or overt discrimination (or

both) against one tester. Each decision was associated with a

description of evidence for subtle or overt discrimination or

no subtle or overt discrimination, as identiﬁed by coders. If

consensus on the discrimination status could not be reached

between the two coders, a third, blinded member coded the

e-mail to resolve the difference. By constant comparison of

these qualitative data, the ﬁndings were aggregated into themes.

These themes then were used as criteria for determining subtle

or overt discrimination.

Converting Qualitative Findings to

Summary Frequency Counts: E-mail,

Telephone, and In-Person Testing

Qualitative data were assigned a quantitative descriptive

frequency count—if subtle, overt, or no discrimination was

indicated—an assessment of whether the response showed

or did not show an openness for further communication with

the landlord. All quantitative data were entered into an SPSS

17.0. Descriptive analysis and cross-tabulation analysis were

conducted to examine the frequency and proportion of different

categories indicating subtle discrimination. McNemar’s Test

was used to analyze the difference in treatment between the

control and protected groups (McNemar, 1947), with a p-value

of less than .05 used to indicate signiﬁcantly different treatment

between control and protected testers.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Results

E-mail Testing

Quantitative Analysis of Keywords Related to

Subtle Discrimination

Quantitative analysis of keywords for positive and negative

language shows signiﬁcantly different treatment in several

areas. The control group was more frequently given a speciﬁc

or range of time to view an available housing unit or to meet

the landlord (p < .05) when compared with protected testers

with MD. E-mail responses to testers with MD showed a

signiﬁcantly higher use of discouraging words (p < .05). A

trend also emerged of using positive language more often with

control group only and negative language more often with

the protected group only; however, the difference was not

statistically signiﬁcant in some cases. See Table A-3.

As shown in Table A-4, an analysis of positive and negative

language used in e-mail responses showed that 34.1 percent

of landlords practiced favorable treatment to people without

disabilities by using at least one of the seven positive language

categories. A total of 17.0 percent of landlords practiced

unfavorable treatment to people with MD by using at least

one of the seven negative language categories. Combined, 37

e-mails (45.1 percent) showed at least one category of subtle

discrimination present.

Qualitative Themes and Descriptive Frequency

Counts on Subtle Discrimination

During e-mail testing, seven themes emerged related to subtle

discrimination against the protected testers with MD. These

themes were (1) more encouraging to control tester or more

discouraging to protected tester; (2) more friendly, informal,

or personalized with control tester or more formal or less

personalized with protected tester; (3) more urgency, priority,

or prompting of further action for control tester to view the

unit; (4) agent made him or herself more available to control

tester; (5) agent requested more effort or information from

protected tester before moving forward; (6) agent (potentially

inadvertently) discriminated against protected tester by asking

about disability or discussing disability in a discouraging way;

and (7) agent did not agree to or acknowledge reasonable

accommodation requests related to MD. Each of the seven

themes that emerged from the data is outlined in the following

sections, followed by examples from ﬁeld notes that epitomize

Table A-3. Response Difference in E-mail Content

Response Present in Neither Present in Both Control Group Only Protected Group Only McNemar p-Value

Positive language

Positive descriptive 45 (77.6%) 8 (13.8%) 2 (3.4%) 3 (5.2%) 1.000

View unit 14 (24.1%) 31 (53.4%) 8 (13.8%) 5 (8.6%) 0.581

Tour time offered 26 (44.8%) 18 (31.0%) 12 (20.7%) 2 (3.4%) 0.013*

Prompts 14 (24.1%) 37 (63.8%) 3 (5.2%) 4 (6.9%) 0.227

Greeting 12 (20.7%) 34 (58.6%) 7 (12.1%) 5 (8.6%) 0.774

Polite 8 (13.8%) 44 (75.9%) 4 (6.9%) 2 (3.4%) 0.687

Other units 56 (96.6%) 1 (1.7%) 0 (0.0%) 1 (1.7%) 1.000

Negative language

Disability 52 (89.7%) 1 (1.7%) 0 (0.0%) 5 (8.6%) 0.063

Discouraging 49 (84.5%) 3 (5.2%) 0 (0.0%) 6 (10.3%) 0.031*

Background 57 (98.3%) 0 (0.0%) 0 (0.0%) 1 (1.7%) NA

Employment 54 (93.1%) 3 (5.2%) 0 (0.0%) 1 (1.7%) 1.000

Fees 47 (81.0%) 5 (8.6%) 2 (3.4%) 4 (6.9%) 0.687

Negative descriptive 58 (100.0%) 0 (0.0%) 0 (0.0%) 0 (0.0%) NA

History 58 (100.0%) 0 (0.0%) 0 (0.0%) 0 (0.0%) NA

NA = not applicable.

Notes: n = 58 pairs. p-value < 0.05 indicates signiﬁcantly different treatment between the two groups. Identical e-mail replies (n = 24) were not

coded with this criterion. n = 9 did not have qualitative data available to code.

Table A-4. Subtle Discrimination Responses in Multiple Categories

Response

Number of Categories

0 1 2 3 4 5

Positive language to control class only 54 (65.9%) 21 (25.6%) 6 (7.3%) — 1 (1.2%) —

Negative language to protected class only 68 (83.0%) 11 (13.4%) 2 (2.4%) 1 (1.2%) — —

Negative and positive language against protected class only 45 (54.9%) 26 (31.7%) 8 (9.8%) 2 (2.4%) — 1 (1.2%)

Notes: n = 82 pairs. n = 9 did not have qualitative data available to code.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

it. In many cases, quotations were taken directly from landlord

responses. In some cases, the researcher’s analysis is summa-

rized given the data supporting subtle discrimination, such as

in the form of tone throughout or timing of the e-mail.

Theme 1: More encouraging to control tester or more

discouraging to protected tester.

Researcher analysis: “Agent used more encouraging

language with Control tester (C) (‘your future home,’

‘your new home’). Agent included list of amenities and

photos of rental community with C [control tester] but

not PC [protected tester]. C received two emails, while

PC got one. The additional email C received had more

information with the range of rates for 1BR and 2BR and

included an easy link to schedule a visit.” (12885)

Theme 2: More friendly, informal, or personalized with

control tester or more formal or less personalized with

protected tester.

Researcher analysis: “Both were invited for a visit. How-

ever, the agent introduced herself (‘my name is Wendy

the property manager’), asked for a contact number,

and showed more willingness (e.g., ‘I look forward to

working with you in the very near future in obtaining an

apartment!’) to assist C and not to Protected Tester (P).”

(12959)

Theme 3: More urgency, priority, or prompting of further

action for control tester to view the unit.

Researcher analysis: “P [protected tester] emailed agent

ﬁrst, but agent responded to C ﬁrst and C was given an

earlier time slot to visit.” (9668)

Researcher analysis: “No invitation to contact the landlord

for P, while C was given the landlord’s phone number.”

(12877)

Theme 4: Agent made him or herself more available to

control tester.

Researcher analysis: “Landlord made himself only

available during the working day to PC, but offered more

availability to C.” (11565)

To protected tester: “I can meet you at the building daily

10am-2pm. The unit is currently occupied till June 1.”

To control tester: “I’m by the building every day, 9am-

3:00pm. What’s a good day/time for you?” (12140)

Theme 5: Agent requested more effort or information from

protected tester before moving forward.

Researcher analysis: “PC asked the agent to suggest available

hours, but the agent did not give PC these hours; instead,

she told PC to call (extra step/effort that PC now needs to

take).” (12133)

Researcher analysis: “Agent offered tour to C immediately,

but for PC the invite was conditional (‘Please introduce

yourself a little bit more before make an appointment’). The

landlord asked several questions about the tester’s request

for an assistance animal accommodation, income, and move

in date and how many people would be renting.” (11914)

Theme 6: Agent (potentially inadvertently) discriminated

against protected tester by asking about disability or

discussing disability in a discouraging way.

Researcher analysis: “Very subtle discrimination. Agent

seems to be trying to be helpful with the parking situation,

but says ‘do you still want to see it?’ as if this may have

deterred PC. Never actually says to PC that unit is still

available, but does to C.” (12491)

Researcher analysis: “Well-meaning agent, but asks questions

about disability and mentions needing to qualify based on

the application to PC but not to C. The agent is polite to

PC (offers to give reminder calls for rent, says would love

to have as part of community, etc.), but still mentions the

need for a formal application for PC but not C.” (12946)

Theme 7: Agent did not agree to or acknowledge reasonable

accommodation requests related to MD.

Researcher analysis: “PC asked about having an assistance

dog as reasonable accommodation, but the agent respond-

ed with the building’s pet policy and that certain breeds

are not allowed in the building regardless of the disability

accommodation request.” (13249)

Researcher analysis: “Agent did not acknowledge the tes-

ter’s reasonable accommodation request. PC email request:

‘…My dog is my assistance animal and I want to make sure

she can stay with me in this apartment.’ No response from

landlord to this request.” (12719)

Using these themes, each e-mail pair was coded as containing

subtle discrimination or not (see Table A-5). This analysis

showed that about 42.7 percent of the protected testers experi-

enced some type of subtle discrimination against them, whereas

only 2.4 percent of the control group did.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Table A-5. Subtle Discrimination in E-mail Testing

Subtle Discrimination Against

the Protected Class

Count

(Percentage)

Yes 35 (42.7%)

No 41 (50.0%)

Unclear 6 (7.3%)

Subtle discrimination against the control group 2 (2.4%)

Notes: n = 82 pairs. n = 9 did not have qualitative data available to code.

Quantitative Analysis of Reasonable

Accommodation Requests

Of the 82 paired e-mail tests, 25 protected testers included a

reasonable accommodation request in their initial inquiry (that

is, asking for either a rent reminder or to have an assistance

animal in the apartment). Of those inquiries, 23 e-mail pairs

could be coded, of which the housing provider did not re-

spond to or acknowledge the request in 16 cases and approved

the request in 3 instances. In the remaining 4 instances, the

landlord did not agree to the request. In other words, they

responded negatively, such as “dogs are not allowed,” or

inappropriately, such as asking about the size of the assistance

animal before deciding whether they would accept the request.

Both of these examples represent discriminatory practices. A

subanalysis was conducted to explore which disability group

or which type of reasonable accommodations were more or

less likely to receive negative responses. See Table A-6.

Quantitative Analysis of the Status of Further

Communication

The status and overall tone of further communication with the

landlord was also coded to examine whether they left an open

door (positive to view unit) or a closed door (negative to view

unit). Although ﬁndings show that that the protected group

had slightly more closed-door episodes, signiﬁcantly different

treatment was not shown between the two groups in e-mail

testing (Table A-7).

Table A-7. Open Door/Closed Door in E-mail Testing

Response

Control

Group

Protected

Group

McNemar’s

p-Value

Open door 53 (91.4%) 50 (86.2%)

0.223

Closed door 2 (3.4%) 4 (6.9%)

Unclear 3 (5.2%) 4 (6.9%) —

Notes: n = 58 pairs. Identical e-mail replies (n = 24) were not coded

with this criterion. n = 9 did not have qualitative data to code.

Telephone Testing Findings

Qualitative Themes With Quantitative Analysis of

Subtle Discrimination

During telephone testing, six themes emerged related to subtle

discrimination against the protected tester. These themes were

(1) more urgency, priority, or prompting of further action

for control tester to view the unit; (2) more encouraging to

control tester or more discouraging to protected tester; (3) more

friendly, informal, or personalized with control tester or more

formal or less personalized with protected tester; (4) agent re-

quested more effort or information from protected tester before

moving forward; (5) agent made him or herself more available

to control tester; and (6) agent (potentially inadvertently)

discriminated against protected tester by asking about disability

or discussing disability in a discouraging way. The themes that

emerged are summarized in the following section, followed by

examples and quotations that epitomize them.

Unlike e-mail testing, telephone testing did not yield rental

agent quotations. Instead, quotations represent ﬁeld note

excerpts or the tester’s own words describing the interaction

with the agent.

Table A-6. Housing Provider Responses to RA Requests in E-mail Testing

Group Type of RA Request Number of RA Requests

Responses

Total

Positive Negative None

I/DD Rent reminder 3 0 1 2 3

Assistance animal 7 2 2 3 7

MI Rent reminder 8 1 0 7 8

Assistance animal 5 0 1 4 5

Totals 23 3 4 16 23

I/DD = intellectual or developmental disability. MI = mental illness. RA = reasonable accommodation.

Notes: n = 23 pairs. n = 2 did not have qualitative data to code.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Theme 1: More urgency, priority, or prompting of further

action for control tester to view the unit.

• Subtheme A: More prompting or urgency for control

tester to view the unit, complete an application, or follow

up.

Quotation (to protected tester): “[The agent said] I do not

give any information over the phone. Good bye.”

Quotation (to control tester): “The agent said would you

like to come in today and see the apartment?”

Researcher analysis: “No information about apartment was

given to PC, while C was given several details and invited

to view the unit.” (16484)

• Subtheme B: Agent shows priority to control tester by

re quest ing more information from him or her during the call.

Researcher analysis: “Agent asked for C’s last name, phone

number and email address to contact them, but did not

ask PC for this.” (12420)

Theme 2: More encouraging to control tester or more

discouraging to protected tester.

• Subtheme A: More units or lease options offered to

control tester.

Researcher analysis: “C offered more apartment options

(with rents ranging between $980 and $1040) and differ-

ent lengths of leases (3-13 months). PC offered one only

($1010) and length of lease wasn’t discussed.” (12420)

• Subtheme B: More positive information given about the

unit, community, or rental process to control tester or

more negative information given to protected tester.

Quotation (to protected tester): “Deborah [landlord] also

explained that Susan [renter] would need to make 3x the

rent and have a credit score of 650 or better.”

Researcher analysis: “Agent mentioned income/credit

requirement to PC tester only.” (12437)

Researcher analysis: “Agent told C about six amenities and

did not tell PC about any. Also, PC was told the applica-

tion process would take 7-10 days, while C was told it

would only take 3-5 business days.” (16235)

Theme 3: More friendly, informal, or personalized with

control tester or more formal or less personalized with

protected tester.

Quotation (to control tester): “I asked if I would just

come in and see the place and if I like it, I can rent it. He

laughed and said, ‘if we like each other, then you can rent

it. It’s not just about you.’ I said fair enough. … He asked

if I was moving with someone else. I said no. He said that’s

beautiful.”

Researcher analysis: “Agent was friendlier/joking with con-

trol tester and did not say anything like this to protected

class tester. Also, agent cut off protected class tester when

they were trying to justify their need for a reasonable

accommodation.” (12115)

Theme 4: Agent requested more effort or information from

protected tester before moving forward.

Quotation (to protected tester): “[Agent said tester’s

granddaughter with a developmental disability must call

to set up an appointment] “if she is really interested in an

apartment.”

Researcher analysis: “The agent would not accept the pro-

tected class tester calling for her granddaughter. Instead,

they required that the granddaughter call to set up the

appointment herself, which requires extra effort and may

not be possible for PC to do by self.” (15058)

Researcher analysis: “Agent asked for two pay stubs and/

or proof of income from the protected class tester but not

from the control tester.” (6696)

Theme 5: Agent made him or herself more available to

control tester.

Quotation (to control tester): “He said that he is available

today until 6pm and tomorrow from 9am-5pm to show me

the unit if I was interested. … He said I could give him a

call on his line; he said the number is different than what I

called. He gave me his personal line.”

Researcher analysis: “The control tester was told about the

agent’s availability and ofﬁce hours, whereas the protected

class tester was not. Also, the control tester was given the

agent’s personal/direct line, whereas the protected class

tester was not.” (11955)

Theme 6: Agent (potentially inadvertently) discriminated

against protected tester by asking about disability or

discussing disability in a discouraging way.

• Subtheme A: Agent questions protected tester’s competence.

Quotation (to protected tester): “[The agent] was concerned

that Susan (PC tester) would not be able to care for herself

and may leave on the gas. I assured her that the counselors

would not allow her to go on her own if they felt she would

endanger herself.” (12250)

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Quotation (to protected tester): “Is Denise (PC tester) able

to care for the dog independently and clean up behind the

animal?” (10980)

Quotation (to protected tester): [The protected tester

asked for a rent reminder as a reasonable accommodation.

The agent] “said he was willing to do that but he would

hold me [family member calling] responsible. I explained

that this would be her responsibility; she would just need

the reminder when it was close to the time. He said that I

told him that Denise had a ‘problem’ and she was autistic.

He said that if he gave her the apt he would be willing to

do what I asked. However, because she had a ‘problem’

ultimately I would have to be responsible because he

couldn’t babysit her up there in that apt.” (12115)

• Subtheme B: Agent makes other assumptions about

protected tester due to their disability.

Quotation (to protected tester): [Immediately after disability

disclosure], “the agent asked if she [the tester’s sister with

a developmental disability was on Section 8 or SSI.” (16593)

Quotation (to protected tester): “[The agent said], ‘you

mentioned your brother needs to stay calm.’ She said ‘he

isn’t violent?’ I said ‘no he isn’t.’ She said ‘we are going to

do a credit/criminal background.’ [Agent] said ‘he doesn’t

have anything on his record?’ I said, ‘no not a thing.’”

Researcher analysis: “The agent appears to be linking

mental illness to violence and a criminal record, therefore

insisting on the need for a background check. This appears

to be evidence of stigma and stereotyping.” (11927)

Basing their analysis on these themes, coders summarized

whether subtle discrimination was present (Table A-8).

The results revealed that 63 percent of the protected testers

experienced subtle discrimination, with a higher percentage

(65.7 percent) among testers with I/DD.

Table A-8. Subtle Discrimination in Telephone Testing

Disability Yes No Unclear Total n

I/DD 197 (65.7%) 83 (27.7%) 13 (4.3%) 293

MI 214 (60.8%) 126 (35.8%) 9 (2.6%) 349

Totals 411 (63.0%) 209 (32.1%) 22 (3.4%) 642

I/DD = intellectual or developmental disability. MI = mental illness.

Notes: n = 642 pairs. 10 pairs were not coded because disability was

not disclosed during testing. 16 pairs did not have qualitative data to

code.

Qualitative Themes With Quantitative Analysis:

Overt Discrimination

Three themes emerged related to overt discrimination that

would be covered explicitly under the FHA. These themes were

(1) only the protected tester received a negative response about

availability of the unit, (2) fees or rent rates quoted were higher

for only the protected testers, (3) the agent made overt com-

ments related to disability, and (4) the agent denied reasonable

accommodation requests.

Theme 1: Only the protected tester received a negative

response about the availability of the unit.

Researcher’s analysis: “Initially said unit available then

changed the answer after disability disclosure, saying “‘call

back next week.’” (15339)

Researcher’s analysis: “To PC, agent said that ‘the apartment

would be rented by the end of the week’ because they have

enough applicants whereas to C, the one bedroom was avail-

able and information was given on how to view.” (12698)

Theme 2: Fees or rent rates quoted were higher for only the

protected testers.

Researcher’s analysis: “PC tester was told rent for apartment

was $1,325. Control tester was told and got from ad that

rent was $1,005 for the same apartment.” (12165)

Researcher’s analysis: “Agent required 1 month security

deposit from PC tester, none for C.” (6728)

Researcher’s analysis: “Security deposit for Protected Class

tester is $99 to one month’s rent ($1005), for Control

tester is $99 to $300.” (7528)

Theme 3: The agent made overt comments related to disability.

Quotation: [To a question about the process for applica-

tion], “I have no process, I’m pretty busy I have no more

time to talk” and “if she needs rent reminder she needs to

stay in the group home and I have no more time to talk to

you. Good bye.” (14442)

Quotation: “We need to keep the crazies out” regarding

criminal and credit background check. (14532)

Quotation: “Do you think that she should consider renting

an apartment if someone needs to remind her to pay the

rent?” (15342)

Quotation: “She said no, we are not the type of property

that deals with people with, mental cases. I don’t really

know how to say it, but yeah, mental cases.” (11619)

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Theme 4: The agent denied reasonable accommodation

requests.

• Subtheme A: Agent denies the request directly and

unequivocally.

Quotation: “Absolutely No! That’s up to the family to

handle if you want her to rent an apartment here.” (8768)

Quotation: “She told me they don’t allow pets under any

circumstances.” (13433)

• Subtheme B: Agents not knowing the legal rights of

people with disabilities and needing to ask their managers

or consult with others (for example, homeowners’

association) to provide the testers with the answer.

Quotation: Agent said, “she is not sure about the rent re-

minder but she would inquire with the main ofﬁce about

it and would get back to me.” (12462)

Researcher analysis: “Landlord was unsure about assis-

tance animal and stated that he ‘needs to ask the home

owner’s association.’” (14472)

• Subtheme C: Verbal rent reminder request not accepted,

but alternative methods (for example, e-mail reminder,

written reminder, auto payment plan) offered.

Quotation: “When asked about the rent reminder I was

informed that they only provide a ﬁve day written notice

for those who are in jeopardy of being late.” (12108)

• Subtheme D: Assistance animal request was often denied

based on general no pet policy or weight or size policy, or

accepted but with additional pet fee charges.

Quotation: Agent said “she is sorry but no pets are al-

lowed. I said my brother has his doctor’s documentation.

She said she is sorry but they have a no pet policy.”

(12968)

Quotation: “I asked if the fees could be waived and she

[agent] explained that they could not be waived because

they would have to do it for everyone.” (13345)

Researcher analysis: Agent said that “pet over 25lbs is usu-

ally not allowed but it would be ok with documentation.

However, pet fee cannot be waived.” (21474)

• Subtheme E: Agent not acknowledging the need of an

assistance animal for a person with MD.

Quotation: “Support animal is not permissible unless the

resident was blind.” (10985)

Overt discrimination was determined based on these themes,

with results showing that approximately 57 percent of the

protected testers experienced overt discrimination (see Table

A-9). No signiﬁcant difference was found by type of MD (MI or

I/DD).

Table A-10 summarizes differences in responses between type

of requests and between different disability testers.

Table A-9. Overt Discrimination in Telephone Testing

Disability Yes No Unclear Total n

I/DD 168 (57.3%) 122 (41.6%) 3 (1.0%) 293

MI 202 (57.9%) 141 (40.1%) 6 (1.4%) 349

Totals 370 (57.6%) 263 (41.0%) 9 (1.4%) 642

I/DD = intellectual or developmental disability. MI = mental illness.

Notes: n = 642 pairs. 10 pairs were not coded because disability was

not disclosed during testing. 16 pairs did not have qualitative data

available to code.

Table A-10. Housing Provider Responses to RA

Requests in Telephone Testing

Group

Type of RA

Request

Number of

RA Requests

Responses

Positive Negative

I/DD Rent reminder 147 79 (53.7%) 68 (46.3%)

Service animal 131 70 (53.4%) 61 (46.6%)

MI Rent reminder 182 76 (41.8%) 106 (58.2%)

Service animal 155 90 (58.1%) 65 (41.9%)

Totals 615 315 (51.2%) 300 (48.8%)

I/DD = intellectual or developmental disability. MI = mental illness. RA =

reasonable accommodation.

Notes: n = 615 pairs. 37 protected testers were not included in the

analysis because they did not request RA during their testing.

Quantitative Analysis of Status of Further

Communication

Coding of the status of further communication with the land-

lord revealed that the protected testers experienced signiﬁcantly

more closed-door or negative communications, whereas control

testers experienced more open-door, positive communications.

Table A-11 summarizes the disposition of closed-door versus

open-door responses among the control group testers and

protected class testers.

Table A-11. Open Door/Closed Door in Telephone

Testing

Response

Control

Group

Protected

Group

McNemar’s

p-Value

Open door 571 (88%) 311 (47.7%)

0.000

Closed door 46 (6.4%) 205 (31.4%)

Unclear 35 (5.4%) 136 (20.9%) —

Note: n = 652 pairs.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

In-Person Testing Findings

Qualitative Themes With Quantitative Analysis:

Subtle Discrimination

During in-person testing, six themes emerged related to subtle

discrimination against the protected tester. These themes were

(1) more urgency, priority, or prompting of further action for

control tester to submit an application, follow up, or continue

the rental process; (2) more encouraging to control tester or

more discouraging to protected tester; (3) more friendly, infor-

mal, or personalized with control tester or more formal or less

personalized with protected tester; (4) agent requested more

effort or information from protected tester before moving for-

ward; (5) agent made him or herself more available to control

tester; and (6) agent (potentially inadvertently) discriminated

against protected tester by asking about disability or discussing

disability in a discouraging way.

The themes and accompanying subthemes that emerged from

the data are outlined in the following list and are followed by

quotations that epitomize them. Similar to telephone testing

data, in-person testing data were taken from tester ﬁeld notes

that testers wrote on completion of the test. The quotations

that follow each theme are ﬁeld note excerpts, or the tester’s

own words describing the interaction with the agent. Research-

er analyses are interpretations or summaries of the tester ﬁeld

note excerpts.

Theme 1: More urgency, priority, or prompting of further

action for control tester to submit an application, follow

up, or continue the rental process.

• Subtheme A: More prompting or urgency for control

tester to continue the rental process.

Researcher analysis: “The agent told the control tester to

please give him a call when they make the decision, that

they won’t ﬁnd a better unit, and that the unit may go fast.

The agent did not say these things to the protected class

tester.” (16213)

• Subtheme B: Agent shows priority to control tester by

requesting more contact information from them.

Researcher analysis: “The control tester was asked to provide

their ID, phone number, and email before the tour while

the protected class tester was only asked to provide their

ID. This is evidence of being more interested in following

up with or giving priority to the control tester.” (14056)

Theme 2: More encouraging to control tester or more dis-

couraging to protected tester.

• Subtheme A: More units or amenities shown, more options

offered, or both to control tester.

Researcher analysis: “The control tester’s tour included two

apartments, the business center, mailroom, and gym area

while the protected class tester’s tour only included one

apartment. Second, the protected class tester was offered

one apartment while the control tester was offered three

apartments.” (14082)

• Subtheme B: More positive information given about the

unit, community, or rental process to control tester or more

negative information given to protected tester.

Researcher analysis: “The control tester was given notes on

ﬂoor plans and the protected class tester received verbal

information only. Also, the control tester was told about more

amenities (parking, business center, dry cleaning service,

package receiving service, storage facilities, security deposit,

close to everything including grocery store), while the

protected class tester was told about fewer amenities (close

to metro, gym) and was also given extra discouraging infor-

mation (that utilities are not included in the rent).” (14082)

Researcher analysis: “The protected class tester was told

about more income/fee requirements. Additionally, they

were told that the application process would take 4-7 days,

while the control tester was told it would take only 3-4 days.

Finally, the agent told the protected class tester that there

was crime in the building/on the property and did not say

this to the control tester.” (13939)

Theme 3: More friendly, informal, or personalized with

control tester or more formal or less personalized with

protected tester.

Quotation (to control tester): “[The agent] introduced me to

[another staff member] and said that her name was ‘Shirley.’

… We returned to the Leasing Ofﬁce and Shirley asked me

if I had seen anything that I liked. I said that I had. Lindsay

mentioned to her that I was looking for my ﬁrst apartment.”

Researcher analysis: “The agent introduced the control tester

to another staff member and initiated conversation between

them, which is evidence of being friendly and personalized

with this tester. The agent did not introduce the protected

class tester to any other staff members. Also, the agent

offered the control tester several decorating tips on what to

do with the space and pointed out the track lighting, which

she did not do with the protected class tester, thus showing

friendliness and personalization once again.” (14082)

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Theme 4: Agent requested more effort or information from

protected tester before moving forward.

Researcher analysis: “The agent told the protected class

tester that no ﬂyers were available and asked the tester to

check the website for them. Conversely, when meeting

with the control tester, ‘went over to a ﬁle and retrieved

a ﬂyer which listed information about the apartments,

including their features, rental rates, fees, and what was

needed to apply.’” (14619)

Theme 5: Agent made himself or herself more available to

control tester.

Researcher analysis: “The agent opened the unit door

and left the protected class testers to look around the

(smelly, uncleaned) unit by themselves. The control tester

was accompanied by the agent when viewing the unit.

Additionally, the agent provided the control tester with

her name, phone number, and ofﬁce hours, but only

provided the protected class testers with her name and

phone number but no ofﬁce hours.” (14619)

Theme 6: Agent (potentially inadvertently) discriminated

against protected tester by asking about disability or

discussing disability in a discouraging way.

• Subtheme A: Agent questions protected tester’s competence.

Researcher analysis: “The agent addressed the protected

class companion tester when asking/answering questions,

rather than speaking with the tester with a disability, who

was the one actually seeking to rent the unit.” (15082)

• Subtheme B: Agent makes other assumptions about

protected tester due to their disability.

Researcher analysis: “The agent mentioned Section 8 housing

to the protected testers, who replied by saying they were

actually looking for market rental housing, not Section 8

housing. Even after the testers made this clariﬁcation, the

agent still mentioned Section 8 housing a second time.”

(13783)

Based on these themes, subtle discrimination was coded (Table

A-12). In all, 45 protected testers (54.2 percent) experienced

Table A-12. Subtle Discrimination in In-Person Testing

Disability Yes No Unclear Total

I/DD 23 (47.9%) 24 (50.1%) 1 (2.0%) 48

MI 22 (62.9%) 13 (37.1%) 0 (0.0%) 35

Totals 45 (54.2%) 37 (44.6%) 1 (1.2%) 83

I/DD = intellectual or developmental disability. MI = mental illness.

Notes: n = 83 pairs. 18 pairs did not have qualitative data to code.

subtle discrimination during their visit to the apartment, with

a higher percentage of subtle discrimination (62.9 percent)

experienced by testers with MI.

Qualitative Themes With Quantitative Analysis:

Overt Discrimination

In in-person testing, similar themes to those in telephone testing

emerged related to overt discrimination; however, reasonable

accommodation requests were not done during the in-person

testing.

Theme 1: Only the protected tester received a negative

response about the availability of the unit.

Researcher analysis: “Units were not available for PC to

view. The agent went to make copies of PC’s ID before

showing the units, talked to another staff for 3-4 mins, and

came back out saying that contractors are working in both

units thus units so not available for viewing. No further

invite was offered to come back to see the units after the

work is done. C was able to view units.” (14284)

Theme 2: Fees or rent rates quoted were higher for only the

protected testers.

Researcher analysis: “To C agent said that there is no security

deposit and now they do non-refundable move-in fee

of $200. PC was told that there is a one month security

deposit ($1005).” (14283)

Theme 3: The agent made overt comments related to disability.

Researcher analysis: “Agent inquired about nature and

severity of disability. Agent asked what kind of disability

PC has.” (16521)

Table A-13 shows the analysis of occasions of overt discrimi-

nation. Far less overt discrimination occurred in the in-person

testing than in the telephone testing. Out of 83 testers, 11 pro-

tected testers (13.3 percent) experienced overt discrimination.

The discrimination rate was higher with testers with MI.

Table A-13. Overt Discrimination in In-Person Testing

Disability Yes No Unclear Total

I/DD 4 (2.1%) 43 (89.6%) 1 (2.1%) 48 (100%)

MI 7 (20.0%) 28 (80.0%) 0 (0.0%) 35 (100%)

Totals 11 (13.3%) 71 (85.5%) 1 (1.2%) 83 (100%)

Notes: n = 83 pairs. 18 pairs did not have qualitative data to code.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Quantitative Analysis of Status of Further

Communication

Coding of the status of further communication with the

landlord shows that the protected group had slightly more

occasions when further communication was closed, but the

differences were not signiﬁcant (Table A-14).

Table A-14. Open Door/Closed Door in In-Person

Testing

Response

Count (Percent)

Control

Group

Protected

Group

McNemar’s

p-Value

Open door 81 (97.6%) 78 (94.0%)

0.250

Closed door 2 (2.4%) 5 (6.0%)

Notes: n = 83 pairs. 18 pairs did not have qualitative data to code.

Discussion and Conclusions

This mixed-methods approach to qualitatively analyzing data,

followed by summarizing ﬁndings quantitatively to examine

differences between people with and without MD in market

rental searches, produces new and signiﬁcant ﬁndings that

further document speciﬁc examples and frequencies of both

subtle and overt housing discrimination against people with MI

and I/DD. These ﬁndings, including implications for education

and training of housing providers on the FHA and speciﬁcally

on reasonable accommodations based on MD within rental

housing, are summarized in this section.

Telephone Testing

• Some agents speciﬁcally cited the FHA but misinterpreted

it. They said that they follow the FHA and, therefore, they

cannot give anyone special treatment—they need to treat

everyone the same. This ﬁnding seems to show progress in

that they are aware of the FHA and are trying to follow it,

but more education about disability reasonable accommo-

dations is needed.

• A trend that emerged was that, with some protected testers

with I/DD, the agent appeared excited about renting the

unit to the tester and would cite positive safety amenities

of the unit and building (for example, gated community)

but would later ask for extra eligibility requirements,

fees, or veriﬁcation. This practice seems to show intent to

initially appear fair, encouraging, and open to the tester

but to later subtly discriminate by making it more difﬁcult

for protected testers to rent.

• Some agents did not know about their ability to approve

a reasonable accommodation and had to ask a manager

before providing the protected tester with an answer. In a

few cases, however, the agents asked the manager while on

the call and then approved the reasonable accommodation

request. This ﬁnding shows that, although many agents

are still not trained in fair housing requirements, some are

making the effort to ensure they are giving the prospective

tenant the correct answer. Also, the managers in these

cases are knowledgeable about the FHA and reasonable

accommodations, which demonstrates the positive effect

the law has had in recent years.

• For protected testers who asked for an assistance animal

reasonable accommodation, some agents would ask

questions related to the building’s pet policy. For example,

some agents asked the breed or size of the animal, and,

although the agents ended up approving the reasonable

accommodation request because the tester’s animal ﬁt

the pet policy parameters, this request would likely have

been denied if the tester had a different assistance animal

that did not ﬁt general requirements. This practice shows

potential discriminatory behavior, but it did not play out

in these particular cases due to the tester’s scenario rather

than the agent’s decision or behavior.

Both Telephone and In-Person Testing

• Protected testers were often told that an application fee

would be charged for each adult applicant, whereas control

testers were told about this extra fee less often. Even when

protected testers clariﬁed that only the person with a

disability would be applying for and renting the unit, some

agents still felt the need to reiterate this additional fee.

Differences Detected

• Overt discrimination practice was found to be much more

likely in telephone testing (57 percent) than in in-person

testing (13 percent). One potential explanation of the

lower overt discrimination rate in in-person testing is

that, in in- person testing, the testers had already made an

appointment in advance and knew the unit was ready to

be viewed. In telephone testing, however, many protected

testers were told that units were not available compared

with control testers who were given information about

available units. Open door/closed door analysis supports

that a signiﬁcant difference in unit availability exists

between control and protected testers.

• Subtle discrimination occurred to more than one-half of

the protected testers in both telephone (63 percent) and

in-person testing (54 percent). During in-person testing,

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

agents could not overtly discriminate because the testers

were already there to view the unit, but the agents still

practiced subtle discrimination with the intention of

discouraging the protected tester from taking the next step

or as a result of asking about MD.

• During in-person testing, testers with MI showed higher

discrimination rates than testers with I/DD in both subtle

and overt discrimination, whereas during telephone testing

discrimination rates against testers with MI and I/DD

were similar. Given the fact that the in-person tests were

conducted by persons with MI and I/DD, as opposed to

the telephone tests, which were conducted by proxies, this

observation adds weight to the notion that persons with

MI experience higher rates of housing discrimination when

compared with persons with I/DD.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

Appendix B: References

Ahmed, Ali M., and Mats Hammarstedt. 2008. “Discrimination

in the Rental Housing Market: A Field Experiment on the

Internet,” Journal of Urban Economics 64 (2): 362–372.

American Psychiatric Association (APA). 2000. Diagnostic and

Statistical Manual of Mental Disorders, 4th ed. Arlington, VA:

American Psychiatric Association Publishing.

Braddock, David, Richard Hemp, Mary C. Rizzolo,

Laura Haffer, Emily Shea Tanis, and Jiang Wu. 2011.

The State of the States in Developmental Disabilities 2011.

Boulder, CO: Coleman Institute for Cognitive Disabilities

University of Colorado. https://www.researchgate.net/

proﬁle/ Richard_Hemp/ publication/8377686_The_State_

of_the_States_in_ Developmental_Disabilities_2011/

links/55df7d4808ae6abe6e86538a.pdf.

Brault, Matthew W. 2012. “Americans with Disabilities: 2010.”

Current Populations Report #P70-131. Washington, DC: U.S.

Census Bureau.

Corrigan, P.W., and D.L. Penn. 1999. “Lessons From Social

Psychology on Discrediting Psychiatric Stigma,” American

Psychologist 54 (9): 765–776.

Dalgin, Rebecca Spirito, and James Bellini. 2008. “Invisible

Disability Disclosure in an Employment Interview: Impact

on Employers’ Hiring Decisions and Views of Employability.”

Rehabilitation Counseling Bulletin 52 (1): 6–15.

Denny-Brown, Noelle, Debra Lipson, Matthew Kehn, Bailey

Orshan, and Christal Stone Valenzano. 2011. “Money Follows

the Person Demonstration: Overview of State Grantee Progress,

January to June 2011.” Prepared for the Centers for Medicare

& Medicaid Services, Division of Community Services

Transformation, Disabled and Elderly Health Programs Group.

Cambridge, MA: Mathematica Policy Research.

Hanson, Andrew, Zackary Hawley, and Aryn Taylor. 2011.

“Subtle Discrimination in the Rental Housing Market: Evidence

From E-mail Correspondence With Landlords,” Journal of

Housing Economics 20 (4): 276–284.

Housing Discrimination Law Project of Vermont Legal Aid,

Inc. (HDLP). 2012. Rental Discrimination Audit Report: Housing

Discrimination in Vermont Rental Markets. Burlington, VT:

Vermont Legal Aid.

Kaye, Kimberly E. 2012. “Public Stigma Toward Housing

Applicants With Psychiatric Disabilities.” Master’s thesis,

Department of Psychology, San Diego State University.

MacDonald-Wilson, Kim L., Zlatka Russinova, E. Sally Rogers,

Chia Huei Lin, Terri Ferguson, Shengli Dong, S., and Megan

Kash MacDonald. 2011. “Disclosure of Mental Health Disabilities

in the Workplace.” In Work Accommodation and Retention in

Mental Health, edited by Izabela Z. Schultz and E. Sally Rogers.

New York: Springer: 191–217.

McNemar, Q. 1947. “Note on the Sampling Error of the Difference

Between Correlated Proportions or Percentages,” Psychometrika

12 (2): 153–157.

National Council on Disability (NCD). 2001. Reconstructing Fair

Housing. Washington, DC: National Council on Disability.

National Fair Housing Alliance (NFHA). 2015. Where You Live

Matters: 2015 Fair Housing Trends Report. Washington, DC:

National Fair Housing Alliance. http://www.nationalfairhousing.

org/FairHousingResourceCenter/ReportsandResearch/

tabid/3917/Default.aspx.

Ng, Terence, Alice Wong, and Charlene Harrington. 2013.

“Home and Community-Based Services: Introduction to Olm-

stead Lawsuits and Olmstead Plans.” San Francisco: University

of California, San Francisco, National Center for Personal Assis-

tance Services. http://www.americanbar.org/content/dam/aba/

events/homelessness_poverty/2013_Annual_Meeting_Medicaid/

intro_to_olmstead_lawsuits_and_plans.authcheckdam.pdf.

U.S. Department of Housing and Urban Development (HUD).

2002. Discrimination in Metropolitan Housing Markets: National

Results From Phase I HDS 2000. Washington, DC: U.S. Depart-

ment of Housing and Urban Development.

———. 2005. Discrimination Against Persons With Disabilities:

Barriers at Every Step. Washington, DC: U.S. Department of

Housing and Urban Development, Ofﬁce of Policy Development

and Research. http://www.urban.org/research/publication/

discrimination-against-persons-disabilities/view/full_report.

———. n.d. Live Free: Annual Report on Fair Housing,

FY 2010. http://portal.hud.gov/hudportal/documents/

huddoc?id=ANNUALREPORT2010.PDF.

RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: RESULTS OF PILOT TESTING

STUDY OF RENTAL HOUSING DISCRIMINATION ON THE BASIS OF MENTAL DISABILITIES: FINAL REPORT

———. 2013a. An Estimate of Housing Discrimination Against

Same-Sex Couples. Washington, DC: U.S. Department of Hous-

ing and Urban Development, Ofﬁce of Policy Development

and Research. https://www.huduser.gov/portal/publications/

fairhsg/discrim_samesex.html.

———. 2013b. “Service Animals and Assistance Animals

for People With Disabilities in Housing and HUD-Funded

Programs.” Press release. Washington, DC: U.S. Department

of Housing and Urban Development. http://portal.hud.

gov/hudportal/HUD?src=/press/press_releases_media_

advisories/2013/HUDNo.13-060A.

———. 2014a. Annual Report on Fair Housing: FY 2012–2013.

Washington, DC: U.S. Department of Housing and Urban

Development.

———. 2014b. “Fair Housing Enforcement Organizations Use

Testing To Expose Discrimination.” Evidence Matters, Spring/

Summer 2014. http://www.huduser.org/portal/periodicals/em/

spring14/highlight3.html.

———. 2015. Discrimination in the Rental Housing Market Against

People Who Are Deaf and People Who Use Wheelchairs: National

Study Findings. Washington, DC: U.S. Department of Housing

and Urban Development, Ofﬁce of Policy Development and

Research. https://www.huduser.gov/portal/sites/default/ﬁles/

pdf/housing_discrimination_disability.pdf.

Wagner, Karen L. 2008. “Tools of the Online Trade,” Journal of

Property Management. 73 (4): 32–37.