FMR Literature Review

!"#$%&"$'()%*(+)%,-$.(/%0(1#2+3%
4%5#)($")-$(%*(.#(6%"+7%,)-7/%0(1#2+%
89"+%
Frederica R. Conrey, PhD
Randal ZuWallack
ICF Macro
Leslyn Hall
Redstone Research
Doray Sitko
Ryan Sullivan
Fred Eggers, PhD
Econometrica

!

Table of Contents!
"#!

$%&'()*+&,(%!##########################################################################################################################################!"!

-#!

./)/'01!233(')041/!5(*6,%7!8'(7'096!##################################################################################################!:!
-#"#!

:#!

=#!

-#"#"#!

5(*6,%7!>?(,+/!@(*+?/'!8'(7'09!########################################################################################!A!

-#"#-#!

8*41,+!5(*6,%7!#######################################################################################################################!A!

-#"#:#!

5BCD!####################################################################################################################################!E!

-#"#=#!

5(*6,%7!F'*6&!.*%)!###############################################################################################################!G!

-#"#A#!

H*II('&,J/!5(*6,%7!8'(7'09!################################################################################################!G!

-#-#!

K(L!$%+(9/!5(*6,%7!F0M!>'/),&6!##################################################################################################!N!

-#:#!

;%,&/)!H&0&/6!(J/'07/!D''('!####################################################################################################################!":!

=#"#-#!

H09I1,%7!D''('!#####################################################################################################################!-"!

=#"#:#!

;%,&!T(%U'/6I(%6/!D''('!#####################################################################################################!--!

=#-#!

O/6I(%6/!2++*'0+P!D''('6!###########################################################################################################!-N!

=#-#"#!

$&/9!T(%U'/6I(%6/!D''('!####################################################################################################!-N!

=#-#-#!

C/06*'/9/%&!D''('!<*/!&(!&?/!O/6I(%)/%&!######################################################################!:Q!

=#-#:#!

C/06*'/9/%&!D''('!<*/!&(!$%&/'J,/L/'6############################################################################!::!

=#:#!
A#!

5;(11/+&,(%!810%!####################################################################################################################!:G!

A#-#"#!

K(+0&,(%6!#############################################################################################################################!:G!

A#-#-#!

O<(11/+&,(%!############################################################################################################!:N!

A#-#:#!

2RH!<0&0!>(11/+&,(%!#############################################################################################################!:V!

A#:#!

2%01P6,6!810%################################################################################################################################!="!

A#:#"#!

W/P!C/06*'/6!######################################################################################################################!="!

A#:#-#!

2%01P&,+!2II'(0+?!###############################################################################################################!=-!

A#:#:#!

O/6/0'+?!X*/6&,(%6#############################################################################################################!==!

O/3/'/%+/6!##################################################################################################################################################!=N!
2II/%),M!2Y!!C0,1!H*'J/P!$%6&'*9/%&ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZ#A:!

Table of Exhibits
F041/!"Y!
./)/'01!8'(7'096!H*II('&,%7!T/L!233(')041/!5(*6,%7!###########################################################!=!
F041/!-Y!
8?(%/!H&0&*6!4P!5(9/!BL%/'6?,I!H&0&*6!3('!29/',+0%!.09,1,/6!-QQV!#################################!"=!
F041/!:Y!
VA[!>(%3,)/%+/!$%&/'J01!3('!&?/!=Q&?!8/'+/%&,1/!####################################################################!-"!
F041/!=Y!
(6&!#########################!-A!
.,7*'/!"Y! .CO!H*'J/P!DMI/',9/%&!H&'*+&*'/!##########################################################################################!:E!
F041/!AY!!!!!!!!>?0'0+&/',6&,+6!(3!$)/01!K(+0&,(%!80,'6!####################################################################################!:N!
F041/!EY!! O<(11/+&,(%!H+?/)*1/!################################################################################################!:V!
F041/!GY!! 2RH!<0&0!>(11/+&,(%!H+?/)*1/!#################################################################################################!=Q!
.,7*'/!-Y!! K/J/16!(3!2%01P6,6!####################################################################################################################!=:!

1. INTRODUCTION
The mission of the Department of Housing and Urban Development (HUD) is “to create strong,
sustainable, inclusive communities and quality affordable homes for all.” (HUD, “About HUD”). The
Housing Choice Voucher (HCV) program is an extension of this mission and a response to the need for
affordable rental housing. The program has changed several times since its inception as Section 8 of the
U.S. Housing Act of 1937. Today, “through tenant-based vouchers, HUD provides rental subsidies for
standard-quality units that are chosen by the tenant in the private market. The subsidy amount is based on
a payment standard set by the Public Housing Authority (PHA) between 90 percent and 110 percent of
the fair market rent (FMR).” (HUD, “Housing Choice”). The Code of Federal Regulations (CFR) defines
an FMR as “the rent, including the cost of utilities (except telephone), as established by HUD… for units
of varying sizes (by number of bedrooms), that must be paid in the market area to rent privately owned,
existing, decent, safe, and sanitary rental housing of modest (non-luxury) nature with suitable amenities.”
(24 CFR 888.111).
The U.S. Housing Act, as amended by the Housing and Community Development Act of 1974, requires
HUD to establish and publish FMRs at least annually. Because rents vary considerably between localities,
FMRs are area-specific. As noted by HUD in the Overview of Fair Market Rents:
The U.S. Department of Housing and Urban Development (HUD) annually estimates FMRs for
530 metropolitan areas and 2,045 non-metropolitan county FMR areas… HUD defines FMR
areas as metropolitan areas and non-metropolitan counties. With a few exceptions, the most
current Office of Management and Budget (OMB) definitions of metropolitan areas are used
(HUD, “Fair Market Rents”).
HUD calculates FMRs to support public-subsidized housing programs. The FMRs are used for HUD
programs, IRS programs, and local programs to help determine appropriate subsidy levels for families
based on their income to ensure their ability to secure housing in the areas they want to live. FMRs
include both cost of the rental along with utilities (but excluding internet, cable or satellite television, and
telephone service). In the case of the HCV program, HUD allocates monies to local PHAs. PHAs in turn
administer the HCV program and provide the rental assistance payment to the landlord on behalf of the
voucher holder. These Congressionally appropriated funds originate as taxpayer dollars, and HUD is
responsible for maintaining fiscal stewardship over the funds. Therefore, FMRs must be set at a value
high enough to grant access to decent, safe, and sanitary housing of a modest nature with suitable
amenities, but also low enough to ensure that the greatest number of eligible persons benefit from
availability of the vouchers.
The process by which HUD estimates an FMR has been relatively consistent since it was introduced.
Historically, HUD has used Decennial Census data (projected forward) to estimate FMRs. Since 2005,
HUD has supplemented the Decennial Census data with data from the ACS. Ultimately, because ACS
data are collected continuously, the ACS data will supplant Decennial Census data in the estimation.
In addition to Census data, HUD has used custom survey data to estimate FMRs in some areas. The
surveys, the subject of the present review, have been RDD surveys to landline numbers of people who
have moved in the past 24 months and were presently living in two-bedroom units in buildings at least
two years old and not owned by a PHA (HUD, 2007). Overtime, the RDD surveys expanded the
definition of an “eligible” survey participant to those meeting the above criteria and who did not live in a
vacation or seasonal home and who did not do work for a landlord in exchange for a reduction in rent.
While the process for collecting the data to calculate an FMR has been relatively consistent, the
affordable housing market and the American communications landscape have changed significantly.
!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%:%

Furthermore, best practices in survey research have also changed and continue to evolve. A quarter of
American adults now rely exclusively on cell phones, and a further 18 percent use their cell phones for
“most or all” calls (Blumberg & Luke, 2010). The numbers are higher for the renter population where 46
percent are cell-only. That means that almost half of American adults are now largely unreachable on
conventional landlines.
The affordable rental housing market also offers more choices than it did 40 years ago. Tenants can
access affordable rental housing by living in a subsidized building or using a voucher to rent a marketbased apartment. Developers are provided incentives and loans to build rental housing where all or a
portion of the rental units are “affordable.” Some of these projects are managed by local housing
authorities; some are privately undertaken. It is more and more difficult to identify housing that is rented
for the “market rate”.
To ensure that the FMR Surveys and the FMRs based on them accurately represent the cost of housing to
families with vouchers, HUD is undertaking an evaluation of new survey methodologies to address the
changes in telecommunications, affordable housing, and survey research best practices. In this report, we
offer some background to the affordable rental housing market. It is necessary to understand the specific
housing options in order to identify the precise set of rents eligible for the FMR survey. We review the
literature relevant to the accurate collection of rental data using surveys and the accurate estimation of
FMRs. Finally, we offer an evidence-based study design plan for an experiment to determine the optimal
FMR survey design.

!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%;%

2. FEDERAL AFFORDABLE HOUSING PROGRAMS
The United States first acknowledged its role in protecting and providing affordable housing in U.S.
Housing Act of 1937’s Declaration of Policy which stated in part, “It is the policy of the United States to
promote the general welfare of the Nation by employing its funds and credit, as provided in this Act, to
assist the several states and their political subdivisions to remedy the unsafe and unsanitary housing
conditions and the acute shortage of decent, safe, and sanitary dwellings for families of lower
income…”. 1 A report from the Joint Center for Housing Studies of Harvard University (JCHSHU)
reiterates that need for affordable housing has risen and has seen a sharp increase in these first years of
the new millennium (JCHSHU, 2001).
HUD’s Worst Case Housing Needs 2009 Report to Congress found “dramatic increases in worst case
housing needs…that cut across demographic groups, household types, and regions.” (Steffen, Keith,
Martin, Teresa, Vandenbroucke, & Yao, 2011). The report also found that, since 2003, vulnerable renters
had faced the “tightest market” for affordable housing than at any time since 1985 (Steffen, et al., 2011).
A 1999 HUD report on the rental-housing crisis indicated that at that time, there were approximately one
million households on Section 8 waiting lists, with an average waiting time of 28 months (HUD, 1999). A
preliminary assessment by the National Low Income Housing Coalition (NLIHC) notes the following
regarding available, affordable rental units:
A preliminary analysis of the 2008 American Community Survey (ACS), with comparison to the
2007 ACS, shows that the shortage of housing affordable for extremely low-income households
has increased. For every 100 extremely low income (ELI) renter households, in 2008 there were
just 37 rental units that were both affordable and available to them. There were 39 such units in
2007 (Pelletier, 2001).
Clearly, no one disputes the need for affordable housing. Rather, the question is how to best address this
shortage. Some of the many proposed (and existing) means include:
x
x
x
x

Publicly funded public housing; government subsidies for new private housing (e.g., Low
Income Housing Tax Credits (LIHTC); local tax incentives);
Government funding of existing private housing (e.g., HCV);
Subsidies for goods and services beyond housing (e.g., mental health services, food
assistance, etc.); and
Private funding of alternate solutions.

Since the 1930s, the Federal government has directly funded the development of new, affordable housing,
originally as public housing owned and operated by public or quasi-public entities. When HUD first
started using FMRs in the 1990s, housing assistance primarily came in one of two forms: (1) subsidized
apartments in buildings owned and operated by local public housing authorities (project-based
assistance), or (2) Section 8 vouchers that allowed individuals to rent elsewhere (tenant-based assistance).
Today, a multitude of Federal programs exists to subsidize the development and construction of new,
1

United States Housing Act of 1937, Pub. L. 93-383, 88 Stat. 653 (codified as amended at 42 U.S.C. 1437 et.
seq.)(hereinafter the 1937 Act) as it was amended by the Quality Housing and Work Responsibility Act of 1998,
Pub. L. 105-276, 112 Stat. 2518(enacted October 21, 1998). This document also reflects amendments made to the
1937 Act by the Departments of Veterans Affairs and Housing and Urban Development, and Independent Agencies
Appropriations Act, 1999, Pub. L. 105-276, 112 Stat. 2461 (enacted October 21, 1998)(hereinafter Appropriations
Act, 1999) and the Omnibus Consolidated and Emergency Supplemental Appropriations Act, 1999, Pub. L. 105-277
(enacted October 21, 1998)(hereinafter Omnibus Act, 1999). http://www.hud.gov/offices/ogc/usha1937.pdf
(accessed May 2, 2011).

!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%<%

affordable housing—primarily by partnering with private developers. As one might expect, many of these
programs are administered by HUD, though the Internal Revenue Service (IRS) and U.S. Department of
Agriculture (USDA) also have roles in funding affordable rental housing (Table 1). The increase in
number and variety of programs designed to provide affordable housing make it more difficult to
accurately identify if rental units are market rate units or not.
Table 1:

Federal Programs Supporting New Affordable Housing

Administering
Agency

Program

Description

Relies in Part
on FMR?

HUD

Public Housing

Funds to develop public housing units

Yes

HOME

Formula grant, many eligible uses

Yes

HTF

Formula grant, primarily for rental housing

SHP

Competitive grant to house homeless individuals
and families

Yes

IRS

LIHTC

Tax credits for new affordable rental housing

Yes

USDA

Rural Rental
Housing

Loans for low-income rental housing in rural
areas

No

Section 538
Guaranteed Loans

90% loan guarantee for low-income rental
housing loans

No

Farm Labor
Housing

Loans and grants to develop housing for
domestic farm laborers

No

According to the National Council of State Housing Finance Agencies, an estimated two million
Federally subsidized apartments for low-income tenants were built between 1986 and 2006 (NCSHFA,
2008). In fact, “by 2008, there were nearly 33 percent more homes built under new government lowincome housing programs (after 1986) than there were subsidized apartments built by all the HUDsponsored programs dating back to the 1960s” (Erickson, 2009).
The proliferation of housing subsidy programs has made it increasingly difficult to identify and collect
rent information from eligible housing units. In the past, when there were fewer programs and more
straightforward financing of housing assistance, it was easier for survey researchers (and for respondents
themselves) to reliably determine whether respondents' housing units were subsidized in whole or part.
With the advent of programs that provide subsidies during the loan and building phase as well as in
varying modes during occupancy, it is more difficult to determine whether the rental cost of the unit
reflects a market rent or not.

2.1. HUD PROGRAMS
HUD administers a number of programs to support the development of new, affordable rental housing.
HUD has four mission-oriented divisions: Housing, Community Planning and Development (CPD),
Public and Indian Housing (PIH), and Fair Housing and Equal Opportunity (FHEO). Of these, only
FHEO does not provide funds to support new, affordable housing. There are many more programs that
may be used to support new housing units as well. For example, Community Development Block Grant
!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%=%

(CDBG) funds may be awarded by a grantee to a Community Based Development Organization (CBDO)
to construct new housing units as a part of neighborhood revitalization, community economic
development, or energy conservation, but Federal regulations prohibit the use of CDBG funds to
construct housing in almost any other circumstance. Neighborhood Stabilization Program (NSP) funds
may be used to construct new housing as part of an effort to redevelop demolished or vacant properties,
but the primary purpose of NSP is to remove or return foreclosed and abandoned properties to the housing
market. To effectively establish the FMR for a region, it is necessary to identify whether and through
what program survey respondents’ rents are subsidized. The following subsections discuss those
programs for which the primary purpose is to support new, affordable housing.

2.1.1. Housing Choice Voucher Program
HUD describes the Housing Choice Voucher Program (HCV) as “the Federal government’s major
program for assisting very low-income families, the elderly, and the disabled to afford decent, safe, and
sanitary housing in the private market” (HUD, “Housing Choice Vouchers”). It is the largest of the
various programs authorized by Section 8 of the United States Housing Act of 1937 for the payment of
rental housing assistance to private landlords; it pays a large portion of the rents and utilities for about 2.1
million households annually. With the HCV program, people are free to choose any housing that meets
program requirements. People do not have to live in subsidized housing projects; they can choose
apartments, mobile homes, townhouses, and even single-family homes. HUD’s PIH allocates vouchers to
local public housing authorities to administer.
According to HUD’s Website, local PHAs determine who is eligible for a housing voucher based on a
family’s total annual gross income and number of family members. A family's income may not exceed 50
percent of the median income for the county or metropolitan area in which the family chooses to live. “By
law, a PHA must provide 75 percent of its voucher to applicants whose incomes do not exceed 30 percent
of the area median income” (HUD, “Housing Choice Vouchers”). The program is limited to U.S. citizens
and specified categories of non-citizens who have eligible immigration status.
Once a family has received a voucher, it is their responsibility to identify and chose a place to live where
the owner will agree to rent under the HCV program. If a landlord agrees to rent as part of the HCV
program, the local PHA will determine if the rental unit meets HUD’s minimum standards for health and
safety. If the unit meets these standards, then the PHA will pay the housing subsidy directly to the
landlord for the voucher holder. The participating family then pays the difference between the actual rent
changed by the landlord and the amount subsidized by the program.

2.1.2. Public Housing
In addition to administering the HCV program, which provides subsidies to low-income renters to allow
them to rent existing housing units on the private market at affordable rates, PIH also provides funds to
PHAs to develop new public housing units. PHAs may use HUD funds to develop public housing units in
any generally accepted way, including (but not limited to) the following:
x
x
x
x
x

Conventional: The PHA owns and designs the project and bids out development to a
contractor.
Turnkey: A developer prepares and develops the project, and the PHA purchases the
completed project from the developer.
Acquisition: The PHA purchases an existing property (with or without rehabilitation).
Mixed-finance entities: Organizations other than the PHA may own part of the project.
Force account: The PHA prepares and develops the project itself.

!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%>%

HUD imposes a limit on the number of public housing units that a PHA may develop. Each PHA may not
develop units beyond the number of units that it had on August 21, 1996, or the number of units for which
it was receiving operating subsidy on that date, whichever is less.
Residents of public housing–whose income must not exceed 80 percent of area median income (AMI)–
pay rent based upon their income. Rent may be set to any of the following standards:
x
x
x
x

Thirty percent of a resident’s monthly adjusted income;
Ten percent of a resident’s monthly gross income;
Welfare rent (not applicable in all states); or
A fixed rent of $25–50 set by the PHA.

2.1.3. HOME
The HOME program has created approximately one million new, affordable rental and homeownership
units since 1992 (Federal Register, 2010). HUD’s Office of Community Planning and Development
(CPD) administers the HOME program, a formula grant designed specifically to increase the supply of
affordable housing. HUD awards HOME funds annually by formula to participating jurisdictions –cities,
counties, states, and consortia of local governments. Participating jurisdictions may use HOME funds to
carry out a number of eligible activities within four main categories:
x
x
x
x

Rental housing (including acquisition, construction, and rehabilitation);
Rehabilitation of owner-occupied housing;
Homeownership assistance (including secondary mortgages, down-payment assistance, and
assistance for new construction); or
Tenant-based rental assistance.

Participating jurisdictions have broad discretion to fund new, affordable rental housing in a variety of
ways (e.g., soft costs, hard costs, relocation costs, etc.). Restrictions are minimal, but a participating
jurisdiction may not use HOME funds to assist public housing units. Projects receiving HOME funds
must receive, on average, no less than $1,000 per unit. 2 Additionally, within two years of receiving its
annual HOME entitlement grant, a participating jurisdiction must commit the entire grant amount by
entering into written agreements with developers, owners, subcontractors, or sub-recipients. Fifteen
percent of the grant amount must be committed to Community Housing Development Organizations
(CHDOs), which are private, non-profit organizations meeting special requirements within the HOME
governing regulations (24 CFR 92.2).
For each HOME entitlement grant it receives, a participating jurisdiction must make sure that 90 percent
of the rental units it funds are occupied by households at no more than 60 percent of AMI. Additionally, if
a rental housing project receives HOME funds and has at least five units, then at least 20 percent of the
units must be occupied by households at no more than 50 percent of AMI. The latter requirement applies
to each project for a minimum of 20 years. HOME program regulations require that all tenants recertify
their income annually. HUD allows maximum HOME rents at the lesser of the following:
x
x

Rent less than or equal to 30 percent of the adjusted income of a family at 65 percent of AMI;
or
Area FMR.

2

HUD also sets limits to the amount of HOME funds that a project may receive. These limits, known as Section
221(d)(3) program limits, vary by metropolitan area and fiscal year.

!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%?%

The rents are part of the universe of rents eligible for the FMR Survey. Furthermore, HUD’s regulations
for the HOME program require that “Owners [of HOME-funded rental housing] may not refuse to lease
HOME-assisted units to a certificate or voucher holder under the Section 8 Program, or to a holder of a
comparable document evidencing participation in a HOME tenant-based rental assistance (TBRA)
program, because of the status of the prospective tenant as a holder of such certificate, voucher or
comparable HOME TBRA document” (HUD, 2006). 3

2.1.4. Housing Trust Fund
Recently, HUD proposed a new subpart to the regulations governing the HOME program to codify the
Housing Trust Fund (HTF), which was created by the Housing and Economic Recovery Act of 2008. The
purpose of HTF is to increase and preserve the supply of rental housing for households earning up to 50
percent of AMI and also to increase homeownership for such households. HUD will award HTF grant
funds to states on a formula basis. HTF, as proposed, will function similarly to HOME, and many of the
same requirements will apply. For example, states must commit HTF funds within two years of receipt.
Though states may use HTF funds for both rental and owner-occupied housing, the program clearly
prioritizes rental housing. States must use no less than 80 percent of their HTF funds to develop,
rehabilitate, or preserve affordable rental housing, and no more than 10 percent for homeownership.
Unlike HOME, HTF allows states to set per-unit subsidy limits, requires a 30-year period of affordability
(though states are free to increase the affordability period), and mandates that assisted units meet energy
and water efficiency standards.

2.1.5. Supportive Housing Program
HUD offers a number of programs to support housing and services for homeless individuals and families.
Established in 1987 under Title IV, Subtitle C, of the McKinney-Vento Homelessness Assistance Act of
1987 (HUD, “Supportive Housing Program”), the Supportive Housing Program (SHP) is a competitive
grant that allows recipients to construct new housing units, among other eligible activities. Funds for new
construction are limited to $400,000. Recipients may use SHP funds to construct permanent housing (i.e.,
long-term housing for persons with disabilities) or transitional housing (i.e., housing that facilitates the
movement of homeless individuals and families to permanent housing). SHP funds may not be used for
emergency shelters. Residents of SHP-funded units must be homeless (HUD, 2008). SHP recipients may
charge rent, though doing so is not required. If a recipient chooses to charge rent, then rent must be no
more than 30 percent of the tenant’s monthly-adjusted income or 10 percent of monthly gross income.
Tenants of SHP-funded housing who pay rent must have their income reviewed at least annually by the
SHP recipient.

3

See also 24 CFR 92.252(d). Note also that “In accordance with the Section 8 program rule at 24 CFR
982.352(c)(6), Section 8 rental assistance voucher and certificate holders cannot also receive TBRA under the
HOME Program because the two programs would provide duplicative subsidies. HOME TBRA recipients who are
offered a Section 8 voucher or certificate must relinquish HOME assistance, if they wish to accept the Section 8
assistance. Similarly, a family currently receiving Section 8 rental assistance may not accept HOME TBRA without
relinquishing the Section 8 assistance. However, a Section 8 rental assistance recipient may receive HOME-funded
security deposit and utility deposit assistance. Similarly, a family cannot receive HOME TBRA if they are receiving
rental assistance under another Federal program (e.g., Section 521 of the Housing Act of 1949 provided through the
Rural Housing Service) or a State or local rental assistance program, if the HOME subsidy would result in
duplicative subsidies to the family” (HUD, 1996).

!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%@%

Recipients may also use SHP funds to lease buildings or units to provide supportive housing to homeless
individuals and families. A recipient may lease a unit directly from a landlord and then place an eligible,
homeless tenant in the unit. The recipient may charge an affordable rent to the homeless tenant. HUD
requires that the recipient pay no more than FMR to the landlord.

2.2. LOW INCOME HOUSING TAX CREDITS
One of the most significant subsidies for affordable housing is the LIHTC program. HUD does not
administer the LIHTC program. Rather, it is a tax credit authorized as part of the Federal income tax code
(26 U.S.C. §42). The IRS allocates these tax credits to states. States then award tax credits to housing
developers on a competitive, project-specific basis. Developers, in turn, sell the tax credits to raise funds
for affordable housing projects.
The Joint Center for Housing Studies points out that:
Market-rate rentals accounted for little more than half of the 300,000 new multifamily units
completed each year from 1995 through 2009. Of the remainder, 23 percent were assisted rentals
produced through the Low Income Housing Tax Credit Program, and the other 24 percent were
intended for sale as condominiums (JCHSHU, 2001).
To be eligible to receive LIHTC assistance, projects must be residential and developers must commit to
an affordable rent ceiling as set by HUD. Rent ceilings reflect rents that are affordable (i.e., no more than
30 percent of a household’s income) to households earning 60 percent of AMI. Developers commit to an
affordability period of at least 30 years (more in some states), during which time rent ceilings apply to at
least 20 percent of units. If a developer sets aside 20 percent of units as affordable, then they must be
occupied by households at no more than 50 percent of AMI. If a developer sets aside 40 percent of units
as affordable, then they must be occupied by households at no more than 60 percent of AMI. Of course,
projects that are 100-percent affordable are also eligible. Tenants of LIHTC-funded housing must report
and recertify their income annually.
The LIHTC program uses FMRs to identify Difficult Development Areas (DDA). By definition, DDAs
have high construction, land, and utility costs relative to AMI. DDAs are limited in size and may contain
no more than 20 percent of the Metropolitan Statistical Area (MSA) or non-metropolitan counties. DDAs
are eligible for a greater share of LIHTCs than other areas (130 percent of qualified basis).

2.3. UNITED STATES DEPARTMENT OF AGRICULTURE
HUD is not the only department in the United States government that provides incentives for building or
providing affordable housing or subsidies for individuals to access housing. The USDA also offers
programs specifically aimed at supporting rural residents. These programs do not rely on the local FMR.
USDA offers four main sources of funds in support of affordable rental housing:
x
x
x
x

Section 515 Rural Rental Housing loans,
Section 538 Guaranteed Loan program,
The Farm Labor Housing program, and
The USDA Rental Assistance Program.

Section 515 Rural Rental Housing loans are direct loans awarded to owners of affordable rental housing
on a competitive basis. Loans are issued at terms of up to 50 years with a one percent interest rate. Forprofit and non-profit developers are eligible to apply. To be eligible for Section 515 Rural Rental Housing
loans, new housing must be in a rural area and 95 percent of units must be occupied by very-low-income

!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%A%

households (Housing Assistance Council, 2010). Generally, recipients of Section 515 Rural Rental
Housing loans also utilize USDA Section Rental Assistance (Cowan, 2008).
The Section 538 Guaranteed Loan program allows USDA to guarantee market rate loans for new or
existing housing in rural areas. To be eligible, average gross rent for all units must not exceed 30 percent
of AMI, adjusted for family size. USDA guarantees up to 90 percent of a qualifying loan, with a
repayment term between 25 and 40 years (Housing Assistance Council, 2008).
Tenants can use HCV to rent units built with Section 515 or 538 loan funds provided that the property
owner agrees to accept them.
The Farm Labor Housing program provides loans and grants to develop housing for domestic farm
laborers. Grant funds and loans are available to farm worker associations, non-profit organizations, Indian
tribes, and public agencies. USDA issues Farm Labor Housing loans at 33-year terms with one percent
interest. Grants may be used to fund up to 90 percent of development costs. Individual farmers,
associations of farmers, and family farm corporations may apply for loans but are not eligible for grants.
Tenants of Farm Labor Housing-funded units must be U.S. citizens or permanent residents and must earn
more than 50 percent of their income from farm work.
The USDA rental housing assistance program is similar to HUD’s Section 8 new construction projectbased vouchers (Section 521), providing an additional subsidy for tenants in Section 515- or 514/516financed rental housing with incomes too low to pay the U.S. Department of Agriculture’s Rural
Development Housing and Community Facilities Programs office (RD) subsidized rent from their own
resources. RD pays the owner the difference between the tenant’s contribution (30 percent of adjusted
income) and the monthly rental rate, which is calculated based on the owner’s project costs” (Housing
Assistance Council, 2008).
USDA rental assistance housing rents would not be part of the universe of eligible rents for the FMR
surveys, yet rural rent housing loan-funded housing would be.

!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%B%

3. THE EXISTING FMR SURVEY
Again, the FMR is “the rent, including the cost of utilities (except telephone), as established by HUD…
for units of varying sizes (by number of bedrooms), that must be paid in the market area to rent privately
owned, existing, decent, safe, and sanitary rental housing of modest (non-luxury) nature with suitable
amenities.”( 24 CFR 888.111). In order to collect the data to calculate FMRs based on the RDD landline
survey, HUD has used a questionnaire with three subsections that:
x
x
x

Determine the eligibility of the household reached;
Collect rent information; and
Determine which utilities the respondent pays for if any.

For the FMR questionnaire, HUD has historically defined eligible survey participants as:
x
x
x
x
x
x
x
x
x
x

Renters;
Living within the targeted geography;
Not living in public housing;
Residing in a residential home such as an apartment, house, or mobile home ( not a barracks,
dormitory, half-way home, hospital, prison, group home, etc.);
Living in a home with two bedrooms;
Having moved into the home within the last two years;
Living in a building that had been built more than two years ago;
Living in a permanent residence and not a seasonal or vacation property;
Individuals who did not perform work for a landlord for a reduction in rent; and
Individuals who could report the rent on their home including any and all subsidies.

The FMR questionnaire did not require participants to report the monthly costs of their utilities.
Participants indicated only which utilities that they paid for, and then also reported the primary fuel type
used to provide that utility.
Once the FMR survey data was collected, rent information was used in conjunction with utility cost
schedules provided by HUD’s regional offices. The utility cost schedules provided information for the
local area’s monthly average cost for all utilities for different fuel types. Rent and the appropriate utility
cost information were added together to calculate a gross rents for the area. A distribution of these gross
rents would then be calculated, and the FMR would be set at the 40th or 50th percentile. From this twobedroom rent standard, HUD would then adjust the FMR for different sized units using based on CPI
data.

3.1. THE FMR PERCENTILE THRESHOLD
Originally, HUD set FMRs at the 45th percentile (i.e., the dollar value below which 45 percent of
standard quality units were rented). In 1995, though, HUD changed the definition of FMRs to reflect rents
at the 40th percentile. This was done primarily as a cost saving measure (Federal Register, 1995). In
response, individuals voiced many concerns. One of the major concerns raised was that the change would
limit the number of standard quality units available to Section 8 voucher holders and result in areas of
concentrated poverty.
HUD responded with two points. First, HUD noted that it considers only recent movers for the purpose of
calculating FMRs. Because tenants who remain in a unit for an extended time often pay less rent than a
recent mover would pay for a comparable unit, considering only recent movers to some extent inflates

!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%:C%

FMRs. Consequently, HUD argued, voucher holders would have access to more than 40 percent of the
actual number of standard quality rental units on the market.
HUD’s second point was that rent-eligible units were not, in fact, concentrated in high-poverty areas and
that the change in percentile would not affect concentration. According to HUD’s analysis, in
approximately 85 percent of neighborhoods with 10 or more two-bedroom rental units, 30 percent or
more of those units were rented below FMR. HUD concluded that reducing the FMR percentile to 40
percent would not affect the adequacy and choice of housing available to voucher holders.
HUD changed the definition of FMRs again in 2000. Though HUD kept FMRs at the 40th percentile in
most areas, HUD allowed some PHAs to choose between the 40th and 50th percentiles. In other areas,
HUD set FMRs at the 50th percentile.
HUD had two reasons for increasing the FMR to the 50th percentile in some areas. First, despite its earlier
analysis that suggested that the 40th percentile was sufficient to maintain the adequacy and choice of
housing available to voucher holders, HUD found that voucher holders in some areas were not able to
find standard quality rental units at the 40th percentile. PHAs have flexibility to set payment standard
amounts at 90–110 percent of the applicable FMR. In 2000, HUD allowed PHAs that set payment
standards at 110 percent of FMR at the 40th percentile to change the FMR to the 50th percentile if less
than 75 percent of families receiving vouchers within a six-month period were able to find housing. PHAs
choosing apply the 50th percentile retain the ability to set payment standard amounts at 90–110 percent of
FMR.
HUD also found that market conditions in some areas confined voucher holders to areas of concentrated
poverty (Federal Register, 2000). In order to promote choice of neighborhood, HUD increased FMRs to
the 50th percentile in areas meeting the following criteria:
x
x
x
x

Evidence shows that low-income families live in concentrated areas.
The FMR area has no less than 100 census tracts.
In no more than 70 percent of census tracts with 10 or more two-bedroom rental units, 30
percent or more of those units have gross rents at or below the 40th-percentile FMR.
At least 25 percent of voucher holders in the area live in the five percent of census tracts with
the largest number of voucher holders.

If HUD determines that an area qualifies for 50th-percentile FMR, that area retains 50th-percentile FMR
for a period of three years. At the end of the three-year period, HUD reevaluates for continued use of
50th-percentile FMR. HUD publishes a list of 50th-percentile FMR areas each year on the Federal
Register.

3.2. BEDROOM SIZE
Because data on less-common unit sizes may be insufficient or unavailable, HUD generally calculates
FMRs for two-bedroom units–nationally the most common rental unit type, comprising approximately 40
percent of the renter-occupied units in the United States. HUD then sets FMRs for other unit sizes based
on the two-bedroom FMR. One-bedroom units represented approximately 27 percent of the responses,
and three-bedroom units represented approximately 23 percent of the responses. Collectively, one-, two-,
and three-bedroom units account for around 90 percent of renter occupied units according to the AHS.
As HUD’s Office of Policy Development and Research points out regarding the two-bedroom unit size,
“being the most common unit, they are the easiest units for which to obtain data. HUD estimates FMRs
for efficiencies, one-bedroom units, three-bedroom units, and units of other sizes using the two-bedroom

!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%::%

estimate as a base.” However, this may be a changing phenomenon. The 2009 AHS revealed that 341 of
the 941 new construction (four years old or less), renter occupied units were built with three bedrooms as
opposed to 314 of the same type built with two bedrooms(US Census Bureau, 2009). In 2007, 379 of the
1,036 total same type of units were built with three bedrooms as compared to 322 two-bedroom units (US
Census Bureau, 2007). In their July 10, 2010 press release regarding the 2009 AHS survey, HUD noted
that “Most homes have three or more bedrooms (64 percent compared to just 48 percent in 1973). New
homes generally have more bedrooms – 80 percent of them have three or more bedrooms”(HUD, 2010).
A more detailed analysis is needed to determine whether this change represents a shift in the market and
is expected to continue or is merely an anomaly, the effect on rental stock, and implications, if any, for
the FMR calculation.

!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%:;%

4. REVISING THE FMR SURVEY DESIGN
The goal of the present research effort is to re-examine the FMR Survey in light of changes in the housing
assistance market and the communications landscape. Broadly, HUD desires to create the questionnaire,
sampling frame, and survey protocol that lead to the best data quality. Survey data quality is a complex
concept, encompassing elements of the sample, the administration protocol, the cognitive tasks required
of respondents, and respondents’ reactions to the survey itself. Survey conclusions have high quality to
the extent that “error” has been minimized at every survey stage. There are many sources of survey error,
and a useful way of organizing them was introduced by Weisberg (2005):
x
x
x
x
x
x
x
x
x
x
x
x

Respondent selection errors
Coverage error
Sampling error
Unit non-response error
Response accuracy errors
Item non-response error
Measurement error due to respondents
Measurement error due to interviewers
Survey analysis and interpretation errors4
Post-survey errors
Mode effects
Compatibility effects

Precision reflects the degree to which the estimate might change due to natural variation if the study were
done again. Sometimes, this is represented as a margin of error (e.g., “plus or minus three percentage
points”). Accuracy reflects the degree to which the estimate is systematically different from the true value
because something is leading to a bias in results (e.g., over-representation of women in the sample). These
types of survey error may have an impact on precision, accuracy, or both.

4.1. RESPONDENT SELECTION ERRORS
The population for an FMR Survey is the set of all rents of standard-quality units that Housing Choice
Voucher recipients could rent. Note that this population is a population of rental units not a population of
people. The final estimate should generalize to units that are currently rented as well as those that are
currently unoccupied—all those units that voucher recipients could rent. To accurately represent this
population, every market rent should have a known, non-zero chance of being represented in the survey,
and HUD should be able to identify and remove responses from any person whose rent is not in the
market (e.g., a person who lives in public housing). This section discusses the ways in which design
choices affect the completeness of the sample coverage and the accuracy with which we identify people
whose rents are in the target market.

4.1.1. Coverage Error
Coverage error is error associated with incomplete coverage of the population. The goal is to achieve a
probability sample in which every rental unit in the population has some known, non-zero chance of
being included in the sample. Very often in practice, a perfect probability sample cannot be achieved.

4

Weisberg refers to this category as “administration issues”, but its contents concern how data are treated,
combined, and interpreted after collection.

!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%:<%

There are two kinds of coverage errors that can happen: some units can have zero chance of being
included and some units can have some chance but the actual probability is unknown.
In the current FMR methodology, the sources of under-coverage (some units in the population have zero
chance of being included are):
x
x
x
x

Unoccupied units.
Units occupied by people with no telephones at all.
Units occupied by people accessible exclusively or primarily by cell phone.
Units that screen out of the survey based on questions that ask about housing subsidies.

The first three of these have to do with the source of sample—RDD telephone numbers. The last has to do
with the structure and content of the questionnaire. We discuss coverage error associated with the sample
frame first.

Scope of Current Sample Frame Coverage Error
Coverage error introduces bias when the units we cannot reach differ from those we can reach. If
unoccupied units have exactly the same rent as occupied units, then their absence from the survey sample
does not bias the estimates. If their rent is much higher or lower, then the coverage error introduces bias
in the estimates.

Unoccupied Units
Currently, the FMR Surveys only call people’s homes, so units that are available for rent but not occupied
by someone with a telephone are not included in rent estimates. Especially in rent-controlled areas, rents
on units that are currently rented may rise at a lower rate than rents on unoccupied units. That is, a family
who rented a unit 24 months ago may have seen an increase of four percent per year on their initial rent,
but a new family renting a unit in the building might find that the market rate had risen by 15 percent in
24 months.
This is a noteworthy limitation for FMR Surveys, but impractical to overcome. FMR Surveys must focus
on collecting rent data for occupied units and develop appropriate estimation models to accommodate the
unoccupied units.

Units Occupied by People Without Telephones
According to data from the Centers for Disease Control and Prevention (CDC), in 2009, an estimated 2.1
percent of families did not have a phone. Families without a phone were more likely to rent their homes,
but the total proportion of all renters who did not have a phone was low at 3.9 percent (Table 2) (CDC,
2009).
Table 2:

Phone Status by Home Ownership Status for American Families 2009

!
/#!01#)-!
2-&&3#)&4!
5%)6&7)-!')&4!
5%)6&7)-!%)6!2-&&!89:%&3;*-<=!

"#$%&! '()*!+#,-! .-)$*!+#,-!
-#"[!
"#Q[!
:#V[!
-A#G[!
"=#=[!
=A#N[!
"A#-[!
"=#E[!
"A#N[!
AG#"[!
GQ#Q[!
:=#A[!

Most researchers seem to agree that coverage error attributable to people without any phones is not a
major concern for most general population surveys. Although these people are especially likely to be

!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%:=%

renters, they represent a very small proportion of all renters, so their absence from FMR samples may not
be a major problem.

Units Occupied by People Accessible Exclusively or Primarily by Cell Phone
As we noted above, a quarter of the adult population has only a cell phone, and more have a landline but
primarily use cell phones for voice communication (Blumberg & Luke, 2010). In 2009, almost half (46
percent, Table 2) of people who rented their homes had only cell phones. That means that FMR Surveys
are missing at least that much of the renting population nationally.
Most researchers agree that under-coverage of cell phone users could pose a substantial threat to survey
estimates of all kinds. Where FMR Surveys are concerned, it is important to note that coverage bias can
impact the means and proportions that are generally studied, but it can also impact variances and
covariances (Peytchev, Carley-Baxter, & Black, 2011). Since the distribution of rents measured in FMR
Surveys is as important as the mean or median (the critical measure is the 40th percentile), this is a
particular concern for HUD.

Sample-Frame Construction Approaches
Researchers have proposed several approaches to addressing the coverage error associated with wireless
substitution:
x
x

x

x

Continue to conduct RDD surveys using landlines only. (Blumberg & Luke, 2009) argue
that landline-only RDD surveys are still an acceptable alternative to dual-frame surveys
(containing both landline and cell phones) since bias is relatively low from their perspective.
Use white pages lists and a cell phone RDD frame. This recently proposed alternative
(Guterbock, Diop, Ellis, Holmes, & Le, 2011) makes use of the efficiency realized by
switching from landline RDD to listed phone numbers to offset the increased cost of
conducting cell phone interviews.
Conduct RDD surveys using both cell phone and landline numbers. This increasingly
popular approach has two subvariants: dual-frame RDD surveys where the cell phone
component includes only users who rely exclusively on their cell phones and dual-frame
RDD surveys where both frames contain dual-users.
Employ an Address-Based Sample. Address-based sampling (ABS) is increasing in
popularity because everyone, or nearly everyone, in the population has one address that is
their primary residence.

The first of these options may have been feasible in 2009, but in 2011 there is a growing body of evidence
that weighting adjustments alone cannot account for all the differences between landline and cell phone
users (e.g., Blumberg & Luke, 2009; Peytchev, et al., 2011). Over time, this bias may decrease as
wireless-only users become more like the general population. However, the rapid rate of wireless
substitution suggests that landline users will ultimately be the minority and will differ in important ways
from the general population. Continuing to conduct RDD surveys using landline phone numbers only is a
short-term solution; another alternative is needed for the FMR Surveys in the long-term.
The approach assumes that cell phones are the base sample frame, and landlines are needed to “patch” a
hole in the frame that will disappear over time. However, the frame coverage is still not complete. The
authors found slight bias in their estimates, but note that the bias will decrease over time as more homes
are covered by cell phones. There are two reasons that a full RDD dual-frame is preferable. First, it is not
clear what proportion of recent renters would have listed their landline phone numbers. Second, FMR
Surveys require substantial geographic precision, so the primary driver of cost is not the landline
interviews but the effort and screening involved in completing the cell phone interviews. We have not
!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%:>%

proposed to test this sampling approach in the 2011 experiment, but we do see it as a marked
improvement over the landline-only RDD. Thus, it is one alternative to consider if, in future, HUD finds
implementing the more rigorous sampling approaches cost-prohibitive.
Either a combined RDD frame or an ABS frame could provide complete or close-to-complete coverage.

Sample Frames for FMR Survey Experiment
Both an ABS frame and a dual-frame RDD will provide near 100 percent coverage of the household
population. Another consideration when evaluating sampling frames is efficiency. While efficiency is not
related to quality, it does have cost implications. Two critical aspects of efficiency related to FMR
Surveys are: 1) the ability to geographically target specific areas, and 2) the efficiency in reaching renteroccupied units.

Landline RDD Frame
Most RDD samples are drawn from a list-assisted sampling frame that is constructed from telephone
exchanges associated with residential landlines. All possible numbers in this set of exchanges are grouped
into blocks of 100 as in: ZZZ-XXX-XX00- ZZZ-XXX-XX99 where ZZZ is the area code and XXX-XX
is the five-digit exchange. These 100 blocks are checked against telephone directories, and blocks with no
listed numbers (zero-blocks) are dropped or truncated. The blocks with at least one listed number are
considered 1+ blocks, or working banks.
Truncating the RDD sampling frame is done to increase efficiency of the sample, but it opens the
possibility of sample under-coverage. If all the numbers in all zero blocks really are unused by
households, then there is no sample under-coverage associated with dropping the blocks. If, however,
some zero blocks actually have unlisted households in them, then dropping those blocks means dropping
some unlisted households.
One recent study estimated that up to 20 percent of landline households are excluded from landline RDD
samples in this way (Fahimi, Kulp, & Brick, 2009), but other recent studies have estimated the undercoverage rate at five percent (Boyle, Bucuvalas, Piekarski, & Weiss, 2009) and seven to 14 percent
(Barron, et al., April, 2010). The winning perspective in this controversy seems still to be undecided in
the literature.
!"#$%&'()*+,-%&-).)*&-)#/+
Since the RDD telephone numbers are cross-referenced to telephone directories, there exists a link
between telephone number and geography. The directory-listed telephone numbers are mapped and
assigned to a specific geographic location (such as a census block group, a census tract, or a ZIP code).
Telephone lines are not restricted by geographic borders, but are generally associated with finite
geographic areas. The mapping results in a many-to-many association between telephone exchanges and
geographic boundaries (i.e. many exchanges associated with many geographic areas). The association
between geographic area and telephone exchanges is quantified by tallying the number of directory-listed
households in each geographic area by exchange combination. The geographic area is assigned to the
telephone exchange with the most number of listed telephones (the rule of plurality). After each
geographic area has been assigned to an exchange, the exchanges inherit the demographic and
socioeconomic characteristics of the geographic areas. These exchange characteristics can be used for
targeting certain geographic areas such as an FMR area or targeting geographic areas with high
concentrations of renters.

!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%:?%

Cell Phone RDD Frame
Similar to the landline sample, a cell phone RDD sampling frame is constructed from telephone
exchanges associated with cellular telephones. The North American Numbering Plan Administration
governs the assignment of area codes, exchanges and 1000-blocks (ZZZ-XXX-XX00- ZZZ-XXX-X999 )
of telephone numbers in the United States. The cell phone sample is selected from the frame of all 100
blocks assigned for cellular service.
!"#$%&'()*+,-%&-).)*&-)#/+
The association with geography is much weaker for cell phones than for landlines. There is no directory
listing of cell phones that ties the cell phone to a place of residence. Cell phone geographic stratification
is limited to county and is based on the physical location of the “switch-center” where the cell phone is
first activated for service. This relationship is tenuous for three main reasons:
x
x
x

Cell phones are portable and many people will move out of county, but keep their cell phone
number;
The county where the cell phone is purchased and activated may not be the county where the
user lives; and
Switch-centers are located around larger population centers; over half of the counties in the
U.S. do not have a switch-center.

Address-based Sampling Frame
ABS sample frames come from mailing lists provided by private vendors. Most of these lists are based on
the USPS Delivery Sequence File (DSF), although many vendors use proprietary systems to update and
clean the lists so that performance may differ across company sources. Norman and Sigman (2009)
mention a study (Link, Battaglia, Giambo, Frankel, Mokhad, & Rao, 2005) that compared coverage of
mailing lists obtained from five address vendors and found that Marketing Systems Group (the sampling
arm of which is known as Genesys) had notably better coverage than three other vendors and better
precision in method than the remaining company.
More generally, the list vendor should have a license to use the Computerized DSF because coverage and
accuracy of DSF lists updated using this system are higher (Dohrmann, Han, & Mohadjer, 2006). Also,
some addresses, especially those in rural areas are “simplified” and do not contain unit or street numbers.
Some vendors, including MSG, can use proprietary databases to augment address information and make it
usable (Norman & Sigman, 2009).
The DSF contains information about all the deliverable mail addresses in the United States including
(Norman & Sigman, 2009) addresses that:
x
x
x
x

Receive or have received mail delivery,
Receive seasonal mail delivery,
Are city route street addresses but receive mail at a PO Box, and
Are city route street addresses but do not receive mail because they are vacant.

Studies comparing Census household counts to address list counts have found that mail address list
coverage is very high in places with fairly dense populations and much lower in rural areas with sparse
populations. Staab and Iannacchione (2003) estimated that national coverage was 97 percent but coverage
was only 83 percent in local areas (defined by city names) with populations less than 10,000. Other
studies that have used a match method based on trained observer enumeration of dwellings have similarly
observed that coverage in urban areas is near 100 percent while coverage of rural areas can be much

!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%:@%

lower. One study found that mailing lists and enumerated lists matched only 72 percent of the time in a
rural South Atlantic community (Dohrmann, Han, & Mohadjer, 2006).
There are two kinds of households that are not covered by the DSF (Staab & Iannacchione, 2003):
x
x

Households that are not on mail routes and receive mail only through PO Boxes, and
Rural route and highway boxes.

Rural route and highway boxes must be excluded from survey samples because we cannot know how
many households are served by each drop point. However, the number of these boxes will continue to
decrease in number as communities move to assign city-style addresses to rural homes to support
enhanced 911 systems (Staab & Iannacchione, 2003). In 2007, about three percent of the adult population
reported receiving mail at “an address with a rural route number” 5 but not also at a street address (Data
from (National Cancer Institute, 2007).
PO Boxes are more problematic. If HUD chooses to include PO Boxes, it will have complete coverage of
people who receive mail at PO Boxes but not at home, but people who receive mail in both places may be
counted twice in the survey. If HUD chooses to exclude PO Boxes, then each included household will
only have one chance of being selected, but no one without a city route street address will be included.
Further, it is difficult to locate PO Boxes within a specific geographic location, since the PO Box and the
PO Box owner are not necessarily co-located.
Iannaccione, Staab, and Redden (2003) found that in Dallas County a large majority of PO Box users also
received mail at home. However, in a later study the result was not found to generalize to a larger
geographic area (Staab & Iannacchione, 2003). Nationally, 88 percent of adults said they received mail at
a street address. Seven percent received mail at a PO Box, but 28 percent of those said they also received
mail at a street address (Data from (National Cancer Institute, 2007). So, in general, about five percent of
the population of adults would be excluded by a DSF frame that excluded PO Boxes.
The threat of under-coverage, however, varies with the mix of geographies in the survey region. Studies
that specifically enumerated housing on DSF routes that do not appear in the DSF (which could be houses
that receive their mail through a PO Box) have found that the incidence of such units is low (e.g., 1.8
percent; (Staab & Iannacchione, 2003).
For the most part, geographies where FMR Surveys are conducted will not be majority rural where the
threat of under-coverage of locations with no mail service is high. However, in cases where this may be a
problem (e.g., when the frame contains many rural routes), HUD could supplement the survey data with
data from the ACS, which is not based on a DSF frame. In such a case, the primary source would be the
survey data, but any available ACS data could be combined with the survey data using small area
estimation techniques. Thus, it seems that the best option for the FMR Survey is to exclude PO Boxes
from the DSF-based frame.
!"#$%&'()*+,-%&-).)*&-)#/+
ABS is a selection of addresses, each with a geographic location. Since the sample is directly associated
with a location, the address frame can be geographically stratified down to the census block level. This
information can be used to precisely locate neighborhoods with high concentrations of renters and
ultimately increase the efficiency of reaching respondents qualified for the FMR Surveys.
Further, an ABS sample can be selected for very small geographic areas, such as census tracts with 100
percent geographic accuracy. This is a distinct advantage over RDD sampling.

5

This estimate from a mail survey that used a DSF frame but included all PO Box and rural route addresses.

!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%:A%

Coverage Error in FMR Surveys
Modifying the sampling frame will go a long way toward addressing the coverage error currently
associated with the FMR Surveys. However, there are some possible threats to coverage that can be
addressed by special analyses of the 2011 test data:
x

Coverage

x

Unoccupied units. If the challenge here is faster-rising rents for unoccupied than for
occupied rents, then respondents who moved in one year prior should have higher
rents than respondents who moved in two years prior. By comparing rents reported
by renters with different tenures, we can estimate the rate of change of rents in
unoccupied units and determine whether an adjustment to the FMR is needed.
o Households without a telephone. The number of these units is small. Only if these
rents are very different from the rents of other respondents, and these differences
cannot be addressed through adjustment, is this coverage error a serious problem.
The FMR Survey experiment can address this by collecting phone usage information
from ABS sample respondents and comparing those without a phone to those with a
phone.
o Cell-only households. Both dual-frame and ABS designs cover the cell-only
population, but the mode of data collection differs between the two designs. It is
highly unlikely that we will have a phone number and an address for a cell-only
household. Therefore, a mail/Web administration will be required for ABS, while a
telephone/Web administration will be required for dual-frame.
o No address. The DSF, the source of ABS sample, is maintained by the USPS and
contains physical addresses as well as PO Boxes and central distribution boxes. We
discuss the DSF and its strengths and weaknesses as a sample source further below in
“Address-based Sampling Frame ”. Here, it is useful to note that we can ask about
how each household contacted in the RDD survey receives mail to evaluate the scope
of coverage error associated with ABS sampling in this population.
Efficiency
o

o

o

o

Targeting small geographic areas. FMR Surveys are conducted in a defined
geographic area. Sometimes these areas are large, such as a metropolitan area, but
may be as small as a tract or a group of tracts. The ability to target small geographic
surveys with cell phone samples is limited. We will evaluate the geographic
incidence of reaching a cell phone respondent in a small geographic unit, such as a
group of tracts.
Reaching renter-occupied units. Over 50 percent of renters are cell-only. Albeit
expensive, cell phone samples are an effective way to reach renters. ABS samples are
more cost-effective in general, but the ability to reach renters may require more mailbased screening, and thus increased printing and postage costs. The survey
experiment will determine cost efficiency in reaching renters.
Field period. ABS samples will generally require a longer field period than a dualframe survey. The longer field period for ABS allows for multiple contacts via
mail—a longer process than multiple contacts via telephone. The anticipated
production rate for renters will be known within days of starting the telephone survey
and the sample size is easily adjusted to meet the target number of respondents. It
will take several weeks to realize the production rate for a mail survey and the
number of completed interviews will be more variable.

!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%:B%

Screening Households out of the FMR Survey
As we mentioned above, housing assistance programs have proliferated since the inception of the FMR
survey. It is now more difficult than ever for HUD and for respondents themselves to determine whether
the rents they pay are “market rents”. To determine which rents are eligible, it is useful to recall that the
FMR survey should produce an estimate of how much a new family receiving a voucher would have to
pay to obtain housing. From this perspective, the programs themselves fall into three basic categories:
x
x
x

Rents on units that are ineligible for the market rate survey. This group includes public
housing because HCV recipients do not obtain housing in buildings owned by the PHA or by
the USDA
Rents on units that are eligible for the market rate survey. This group includes housing that
was funded with special loans or other money from HUD or USDA. Families receiving HCV
could rent these units.
Rents on units that are eligible for the market rate survey whose residents do not pay the
market rate. This, the most problematic group, includes units whose current residents already
receive HCV or another subsidy. These units themselves probably have market rates that are
paid by a combination of HUD and the resident. A new family with HCV could rent the unit
for the total rent being paid. The problem is that the resident does not personally actually pay
the market rate.

The survey currently includes several questions designed to identify residents of public housing. The
internal consistency of these questions can be evaluated to determine whether a subset of these questions
could be as effective as the full set. Rents reported by respondents who are identified as living in housing
owned by the government should not be included in the FMR calculation.
The survey also asks whether the respondent receives a subsidy and how much the rent would be if the
subsidy were not in place. If these questions are effective in eliciting the true total rent, then this is the
best solution to covering rents paid by people who receive rental assistance. However, there is good
reason to believe that many respondents will have difficulty answering the question about what their rent
would be if they paid the whole thing.
This question calls for the respondent to construct a “counterfactual”, an imaginary alternative to reality.
This is a special kind of cognitive process, and a counterfactual constructed in the service of answering a
survey question requires special cognitive effort. There is evidence that this question would is easier to
answer for people whose “true” rent is closer to the amount they actually pay than for people whose rent
is more heavily subsidized (see Roese, Sanna, & Galinski, 2005 for a review), and the values that people
give may be influenced by their actual rent paid through priming or context effects. In short, there is a
significant possibility that asking respondents to report the market rent for a unit for which their own rent
is subsidized produces biased results.
HUD can explore this issue in the experiment survey data by examining:
x
x
x

Item non-response rates for the counterfactual question
Variance of responses to the counterfactual compared to variance of responses made by
individuals whose rent is unsubsidized
Correlation of self-paid rent with reported subsidy amount

If there is evidence that the question is introducing bias, then the alternative is to exclude the responses of
people whose rent is subsidized from the FMR calculation. This exclusion will likely bias estimated FMR
upward since rents of subsidized units are almost certainly systematically lower than unsubsidized rents
(so excluding them would leave more higher rents in the analysis). The scope of this bias can be evaluated
!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%;C%

by examining what proportion of all rents are subsidized (as reported in the survey). If only a small
proportion of rents is subsidized, then eliminating this difficult question from the survey could be
warranted.

4.1.2. Sampling Error
Sampling error is the error, or noise, associated with measuring a sample rather than the whole
population. Sampling error is inherent in all probability-based samples (samples where everyone in the
population has a known, non-zero probability of being included). Because the probabilities are known, the
sampling error can be estimated and the magnitude of the error can be reported--as the sample size goes
up, sampling error goes down (Weisberg, 2005), meaning the estimate is more precise.
Currently the value of interest is the 40th percentile (or another percentile in some cases) of rents. One
valuable way to look at this information is as reflecting the access that families have to housing. A family
with a voucher is supposed to have access to 40 percent of the local housing. Because of sampling error,
the value that is calculated as the 40th percentile could really be the 35th percentile or the 45th percentile.
We want a sample size large enough to ensure that we do not accidentally set the FMR substantially
below the true 40th percentile.
There are several advanced variance estimation techniques for computing standard errors and confidence
intervals for a median or other percentile. A complete review of approaches to variance estimation for
estimates of 40th percentiles is beyond the scope of this review (but see Kovar, Rao, & Wu, 1988 for a
summary). There is some evidence that different approaches perform similarly in the estimation of home
sales prices from complex survey data (Thompson & Sigman, 2000).
It is important to have some sense of what the sampling error might be for FMR Surveys so that we can
evaluate the ideal sample size. Historically, FMR Surveys have used the Woodruff (1952) method for
calculating confidence intervals around the 40th percentile. The method uses the observed distribution of
values from the sample and the fact that the percentile is a value that represents a position in the
distribution—the 40th percentile is the value in which 40 percent of the values are less than or equal. An
estimate of 40 percent of the population has variance that can be calculated for simple random samples as
demonstrated below as well as for complex samples such as dual-frame RDD and ABS. Thus the 95
percent confidence interval is the 40%±1.96!se(40%)RU“¥îQIRUVLPSOHUDQGRP
samples. The confidence interval for the 40th percentile is bounded by the percentile for the lower
confidence limit and the percentile for the upper confidence limit.
Table 3:

95% Confidence Interval for the 40th Percentile

>?@!2A!B#?@!2A!B#(%'/P]!-Q""c#!
H/J/'01!6&*),/6!?0J/!3(*%)!&?0&!&?/!,%+/%&,J/6!
0'/!+(6&!/33/+&,J/!b6//!H,%7/']!-QQ-!3('!0!
'/J,/Lc]!4*&!&?/!,%+,)/%+/!,%!&?,6!6&*)P!,6!30,'1P!
1(L#!F?0&!,6]!9(6&!I/(I1/!L?(!'/+/,J/!6*'J/P6!
L,11!%(&!4/!d*01,3,/)!'/%&/'6#!C0%P!I/(I1/!L?(!
'/+/,J/)!&?/!,%+/%&,J/!L(*1)!%(&!4/!/1,7,41/!3('!
&?/!6*'J/P#!

8"2(%;>%

2#)67$7#)%&!
A)E-)$7N-!

014*7E%&!
M:?0%+/!(3!'/6I(%6/!,%+'/06/6!L,&?!&?/!
*6/!(3!6I/+,01!)/1,J/'P!6/'J,+/!bD)L0')6]!/&!01#]!
-QQVci!,9I0+&!(%!%(%U'/6I(%6/!4,06!*%e%(L%#!

/-H%$7N-#!B))6!(3!'/6I(%6/!,%+'/06/!
6*46&0%&,011P!L?/%!6I/+,01!)/1,J/'P!6/'J,+/!,6!
*6/)!bD)L0')6]!/&!01#]!-QQVc]!4*&!&?/!+(6&!,6!
I'(?,4,&,J/]!6(!&?/!&/+?%,d*/!,6!*6*011P!*6/)!
L,&?!I(I*10&,(%6!&?0&!0'/!I0'&,+*10'1P!),33,+*1&!
&(!'/0+?!6*+?!06!)(+&('6!b/#7#]!%

5. FMR STUDY DESIGN PLAN
Based on HUD’s goals and the literature review above, we here propose an experiment that will
determine the optimal FMR survey design. HUD’s priorities are for this research endeavor are quality,
cost, and speed. In other words, the overarching goal is to create a design with the survey sample and
contact protocol that produces the best possible data as quickly as possible for as little cost as possible.

5.1. EXPERIMENT STRUCTURE
The core of the study design is an experiment to test the marginal utility of various protocol elements
(Figure 1). Marginal utility refers to the value (in terms of data quality) that HUD gets in return for the
additional cost and time invested in each element. We will test:
x
x
x
x
Figure 1:

Pre-notification letters. Although, these are known to increase response rates (See Section
4.1.3), they add a week to the data collection schedule and cost to the project budget.
Web response options. These are known to decrease response rates (See Section 4.1.3), but
the cost of a Web option is low, and it could decrease non-response bias.
Second survey mailings. These are known to increase response rates (See Section 4.1.3), but
they have a significant impact on the data collection schedule. HUD can test whether the
second survey mailing is more cost effective than early telephone follow-up.
Non-response letters to RDD respondents. These could help drive response on the phone or
the Web.
FMR Survey Experiment Structure

!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%%

x

Non-response bias calculated for the condition that has the lowest cost-per –interview..

If the conditions with the lowest non-response bias or cost-per –interview are different for different sites,
we will evaluate all the candidate conditions. If the analysis in the first two sections reveals that a
particular treatment had no impact on non-response bias, we will collapse the analysis with respect to that
treatment to increase analytic power.

Questionnaire Content Analysis
Mail- vs. Phone-Based Surveys
To compare the quality of the data we collect, we will determine whether:
x
x
x

Item non-response is higher on mail or phone, especially on the rent question.
Results from unfolding brackets substantially improve the final estimates by improving
imputation of rents.
Respondents tend to round their rent amounts and whether this effect appears higher among
respondents with higher rents. To the extent that rent is a socially sensitive topic, evidence
suggests that rounding should be less in the self-report rather than the intervieweradministered context (Tourangeau & Smith, 1996).

Sampling Error
The analysis presented in Section 4.1.2 suggests that the percentile calculated from a sample of 200 twobedroom recent movers could be as low as the 33rd percentile of the true population. The ACS provides
the optimal dataset for exploring this issue, but we will calculate the margins of error around the
estimated 40th and 50th percentiles for HUD’s reference in interpreting other analytic conclusions.

Eligible Population: Bedrooms
Restricting the eligible population to recent movers who rent two-bedroom units substantially decreases
the survey incidence. We will evaluate:
x
x
x

Comparability of rents estimated from all one-, two-, and three-bedroom respondents to rents
estimated using the conventional conversion models;
Potential cost savings of including all of these respondents in the eligible survey population;
and
Potential increase in precision of including all of these respondents in the eligible survey
population.

Eligible Population: Public Housing
There are several survey questions concerned with determining whether respondents live in public
housing. We will evaluate:
x
x
x
x

Item non-response on each question;
Correspondence among questions;
Self-reported “you pay” and total rents for question respondents; and
Self-reported subsidy or vouchers for public housing respondents.

The ideal set of questions will have low non-response, relatively low redundancy, and show a markedly
lower reported rent than the population in general. People living in public housing should also not report
receiving vouchers or subsidies.

!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%=?%

Eligible Population: Subsidies
Units occupied by people receiving subsidies are eligible for the survey, but families receiving subsidies
may not be able to accurately report total rents. We will evaluate:
x
x
x
x

Item non-response on the counterfactual question;
Correspondence between the “what if” rent, “you pay” rent, and total rent;
Variance and apparent rounding in “what if” rents; and
Total estimated FMR including and excluding subsidized renters.

If item non-response is high or the response to the “what if” question corresponds closely to the “you
pay” rent, then respondents may not understand the question. If responses are variable or subject to
rounding or digit preference, then respondents may be guessing at the total rent they would pay if they did
not receive subsidy. If the question does prove problematic, then, ideally, the estimated FMR including
and excluding these people would be about the same, and subsidy-receivers could be excluded from FMR
analysis.

!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%=@%

REFERENCES
Baker, M. (1992). Digit preference in CPS unemployment data. Economics Letters , 117-121.
Barron, M., Kelly, J., Montgomery, R., Singleton, J., Shin, H.-C., Skalland, B., et al. (April, 2010). More
on the extent of undercoverage in RDD telephone surveys due to the omission of 0-banks. Survey
Practice , http://www.surveypractice.org.
Bassili, J., & Scott, B. (1996). Response latency as a signal to question problems in survey research.
Public Opinion Quarterly , 390-399.
Blumberg, S. J., & Luke, J. V. (2009). Reevaluating the need for concern regarding noncoverage bias in
landline surveys. American Journal of Public Health , 1806-1810.
Blumberg, S., & Luke, J. (2010, December). Wireless substitution: Early release of estimates from the
National Health Interview Survey, January - June 2010. Retrieved January 2011, from Centers
for Disease Control and Prevention:
http://www.cdc.gov/nchs/data/nhis/earlyrelease/wireless201012.htm
Boyle, J., Bucuvalas, M., Piekarski, L., & Weiss, A. (2009). Zero banks: Coverage error in list assisted
RDD samples. Survey Practice , http://surveypractice.files.wordpress.com/2009/02/surveypractice-january-2009.pdf.
Browning, M., Crossley, T. F., & Weber, G. (2003). Asking consumption questions in general purpose
surveys. Economic Journal , F540-567.
CDC. (2009). National Health Interview Survey. Retrieved 5 2, 2011, from data release:
http://www.cdc.gov/nchs/nhis/nhis_2009_data_release.htm
Cowan, T. (2008). An overview of USDA rural development programs in rural development issues. (A. V.
Burlingham, & W. N. Townsend, Eds.) New York: Nova Science Publishers, Inc.
de Leeuw, E. D. (1992). Data quality in mail,telephone, and face-to-face surveys. Amsterdam: TT
Publications.
Del Valle, M., Morgenstern, H., Rogstad, T., Albright, C., & Vickrey, B. (1997). A randomised trial of
the impact of certified mail on response rate to a physician survey, and a cost-effectiveness
analysis. Evaluation & the Health Professions , 389–406.
Deutsch, R., Strack, F., & Gawronski, B. (2006). At the boundaries of automaticity: negation as reflective
operation. Journal of Personality & Social Psychology , 385-405.
Dillman, D. A., Smyth, J. D., & Christian, L. M. (2009). Internet, mail, and mixed-mode Surveys: The
tailored design method. Hoboken, New Jersey: John Wiley and Sons.
Dohrmann, S., Han, D., & Mohadjer, L. (2006). Residential address lists vs. traditional listing:
Enumerating households and group quarters. Proceedings of the Survey Research Methods
Section of the American Statistical Association, (pp. 2959-2964).
Dubray, P., & Conrey, F. R. (2011). Advance letter and pre-incentive effect on BRFSS survey response.
Poster Presented at the Annual Conference for the Behavioral Risk Factor Surveillance Survey .
Edwards, P., Roberts, I., MJ, C., DiGuiseppi, C., Wentz, R., Kwan, I., et al. (2009). Methods to increase
response to postal and electronic questionnaires (Review). In T. C. Collaboration, The Cochrane
Library. JohnWiley & Sons, Lt.
Erickson, D. (2009). The housing policy revolution: Networks and neighborhoods. Washington DC: The
Urban Institute Press.

!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%=A%

Fahimi, M., Kulp, D., & Brick, M. (2009). A reassessment of list-assisted RDD methodology. Public
Opinion Quarterly , 751-760.
Federal Register. (1995). 60 (157), 42222.
Federal Register. (2000). 65 (191), 58871.
Federal Register. (2010). 75 (209), 66979.
Ferber, R. (1956). The effect of respondent ignorance on survey results. Journal of the American
Statistical Association , 576-586.
Gordon, E. L., Chipungu, S., Bagley, L. M., & Zanakos, S. I. (2005). Improving housing subsidy surveys:
Data collection techniques for identifying the housing subsidy status of survey respondents.
Washington DC: U.S. Department of Housing and Urban Development Office of Policy
Development and Research.
Griffin, D. H., Fischer, D. P., & Morgan, M. (2001, May 17-20). Testing an internet response option for
the american community survey. Paper presented at the Annual conference for the American
Association for Public Opinion Research .
Groves. (2006). Non-response rates and non-response bias in household surveys. Public Opinion
Quarterly , 646-675.
Groves, R. M., & Peytcheva, E. (2008). The impact of non-response rates on non-response bias: A metaanalysis. Public Opinion Quarterly , 167-189.
Guterbock, T. M., Diop, A., Ellis, J. M., Holmes, J. L., & Le, K. T. (2011). Who needs RDD? Combining
directory listings with cell phone exchanges for an alternative telephone sampling frame. Social
Science Research , 860-872.
Housing Assistance Council. (2008, June). USDA guaranteed rural rental housing program. Retrieved
June 13, 2011, from http://www.ruralhome.org/storage/documents/rd538guaranteed.pdf
Housing Assistance Council. (2010, April). USDA rural rental housing loans (Section 515). Retrieved
June 13, 2011, from http://www.ruralhome.org/information-and-publications/informationsheets/27-rd-programs/115-rural-rental-housing-section-515
Houtkoop-Steenstra, H., & Bergh, H. V. (2000). Effects of introductions in large-scale telephone survey
interviews. Sociological Methods and Research , 281-300.
HUD. (n.d.). About HUD: Mission. Retrieved May 25, 2011, from
http://portal.hud.gov/hudportal/HUD?src=/about/mission
HUD. (n.d.). Fair market rents: Overview. Retrieved May 27, 2011, from
http://www.huduser.org/portal/datasets/fmr.html
HUD. (n.d.). Housing choice vouchers fact sheet. Retrieved June 13, 2011, from HUD.gov:
http://portal.hud.gov/hudportal/HUD?src=/program_offices/public_indian_housing/programs/hcv
/about/fact_sheet
HUD. (n.d.). Housing choice voucher program. Retrieved May 25, 2011, from
http://portal.hud.gov/hudportal/HUD?src=/hudprograms/hcvp
HUD. (n.d.). Supportive housing program. Retrieved June 13, 2011, from HUD.gov:
http://portal.hud.gov/hudportal/HUD?src=/program_offices/comm_planning/homeless/programs/
shp
HUD. (1996, November 1). CPD Notice 96-07. Retrieved June 3, 2011, from
http://www.hud.gov/offices/cpd/lawsregs/notices/1996/96-7.pdf
!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%=B%

HUD. (2006). Building HOME: A HOME program primer. Retrieved June 13, 2001, from Homes &
Communities: http://www.hud.gov/offices/cpd/affordablehousing/training/materials/building/
HUD. (1999). Waiting in vain: An update on America’s rental housing crisis. Washington DC:
Department of Housing and Urban Development.
HUD. (2007, July). Fair market rents for the Section 8 housing assistance payments program. Retrieved
5 2, 2011, from Hud.Gov: www.huduser.org/datasets/fmr/fmrover_071707R2.doc
HUD. (2008). Homelessness Resource Exchange. Retrieved June 13, 2011, from Supportive housing
program desk guide: http://www.hudhre.info/index.cfm?do=viewShpDeskguide
HUD. (2010, July 10). HUD.gov. Retrieved June 13, 2011, from Press release: HUD releases 2009
American Housing Survey:
http://portal.hud.gov/hudportal/HUD?src=/press/press_releases_media_advisories/2010/HUDNo.
10-138
Iannacchione, V. G., Staab, J. M., & Redden, D. T. (2003). Evaluating the use of residential mailing
addresses in a metropolitan household survey. Public Opinion Quarterly , 202-210.
Israel, G. (2009). Obtaining responses by mail or Web: Response rates and data consequences.
Hollywood, FL: Paper presented at the annual meeting of the American Association for Public
Opinion Research.
JCHSHU. (2001). America’s rental housing: Meeting challenges, building on opportunities.
Juster, F., & Smith, J. P. (1997). Improving the quality of economic data: Lessons from the HRS and
AHEAD. Journal of the American Statistical Association , 1268-1278.
Keeter, S., Kennedy, C., Dimock, M., Best, J., & Craighill, P. (2006). Guaging the impact of growing
non-response on estimates from a national RDD telephone survey. Public Opinion Quarterly ,
759-779.
Kovar, J. G., Rao, J. N., & Wu, C. F. (1988). Bootstrap and other methods to measure errors in survey
estimates. The Canadian Journal of Statistics , 25-45.
Krosnick, J. A. (1991). Response strategies for coping with the cognitive demands of attitude measures in
surveys. Applied Cognitive Psychology , 213-236.
Krosnick, J. A., Holbrook, A. L., Berent, M. K., Carson, R. T., Hanemann, W. M., Kopp, R. J., et al.
(2002). The impact of 'no opinion' response options on data quality: Non-attitude reduction or an
invitation to satisfice? Public Opinion Quarterly , 371-403.
Kupek, E. (1998). Determinants of item non-response in a large national sex survey. Archives of Sexual
Behavior , 581-594.
Link, M., Battaglia, M., Giambo, P., Frankel, M., Mokhad, A., & Rao, S. (2005). Assessment of address
frame replacements for RDD sampling frames. Paper presented at the 2005 American
Association for Public Opinion Research Annual Conference. Miami, FL.
Messer, B. L., & Dillman, D. A. (2010). Using address based sampling to survey the general public by
mail vs. Web plus mail. The National Science Foundation Division of Resource Statistics.
National Cancer Institute. (2007). Health information national trends survey. 5HWULHYHGɣɢɥ-May
from http://hints.cancer.gov/
NCSHFA. (2008). State HFA factbook: 2006 NCSHA annual survey results. Washington DC: National
Council of State Housing Finance Agencies.

!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%>C%

Norman, G., & Sigman, R. (2009). Using addresses as sampling units in the 2007 Health Information
National Trends Survey. Proceedings of the Joint Statistical Meetings, (pp. 4741-4752).
Washington DC.
Paulin, G. D., & Ferraro, D. L. (1994). Imputing income in the Consumer Expenditure Survey. Monthly
Labor Review , 23-31.
Pelletier, D. (2001). Preliminary assessment of American Community Survey data shows housing
affordability gap worsened for lowest income households from 2007 to 2008. Retrieved May 11,
2011, from http://www.nlihc.org/doc/Prelim-Assess-Rental-Affordability-Gap-State-Level-ACS12-01.pdf
Peytchev, A., Carley-Baxter, L. R., & Black, M. C. (2011). Multiple sources of nonobservation error in
telephone surveys: Coverage and non-response. Sociological Methods & Research , 138-168.
Poe, G. S., Seeman, I., McLaughlin, J., Mehl, E., & Dietz, M. (1988). "Don't know" boxes in factual
questions in a mail questionnaire. Public Opinion Quarterly , 212.
Roese, N., Sanna, L., & Galinsky, A. (2005). The mechanics of imagination: Automaticity and control in
counterfactual thinking. In U. J. Hassin RR (Ed.), The new unconscious (pp. 138–170). New
York: Oxford University Press.
Rubin, D. B., & Little, R. J. (2002). Statistical analysis with missing data (2nd ed.). New York: Wiley.
Schaeffer, N., Dykema, J., & Maynard, D. (2010). Interviewers and interviewing. In J. Wright, & P.
Marsden (Eds.). Emerald Group Publishing.
Schwabish, J. (2007). Take a penny, leave a penny: The propensity to round earnings in survey data.
Journal of Economic and Social Measurement , 93-111.
Schwarz, N., Bless, H. S., F., K. G., Rittenauer-Schatka, H., & Simons, A. (1991). Ease of retrieval as
information: Another look at the availability heuristic. Journal of Personality and Social
Psychology , 195-202.
Schwarz, N., Hippler, H. J., Deutsche, B., & Strack, F. (1985). Response scales: Effects of category range
on reported behavior and comparative judgments. Public Opinion Quarterly , 388-395.
Schwarz, N., Knäuper, B., Hippler, H. J., & Noelle-Neumann, E. (1991). Rating scales: Numeric values
may change the meaning of scale labels. Public Opinion Quarterly , 570-582.
Sherman, R. P. (2000). Tests of certain types of ignorable non-response in surveys subject to item nonresponse or attrition. American Journal of Political Science , 362.
Shiffman, S. (2009). How many cigarettes did you smoke? Assessing cigarette consumption by global
report, time-line follow-back, and ecological momentary assessment. Health Psychology , 519526.
Shoemaker, P. J., Eichholz, M., & Skewes, E. A. (2002). Item non-response: Distinguishing between
don't know and refuse. International Journal of Public Opinion Research , 193-201.
Singer, E. (2002). The use of incentives to reduce non-response in household surveys. In D. A. In R.
M.Groves, Survey Non-response (pp. 163-177). New York: Wiley.
Smith, E. R., & DeCoster, J. (2000). Dual-process models in social and cognitive psychology: Conceptual
integration and links to underlying memory systems. Personality and Social Psychology Review ,
108-131.
Staab, J. M., & Iannacchione, V. G. (2003). Evaluating the use of residential mailing addresses in a
national household survey. Proceedings of the Joint Statistical Meetings, (pp. 4028-4033).

!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%>:%

Steffen, B. L., Keith, F., Martin, M., Teresa, S. M., Vandenbroucke, D. A., & Yao, Y. G. (2011,
February). Congress, worst case housing needs 2009: Report to Congress. Retrieved May 10,
2011, from http://www.huduser.org/portal/Publications/pdf/WorstCaseNeeds2009Summary.pdf
Thompson, K., & Sigman, R. (2000). Estimation and replicate variance estimation of median sales prices
of sold houses. Survey Methodology , 153-162.
Tourangeau, R., & Smith, T. W. (1996). Asking sensitive questions: The impact of data collection mode,
question format, and question context. Public Opinion Quarterly , 275-304.
Tourangeau, R., Rips, L., & Rasinski, K. (2000). The Psychology of Survey Response. Cambridge
University Press.
US Census Bureau. (2007). American housing survey for the United States. Retrieved June 13, 2011,
from Table 4-3: Size of Unit and Lot in Renter Occupied Units:
http://www.census.gov/hhes/www/housing/ahs/ahs07/ahs07.html
US Census Bureau. (2009). Table 4-3: Size of unit and lot in renter occupied units. Retrieved June 13,
2011, from American Housing Survey for the United States:
http://www.census.gov/hhes/www/housing/ahs/ahs09/ahs09.html
Weisberg, H. F. (2005). TheTotal Survey Error Approach: A Guide to the New Science of Survey
Research. Chicago, IL: University of Chicago Press.
Winter, J. (2004). Response Bias in Survey-Based Measures of Household Consumption. Economics
Bulletin , 1-12.
Wood, M. (2005). Application of a test for evaluating ignorability of missing survey income data.
Conference Papers -- American Sociological Association, (pp. 1-16).
Yan, T., & Tourangeau, R. (2008). Fast times and easy questions: The effects of age, experience and
question complexity on Web survey response times. Applied Cognitive Psychology , 51-68.

!&*%,-$.(/%,)-7/%0(1#2+%89"+%

8"2(%>;%

!

APPENDIX A: MAIL SURVEY INSTRUMENT
1.

Is this house, apartment, or mobile home –

Rented
Occupied without payment of rent
Owned by you or someone in this household with a
mortgage or loan
Owned by you or someone in this household free
and clear (without a mortgage or loan)
This is NOT a residential house, apartment, or
mobile home.

If you checked “Rented” please continue to QUESTION 2.
If you checked any other box, you’re done! Please return the survey in the envelope
provided.

2.

How long have you lived at this house,
apartment or mobile home?

Less than 2 years
2 years or more

3.

If this house, apartment, or mobile home were
on the market for sale or rent, how many
bedrooms would you say it has?

No bedrooms; it is an efficiency or studio
1 bedroom
2 bedrooms
3 bedrooms
4 or more bedrooms

4

Which best describes this building?

A mobile home
A one-family house detached from any other house
A one-family house attached to one or more houses
– for example, a townhouse or rowhouse
A building with 2 to 4 apartments
A building with more than 5 apartments
Other (please specify)

!"#$%&'()$*+$&+($,-./0&
5.

Was this house, apartment or mobile home built
within the last two years?

Yes, it was built within the last two years
No, it is more than two years old

6.

Do you or does any member of this household
live or stay at this address year round?

Yes
No

If this is your vacation or seasonal home, apartment,
or mobile home where you do not live year round,
please select “No”.
7.

Is this house, apartment, or mobile home owned
by a relative?

Yes
No

If this home, apartment, or mobile home is owned by
your spouse or another individual living at this
address then please select “No.””.
8.

Do you do any work for your landlord so you can
pay less rent?

Yes
No

!
!"""#$%&'()""""!!
!"#$%&'()*$%+&,*$-)./01$2341$$$$$$$$$$$$$$$$$$$$$
!

!
$

240)$56

9.

Apart from utility costs, does this house,
apartment, or mobile home rent for the same
amount every month of the year?

9a.

Yes
No ¨ Skip to 10

(If the rent changes month-to-month) Why does
the rent change during the year?

The owners’ charge different rents during the summer
and winter seasons
The owners’ charge a different rent during the school
year
The owners’ costs – like heating and air conditioning
– change during the year so they charge different rent
Some other reason (please specify)

!"#$%&'()$*+$&+($,-./0&
10.

As part of your rental agreement, do you have to
recertify your income every year to determine
how much rent you pay?

Yes
No

11.

Is your rent amount lower because you are in a
government-housing program?

Yes
No

12.

Is the housing authority your landlord?

Yes
No

13.

Does a local housing authority own this house,
apartment, or mobile home?

Yes
No

14.

How much do YOU or does YOUR FAMILY pay
for rent each month for this house, apartment,
or mobile home?

15.

Do you have any roommates or housemates
that pay part of the rent who are not members
of your household?

16a.
17.

.

0

0

.

0

0

Yes
No

How much is the TOTAL monthly rent for this
house, apartment, or mobile home?
This should be how much you pay (Q15), what all
other roommates pay, and any assistance you
might receive. Do not include separate parking
fees or utility costs.

16.

,

The rent you pay each month

Do not include separate parking fees or utility
costs.
14a.

$

Does your household have a voucher that
allows you to choose where you live and pays
for part of your rent?
(For voucher holders) How much would your
rent be IF you had to pay it all yourself?
Are utilities included in your rent, or do you pay
for them separately?

$

,

Total rent per month

Yes
No ¨ Skip to 16a

$

,

.

0

0

Yes, utilities are included in the rent
No, pay separately for utilities

Utilities are heat, air conditioning, lights, water,
sewage, cooking fuel, or trash collection.

!

!

!"#$%&'()*$%+&,*$-)./01$2341$$$$$$$$$$$$$$$$$$$$$
!

$

240)$57

If you checked “Yes, utilities are included in the rent,” skip to QUESTION 28.
If you checked “No”, please continue to QUESTION 18.
!
18.

Is the cost of heating included in the rent, or do
you pay separately for heat?

18a.

Which FUEL is used MOST for heating this
house, apartment, or mobile home? (check
one)

Heat is included in the rent
Pay separately for heat
Your home does not have heat
Gas: from underground pipes serving the
neighborhood
Gas: bottled, tank, or LP
Electricity
Fuel, oil, kerosene, etc.
No fuel used
Other fuel (please specify)

!"#$%&'()$*+$&+($,-./0&
19.

Is air conditioning included in the rent, or do you
pay separately for air conditioning?

19a.

Which FUEL is used MOST for air
conditioning this house, apartment, or mobile
home? (check one)

Air conditioning is included in the rent
Pay separately for air conditioning
Your home does not have air conditioning
Gas: from underground pipes serving the
neighborhood
Gas: bottled, tank, or LP
Electricity
No fuel used
Other fuel (please specify)

!"#$%&'()$*+$&+($,-./0&
20.

Is cooking fuel included in the rent, or do you
pay separately for cooking fuel?

20a.

Which FUEL is used MOST for cooking fuel in
this house, apartment, or mobile home?
(check one)

Cooking fuel is included in the rent
Pay separately for cooking fuel
Your home does not have cooking facilities
Gas: from underground pipes serving the
neighborhood
Gas: bottled, tank, or LP
Electricity
No fuel used
Other fuel (please specify)

!"#$%&'()$*+$&+($,-./0&

21.

Is hot water included in the rent, or do you pay
separately for cooking fuel?

21a.

Which FUEL is used MOST for cooking in this
house, apartment, or mobile home? (check
one)

Hot water is included in the rent
Pay separately for hot water
Your home does not have hot water
Gas: from underground pipes serving the
neighborhood
Gas: bottled, tank, or LP
Electricity
No fuel used
Other fuel (please specify)

!"#$%&'()$*+$&+($,-./0&

!"#$%&'()*$%+&,*$-)./01$2341$$$$$$$$$$$$$$$$$$$$$
!

$

240)$55

22.

Is water included in the rent, or do you pay
separately for water?

Water is included in the rent
Pay separately for water
Your home does not have water

23.

Is the cost of lighting included in the rent, or do
you pay separately for the cost of lighting?

The cost of lighting is included in the rent
Pay separately for lighting
Your home does not have any lighting

24.

Do you pay a separate monthly rental fee for a
range or refrigerator?

Yes
No

25.

Is a sewage fee included in the rent, or do you
pay separately for sewage?

Sewage is included in the rent
Pay separately for sewage
Do not have municipal sewage – have a septic field

26.

Is trash collection included in the rent, or do
you pay separately for trash collection?

Trash collection is included in the rent
Pay separately for trash collection
The city or government provides trash collection
Your home does not have trash collection

27.

How many different residential landline
telephone numbers, not extension phones, ring
into your household?

28.

Do you or does anyone in your family have a
working cell phone?

28a.

(If you have a cell phone) How many cell
phones do you or the people in your family
have?

28b.

(If you have a cell phone) Of all the telephone
calls that you or your family receives are…

29.

Total number of landline telephone
numbers
Yes
No ¨ Skip to 29
Total number of cell phones

All or almost all calls received on cell phones
Some received on cell phones and some received as
regular calls
Very few or none received on cell phones

What is your town’s or city’s name?
Town or City Name

30.

What is your ZIP code?
ZIP code

Thank you! We appreciate your help!
If you have any questions about this study, please contact ICF Macro, the
company HUD hired to collect this information at
1-800-xxx-xxxx or [email protected]
!

!"#$%&'()*$%+&,*$-)./01$2341$$$$$$$$$$$$$$$$$$$$$
!

$

240)$58