«Moving to opportunity voluMe 14, nuMber 2 • 2012 U.S. Department of Housing and Urban Development | Office of Policy Development and Research ...»
Approximately two-thirds of MTO participants continued to rely on housing assistance at the time of the final impacts evaluation. Because voucher holders did not experience any improvements in their employment or earnings over the course of the demonstration, the fact that they are still as likely to rely on housing assistance as the control group is not surprising.
We found limited evidence that providing vouchers results in more housing instability compared with not offering vouchers. Receiving a voucher did not have any differential effect on experiencing homelessness compared with that of the control group; the Section 8 group (those who were offered a voucher without any geographic restrictions and no additional counseling), however, experienced more instances of doubling up with friends and family than the control group did.
Vouchers also had no disproportionate effect on housing costs, which were very high for all MTO participants, even those still receiving housing subsidies. The recent housing boom may have played a factor in explaining these high housing costs, causing all MTO voucher holders who stayed in place to pay more out of pocket for rent.
Voucher holders were more likely to ensure they were not late paying their rent, reflecting the requirements of living in the private market. It appears, however, that they were also more likely to make a tradeoff by paying their utilities late or not at all, which resulted in the utilities being turned off. Again, this finding is consistent with other research on families moving from public housing to the private market and suggests a need for greater attention to helping voucher holders meet the costs of utilities in private-market units.
Data and Methods For most of our analysis in this article, we rely on the survey data collected by the University of Michigan’s Institute for Social Research between June 2008 and April 2010 under its contract with the National Bureau of Economic Research for the final impacts evaluation. The survey used for the MTO final impacts evaluation (Sanbonmatsu et al., 2011) collected information from 3,273 adults and 5,101 targeted youth, covering a wide variety of outcomes and mediators in six domains, with response rates of 89.6 percent for adults and 88.7 percent for youth. In this article, we focus on the following housing outcomes and mediators self-reported by the MTO participants: housing quality, housing assistance, homelessness and doubling up, housing costs (rents or mortgages plus utilities), housing cost burdens, difficulty paying rents or mortgages plus utilities on time, eviction because of late rent or mortgage payments, and utilities being turned off because of late payments.
Most of our analysis in this article focuses on the intention-to-treat (ITT) effects, comparing the average housing outcomes of the experimental group with those of the control group. We also compare the average outcomes of the Section 8 group with those of the control group. All ITT effects are listed as such in the exhibits. The exhibits also include the treatment-on-the-treated (TOT) effects, which capture the effect of moving with either an MTO low-poverty or a traditional Section 8 voucher.2 We used the same sampling weights and regression models as described in Sanbonmatsu et al. (2011) to report the same ITT and TOT results as published in the final impacts evaluation.
Because we are interested in assessing whether offering vouchers improves low-income families’ outcomes, we look at how both the experimental and Section 8 groups fared compared with the control group.
Identifying Instances of Homelessness and Doubling Up Analysts tracked MTO participants’ residences throughout the duration of the demonstration, including addresses identified between surveying periods. The final impacts evaluation survey asked MTO participants to confirm each past address and report the month and year that they first moved to and left that address. In addition, the survey asked participants if, at any time between these addresses, they did not have a place of their own to stay. It also asked those responding in the affirmative with whom or where they stayed—such as, with friends or relatives, on the street, in a shelter, in an abandoned building, in a car, or in a hotel or motel, among other options—how long they were without a place of their own, and if their child(ren) was (were) living with them at the time. Participants were identified as having an instance of doubling up if they reported that they did not have a place of their own to stay and lived with their friends or family. People who reported being doubled up are, by definition, unstably housed. Heads of household were identified as having been literally homeless3 if they reported that they did not have a place of their own to stay and lived on the street or in shelters, abandoned buildings, cars or vans, movie theaters, or laundromats—essentially anywhere that is not deemed fit as a typical residence. Neither definition included participants staying at a hotel or motel, even when the respondent did not have a place of his or her own.
Identifying Housing Assistance Status Using Multiple Data Sources We employed a new multistep, multisource process to identify more accurately whether each MTO head of household was receiving any federal rental assistance4 and to determine the specific type of assistance received among those who were assisted at the time of the final impacts evaluation.5 Although housing assistance status is a key outcome of the MTO demonstration, determining whether a household is still receiving a subsidy and, if so, what type of subsidy it is receiving has been surprisingly difficult to determine. Other research has documented that recipients often misidentify the type of housing assistance they receive or erroneously report not receiving any assistance at all (see the appendix of Shroder, 2002). For instance, those using housing vouchers often misreport that the PHA is their landlord or simply say that they pay their own rent. Residents in all types of assisted housing often just respond that they live in “housing” without being able to specify which type. Relying on administrative housing assistance data can also be unreliable, because resident HUD uses the term literally homeless to differentiate between families living in places not fit for everyday residence (such as on the street, in abandoned buildings, in cars or vans) and families who are precariously housed, such as those who are doubled up. The full definition is available in HUD (2006).
Federal rental assistance, also referred to as deep subsidy, is defined as participating in a program that cuts housing costs to 30 percent of income (or some specified flat cost) for all participants in that program.
The specific types of assistance include public housing, tenant-based federal rental assistance, project-based nonpublic housing federal rental assistance, and no federal rental assistance (including owners, unassisted renters, the homeless, and those with other statuses).
annual recertification records are not always entered into the appropriate databases (Olsen, Davis, and Carrillo, 2005). To solve this problem, other researchers determined housing assistance status only where survey and administrative data match (Verma and Riccio, with Azurdia, 2003). This methodology, however, can exclude a significant proportion of those receiving housing assistance.
At the time of the interim survey, Orr et al. (2003) reported two housing assistance statuses of MTO participants: one based on survey responses and the second from administrative data. Only a 78-percent agreement existed between the two data sources.
To reduce misreporting by MTO participants, the survey for the MTO final impacts evaluation included a new series of questions to assess MTO participants’ housing assistance status.6 We then compared the survey responses with two annually collected administrative sources—Multifamily Tenant Characteristics System (MTCS)/Public Indian Housing and Information Center (PIC) and Tenant Rental Assistance Certification System (TRACS)/Multifamily data—to identify each MTO participant’s type of housing assistance. MTCS/PIC data contain longitudinal information on families living in public housing or receiving tenant-based housing vouchers (Form 50058), whereas TRACS/Multifamily data contain longitudinal information on families living in project-based Section 8 housing (Form 50059). The U.S. Department of Housing and Urban Development (HUD) Office of Policy Development and Research successfully matched approximately 90 percent of MTO heads of household to one or both longitudinal administrative data sources using a combination of first and last names, date of birth, and Social Security number.
In the first step of this new process to identify participants’ housing assistance status, we analyzed the series of housing assistance survey responses (step 1 in exhibit 1). The researchers coded respondents’ answers to each survey question as either eliminating or not eliminating each of eight possible housing assistance statuses tracked in this first step.7 As a result, MTO participants could have more than one possible assistance status at this point. Researchers chose this elimination method, as opposed to identifying affirmative answers to questions, to remove the nonresponse bias, particularly from inconsistently applied skip patterns. It also enabled the analysts to confirm participants’ multiple possible responses against the two administrative data sets. For instance, if we had used an affirmative method, a head of household who answered that the PHA is his or her landlord, even if that was not the case, would eliminate all housing assistance statuses except public housing, which often mistakenly occurs. Another example is that nine heads of household in the MTO final survey affirmatively answered that they received housing vouchers but denied that their landlords required proof of income for housing. By keeping the possibility that the person was using a voucher, we later were able to use the administrative sources to further hone down the possible housing assistance types.
Researchers based the new questions on the MTO interim survey (Orr et al., 2003) and the HOPE VI Panel Study, a fivesite study that tracked outcomes for 887 residents of public housing developments targeted for redevelopment. See Popkin et al. (2002) for a full description of the study.
The eight possible housing categories are renter with tenant-based assistance, renter in public housing, renter with projectbased assistance, renter without housing assistance, homeowner, homeless individual, individual who lives with family or friends and does not pay rent, and individual with another housing arrangement. The researchers could not determine assistance status for owners, because most owners were not asked any questions about housing assistance. For this reason, the final categories include information only on rental assistance, not on homeownership assistance.
Researchers then separately analyzed MTO heads of household who were successfully linked to the MTCS/PIC and TRACS/Multifamily data to determine the housing assistance status for any head of household on the administrative files (step 2). Researchers then compared the identified type of housing assistance from the survey responses and the administrative sources (step 3). If one status from the survey analysis matched a status from the administrative data, analysts assigned the respondent that housing assistance status.
Survey responses and the administrative sources did not match for 14 percent of MTO participants.
Analysts compared those participants’ residences at the time of the final evaluation survey with the known addresses of the PHA’s housing developments and project-based assistance buildings (step 4).
Also, they compared MTO participants’ addresses (ZIP+4) at the time of the final survey with both the survey responses and administrative data (step 5). For the 7 percent of MTO participants who still had conflicting housing assistance statuses after step 5, analysts selected the housing assistance status from the administrative data if the participant’s administrative records matched residents’ characteristics from the survey file and they found no duplicate records (step 6). Otherwise, analysts assigned participants a status based on the survey result. Exhibit 1 summarizes the process.
The following sections describe the housing-related findings from the survey for the MTO final impacts evaluation.
Findings The experimental and Section 8 group households experienced improvements in housing quality, but findings on housing stability were mixed. Changes in the housing market affected all households in the MTO demonstration. Neither the experimental nor Section 8 group experienced any differences in housing affordability or housing assistance.
Cityscape 93Comey, Popkin, and Franks
Sustained Housing Quality Improvements The MTO demonstration improved the housing quality of households that formerly lived in rundown public housing projects, consistent with findings from the HOPE VI Panel Study (Comey, 2004;
Popkin, Levy, and Buron, 2009). MTO participants started off at baseline living in very dire housing conditions. For instance, 25 percent reported their housing to be in poor condition, 58 percent reported problems with plumbing, and 61 percent reported problems with rats or mice. At baseline, MTO participants identified wanting better housing quality or a bigger sized unit (or wanting to leave their unsafe neighborhoods) as one of the main reasons for wanting to move (Orr et al., 2003).
At the time of the interim survey, evidence suggested that the demonstration positively affected both the experimental and Section 8 groups’ housing quality: 52 percent of the control group rated their housing as excellent or good compared with 62 percent of the experimental group and 59 percent of the Section 8 group (Orr et al., 2003). The experimental and Section 8 groups also reported fewer problems with vermin and peeling paint.
Exhibit 2 shows that, at the time of the final impacts evaluation, the positive effects on housing quality were sustained, particularly for the experimental group. Participants in the experimental group were more likely than those in the control group to rate their housing as excellent or good.