«Mixed Messages on Mixed incoMes Volume 15, Number 2 • 2013 U.S. Department of Housing and Urban Development | Office of Policy Development and ...»
The initial draft of the survey questionnaire was based on the study’s goals and informed by previous studies in the literature. LHA staff reviewed the questionnaire and provided suggestions to improve its clarity and content; a revised draft questionnaire was prepared for the formal pretest. The LHAs assisted in recruiting current voucher holders to participate in focus groups and to pretest the questionnaire. After the focus groups, slight revisions were made to the questionnaire, which was then finalized for implementation in the field. At this point, guided by the available demographic information for the voucher-holder population in Orange County, survey materials were translated into Spanish and Vietnamese and, to ensure accuracy, bilingual LHA staff reviewed the translations.
The survey design followed methods recommended by Dillman (2000) to optimize the response rate. For example, an introduction letter, describing the study and signed by the researcher and an LHA representative, was mailed to the respondents with the survey. The letter provided the same information in three languages—English, Spanish, and Vietnamese—and offered to provide the questionnaire in the voucher holder’s preferred language.9 The initial mailing was followed by a reminder postcard and, for nonrespondents, letters with a copy of the survey were mailed again, twice if necessary, at 2- to 3-week intervals.
The survey field period lasted for 5 months and concluded in August 2002. Response rates for the two areas were a concern from the planning stage of the project, because the literature suggests that certain characteristics associated with the voucher population, such as race, low incomes, and lower educational achievement, may affect survey response rates, although these results have varied across studies (DeMaio, 1980; Hennigan et al., 2002; Krysan et al., 1994). The response rates, however, were good for both LHAs, with 63 percent (n = 1,268) from the OCHA group and
56.3 percent (n = 467) in the SAHA sample.
These data are contained in the “Resident Characteristics Report,” which can be viewed and downloaded at http://portal.
Survey questionnaires were completed in Spanish, Vietnamese, and, in one instance, Farsi (special request by a respondent).
Across the two LHAs, however, less than 2 percent of the questionnaires were completed in a language other than English.
The current addresses of all voucher holders and the previous addresses for voucher holders who moved in the past 3 years were geocoded using Geographic Information Systems. Census-tract information from the 2000 census, summary files 1 and 3, was attached for all current and previous addresses to capture neighborhood characteristics, such as the poverty rate. School quality was measured by the Academic Performance Index (API), which is produced by the California Department of Education.10 The public schools were linked to each address using district boundary maps and school-locator search engines available through individual school districts; the schools and their API scores were then added for every address in the database. Finally, select LHA administrative data (for example, contact rent) were merged with the final dataset.
The full dataset contains 1,706 cases and is used when comparing movers with nonmovers. An exception is the analysis of public school quality, which has 1,522 cases for analysis. The sample contains 570 movers, but the lack of reliable previous address and school information resulted in fewer cases for analysis, and the number varies by the focus of the analysis. In these instances, the number of cases in an analysis is shown in the corresponding exhibit.
A caveat concerning the generalization of the results from these data is necessary for several reasons. First, the analyses use unweighted data.11 Second, some cases could not be confidently geocoded because of incomplete address information and had to be dropped from the analyses.12 Third, school-quality data used in several of the analyses were unavailable for approximately 12 percent of the addresses in the sample.
Analysis Background on LHAs The LHAs are in Orange County, in Southern California. Orange County was historically a suburban county, with residents often referring to the Orange Curtain, a sociodemographic line separating Orange County from the urban conditions of Los Angeles County, its neighbor to the north. Moreover, Orange County has and had a majority White population, with a higher median income and lower poverty rate than the state of California as a whole.13 Orange County has been changing, however, as its communities age and is experiencing increasing racial, ethnic, and economic diversity.
The California Public Schools Accountability Act of 1999 created the API score, a measure used to assess and track school performance over time. The score, which ranges from 200 to 1,000, is based on student performance on statewide testing.
The California Department of Education publishes these scores annually for public elementary, middle, and high schools throughout the state. For this research, I calculated the average API for the schools serving each address; however, in some cases, it was not clear whether a school served a specific address, so that school was not included in the calculation. For most cases, I had API scores for all three schools (elementary, middle, and high), but in some cases I had to average scores for only two schools, and in even fewer cases (less than 1 percent), I had only one school’s API score.
Weighting to address oversampling in each area was not done, because the samples were combined and response analyses showed some sociodemographic differences for responders in both samples (see Basolo and Nguyen, 2009, for a more detailed discussion of the response bias analyses for these data).
For these analyses, I filled missing values with the mean or mode of the variable, including missing values for the neighborhood poverty rate, which were a result of incomplete address information (1.2 percent of the cases).
See Basolo and Nguyen (2009) for Orange County demographic information in 2000.
The LHAs are also operating in one of the most expensive housing markets in the state and in the country. Despite the significant price downturn in housing markets throughout Southern California, Orange County has remained a relatively high-cost housing market and now appears to be rebounding in sales volume and median price for single-family homes (Lazo, 2012). The median contract rent in Orange County consistently exceeds the state and national figures.
Orange County has four LHAs that administer the HCVP. In general, the OCHA administers approximately 50 percent more vouchers than the next largest LHA (in Anaheim) and more than four times as many vouchers as the smallest LHA in the county (in Garden Grove). Its relative size among the LHAs in the county was the reason it was selected for the study. The SAHA has the third largest voucher program in the county, but was chosen for its location in the central city of Orange County, based on size, age, demographics, and, to a lesser extent, its role as the government center for the county.
Exhibit 1 shows voucher program characteristics for the OCHA and SAHA. The characteristics for 2000 and 2004 are presented primarily because the 2002 data (the year of the survey) were unavailable from the “A Picture of Subsidized Households” dataset on the HUD website. Also, however, these data allow for temporal comparisons for each LHA and comparisons between LHAs. In doing so, I recognize that the occupancy and reporting rates vary across the years and for the LHAs.14 The indicators provide a snapshot of the occupants using voucher assistance in 2000 and how voucher holders changed during the 4-year period. Both LHAs show an increase over time in seniors (defined as residents age 62 or older) with vouchers, although the OCHA provides about 5 percent more of its vouchers to seniors than does the SAHA. In 2000, SAHA vouchers served significantly more minorities proportionately (88 percent) than did OCHA vouchers (56 percent), with only a 1-percent decrease in minority voucher holders for the SAHA and a 1-percent increase for the OCHA from 2000 to 2004. Both LHAs showed a reduction in the average number of people
The variation in reporting rates is one criticism of using the administrative data from this source, because incomplete reporting can introduce bias into analytic results.
in the units subsidized by vouchers during the 4-year period, with the SAHA exhibiting a slightly greater decrease. Interestingly, although voucher holders’ average annual household income increased for both LHAs during the period, the SAHA showed a much greater increase than the OCHA ($900 versus $1,600).15 Given the average income data, it is not surprising that the tenant portion of the rent, on average, increased more for the SAHA ($53) than for the OCHA ($1) over time.
Descriptive Statistics and Analyses The variables used in the analyses and their measurement are presented in exhibit 2. Most of the variables come from the household survey. The exceptions are the neighborhood poverty rate, which was downloaded from the U.S. Census Bureau website; the annual household income and monthly rent, which came from the LHA’s client file database; and the public school quality measure (average API score). Note that the average neighborhood poverty rate for the sample is 14.8 percent, which is relatively low compared with the conventional rate of 40 percent used to identify concentrated poverty. The average API score for the sample is 665, however, much less than the California Department of Education goal of 800 for all schools.
Moves in the HCVP, in nearly every case, are the choices of voucher holders.16 Although this research is primarily interested in the outcomes associated with those moves, it is helpful to briefly consider whether moving choices are associated with any basic voucher-holder characteristics.
This difference may be attributable to the larger population of seniors in the OCHA, because they are less likely to be working and more likely to be on a fixed income.
An involuntary move can occur through eviction or if the housing unit falls below the quality or affordability standards of
Exhibit 3 presents a range of sociodemographic characteristics and one locational item (lives in central city) for movers and nonmovers in the sample. For the most part, movers and nonmovers appear to be very similar sociodemographically and in relation to location within the central city.
They differ on only age and rent. On average, movers tended to be younger and to pay more in monthly rent.
The main analyses focus on a move within the past 3 years. Voucher holders may have moved multiple times during this period, however, or they could have been anticipating a move in the coming year. Because intentions to move and the frequency of moving are relevant to understanding mobility and also because these characteristics are rarely discussed in studies of the HCVP, in exhibit 4, I present this information for the voucher holders in the study. Voucher holders who had not moved and voucher households that had moved three or more times in the past 3 years were less likely to be planning a move in the upcoming year; however, a chi-square analysis found no statistically significant association between frequency of moves and intention to move among these voucher holders.
The next set of analyses explores the factors associated with neighborhood poverty levels, employment status, and school quality, with a particular interest in the effect of moving on these outcomes.
The degree of poverty in a neighborhood is thought to affect individual outcomes in numerous ways. Mobility out of poverty is one approach to addressing negative outcomes, but although the HCVP is designed to enable mobility, it does not require voucher holders to move in general or to move to lower poverty neighborhoods. Thus, it is unclear whether a policy of residential choice can achieve lower neighborhood poverty rates for voucher holders. To investigate this question, a linear regression model was specified with neighborhood poverty level17 as the dependent variable and a set of voucher-holder sociodemographics—rent, central city location, and, the primary variable of interest, whether the voucher holder had moved in the past 3 years—as the independent variables. The results of the analysis are shown in exhibit 5.
The coefficient for “mover” has a negative sign but is not statistically significant. Thus, the analysis indicates that movers, as compared with nonmovers, did not live in neighborhoods with lower poverty levels. Six variables in the analysis are associated with the neighborhood poverty rate, however. As the age of voucher holders increases, neighborhood poverty levels tend to decrease on
average. This result may be because of more knowledge, based on additional years of life experience, or it may be that older people experience a greater degree of landlord acceptance in more affluent neighborhoods. Males tended to locate in neighborhoods with higher poverty rates, which may reflect a tendency to conflate poverty with personal safety and the different perceptions of safety by men and women. Minorities generally lived in neighborhoods with higher poverty rates.
This finding is consistent with the existing literature and may be because of several factors, including discrimination, lack of information, or availability and location of support networks. Voucher households with children also lived in neighborhoods with higher poverty rates. Again, discrimination and the availability and location of support networks might help explain this result. Not surprisingly, as voucher holders’ rents increase, on average, neighborhood poverty levels decrease. We would expect rent to reflect not only housing unit attributes, but also neighborhood characteristics.
Finally, as found in previous analyses and generally accepted in the literature, living in a (lowincome) central city is associated with higher neighborhood poverty rates.