Support Statement Part B

Supporting Statement to Accompany
OMB Clearance of Customer Satisfaction Surveys:
Phase Two 2009 HUD Partners Surveys

Part B: Statistical Methods of Data Collection

U. S. Department of Housing and Urban Development

February 2010

PART B:
STATISTICAL METHODS OF DATA COLLECTION
B1

Potential Respondent Universe

For three of the four HUD partner groups that will be surveyed in Phase Two of
the 2009 HUD Partners Surveys, the universe is similar to that of the 2001 and 2005
HUD Partner Surveys. It consists of:
1. All entities that own Section 202 or Section 811 multifamily properties,
estimated to be approximately 6,926.1
2. All entities that own HUD-insured, unsubsidized multifamily properties,
estimated to be approximately 4,468.
3. All entities that own HUD-assisted, subsidized multifamily properties,
estimated to be approximately 12,500.
The fourth partner group that will be surveyed in Phase Two had not been previously
surveyed for either the 2000 or 2005 HUD Partners surveys. It consists of all HUD-FHA
insured single family mortgage lenders. There are 9,606 such entities.
B2

Sampling Method and Respondent Universe

Samples will be drawn from the universe of multifamily property owners and
single family mortgage lenders—with the objective of obtaining 1,250 completed
responses from each group.
•

Previous experience with surveying owners of HUD-related multifamily
properties using extensive and state-of-the-art survey techniques indicates
that this particular partner group, as compared to others being surveyed in
Phase One, is the most difficult to survey.2 To achieve the objective of an

1

Note that these numbers are estimated because HUD’s data files consist of property lists; owners (the
sampling unit) may own more than one property, however. Property lists, therefore, must be converted to
owner lists, and this will be done at the point when the sample will be drawn to ensure that the latest, up-todate, information is used.

2

Compared to response rates of 80 percent to 91 percent for other HUD partner groups surveyed in 2005,
the response rate for multifamily partners averaged 63 percent. A primary reason is the accuracy of contact
information in HUD’s multifamily property databases. HUD faces significant challenges in maintaining
complete, up-to-date lists of names, addresses, and telephone number of ownership entities. Of the groups
surveyed (by mail) in 2005, only those using contact information from the multifamily databases resulted in
numerous returned (“addressee unknown”) letters/questionnaires by the U.S. Postal Service, or contained

estimated 60 percent response rate, the sample sizes for each group will be
somewhat more than 2,000 ownership entities—as shown in Section A12,
Exhibit 1. As was the case for the HUD Partners Surveys conducted in 2000
and 2005, all ownership entities in the universe will have an equal and known
probability of selection.
•

The universe of FHA-insured single family mortgage lenders will be stratified
by number of FHA-insured mortgage originations associated with each
lender, as indicated below.

Stratum
1
2
3
4

Number of
Loan
Originations
2,000+
1,000-1,999
100-999
1-99
TOTAL

Universe Size
131
141
2,168
7,166
9,606

Sample Size
131
141
1,610
200
2,082

Expected
Number of
Respondents
79
85
966
120
1,250

Strata 1 and 2 consist of lenders that originate (1) 2,000 or more FHA-insured
loans and (2) 1,000 to 1,999 FHA-insured loans, respectively; these strata will
be sampled with certainty (probability = 1). Stratum 3, which consists of
lenders that originate 100 to 999 FHA-insured loans, will be sampled with a
probability of .74; and stratum 4, which consists of lenders that originate
between 1 and 99 FHA-insured loans, will be sampled with a probability of
.028. The rationale for disproportionate selection is that the distribution of
FHA-insured lenders is such that a relatively small number of lenders account
for an extremely large proportion of all FHA insurance originations, and a very
large number of lenders originate fewer than 100 FHA-insured loans. The
selection probabilities reflect an interest in ensuring that lenders with the most
contact with HUD/FHA are sampled with a higher probability than others, and
that there will be a sufficient number of respondents to draw reliable
estimates. While some analysis will mirror the division by strata, shown
above, sample weights will be applied to permit appropriate estimation to the
full universe of FHA-insured single family lenders.

missing or inaccurate telephone numbers, resulting in failure to make contact. According to HUD’s Office of
Housing, some owners have little interest in maintaining a continuing relationship with HUD beyond that of
their currently insured or assisted properties and, therefore, little motivation or incentive to keep their contact
information up to date. Likewise, there is limited functional benefit for HUD to allocate scarce resources to
maintaining such contact information when it is not used on a regular basis. When cases where contact
could not be made are removed from the sampling frame for purposes of calculating an adjusted response
rate (i.e., based on those who could be contacted), the rate increases to 72 percent.

B3

Procedures to Deal with Non-Response

Extensive efforts are being made to ensure that lists of potential respondents are
as accurate and up-to-date as feasible, to permit delivery of mail questionnaires and
receipt of follow-up telephone calls. Complete, accurate, and up-to-date lists serve as
the starting point.
•

HUD maintains a list of multifamily properties insured or assisted under HUD
programs. The list, which includes contact information regarding the entity
that owns the property, will be used to identify all entities that own such
properties, such that the ownership entity (not the property) will be the basis
for sampling. The sample of ownership entities, therefore, will consist of
those that own a single property insured or assisted under HUD programs as
well as those that own multiple properties; and sampled ownership entities
should receive only one questionnaire regardless of the number of properties
owned.

•

HUD maintains a list of all HUD/FHA-insured single family lenders, which
contains contact information. The list will be used as the basis for sampling
lenders.

The successful practices used for the 2001 and 2005 HUD Partner Surveys will
be used in the 2009 surveys to optimize response rates—following the Dillman method
for conducting surveys by mail. Initially, cover letters and questionnaires will be sent to
all respondents. A reminder follow-up post card will be sent if responses are not
received within a reasonable amount of time and a second cover letter and
questionnaire mailed to those not responding to the previous mailings.3 If there is no
response to these mailings, respondents will be phoned and encouraged to respond,
offered a re-mailing or fax copy of the instrument, and given the opportunity to respond
by phone. A minimum of six attempts will be used to secure telephone interviews—on
different days and at different times. Potential respondents for whom voice mail
messages must be left will be provided with a toll free 1-800 number, with
encouragement to return the call. When contact is made, potential respondents will be
offered the opportunity to schedule a time for the interview at their convenience.
Experienced interviewers will be used who have sufficient knowledge of the study to
explain its purpose and importance. These procedures should maximize the rate of
response.

3 If appropriate, certified letters may be sent on a targeted basis.

B4

Pre-Testing of Procedures

The proposed 2009 HUD partner surveys are a follow-up to the 2005 surveys
and, as such, the 2005 questionnaire served as the basis for the 2009 questionnaire.
Staff of the Urban Institute and Silber & Associates met with staff of HUD’s Office of
Housing to consider if any updating was appropriate; as a result, only minor
modifications were made. Therefore, since the questions on the 2009 survey
questionnaire for HUD’s multifamily owner partners are by-and-large the same as the
questions used in the 2005 partners survey, pre-testing is not considered necessary.
The survey questionnaire for HUD/FHA-insured single family lender partners,
which has not previously been used, will be pretested on a small sample of nine or fewer
such entities. Such pre-testing will consider wording, question flow, and meaning and,
where appropriate, include cognitive discussions with respondents to understand their
interpretation of the questions and the reasoning behind their answers.
B5

Person Responsible for Statistical Aspects of the Design

Martin Abravanel and Chris Hayes of the Urban Institute, and Bohne Silber of
Silber & Associates, are responsible for the statistical aspects of the survey design.
Appendices
¾ Correspondence with prospective respondents
¾ Survey instruments