In this section:

• Minutes of Meeting of the 2001 Residential Finance Survey Working Group
• Minutes of Second Meeting of the 2001 Residential Finance Survey Working Group
• Minutes of Third Meeting of the 2001 Residential Finance Survey Working Group
• Minutes of Fourth Meeting of the 2001 Residential Finance Survey Working Group

Bureau of the Census
Suitland, Maryland
November 13, 1998

Attendees: Fred Eggers, Bob Knight, and Ron Sepanik (Department of Housing and Urban Development); Tom Blatt, Walter Busse, Barb Forsyth, Pete Fronczek, Nancy Gordon, Len Norry, Eileen O'Brien, Howard Savage, Mary Schwartz, Dan Weinberg, Ellen Wilson (Census Bureau); Kathy Dillard and James Mikesell (Department of Agriculture); Jim Follain and Frank Nothaft (Freddie Mac); Jack Goodman (National Multi Housing Council); Stephen Zabrenski (Office of Thrift Supervision); Paul Emrath and Michelle Hamecs (National Association of Home Builders); Mark Calabria and Forrest Pafenberg (National Association of Realtors); Dan Quan (Federal Reserve Board); Bob Barr (Fannie Mae); Doug Duncan (Mortgage Bankers Association); Lynne Sabatiuk (Department of Veterans Affairs); George Smith and Paula Young (Bureau of Economic Analysis).

The group was formed to advise on the content of the 2001 Residential Finance Survey (RFS). Membership in the group is comprised of government agencies - Federal Reserve Board, Bureau of Economic Analysis, the Department of Agriculture, Office of Thrift Supervision, the Veterans Administration, and the Department of Housing and Urban Development; government-sponsored enterprises - Fannie Mae, Freddie Mac; and trade associations - National Association of Realtors, National Association of Home Builders, and Mortgage Bankers Association.

A similar group was instrumental in questionnaire redesign prior to the 1991 RFS. Indeed, input from this group resulted in changes to nearly 40% of the questionnaire prior to the 1991 RFS. The group was also instrumental in identifying emerging issues, such as the need to cover mobile homes and condominiums, which appeared in the survey for the first time in 1991.

The first meeting of the 2001 RFS group was held November 13 from 9:30 to 12:00 at the Bureau of the Census.

Pete Fronczek thanked everyone for coming. Guests introduced themselves. Fronczek talked about the partnership between the Department of Housing and Urban Development (HUD) and the Census Bureau in making the RFS possible. Fronczek indicated that a year ago, the RFS seemed unlikely to occur. The survey, however, was revived by the many guests who had made a convincing case about the need for the RFS, and in particular, HUD which is funding the survey.

Fronczek introduced Fred Eggers, Assistant Secretary for Economic Affairs at HUD. Eggers spoke about the considerable amount of work and research that was undertaken to ensure the RFS was in the HUD budget. Further, he indicated that this commitment to the RFS required reprogramming of HUD's resources. Consequently, while HUD believes the American Housing Survey Metropolitan Sample (AHS-MS) is important, it believes that the RFS is a better use of its funds this year.

Nancy Gordon, Associate Director for Demographic Programs at the Census Bureau, agreed that the road to the RFS was not a smooth one. The Residential Finance Survey was not funded in the Census Bureau's budget because of the cost-cutting for the decennial census mandated by Congress. Thus, the Census Bureau, researchers, public policy makers, etc. are grateful to HUD for reprogramming its budget to include the RFS.

While the data products of the RFS were well-known to the group, the process itself was less so. To shed some light for the group, Fronczek gave an overview of the operational side of the survey. He explained that the sample of housing units is drawn from the decennial census' master address file. In 1991, the 70,000 sampled units were contained in the nearly 60,000 properties that comprised the RFS sample. (A mortgaged property is defined as all land and buildings covered by the first mortgage. A nonmortgaged property is defined by the owner.) After the sample is drawn, a decision is made as to whether each property in which the sample unit is contained is a "homeowner" or "rental and vacant" property. "A homeowner property", according to the RFS definition "is one with one, two, three, or four housing units, at least one of which is occupied by the owner." Similarly, "the term rental and vacant applies to properties with five or more housing units (even though one of the units may be occupied by the owner) and to properties with 1 to 4 housing units, all of which are either renter-occupied or vacant." Homeowner questionnaires are mailed to homeowner properties. "Owner-seeker" letters are mailed to rental and vacant properties. Once the owner or owner's agent is identified, a rental and vacant property questionnaire is then sent to the rental property owner/agent. (Follow-up is done for all nonrespondents to the owner seeker letter.)

Both property questionnaires contain a question to determine whether a property is mortgaged or owned free and clear. Respondents who indicate that their properties are mortgaged are also asked to whom they send their mortgage payments. For mortgaged properties, a lender questionnaire is mailed to the mortgagee with detailed questions about the characteristics of the mortgage. Information on the property questionnaire and the lender questionnaire (if applicable) are linked to form the residential finance property record.

Planning, which was anticipated to start in 1995 or 1996, is just beginning, according to Fronczek. Thus, time and budget considerations will limit the number of innovations in sample design, data collection methods, and processing methods for the 2001 RFS. Some innovations considered but subsequently rejected included using administrative records from various sources rather than surveying; and changing the universe to lenders rather than nonfarm, privately owned, residential properties, thereby working in reverse - from lenders to property owners. Notwithstanding time and budget considerations, according to Fronczek, there will be several innovations to the RFS, including:

1. Web Page

The objective of the web page is to answer surveyed property owners' and lenders' questions thereby reducing the amount of time spent by the Census Bureau staff responding to routine inquiries. The site will include the following: explanation of what the RFS is and an overview of the survey; minutes from Working Group and RFS-related meetings; informational copies of the questionnaires; answers to frequently asked questions; information on , the legal authority for taking the survey and the basis for the strong protection of the confidentiality of the data; statistical briefs and press releases from the 1991 survey; research that has been done using the RFS; and testimonials on the usefulness of the data. The web site will also refer to an 800 number that can be used for nonroutine inquiries. The site will be updated frequently and monitored for the number of "hits".

1. Lender Database

The thrust behind the lender database, which was started for the 1991 RFS, is to control and monitor the flow of the lender questionnaires. Many of the lenders are very large, multinational, and often international, organizations. Thus, the questionnaires often "get lost" within the organizations. This problem will likely be even more pervasive in the 2001 RFS due to massive consolidation in the banking industry during recent years. The identification of appropriate contacts (individuals most qualified to answer the questionnaire) at lending institutions before the questionnaire is sent will allow the Census Bureau to track the questionnaires and reduce time required for lender nonresponse follow-up.

The lender database for the 1991 survey was based on the Standard Statistical Establishment List from the economic directorate of the Census Bureau. This list was supplemented by lists of lenders and members, provided by Fannie Mae, Freddie Mac, and the Mortgage Bankers Association. The Census Bureau's Housing and Household Economic Statistics (HHES) Division will again ask these organizations, as well as some newly identified ones whose members represent nontraditional, institutional mortgage lenders (Real Estate Investment Trusts [REITs], pension funds, mortgage conduits, etc.), for their member lists. Duplication of lenders from various lists will be eliminated by HHES staff. After Office of Management and Budget (OMB) clearance is obtained, there will be a presurvey of lenders in the database 12 to 18 months prior to going into the field.

Fronczek told the group that the Census Bureau's Center for Survey Methods Research (CSMR) will be assisting in developing and implementing an overall data collection strategy. In this data collection strategy, several innovations will be considered, including alternative data collection methodologies (primarily for lenders via electronic media) and additional versions of survey instruments.

Content development, one element of the larger data collection strategy, was the area in which the attendees were being called on to help. Fronczek told the group that the current meeting was the first step in the content development plan. Fronczek gave the floor to Eileen O'Brien from CSMR to talk about the content development plan and the important role of the attendees in the development.

The identification of measurement goals and policy uses was critical to designing a survey instrument that would meet the needs of the data users, according to O'Brien. The assembled group then would be instrumental in determining the content since they represented the researchers, policymakers, and data users. While the instrument needed to be effective in capturing policy-relevant questions, it also needed to be efficient, thereby minimizing respondent burden. Once content is finalized, survey questions need to be evaluated through various pretesting techniques such as focus groups and cognitive testing. This will ensure respondents understand the proposed survey questions and, therefore, are able to provide correct answers. Then the data collection protocol and survey instruments will be finalized.

Fronczek then presented the group with the timing of activities:

1. Content Development Activities:
   • Working Group Meetings 4/30/99
   • Cognitive Testing/Focus Group Activities 7/30/99
2. Final Content Determined 9/30/99
3. Final Collection Instruments (Questionnaires) 3/30/00
4. Pre-survey contact of Lenders 2/01/00-3/30/00
5. Interviewing Property Owners 2/01/01-8/31/01
6. Interviewing Mortgage Lenders 7/23/01-12/28/01
7. First Data Products Released 4/30/03

Fronczek then turned the floor over to Howard Savage to lead the discussion of the major objective of the meeting - the content issues.

According to Savage, the current working group needed to answer two questions. What do we need to know about residential finance that we currently do not know? Are there other sources of that information already available? We do not need to collect information already available.

In response to Savage's question whether the RFS still needed to collect information on mobile homes, there was a resounding "yes" for various reasons. One compelling reason to keep mobile homes within the universe is that mobile homes (or manufactured housing) are an important type of affordable housing. In addition, with over one-half of the mobile homes in the country being in rural areas, their omission may greatly compromise the value of the RFS for rural housing policy and program decision-making. Moreover, these data are not available from other sources.

Although difficult in the past, it is now getting easier to identify mobile home lenders, according to several members of the group. Indeed, the industry is highly concentrated with 12 lenders doing about 90 percent of the lending. Forrest Pafenburg, Eggers, and Frank Nothaft said that they would supply lists of mobile home lenders to assist in the development of the lender database (discussed above). Eggers asked whether these 12 lenders have both secured and unsecured loans in their portfolios; to which Nothaft replied that he was not certain. Eggers is concerned that personal property loans for mobile homes may be considered mortgage loans, when in fact they are not, confusing the issue.

Coverage of mobile homes can be problematic, according to Dan Weinberg, since the Census Address File may not include all mobile home parks. Weinberg went on to say that proper weights developed from the Housing Unit Coverage Survey could compensate for any underrepresentation on the Census Address file.

An issue of growing importance because of the aging baby boomers and one of great concern to the group was the coverage of seniors housing. There seemed to be some concern that these properties may not be covered well, because they violate the traditional definition of a housing unit (i.e. don't have individual kitchens; thus may not be treated as housing units and may not get into the RFS sample, etc.) Len Norry allayed this concern by indicating that a housing unit does not have to have a kitchen in order to be considered a housing unit. Moreover, Fronczek indicated that the Census Address File , from which the RFS sample is drawn, captured many of these units in the last decennial census and should continue to do so for Census 2000.

Nothaft indicated a need to separate out the payment for residential rent on these units from payment for long term care, etc. Otherwise, the actual housing costs for seniors would be distorted.

Jack Goodman raised the question of how to deal with corporate hotels. Fronczek explained that in past RF surveys, if the property contained more housing than other uses, it was treated as residential; otherwise, it would be defined to be a commercial property and out of the universe of the survey. In particular, the respondent is asked the following: "Is it a hotel or motel with 50 percent or more of the accommodations reserved for transient use?" and "Is the property one in which more than one-half of the floor space is used for nonresidential purposes"? If the respondent answered "yes" to either of these questions, the property was excluded from the survey. Thus, corporate hotels, even extended stay hotels, would not be included in the survey.

Weinberg pointed out that properties built after the address listing for the decennial census is completed will be missed by sampling from the Census Address File. This is particularly true for many smaller single family properties that take little time to plan and build. He suggested updating the Census Address File with building permits or using the American Community Survey (ACS) to update the RFS sample. Fronczek agreed; however, he indicated a time constraint - sample selection must take place in September/October 2000. The RFS sampling staff is looking into several possibilities, including using the ACS, for providing the most complete sample for the RFS.

Savage indicated that a limited set of questions about home equity loans had been added to the 1991 RFS. Did we still need this information? Did we need more information, particularly regarding home equity loans for more than 100 percent of equity in the property? Was this information attainable from other sources?

While there was consensus on retaining and adding questions on home equity loans for the 2001 RFS, reaction to importance of special treatment of high LTV (loan-to-value) mortgage lenders was mixed. Pafenburg mentioned that many of these very high LTV loans are really credit card consolidation vehicles as opposed to true mortgages. Nothaft adds that "125 LTV" loans are a small, albeit growing, share of the market. Thus, they may have less relevance for residential finance than lower LTV home equity loans. However, Pafenburg went on to say that if they are covered, the characteristics and servicing of these high LTV, home equity loans are very different from traditional mortgages (e.g., servicing fees are 150 basis points above other loan types, etc.) and this may have to be reflected in the design of the home equity loan questions. Pafenburg referred to a book which John Weicher wrote recently about home equity loans.

Stephen Zabrenski said that his organization (OTS) is very interested in high LTV loans and that they need to be identified since they are at a higher risk of delinquency, default, and foreclosure. Zabrenski continues, that he has found that many lenders who deal in BC loans (loans of lesser quality, which generally cannot be sold on the secondary market) are also high LTV lenders. Zabrenski may be able to supply a list of some BC (high LTV) lenders. Jim Follain indicated that he would send the name of a colleague who works on Wall Street and follows the home equity loan market closely. This Wall Street contact could perhaps be of assistance in helping identify BC lenders for the lender database and designing home equity loan questions.

Michelle Hamecs asked if subprime loans (loans of subprime credit quality) could be identified on the RFS and expressed some interest in using the RFS as a vehicle to do so. While the survey as it is currently designed does not lend itself to identifying subprime loans, some group members indicated that interest rates that were significantly above "market" may be an indicator of a subprime loan. Working group members expressed some reservations about using above market interest rates as proxies for subprime loans, particularly in light of risk-based pricing, where high interest rates could lead to erroneous conclusions about loan quality. Nothaft raised the possibility of using the fees paid by the borrower per the HUD1 Settlement Form as an indicator of credit quality for HUD-insured loans (i.e., the higher the fees, the lower the credit quality).

Paul Emrath indicated that the Consumer Expenditures Survey has information on home equity loans.

Some adjustable rate mortgage information is necessary for several reasons, including ARMs' large share of the jumbo market, according to Nothaft. Regarding the adequacy of the questions on adjustable rate mortgages on the 1991 questionnaire for advancing policy research, Pafenburg indicated the importance in being able to identify loans that switch from one type to another, e.g. from fixed to adjustable rate, which the current questionnaire is unable to do. He will send a residential finance survey that his organization (NAR) does that captures this switching information. Goodman indicated a need to have a 10-year yield maintenance question for ARMs on the questionnaire. Otherwise, the group members (Follain and Nothaft) seem to agree that not much additional information is needed because at the current low level of interest rates, ARMs are not competitive and few are being made. This may change, however, by the time the survey is conducted. If rates go up, there will likely be an increase in ARMs' share of market.

This did not elicit much input from the group. Dan Quan suggested that buydowns (payment of points upfront by the developer/homebuilder on behalf of the borrower to reduce the borrowers' mortgage interest rate below market rate for a specified period, often five years) are important now, and that the Census Bureau should identify these loans but not collect much information on them.

The reluctance of the private sector to leap at the RAMs market elicited some group members to question the need to collect information on this sector. Pafenburg cites that only 28,000 reverse mortgages have been made so far. Tom Blatt added that there are only 8 in the AHS design. Eggers seemed intent on getting information about this sector, however. Despite its small size now, future legislation could in fact change the market for RAMs considerably, adds Eggers. Eggers did think that some of the policy questions regarding RAMs could be answered by administrative records from the FHA. In fact, Eggers adds, these administrative records could be used to enrich the database. Nothaft indicated that since there were so few, perhaps the Census Bureau could get a file of them and match back to the borrowers and include them in the survey.

Initial discussion focussed on the near impossibility of getting the mortgage holder or ultimate investor in the current complex financing environment. To determine the "ultimate investor", one would have to determine who gets the cash flows from the mortgage, which would be very difficult at best. Savage indicated that he believes getting at the ultimate investor is beyond the scope of the RFS. The goal of the RFS then, according to Savage, is to get at the person/institution that is most knowledgeable about the mortgage, be it the holder, servicer, or holder/servicer. (Pafenburg said that there is a MERS file and perhaps the Census Bureau could use information from it to ascertain who has servicing rights for a particular mortgage.)

Nothaft said that while the ultimate investor of securities could not be determined, the RFS could be used to determine who is legally responsible for the security (i.e. who is the fiduciary?) by asking the lender "Who do you remit payment to?" rather than asking "Who is the holder of the mortgage?" (Nothaft indicated that the 1991 RFS underreported the volume of loans held by Fannie Mae and Freddie Mac perhaps because servicers did not know Fannie and Freddie owned the loans.)

A lengthy discussion ensued about the necessity of finding out who the mortgage originator was. There seemed to be general agreement that this information would be very helpful. Follain indicated that this information was necessary to satisfy new policy laws regarding the government-sponsored enterprises (GSEs). Nothaft concurred that the RFS should find out about the originator. This is particularly true for multifamily loans, Nothaft adds, since single family loans are pretty well covered by the Home Mortgage Disclosure Act (HMDA) data.

In addition to the necessity of finding out who the originator was, the group members generally agreed that finding out may be difficult and test the boundaries of data collection. Nothaft said that brokers usually sell both the loan and the servicing rights, making it difficult in such cases to find the originator. James Mikesell's solution, to which others agreed, was to ask the borrower where he/she got the loan; that is, who originated the loan. Moreover, Mikesell added, the Census Bureau could use information from the economic side to assign a "type" to the originator eliminating the need for the borrower (who is generally unqualified) to indicate a originator type. Pafenburg indicated that there may still be difficulty in determining who originated the mortgage even if the borrower is asked since the individual/organization with whom the borrower dealt may have been a broker and used someone else's money. Thus, the borrower's answer to who is the originator may be a broker, not the true originator.

Savage asked the group to send information and comments on what/how this should be measured.

There seemed to be little interest in finding out the number of times a property had been refinanced. Indeed, other data sources on abusive practices (churning of loans) do exist, according to Pafenburg, that likely measure the number of refinancings.

The Federal Reserve is interested in cash outs since this represents a source of spending on other items. Quan said that the RFS should determine how much cash came out of the property when the loan was refinanced. How much was saved from lower payments on refinancing that could conceivably be spent in other ways or saved in other vehicles? This would mean the RFS would have to ask about the previous payments on the loan so that they could be compared to current mortgage payments. (Hamecs agreed that information about the previous loan would be beneficial.) Quan was also interested in determining how many mortgages changed from 30 to 15 years at time of refinancing.

Nothaft was more interested in the reasons for the refinancing rather than the number of times refinanced or the characteristics of the refinancing itself. Nothaft said that the RFS should ask, "if the current loan is a refinance, why was it done?" Goodman dissented saying the consumer motivations about refinancing vary a great deal, and are being measured adequately elsewhere.

The determination of the LTV at the time of refinancing is an important issue, said Pafenburg. This would mean collecting the appraised value at the time of refinancing, which should be in the lender's files. Thus, this item should be added to the lender questionnaire.

Prior to the 1991 survey, there was a question on the lender questionnaire regarding the current loan status (e.g. delinquent, in default, in foreclosure). This question was eliminated in the 1991 survey because members of the working group for the 1991 RFS indicated that delinquency data were readily available from other sources. However, Follain suggests that the Census Bureau may wish to add this question again in the wake of renewed interest in using data from the RFS in conducting mortgage performance research. He added that this type of information about multifamily properties is critical to expanding securitization of multifamily loans and thereby increasing the funds available for affordable rental housing. While some of the determinants of default are already in the RFS, Follain believes that by asking about the current loan status the final piece of the puzzle would be in place.

Since most studies show that LTV plays a significant role in current loan status, Hamecs believes that it is important to be able to calculate LTV so that one can examine delinquencies by LTV. With the current questions, one can calculate LTV at the time the property was purchased based on the purchase price and mortgage amount at the time the loan is placed. LTV at the time of the survey can be determined based on questions on the 1991 survey. The 1991 lender questionnaire can supply the unpaid mortgage balance, and the property questionnaire the value of the property. Quan asked if the questionnaire contained a question regarding the lender's estimate of value of property at the time of the survey, which it does not. Quan recommended that the Census Bureau ask the lenders for the property price at the time of purchase. He added that since this price information is on the loan application it would not be a burden for lenders to provide. Another good question to add to the lender questionnaire is for the lenders' estimate of home value at the time of the survey. Fronczek indicated that lenders would necessarily have an estimate of value at the time the mortgage is made; the estimate of property value by the lender at the time of the survey could be imputed by bumping up the value at the time the loan was made by some inflation factor. (Follain suggested that many lenders to large multifamily rental property do have a current estimate of property value based on the capitalized value of net operating income for the property. However, Fronczek expressed some reservations about alienating property owners by asking lenders to give detailed information on the capitalized value of the property's net operating income since owners are already asked about the value of their properties.) The imputed lenders' estimate of property value coupled with the unpaid balance on the mortgage (which is already on the 1991 questionnaire) would allow for the determination of the LTV at the time of the survey which could be used in delinquency research.

In addition to the need to determine LTV at the time of the survey, Pafenburg also expressed an interest in determining LTV at the time of refinancing. This would mean collecting the appraised value at the time of refinancing, which should be in the lender files.

(Note: GAO did a study recently on high LTV mortgages; however, no information on foreclosures is included.)

(Note: Pafenburg also suggested that Freddie Mac's data base on property resales, which can be used to measure home price appreciation/depreciation [one element of risk], may be able to be merged with the RFS to further mortgage performance research.)

In addition to property characteristics, borrower characteristics are critical to loan performance research. Follain said that loan approvals are based on the credit of the borrower, not the property, so the Census Bureau should consider getting information on the credit history of the borrowers, although Follain admits, this would be difficult.

Rural Housing - How Well is It Represented in the Sample and How Well Do Final Tabulations Cover It?

Kathy Dillard was interested in the sampling plan. Does the current sample represent rural properties adequately? Dillard asked if the Census Bureau could over sample for rural areas. Fronczek/Savage indicated that while this could be done, the current RFS budget would demand a reduced sample somewhere else.

Mikesell indicated that there were no rural identifiers in the 1991 data, to which Savage responded that he would work with USDA in getting the tabulations it needed with regard to rural housing in the final products for the RFS.

Dillard also expressed some concern that the 1991 survey questionnaire may not be adequate in capturing the high leveraging (including leveraged second mortgages) of some rural properties. Savage asked that Dillard send details of some of these highly-leveraged loans so that he could incorporate this issue into questionnaire design.

Ron Sepanik asked whether the Census Bureau could sample at different rates -- for example, large versus small properties and rural versus metropolitan area properties. Fronczek indicated that the 1991 sample was stratified based on various characteristics and this was likely to continue in 2001.

Nothaft asked about the possibility of two lender questionnaires - one for 5+ unit properties, and one for smaller rental and vacant properties, since the two types are very different with regard to their operating characteristics. Fronczek responded that this would be difficult to administer. Goodman said that perhaps just a really well designed skip pattern on one lender questionnaire would suffice. Mark Calabria stressed the need to keep the skip pattern simple.

Many data initiatives have been undertaken since the 1991 RFS that collect detailed property operating data, according to Goodman. (Follain added that the Mortgage Information Corporation of California have these data.) Thus, it is unnecessary for the RFS to collect detailed operating information; summary data will do. Emrath concurred and suggested a "tradeoff", collecting information on expenses for capital improvements by owners rather than collecting detailed operating expenses, which are available elsewhere.

Pafenburg alerted us to the fact that loans from state housing agencies are often used in conjunction with FHA mortgage insurance. This situation may create problems with question 6 of the 1991 lender questionnaire, which seems to imply that the respondent/lender should choose one agency type that insures or guarantees the mortgage. The possibility of having a mortgage insured/guaranteed by more than one of the agency types needs to be addressed, according to Pafenburg.

Nothaft asked whether the add one question/delete one question rule that applied to the 1991 RFS was again applicable. Fronczek responded that while respondent burden does still have to be minimized, there is not necessarily a one-to-one relationship. Fronczek stressed the need to focus on needed content, then "trim" later.

The second meeting is scheduled for February 4, 1999 from 9:30 to 12:00 a.m. at the Census Bureau. Eggers asked that suggestions for additional users/analysts who should attend be forwarded to the Census Bureau. Finally, Savage and Fronczek asked that participants send questions, modifications, and drops/adds to HHES before the meeting. A discussion document will be developed from the input and sent to the group prior to the February 4 meeting.

Bureau of the Census
Suitland, Maryland
February 4, 1999

Attendees: Fred Eggers and Ron Sepanik (Department of Housing and Urban Development); Tom Blatt, Marie Butler, Walter Busse, Pete Fronczek, Eileen O'Brien, Howard Savage, Mary Schwartz, Dan Weinberg, Ellen Wilson (Census Bureau); Michael Feinberg and James Mikesell (Department of Agriculture); Jim Follain and Frank Nothaft (Freddie Mac); Jack Goodman (National Multi Housing Council); Stephen Zabrenski (Office of Thrift Supervision); Paul Emrath and Michelle Hamecs (National Association of Home Builders); Mark Calabria and Forrest Pafenberg (National Association of Realtors); Dan Quan (Federal Reserve Board); Bob Barr (Fannie Mae); Amy Bogdon (Fannie Mae Foundation); Brian Carey (Mortgage Bankers Association); Lynne Sabatiuk (Department of Veterans Affairs); George Smith (Bureau of Economic Analysis); and Shaun Brady (Multifamily Housing Institute).

The group was formed to advise on the content of the Residential Finance Survey (RFS), designed to look at the financing of the nation's owner-occupied and renter homes. Membership in the group is comprised of government agencies - Federal Reserve Board, Bureau of Economic Analysis, the Department of Agriculture, Office of Thrift Supervision, the Veterans Administration, and the Department of Housing and Urban Development; government-sponsored enterprises - Fannie Mae, Freddie Mac; trade associations - National Association of Realtors, National Association of Home Builders, Mortgage Bankers Association, the National Multi Housing Council, and the Multifamily Housing Institute; and the Fannie Mae Foundation.

Peter Fronczek introduced himself and welcomed a few new members to the group.

Fronczek then briefed the working group about what the Housing and Household Economic Statistics Division (HHES) had done on the RFS since the first meeting in the fall, including the following.

• Lender reference Database and Presurvey Contact of Lenders

HHES is again developing a lender reference database, as one element of a comprehensive automated control system development for the 2001 RFS. Fronczek told the group that HHES staff had already contacted many residential finance-related organizations, many of which were represented at the meeting, to obtain lists of their members, customers, and so forth. Fronczek indicated that the response to the requests had been favorable, with several organizations having already provided lists.

Fronczek then explained that once the database is developed, the Census Bureau will conduct a presurvey of all lenders on the database. The presurvey will be sent to a senior member of the lending institution (either by name or by title), who will be asked to provide the name and address of the appropriate respondent for the lender questionnaires, should one or more of the lending institution's properties fall into the RFS sample.

To answer many frequently-asked questions about the survey, the presurvey would also contain a brief overview of the survey, a reference to Title 13 (the legal authority for taking the survey and the basis for the strong protection of the confidentiality of the data), and Title 12 (which applies to lenders, but which does not exempt them from answering the questionnaire), and a question about the lender's preference for paper or electronic reporting.

• Web Page

The objective of the web page is to be a source of information on the RFS to all those interested in the survey, including respondents and working group members. The site will include the following: explanation of what the RFS is and an overview of the survey; minutes from working group and RFS-related meetings; informational copies of the questionnaires; answers to frequently asked questions; information on Titles 12 and 13; statistical briefs and press releases from the 1991 survey; research that has been done using the RFS; testimonials on the usefulness of the data; and a "glossary" of terms used in residential finance to provide some guidance to respondents in answering the questionnaire. The site will be updated frequently and monitored for the number of "hits".

Fronczek told the group that at the end of this meeting, HHES will draft questionnaires for Eileen O'Brien in the Center for Survey Methodology Research (CSMR) at the Census Bureau to begin testing. Anything added to questionnaires at this point is for testing purposes only. Decisions on the final content for the 2001 RFS will wait for the results of the testing. He then turned the floor over to Howard Savage to discuss HHES' interpretation of the working group's suggestions for questionnaire content from the November 13 meeting and how HHES may implement them. (Refer to minutes from the November 13 meeting for the Working Group's suggestions.)

Both Michael Feinberg and Fred Eggers took a strong stand on wanting to collect information about consumer installment loans used to finance mobile homes. Eggers pointed out that this was indeed a residential finance survey and to the extent that consumer installment loans financed homes in the survey they should be covered, although not necessarily in detail. The outcome of the discussion was to add the following questions regarding mobile homes: (1) Is there an installment loan on the mobile home? and (2) Is the mobile home on a permanent foundation?

Asking the property owner "how much was saved in monthly payments as a result of the latest refinancing" elicited some interesting discussion. Group members were quick to point out that not all refinancings result in a reduction in the mortgage payment. Dan Weinberg pointed out that some mortgage payments actually go up as a result of a refinancing, although the property owner is better off as a result. Eggers agreed. The group suggested rewording the question to ask "what was the change in principal and interest payment" as a result of the refinancing. Frank Nothaft and Jim Follain, having done work in the area, suggested that perhaps asking about the change in the interest rate rather than the mortgage payment (principal and interest) would be a reasonable alternative.

Respondents are asked in question 7i what was their main reason for refinancing. Some of the participants expressed concern that question 7i as it is currently worded suggests that cash proceeds came from the refinancing and this should be changed to reflect the fact that many respondents did not receive any cash as a result of the refinancing.

"Simplify", and find out how much, if any, cash came out of the refinancing, was the unanimous conclusion of the group.

Savage indicated that while a question will be added to try and identify the mortgage originator, the goal of the question will be to find out who originated the paper work and was receiving the origination fee rather than where the funds were actually coming from.

Discussion ensued about the purpose of the originator question from a policy perspective. Savage indicated that the ultimate purpose of the question would be to classify mortgage originators into types (banks, other depository institutions, mortgage banker, and so forth). Since the respondent might not be qualified to classify the mortgage originator, the Census Bureau would classify the originators using the Standard Statistical Establishment List (SSEL). While many agreed that this would be very useful information, Eggers and Forrest Pafenberg strongly suggested that this would likely be very difficult to implement given the dynamic nature of the mortgage industry. Follain agreed, adding that while the Home Mortgage Disclosure Act (HMDA) contained mortgage originator data for single family homes, he questioned how good that particular data item was given the difficulty of classifying mortgage originators.

How the mortgage was originated rather than who originated the mortgage was a question Pafenberg wanted to see added to the property questionnaire. The group was enthusiastic about this suggestion since mortgage originations, which used to be done solely via personal contact, are migrating to the web, phone, mail, etc.

The discussion of the Internet raised some concern by James Mikesell about rural access to loans. While the Internet would make loans more accessible to some, many in rural areas do not have computers and Internet access and thus would be unassisted by this new method of mortgage origination. Paul Emrath picked up on the access discussion suggesting that the construction industry had used the distance traveled to obtain a mortgage as a measure of access. Fronczek indicated that while access is an important issue, unfortunately it may be beyond the scope of the RFS.

Relative to rural properties, Feinberg suggested that the Farmers Home Administration (FmHA) has been changed to Rural Housing Service/Rural Development and that all references to FmHA should be changed accordingly. He further recommended that since the change was relatively recent that all references read "Rural Housing Service/Rural Development, formerly Farmers Home Administration" to clarify any confusion that respondents may have. He also mentioned that Rural Housing Service/Rural Development does not insure loans rather it guarantees loans and makes direct loans.

Jack Goodman considered this a very important question since there are several surveys that capture operating expenses but not many that capture capital improvements. Discussion focused on what the time frame should be. Emrath said that in his experience respondents were unable to remember capital expenditures over a very long time period. O'Brien asked whether the time period should correspond to the tax year since capital improvements may have some tax implications. In the end, the group agreed to a three-year period.

The difficulty of getting good data on capital improvements was a theme among the participants. Shaun Brady said that he has been trying to get this information since his organization's survey of apartment expenses began three years ago and said that the data that he was getting did not appear to be very good. Amy Bogdon asked how routine maintenance was treated in the RFS - as an annual operating expense or capital expenditure? (It is currently treated as an operating expense, according to Savage.) She said that unless the questionnaire is very explicit about the categories of maintenance, many respondents will have a difficult time distinguishing between these. O'Brien agreed.

Brady indicated that his survey (2 pages) would be mailed to owners of roughly 50,000 apartments at the end of February and that the Census Bureau could test RFS questions in his survey to see what kind of responses could be expected in the RFS.

Bogdon again expressed concerns similar to those about ambiguous categories of capital improvement and said that categories that were clear to the respondent would improve the validity of the data. O'Brien agreed with the difficulties that arose when respondents are given vague categories. Fronczek suggested that the Census Bureau may be able to clarify the categories by giving the respondents strict guidelines. He also suggested perhaps only having three categories: real estate taxes; property insurance; and "other" operating expenses.

George Smith said that in his experience, even when guidelines are given to respondents, it was nearly impossible to know what he was getting. Therefore, he suggested that there be three or four well-defined categories that constitute the bulk of expenses and the Census Bureau should limit respondent choices to these categories.

Bogdon indicated that it would be useful to know whether there was a professional property manager or management company for the property and suggested that one possible way to capture this may be to make one of the operating expense categories be fees paid to a professional property manager or property management company.

The group agreed that a limited number of clearly defined, operating expense categories for a one-year period would be sufficient.

There was some concern that when rental property owners are asked "how much were the total actual receipts from rent during the past year from residential units" (Question 17 a. (1) on the Rental and Vacant Property Questionnaire) that for some senior housing facilities the rent would be overstated because it would include payments for long-term care. Emrath and other group members allayed this concern saying that generally tenants pay separately for housing and long-term care components. Nonetheless, Nothaft stressed the importance of separating out the long-term care component from the housing component, while Emrath and Goodman believe that this will be difficult to do if the residents do not pay separately. (Emrath and Goodman drafted several questions for identifying seniors housing. The proposed questions for both the homeowner and rental and vacant property questionnaires are designed to identify both housing that is age-restricted and housing that is primarily occupied by senior citizens due to marketing campaigns aimed at this group. They have also proposed a question about whether or not the rental payment includes charges for meals or other services in order to identify rental payments that are "contaminated" by nonresidential components. Goodman and Emrath concurred that to find out how much of the rental payment is not for housing services is beyond the scope of the RFS.)

The group agreed to test the following questions. (1) Is the housing age restricted or targeted to the seniors market? (2) Is long-term care provided? (3) If long term care is provided, is the payment included in the rent?

Emrath stressed the importance of including seniors housing-related questions on both the rental and homeowner property questionnaires.

Savage indicated that in addition to retaining the questions that appeared on the 1991 RFS on the lender questionnaire, a question would be added to identify high loan-to-value (LTV) loans. Is this loan for an amount greater than the appraised value? This identification of high LTV loans would assist in distinguishing between lines of credit and traditional second mortgages.

The matching of mortgage lenders to a list of those lenders known to be in the business of making subprime (B and C) loans could be done at the Census Bureau if necessary, said Savage. Nothaft thought that the Census Bureau should just ask the lender if he/she considered the loan on the sample property a subprime loan. Pafenberg voiced the concern that since there is no single definition of a subprime loan, two lenders may look at the same loan with one identifying it as subprime and the other not identifying the loan as such. Nonetheless, the group generally agreed that it would be good to ask the lender, "Is this a subprime loan?"

One addition to the lender questionnaire related to adjustable rate mortgages will be to ask if non ARMs had already been converted to fixed rate prior to the survey. Nothaft also wanted to know if the terms or length of the mortgage had changed.

Buydowns (payment of points up-front by the developer/homebuilder/seller on behalf of the borrower to reduce the borrower's mortgage interest rate below market rate for a specified period, often five years) again sparked considerable interest among group members. Dan Quan emphasized the need to capture the buydown features of mortgages since builder buydowns in the first years of the mortgage may make the interest rate appear artificially low. Pafenberg agreed, adding that the Census Bureau also needs to capture the terms of the buydown. Stephen Zabrenski said he is also very interested in seller buydowns. But in addition to terms, he is interested in who is actually subsidizing the loan. He is also interested in how much of the loan settlement was contributed by the seller (buydown). And, if there was an interest rate buydown, how long does it last?

Discussion of this centered on simply changing the name from reverse annuity mortgage to reverse mortgage, which is the current vernacular, reflecting the fact that more mortgagors opted for the line of credit rather than the annuity feature.

Renewed interest in mortgage performance research led group members to call for putting the 1981 questions regarding the current loan status back on the 2001 RFS lender questionnaire. Nothaft and Follain indicated the Census Bureau should measure a longer time period than in 1981, when loan payments that were 30 days past due were considered delinquent. For instance, ask the lender if the mortgage was up-to-date (or less than 90 days past due) or delinquent by 90 days or more. If it was delinquent, was it in foreclosure?

Nothaft suggested in the November 13 meeting that the Census Bureau change the wording on the lender questionnaire from "who holds this mortgage?" to "to whom do you remit the mortgage payment?". Savage was enthusiastic about this suggestion and thinks that more lenders will be able to answer the latter question correctly.

Pafenberg pointed out in the November 13 meeting that loans from state housing agencies are often used in conjunction with FHA mortgage insurance. This situation will likely create problems with question 6 of the 1991 lender questionnaire, which requires that the respondent/lender choose one agency type that insures or guarantees the mortgage. The question will be changed to offer the respondent the choices for insurance or guarantees of "state or local housing finance agency only" or "state or local housing finance agency with FHA insurance, VA or Rural Development/Rural Housing Service guarantees" in addition to the other choices offered on the 1991 questionnaire.

Many of the group members were interested in collecting information on the credit history of the borrower. Fronczek was reluctant to do this. Quan indicated that the Census Bureau should ask the owner about the owner's credit history and whether the owner may have defaulted on a loan in the past similar to the way in which the Board of Governors of the Federal Reserve System's Survey of Consumer Finances does it.

On a related issue, Nothaft suggested that the Census Bureau ask the property owner how many times did he/she have to apply for loans prior to obtaining one for this property. Pafenberg was interested in determining from the property owner how the owner remedied the problem to obtain the loan, but this may not be feasible in the RFS.

Goodman indicated the importance of making the categories of type of property owner more consistent with prevailing legal definitions. He suggested adding "pension fund"; substituting "depository institution (commercial bank, savings and loan, credit union)" for "financial institution (other than life insurance company); dividing "REIT (real estate investment trust)" into two categories: "public REIT (stock trades on a public stock exchange)" and "private REIT (stock does not trade on a public stock exchange)"; and combining "real estate corporation" and "corporation (other than real estate)" into a new category "corporation (other than REIT but including limited liability companies)". Goodman feels that "real estate corporation" is confusing because the response in 1991 was a self-identification by respondents based on their primary line of business. A problem can arise when respondents view themselves primarily as real corporations and thus respond "real estate corporation" when in fact they legally fall into another category. Goodman believes that this self-identification seems vague and confusing to both respondents and data users.

The definitions of a property differ between the POMS and the RFS. In the POMS, a property is defined to be all land and buildings covered by a single deed. Even if there were individual mortgages on multiple buildings, as long as one deed covered the multiple buildings, then it was considered to be one property. In the RFS, a mortgaged property is defined to be all land and buildings covered by a single first mortgage. If multiple buildings covered by a single deed each have individual mortgages, then they are considered multiple properties despite the fact that they have a single deed. Goodman expressed interest in reconciling the definitions between the surveys. However, after some limited discussion, it was agreed that the objectives of the surveys were different and that to try and reconcile the two definitions would very difficult to implement.

There was also a lot of interest in being able to classify properties according to the number of buildings on them. Fronczek indicated that this can be done using questions that appeared on the 1991 RFS questionnaire.

The respondent to the RFS questionnaire could affect the reliability of the responses. Bogdon asked if the Census Bureau could find out who the respondent to the property questionnaire is (owner, professional property manager, etc.) so that researchers using the public use microdata can make their own determination about the reliability of some of the responses based on the type of respondent.

For various reasons, many participants wanted to be able to match records from the 2001 RFS to records from Census 2000. Follain wanted to match back to the Census file to find the income distribution of tenants in a multiunit property. Bogdon wants to be able to match to find the building size from the Census file. Fronczek and Savage indicated that this possibly could be done by specific request.

• Question 4.c. - Savage suggested adding two responses, "By purchasing lot first, then building house," and "By divorce." The group agreed.
• Question 4.d. - The group agreed to Savage's suggestion of modifying the first response option to read, "New or vacant lot."
• Question 5.b. - Michelle Hamecs recommended adding a response option: "placed construction loan that turned to permanent loan". Savage indicated that construction to permanent mortgages currently would be captured under the first response option, "Placed new mortgage (land contract, etc.)"
• Question 5.c. - Mikesell and Follain suggested dropping this question. There was no objection from the group.
• Question 5.d. - There were several suggestions for additions to the list of possible major sources of down payment from which the respondent could choose. They include: funds borrowed from a retirement plan; insurance proceeds; nonprofit organizations; and lease purchase agreement.
• Question 7.d. - Group members suggested adding "biweekly", in addition to monthly, quarterly, and other, as a choice for how mortgage payments were made.
• Question 7.i. - "Settlement of divorce" should be added as another main reason for refinancing or placing a mortgage on a property that was owned free and clear. The group had discussed earlier that all refinancings are not done to receive cash. Since the response options to this question seem to suggest that cash proceeds would come from the refinancing, the group concurred that the question should be restructured to include the amount of cash that came out of the refinancing. "To change partnership agreement" (rental property questionnaire only) and "to start a small business" should be added as response options.
• Question 15.a. (14.a. on the Rental and Vacant Property Questionnaire) - An additional response option to "Does this property benefit from - " should be "land trust". (A land trust is a private, non-profit organization whose goal is to acquire and hold land for the benefit of the community and to provide secure affordable access to land and housing for community residents. A more detailed definition can be found at //users.lazerlink.com/~scclt/what.htm).
• Question 15.c. - The Census Bureau will add two additional possible responses for "Are these benefits provided because the property is -": "covered by disaster relief"; and "covered by Homestead Exemption." (Homeowner questionnaire only.)
• Question 19 (Homeowner questionnaire only) - Lynne Sabatiuk recommended changing this question about the property owner's/respondent's history of active duty in the Armed Forces to reflect current guidelines, which she will provide.

A new question should be added to rental and vacant property questionnaire to get the property owners' income (individual property owners only).

Savage also indicated that a question will be added to the lender questionnaire asking for the appraised value of the property at the time the mortgage was placed. However, no question would be added to the lender questionnaire asking for a property value at the time of the survey.

Bogdon would like to pose questions to rental property owners/professional managers to find out if the property accepts tenants receiving Section 8 assistance. And if so, are there currently any Section 8 tenants living in any of the units?

Fronczek told the group that the Census Bureau will develop a set of questionnaires fairly soon. They will be sent to the group for comments. At the same time, they will be sent to our survey methods experts for testing through focus groups, cognitive interviews, and/or whatever other methods they think proper. We will then get back together in August/September to discuss the results of the testing.

Bureau of the Census
Suitland, Maryland
February 18, 2000

Attendees: Fred Eggers, Robert Knight, William Reeder, and Ron Sepanik (Department of Housing and Urban Development); Tom Blatt, Walter Busse, Linda Cavanaugh, Peter Fronczek, Debbie Knoll, Ashley Landreth, Len Norry, Eileen O’Brien, Howard Savage, and Mary Schwartz (Census Bureau); Amy Bogdon (Fannie Mae Foundation); Frank Nothaft (Freddie Mac); Forrest Pafenberg (National Association of Realtors); Steve Andrews and George Smith (Bureau of Economic Analysis); and Jack Goodman (National Multi Housing Council).

The Census Bureau formed the group to advise on the content of the Residential Finance Survey (RFS), designed to look at the financing of residential properties. Membership is comprised of government agencies – Federal Reserve Board, Bureau of Economic Analysis, Department of Agriculture, Office of Thrift Supervision, Veterans Administration, Department of Housing and Urban Development, and the Census Bureau; government-sponsored enterprises – Fannie Mae and Freddie Mac; trade associations – National Association of Realtors, National Association of Home Builders, Mortgage Bankers Association, National Multi Housing Council, Multifamily Housing Institute; and the Fannie Mae Foundation.

Peter Fronczek welcomed the members of the working group and complimented them for coming out on a day with such inclement weather.

Fronczek then briefed the group on two things that had occurred since the last meeting:

• Web Page

The RFS web page is up and can be accessed at //www.census.gov/housing/rfs/. The page currently contains an overview of the survey; minutes from the previous two working group meetings; a "glossary" of terms used in residential finance; information on Titles 12 and 13; and a list of important RFS dates. The web page will be updated periodically.

• Pre-Survey Contact with Lenders

A new initiative for the 2001 RFS is the pre-survey contact with lenders. The Census Bureau will contact about 15,000 lenders about nine months prior to the RFS to alert them to the survey and to determine the appropriate "contact person" for each lending institution. This should be a major improvement from previous surveys in questionnaire control, and should translate into reduced time and cost in conducting the lender portion of the survey.

Eileen O’Brien of the Center for Survey Methods Research (CSMR) provided a detailed report on the results of the Census Bureau’s cognitive testing of the two primary property questionnaires – the homeowner questionnaire and the rental and vacant property questionnaire.

• Homeowner Questionnaire

CSMR conducted in-depth cognitive interviews with 16 homeowners – six single-family owners, three townhouse owners, three condominium owners, and four mobile home owners. Each interview lasted 60 to 70 minutes.

Most respondents believed the survey made reasonable demands on their time, knowledge, and effort. Homeowners were less form and topic savvy than rental property owners/managers and made more errors in interpreting questions and executing skip instructions. In spite of their errors, most thought the questionnaire was generally self-explanatory and easy to follow.

Homeowners had problems with the following:

• The concept of property - Property was interpreted narrowly by condominium and mobile home owners. They sometimes thought of property as land only.
• Cash from refinancing - This was often interpreted narrowly as only the amount of money that the owner actually walked away with, and not any money that had gone to paying off bills or loans on the owner’s behalf.
• Purchase price, original loan amount, and current first mortgage – Respondents perceived these as redundant questions and wanted to make them the same.
• First contact with lender – The intent of this question was not clear to respondents. Is it to determine how they completed their application, or how they heard about the lender?
• Special assessments – Respondents didn’t know what was wanted and will often create their own definitions.
• HUD-1 question series – Respondents were concerned that this is an invasion of privacy. They also had little knowledge as to what the HUD-1 form was, and they had little motivation to go to their records and try to locate the form. The effect on return rates must be considered here.

There were successes with homeowners in the following areas:

• Answers to key questions, such as purchase price, mortgage amount, year purchased, etc., were instantaneously recalled, probably because they are linked with major life events.
• Respondents rarely referred to records, suggesting the questionnaire can be completed in one setting, thus promoting good return rates.
• Respondents had a more sophisticated understanding of mortgage finance than expected. Survey data quality should benefit from this increase in technical knowledge.

The Census Bureau will work to address the problems encountered by homeowners by improving instructions and modifying questions. The HUD-1 series of questions will be dropped.

CSMR conducted in-depth cognitive interviews with owners/managers of six large multi-unit rentals, and two owners of single family rentals. Interviews lasted 45 to 90 minutes.

Large rental property managers sorted the questions into two categories – those they could answer, and those they did not have the knowledge or authority to answer. This will result in the Census Bureau having to contact two (or more) respondents to complete one questionnaire at many rental properties.

Although respondents thought the questionnaire could be answered easily, few believed it would get the attention it deserved unless the mandatory nature of the inquiry was made more visible and a deadline for responding was given.

Rental property owners/managers had problems with the following:

• First contact with lender – This question doesn’t reflect reality, at least with large rental property owners. These owners often have long term business relationships with their lenders.
• Property value – Property managers were reluctant to speculate on property value.
• Special assessments – Same problems as with homeowners.
• Property management and administration and all other operating expenses – There was some evidence that respondents had trouble determining what went in one category versus the other.
• Capital improvements – Respondents used different definitions to identify these improvements.
• Units rented or available for rent and estimated vacancy loss – Respondents were confused as to the intent of these questions.
• HUD-1 question series – There was even a greater concern about privacy and the burden imposed on respondents by rental property owners/managers than by homeowners.

There were successes with rental property owners/managers in the following areas:

• They were far better readers and more form and topic savvy than homeowners.
• They generally referred to their records and were confident the information they extracted from their records was accurate.
• As knowledge sharing is common between the local and central office of management companies, multiple respondents were likely to be available to answer questions.

The Census Bureau will work to address the problems encountered by rental property owners/managers. The HUD-1 series of questions will be dropped.

In addition, the results of the cognitive testing, specifically regarding the need to go to multiple respondents, has led the Census Bureau to consider not mailing to large rental property owners/managers (those owning/managing over 10 units). We will go directly to the field with these cases. This should provide the best opportunity to get to the right respondents with the least amount of difficulty.

Homeowner and Rental and Vacant Property Questionnaires

• The intent of question 8c (first contact with lender) was a problem for many respondents. The consensus was to ask how the respondent applied for the loan (not the pre-qualification application). The following wording will be tried, "How did you apply for the CURRENT first mortgage on this property?"
• Change and simplify the first two response options to question 8h (type of insurance) to be: "FHA insured" and "VA guaranteed".
• Add the following as part of response option 5, question 9c (main reason for refinancing): "To receive cash or pay off debts".
• Add the following as response option 9, questions 9e, 10b, 11f, and 12f (reasons for obtaining various loans): "To pay taxes".
• Add the word "current" to question 13a (name of lender), "To whom do you make your current FIRST mortgage…"
• Questions 20a, b, and c (property benefits) (19a, b, and c on the rental property questionnaire) will be reviewed by Forrest Pafenberg (NAR) and HUD to see if any changes to the response options are required, particularly with regard to equity sharing with local governments.

Rental and Vacant Property Questionnaire

• Contact the owners/managers of the largest rental properties prior to the survey and inform them of the survey, including the mandatory nature of the survey. A list of these owners/managers is available at the National Multi Housing Council web site.
• Add as a parenthetical in question 17I (operating expenses) the following: "…(exclude principal and interest payments and capital expenditures)".
• Make current question 23 (units available for rent) part of question 14 (14c), and change the wording of the question. Members of the group will send their suggestions on wording to the Census Bureau.
• Add a question on whether there is a third party management company managing the property.

Lender questionnaire

• Change the wording of questions 3a and 5a (holder/servicer and type of lender) to, "Do you eventually get…", and "Who eventually gets…"
• Change question 16b (type of buydown) to, "What type of buydown is this?" Forrest Pafenberg will send a suggested question and response options.
• Prepayment penalties are still common with conventionally-financed multi-family properties. The Census Bureau will look into adding a question to capture penalties in the form of yield maintenance and lockout periods.

The Census Bureau will redraft the three RFS questionnaires based on the results of the cognitive testing and the suggestions provided at this meeting. The group members will have one last chance to comment on the documents. The deadline for having a final set of RFS questionnaires will be the end of March 2000.

The Census Bureau will call on the working group for additional guidance and help: (1) to advise on the type of data products and tabulations that should result from the survey; and (2) for those organizations that hold mortgages, to act as a liaison between the Census Bureau and the organization to identify the person who will complete the RFS questionnaires. There will be a fourth meeting of the working group, probably in late spring or early summer to discuss these issues.

Bureau of the Census
Suitland, Maryland
September 14, 2000

Attendees: Fred Eggers, Bob Knight, Bill Reeder, Bill Segal, and Ron Sepanik (Department of Housing and Urban Development); Tom Blatt, Walter Busse, Linda Cavanaugh, Pete Fronczek, Rich Levy, Len Norry, Howard Savage, and Mary Schwartz (Census Bureau); Michael Feinberg and Jim Mikesell (Department of Agriculture); Jim Freund and Frank Nothaft (Freddie Mac); Mark Obrinsky (National Multi Housing Council); Jo Chapman (National Association of Home Builders); Mark Calabria and Forrest Pafenberg (National Association of Realtors); Amy Bogdon (Fannie Mae Foundation); Steve Andrews and George Smith (Bureau of Economic Analysis); and Jack Goodman (Hartrey Advisors).

The group was formed to advise on the content of and data products from the Residential Finance Survey (RFS), which gathered data about the financing of residential properties. Membership in the group is comprised of government agencies - Federal Reserve Board, Bureau of Economic Analysis, the Department of Agriculture, Office of Thrift Supervision, Veterans Administration, Department of Housing and Urban Development, and the Census Bureau; government-sponsored enterprises - Fannie Mae and Freddie Mac; trade associations - National Association of Realtors, National Association of Home Builders, Mortgage Bankers Association, National Multi Housing Council, and Multifamily Housing Institute; private sector participants - Hartrey Advisors; and the Fannie Mae Foundation.

Peter Fronczek (Census Bureau) welcomed the working group and everyone introduced themselves. He also thanked HUD, the sponsor of RFS, for its commitment to the survey. Fronczek specifically recognized Fred Eggers' leadership and said that he would be missed when he retires in November. Fred Eggers (Department of Housing and Urban Development) thanked Fronczek, and told the group that many difficult decisions had to be made to keep the RFS, in particular the postponement of the American Housing Survey (AHS) 2000 metropolitan area survey.

Highlights since the last meeting were mentioned, including the pre-survey of lenders and the electronic reporting option for lenders, both of which were making their debuts in the 2001 RFS.

Mary Schwartz (Census Bureau) provided a report on the pre-survey. The main objective of the pre-survey was to get lending institutions to identify a primary contact to whom to send the lender questionnaire in the event that the lender held a lien on a sampled property. The second objective of the pre-survey was to determine a lender's preference for booklet or electronic reporting since larger lenders will probably be getting many questionnaires. A positive "side effect" of the pre-survey is that it gave lenders a "heads-up" about the RFS, which would take place one year later.

In June 2000, we mailed a brief questionnaire to approximately 11,000 corporate headquarters of lending institutions. The response rate for the mailout was just under 63 percent. About one month after the mailout, we sent the nonresponse cases to telephone follow-up. Of the nonresponding lenders that the telephone center was able to reach, only four refused to complete the pre-survey. Of the participating lenders, about one-fifth said they wanted to answer electronically.

Schwartz thanked the working group for providing lists of members and customers and for their endorsement of the survey.

Howard Savage (Census Bureau) told the group that Federal agencies, many of which hold or service mortgages, were not included in the pre-survey. As such, contacts would need to be identified at these federal agencies, many of which had representatives at the meeting today. He asked each federal agency representative present to provide Walter Busse (Census Bureau) with an appropriate contact for his or her agency. He added that if any formal agency to agency contact were needed to let him know.

Rich Levy (Census Bureau) reported on the electronic reporting alternative, which is being offered for the first time in 2001. Over the past decade, mortgage banks and other residential lending institutions have merged. It is now more likely a lending institution (particularly a large one) will hold or service mortgages for more than one property in the RFS sample.

In order to make it easier for lending institutions to complete the RFS, we gave the institutions in our pre-survey the option to request an electronic version of the questionnaire.

For those who want to respond electronically, our idea is to send a diskette in either spreadsheet or ASCII format, with detailed instructions about where they should put the data for each question.

Informal conversations with staff from lending institutions, most of whom were calling with regard to the pre-survey, suggested that a spreadsheet in Excel or ASCII would work for them. We are also trying to set up structured interviews with staff from lending institutions.

We have already learned that lenders are used to such data requests, and they can work with the format we proposed. However, it will be important for the institutions to receive instructions in advance. Second, while organizations are understandably concerned about the confidentiality of customer data, the Census Bureau's reputation would help mitigate these concerns. A letter accompanying the survey, specifying a contact person and the statutes requiring the collection of data, would be particularly critical in mitigating concerns.

There has been a change from the 1991 RFS in the method for identifying the owners of large (those with 20 or more units) rental properties, Fronczek explained. (This is known as the "owner-seeker" operation.) Rather than send out letters to seven residents of large rental properties to identify the property owner, representatives from the Field Division will call or visit the property to identify the owners. In the survey itself, the same field representatives will then go directly to interview the owners of the properties that they identified in the owner-seeker operation. The expectation is that this direct to field approach for large rental properties will speed up the process and improve the accuracy of owner identification.

A printed report component enjoyed widespread acceptance among group members. Many of the working group members, including Frank Nothaft (Freddie Mac) and Mark Obrinsky (National Multi Housing Council), indicated that they liked having the printed report on their shelves for quick reference. But ideas for other ways in which to disseminate the data emerged, like putting more detailed data on CD-ROMs or an American Fact Finder-type system. Suggestions for new tabulations for the printed report, Internet release, and CD-ROM also emerged. The microdata file (public use file) was another topic of discussion. Fronczek made it clear that the Disclosure Review Board (DRB) will review all proposed products and will ultimately decide what may be released.

• Regions (Four Census Regions) Tabulations.
  George Smith (Bureau of Economic Analysis) uses regional data but suggests putting regional data on a CD rather than in the printed volume, thereby reducing the size of the printed report. Obrinsky concurs; he suggests that the printed report should have just one table containing basic regional data with regions listed across the top of the table in the box head.
• Division (Nine Census Divisions) Tabulations.
  The group would like to have some tabulations compiled at the division level.
• State distributions.
  Mark Calabria (National Association of Realtors) wanted to know the rationale the Census Bureau used in determining the four states for which it presented state data in 1991. Fronczek responded that it was determined after the fact. That is, the RFS is designed to be a national sample (about 70,000 addresses comprising 60,000 properties in both 1991 and planned for 2001); the Census Bureau determined the states for which the RFS could provide reliable estimates but this was after the sample was drawn and data were collected. Calabria wanted to know if we could produce data for perhaps a dozen states. Fronczek said that he would check into this.
• AHS distributions.
  Given his constituency, Jim Mikesell (Department of Agriculture) would like to see an urban/rural distribution in the printed report, similar to that which is done in the AHS. Smith would like to use the same property value classes as those in the AHS in order to be able to impute rents to owner-occupied properties. While there cannot be comparable classes for large rental properties between the AHS and the RFS, there can be comparable classes for home owner properties, said Fronczek. In fact, AHS and the 2000 Census are used as guides when distributions for the RFS are determined.
• Differing distributions depending on tenure (home owner or rental) and property size.
  Since disclosure breaches are more likely for rental properties, which tend to be unique and thus more identifiable, than for home owner properties, Calabria proposed that the Census Bureau consider pushing the geography of home owner properties to a lower level than for rental properties. Similarly, since large rental properties are more likely to be unique and identifiable than smaller rental properties, perhaps the Census Bureau could present data for lower levels of geography for smaller rental properties than for larger rental properties. Fronczek indicated that he would look into these and discuss with the DRB.
• Split Out Single Family Rental Properties From Other Rental Properties.
  Because the operating and other characteristics of single family rental properties are dramatically different from other rental properties, according to Obrinsky, they should be separate from other categories for all tabulations. This can be done.
• Rental Property Characteristics by Owner Type.
  Jack Goodman (Hartrey Advisors) is interested in the characteristics of rental properties by owner type, particularly for large (those with 50 or more units) properties, whose owners are often apartment REITs (Real Estate Investment Trusts).
• Neighborhood Characteristics.
  Amy Bogdon (Fannie Mae Foundation) asked if we can produce tabulations dealing with neighborhood characteristics. (This could be done by linking RFS properties with data about the census tracts in which they are located from the 2000 Census long form data). Nothaft and Bill Reeder (Department of Housing and Urban Development) would like these tabulations on the more detailed CD-ROM. Savage urged them to send specifics to him.
• Seniors' housing.
  Jim Freund (Freddie Mac) asked if the Census Bureau can tabulate data on seniors' housing for the printed report. This way the growing number of analysts interested in seniors housing would have some data on seniors housing at their fingertips. Such tabulations would also make an interesting Statistical Brief.
• Junior financing and HELOCs (Home Equity Lines of Credit).
  Smith would like to tabulate the characteristics of HELOCs separately from other junior financing, and would like to know if there is sufficient interest to do so. (The group unanimously agreed that there was great interest.)
• Mortgage insurance status of and junior mortgages on 2-4 unit home owner properties.
  Nothaft would like to see the two tables on Mortgage Insurance Status and Junior Financing/HELOCs included in the chapter on 2-4 unit home owner properties, even if only on the CD-ROM or Internet version and not on the printed volume.
• Tables that are not reproducible.
  Nothaft suggests that the CD-ROM or printed report would be an excellent way of disseminating tables that an analyst could not reproduce using the public use microdata files because of the use of top coding and other data masking techniques. Perhaps, the bulk of tables in this CD-ROM or printed report would be of this form. Nothaft stresses that we should still publish some reproducible tables so that a user of the microdata file will have a published resource to check his/her own tabulations to make sure he/she has done them correctly.
• Averages of top coded values.
  Calabria liked getting averages of top coded values.
• Explanation of calculation of values.
  In the printed report (and CD-ROM and Internet), Bill Segal (Department of Housing and Urban Development) suggested offering more detailed explanations and better documentation of how values were calculated. This would help analysts using the microdata files to reproduce published values. Fronczek replied that this can be added to the documentation.

Suggestions for method of dissemination included:

• Use CD-ROMs for detailed tables and reserve printed report for summaries, Savage suggested. The entire group concurred. The cost for the CD would be low as the Census Bureau is no longer trying to price it to cover overhead but only to recover direct costs.
• Develop a product like American Fact Finder, in which the user can generate his/her own desired tabulations, recommended Bogdon. This eliminates the need to anticipate in advance what tabulations will be needed.
• Tom Blatt (Census Bureau) suggested putting the entire document on the Internet and printing only a subset. This suggestion had widespread appeal. Calabria said that he wants everything on the Internet; in fact, this is his primary method for accessing the AHS. Ron Sepanik (Department of Housing and Urban Development) also favored putting everything from the RFS on the Internet. Bogdon preferred a spreadsheet format rather than a PDF file, so that she doesn't have to enter data after downloading it from the Internet. Eggers agrees with Bodgon that PDF files are cumbersome to use. Blatt said he will find out if we can put tables into spreadsheet rather than (or in addition to) PDF file.
• Increase the number and variety of statistical briefs as a means of dissemination. This method is particularly effective for the press. Most of the group members responded favorably to the statistical briefs that the Census Bureau had done but wanted a wider variety. Smith particularly liked the statistical brief from the 1991 survey about home equity lines of credit. Freund would like statistical briefs on seniors housing, which was not identified in 1991, but will be identifiable in the 2001 RFS. Forrest Pafenberg (National Association of Realtors) would like to see a Statistical Brief on junior financing, particularly one comparing characteristics of first mortgages to home equity lines of credit. Smith would like one on expenses by type of ownership.

• Rounding in Tables.
  Do not round interest rate on the microdata file, requested Jo Chapman (National Association of Home Builders). She suggested making data available as reported by the owner on the property questionnaire and by the lender on the lender questionnaire. Chapman would like to compare lenders' and property owners' reporting.
• File Structure.
  Goodman asked whether putting all microdata observations into one file rather than having home owner and rental and vacant properties in separate files, as was the case in 1991, made sense. Blatt says that the two, although very similar, have different formats. Therefore, it is probably best to keep them separate.

Bogdon and Goodman agreed that linking all mortgages to the respective property in one record should be done. Computer space was the main reason that they were not in one record in the 1991 files, Blatt reported. Since two-fifths of the properties were not mortgaged, there would be a lot of blank space on each record. Blank space greatly increases the size of the file. But downloading such a large file did not seem to be a problem for anyone in the group.

Blatt asked about the preferred file formats. The group wanted both a SAS file and a comma-delimited, ASCII file.

• Linking properties to census tract characteristics.
  Nothaft, Reeder, and Bogdon initiated a discussion about linking characteristics about the census tract in which it is located to the property. The group concurred that this linked information would be valuable for the GSEs (Government-Sponsored Enterprises), like Fannie Mae and Freddie Mac, the regulators, like HUD, and the research community in general. Information like the percentage of the census tract that is minority and whether the median income for the tract is less than the area median family income would be useful for policymakers without compromising confidentiality.

Fronczek and Bogdon actually did this "linking" with the 1991 RFS data, creating an internal file that was made available to Bogdon, who had been sworn in as a Census Bureau employee. Something similar can be done in 2001 for researchers wanting to use the file. Sepanik indicated that a formal request, along with a research proposal, must be submitted for consideration, however. Fronczek was certain that such a linked file will not be available as a public use file. Savage concluded by asking the group to provide us with guidelines as to what information that they would like to see at the census tract level.

• Inability to reproduce published values using microdata file.
  Analysts have had problems using the microdata file to replicate values that appear in printed reports. The group said that there are often discrepancies between their calculations and published values. Calabria suspected that these discrepancies may be due to top coding. Fronczek suggested that most discrepancies due to top coding should be minor. Savage further stated that aggregates will generally come out the same between the analyst's calculations and the published report.
• Difficulty in using rental and vacant property data with high property value.
  Chapman said that it was sometimes difficult to use the data in the top property value ranges. This problem arises because the Census Bureau takes the average value for a range and then assigns this average value to every property within that range. She asked if the Census Bureau could apply this same method to corresponding variables, like operating costs, and so forth, so that analysts can actually use the average values for ranges in analyses of individual properties. Fronczek responded that we will use the average value in range for all financial variables.
• "Mean value in range" problematic in regressions.
  The method that the Census Bureau used of assigning the mean value in a range to all observations falling within that range in the 1991 RFS, caused problems in the analysis of individual observations, Bogdon stated. She reported differences in regression results when using public use microdata file and printed values as independent variables. Savage asked her to send specific examples of differences she was experiencing.

Goodman wanted to know if the Census Bureau could provide a preliminary release of microdata for some "uncontroversial" variables, even before it had passed all the disclosure reviews. Fronczek said probably not. Sepanik was strongly opposed to having multiple versions of the same file in the public domain. Fronczek did say, however, that the products should be released more quickly this time because there will not be a huge printed report and because of improved processing methods.

We asked the agencies present to provide the names of contacts at Federal agencies and GSEs who should receive the lender questionnaire in the event that the property owner identifies an agency or GSE as the mortgage holder/servicer. We are also awaiting specific guidelines from the group as to what information they would like to see at the census tract level. We are also awaiting examples of inconsistencies between regression results based on published values of variables versus those based on values from the public use microdata file.

Since this was the last formal meeting of the 2001 RFS Working Group, we will be contacting representatives individually, if necessary, as we try to implement the group's suggestions and requests for data products.