«Mixed Messages on Mixed incoMes Volume 15, Number 2 • 2013 U.S. Department of Housing and Urban Development | Office of Policy Development and ...»
Within only a few months, residents received notification of whether they had been approved for a voucher. To actually get the voucher, residents were then required to participate in the Good Neighbor Program. AHA (2008: 7) described this program as “a training series that prepares AHAassisted families to transition successfully from environments of concentrated poverty into healthy mixed-income communities.” The program provided information to voucher-qualified residents about compliance with private rental landlord and neighborhood expectations, including (1) caring for a unit and premises, (2) respecting the rights of neighbors, and (3) compliance with other essential conditions of tenancy. In addition, residents were assigned relocation counselors to assist them in finding a new place (AHA, 2008).
On completion of the Good Neighbor Program, residents begin the waiting-to-move process. Residents can look at single-family rental houses and apartments and express interest with the respective landlords and property managers. Landlords and property managers, however, typically do not allow a lease signing until the residents have evidence that their voucher has been issued. HUD distributes voucher subsidies to state PHAs first, and then allocates them to local PHAs in the state (HUD, 2001).
Thus, at least in the case of our study, residents approved for a voucher received the official paperwork at different times, meaning that some residents were able to move before other residents. On receipt of the official voucher paperwork, the residents had 90 days to find a new place to live.2 The AHA did allow extensions of this deadline.
Data and Methods About 6 months after the AHA’s 2007 announcement, members of the jurisdictionwide Public Housing Resident Advisory Board met with Georgia State University (GSU) sociology department faculty to discuss conducting a survey of residents’ views about relocation and how relocation ultimately affected their lives and overall well-being. GSU formed the Urban Health Initiative to conduct this study.
Of the public housing communities slated for demolition, five were nearly vacant and one was inaccessible because the resident board president had already been relocated when we began developing the survey in early 2008. Thus, we targeted communities that would not begin relocation until September 2008 (which included four family developments and two highrises for seniors and people with disabilities).
We conducted a baseline (prerelocation) survey during the summer of 2008. We intended to compile a disproportionate random sample of 426 participants with equal numbers from each housing community (N = 71). We initially achieved only 49 percent of our goal (N = 208) because of constraints beyond our control, primarily regular interference from the AHA, but not related to characteristics of the public housing residents; thus, no systematic bias was introduced. We then opened the study up to volunteers to increase the sample size. An additional 103 residents volunteered. Our final sample size is 311, or 73 percent of our desired sample size, a major limitation of our study. We tested the random and nonrandom portions of the sample on all variables included in the study and found no significant differences on any variables, however.
All respondents were age 18 or older, more than 90 percent were the leaseholder, and only one member per household participated. Given that we knew which units were occupied in each housing community before sampling, we created postsurvey sampling weights to make the sample representative of the six public housing communities. Nonetheless, apply caution when making generalized inferences from this sample.
We reinterviewed our respondents 6 to 8 months after relocation, from November 2009 to September 2010. The survey was essentially the same as the baseline survey to assess prerelocation-topostrelocation change. For the 6-month followup, we obtained a retention rate of 87 percent across the six relocating sites. At the writing of this article, we had recently completed data collection for the 24-month followup, data that are currently being cleaned.3 For this article, we limited the analyses to relocated respondents who completed both the prerelocation and 6-month postrelocation surveys (N = 248), dropping 13 cases with missing values on the relocation process variable. Two participants who did not participate at 6 months after relocation but did participate in the 24-month postrelocation interview were included in the analyses. Both participants experienced a hard relocation process. We then geocoded the addresses of the original public housing sample and the addresses residents lived in 6 months after relocation using 2010 census boundaries. Using the geocodes, we attached census tract identifiers for each Our retention rate between waves two and three is 91 percent.
participant and merged the survey data with 2005–2009 American Community Survey (ACS) data.
In this process, we limited the sample to those who moved within the Atlanta metropolitan area (dropping five out-of-state cases), giving us a final full sample size of 232.
Before presenting our multivariate analysis, we provide a thematic map of where residents moved, average census tract characteristics from the 2005–2009 ACS, and crime statistics from the Atlanta Police Department 2009 Crime Incident Reports.
Perceived Relocation Satisfaction Constructs Our dependent variable comes from an open-ended question asked at the 6-month postrelocation interview and at 24 months for those few respondents lost to followup at 6 months. Specifically, we asked, “Looking back, how would you describe the relocation process?” Answers ranged from statements like, “It wasn’t a problem because I found a place and everything just went smoothly”;
to, “Kind of stressful” or, “It was a fair process, it was ok”; and then finally to statements like, “The process was stressful because you did not know where you were going and you were just looking around.... It was very stressful” and, “It was horrible. I had to move from a place where all my friends were like family and now I am in the middle of nowhere.” Two authors independently coded responses to this question, met, and came to consensus on the categories to use to describe the process. A consistent theme became apparent. They then recoded the question using the agreed-on categories. After we achieved more than 65 percent agreement on the codes, the two authors met and coded the remaining responses together and refined the category definitions. We created six codes: process was (1) hard, terrible, or traumatic with stress;
(2) stressful; (3) somewhat stressful; (4) OK/alright/fair; (5) good/fine; (6) easy/smooth. Given the difficulty in distinguishing “good/fine” from “easy/smooth” and distinguishing “hard, terrible, or traumatic with stress” from “stressful”, we then collapsed the six categories into three: (1) hard, (2) neutral, and (3) easy relocation experience. For this analysis, we are interested in the characteristics of public housing residents who thought the relocations were easy, so we dichotomized the variable with easy relocation experience coded as 1 and neutral and hard coded as 0.
We controlled for variables that might predict an easy relocation experience. First is a dummy variable for moving from family housing versus from a highrise for seniors. We include this variable for three reasons. First, family housing and senior housing were very different. The family housing was barrack-style, in worse condition, and farther from the city center. Second, the neighborhoods differed: the family housing was located in poorer, more racially segregated, and higher crime areas.
Third, a major finding from our premove baseline survey was that, although the majority of the family housing residents wanted to move (73 percent), the majority in senior housing did not want to move (61 percent). Because there were seniors in family housing, we also included age as a control.
Disability status was coded 1 for individuals who said they did not work because of a disability or who said carrying groceries, walking up a flight of stairs, or walking around the neighborhood without assistance was a significant problem, and coded 0 for individuals who did not indicate any of these problems. A dummy variable for having a chronic condition was coded 1 for respondents who had been diagnosed with at least one of the following conditions: high blood pressure, heart disease, asthma, arthritis, stroke, or cancer. We measured tenure in public housing in years.
Cityscape 179Oakley, Ruel, and Reid
Experiencing financial strain is a dummy indicator, with 1 coded as not having enough money to make ends meet most months in the past year and 0 coded as having more than enough money, some money left over, or barely enough money to make ends meet at the end of most months.
Having no friends in public housing is a dummy variable, with 1 coded as having no friends living in public housing and 0 coded as having at least a few friends living in same public housing community. We asked respondents if they thought the condition of their postrelocation home was excellent, good, fair, or poor. We also asked respondents if they were (1) very satisfied, (2) somewhat satisfied, (3) in the middle, (4) somewhat dissatisfied, or (5) very dissatisfied with their postrelocation neighborhood. We did not include race, gender, marital status, education, or income as controls because of lack of variation.
For the second set of analyses, we used several measures from the ACS 2005–2009 census tractlevel data. We used percentage of residents living in poverty, percentage of households that are female headed, percentage of household heads who are unemployed, percentage of homes that are vacant, percentage of homes that are occupied by renters, percentage of homes that are more than 30 years old, and percentage of residents who are non-Hispanic African American. Turnover in the neighborhood is measured by the percentage of households living in the same place for less than 10 years. Finally, we include a measure we call high former public housing receiving neighborhood. Our receivership categorization is similar to that of Popkin et al. (2012). High-receiving neighborhoods had more than 12 former public housing households move in, medium-receiving neighborhoods had 5 to 12 former public housing households move in; low-receiving neighborhoods had fewer than 5 former public housing households move in. In some analyses we present, we dichotomize our receivership variable so that high receiving is coded 1 and all other is coded 0.
In preliminary analyses, we also included educational attainment and household size as predictors of an easy relocation process, but they were not at all significant and, because our sample size is very small, we decided to drop them.
Multivariate Analysis We ran generalized estimating equations, or GEE, for these analyses using the GENMOD (Generalized Linear Model) procedure in SAS® version 9.2 to deal with the autocorrelation inherent in clustered data (six communities) (Liang and Zeger, 1986). In this case, we use the public housing community as our cluster, because 6 months is not a sufficient time to diminish the autocorrelation of living in a specific public housing community. In addition, we run logistic regression models in which the outcome is the probability of experiencing an easy relocation process versus not experiencing an easy relocation process. We present raw logit estimates and standard errors. Finally, to examine whether destination home and neighborhood conditions differ significantly between those who experienced an easy relocation and those who did not, we use ANOVA procedures to test mean differences in reported home and neighborhood conditions.
Results Exhibit 1 shows the demographic information for our initial sample and the sample in our present analysis. Although we were unsuccessful in locating about 13 percent of those who participated
Of the 232 original respondents, 24 died before the 6-month interview, we were unable to locate 31 for the 6-month interview, b 5 moved out of state before the 6-month interview, and 11 had missing information on one or more of the outcome variables.
in the baseline survey, and another 6 percent died, the population characteristics among the three survey periods are very similar. Comparing those who participated in the baseline survey with only those who participated in the 6-month followup, however, age is different, meaning younger people were less likely to participate in the 6-month followup.
Most of our sample, by far, were African American (96 percent) and female (85 percent); 46 percent were between the ages of 18 and 44 years, another 39 percent were between the ages of 45 and 64 years, and 15 percent were 65 years or older. Nearly three-fourths reported living in public housing for between 2 and 8 years. Only 5 percent reported being married, and the average number of children younger than age 18 in the household was two. Only 55 percent reported having a high school degree or general equivalency diploma (GED), and the average monthly income was $832.41, putting these households, regardless of size, well below the federally established poverty line.
Exhibit 2 presents basic descriptive statistics of the variables used in our analyses. The topmost rows of exhibit 2 provide the distribution for the original coded variable and the final dichotomous variable. Findings indicate that 31 percent considered the relocation process either hard, terrible, or traumatic with stress (15 percent) or stressful (16 percent). Another 15 percent considered it somewhat stressful, 18 percent reported it being ok/alright/fair, and 37 percent considered it either good/fine (18 percent) or easy/smooth (19 percent). On a scale of 1 (excellent) to 4 (poor), the average level of satisfaction with the new home is 1.79. On a scale of 1 (very satisfied) to 5 (very dissatisfied), the average level of satisfaction with the new neighborhood is 2.04. The average tenure in public housing is 6.2 years, and the average age is nearly 46. The means for the computed dichotomous variables (living in a family project, being disabled, having a chronic health condition,