«A Journal of Policy Development and Research HoPe VI Volume 12, Number 1 • 2010 U.S. Department of Housing and Urban Development Office of Policy ...»
3. Neighboring behaviors. An index of six questions related to the degree to which the respondent engages in neighboring activities or behaviors, such as talking to people in the neighborhood, borrowing things, and providing informal childcare, had six answer categories for each of the six neighboring behaviors. The index was a simple average of responses across the six questions and thus could range from 1 to 6. The Cronbach’s Alpha, testing the reliability of the indices, was 0.746 for the intake interview items and 0.738 for the year-3 survey items.
These values exceed the commonly used threshold for the Alpha statistic, suggesting that the items constitute valid indices. The computed change variable ranges from -3.17 to +1.67.
4. Economic security. A question about whether the respondent has enough money to pay for basic needs each month (with three answers: “never,” “sometimes,” and “always”) had answers coded so that higher values mean greater economic security. The computed change variable ranges from -3.0 to +1.5.
5. Employment. A binary variable taking the value of 1 for respondents who were employed and 0 otherwise had a computed change variable that ranges from -1 to +1.
From the census-tract data, the analysis uses the following nine items:
1. Percent of the population that is non-White.
2. Percent of the population that is African American.
3. Percent of the households headed by a woman.
4. Percent of the labor force employed.
5. Median family income.
6. Percent of the population receiving public assistance.
14 HOPE VI Better Neighborhoods, Better Outcomes? Explaining Relocation Outcomes in HOPE VI
7. Percent of the population below the poverty level.
8. Percent of residences that are owner occupied.
9. Median value of owner-occupied housing.
Change variables were computed for each of the nine census measures by subtracting the value for the new neighborhood from the value of the Harbor View neighborhood. Thus, if a resident moved into a neighborhood with more poverty, the change variable would register a positive number.
Findings Neighborhood-Level Outcomes The relocation of families from the Harbor View site took place between April 2003 and August
2004. Most of the families remained in a central city neighborhood: 77 percent stayed in Duluth’s inner city and another 7 percent moved to the inner cities of Minneapolis or St. Paul. Most of the relocated families moved to neighborhoods with significantly lower levels of distress than their original public housing neighborhood (exhibit 3). Unemployment in the new neighborhoods was around 8 percent compared with 12 percent for the Harbor View neighborhood. Poverty rates in the new neighborhoods were roughly one-half that in the original neighborhood, median incomes were almost twice as high, homeownership rates were significantly greater, and the percentage of the population on public assistance was less than one-half (9.6 percent instead of 28 percent).
These findings are similar to those reported in other studies of HOPE VI: families typically remain in the central city, and relocation from HOPE VI sites seems invariably to result in moves to better neighborhoods, as measured by census-tract indicators. The reasons for such consistent outcomes are not a mystery. Most HOPE VI sites are located in what had been the most disadvantaged neighborhoods of their respective cities. The public housing projects subject to the redevelopment were concentrations of poverty in and of themselves, and typically the immediately surrounding communities have similar socioeconomic profiles. Thus, almost by definition, moving out of such neighborhoods means moving to neighborhoods with fewer indicators of economic distress.
One in five families (21 percent) had moved more than once by the time of the survey, in year 3 of the study. Multiple-movers live in neighborhoods that are statistically similar to single-movers’ neighborhoods, with one exception: multiple-movers’ neighborhoods have a significantly lower median income ($27,140 compared with $31,745). Although the two types of neighborhoods have slight differences in poverty, homeownership, and percentage of residents on public assistance (all of which indicate that multiple-movers are in neighborhoods with slightly higher levels of distress), these differences do not reach statistical significance.
Individual-Level Outcomes Exhibit 4 describes the changes that residents reported before and after relocation. The first row of figures indicates that 35 percent reported less neighborhood satisfaction, 40 percent reported the same degree of satisfaction, and 25 percent reported more satisfaction. The difference between the two time points was not statistically significant (either as a difference in mean response or by Wilcoxon Signed Rank test).
On the other hand, statistically significant numbers of respondents reported fewer neighboring behaviors after moving (57 percent engaged in fewer behaviors, 37 percent in more, and 6 percent in the same). This outcome may be a result of the families’ having only recently moved into their new neighborhoods, although other research has indicated that length of time in the new neighborhood was not related to the frequency of neighboring behaviors among relocated people (Goetz, 2003). Residents also reported significantly less economic security after the move, indicating that they more frequently lack enough money to buy basics or more frequently make use of local food banks. The data also show a higher rate of families with health problems after relocation.
Either these health problems are unrelated to environmental conditions (and are thus coincidental to relocation) or the relocation process or the new neighborhood environment is producing negative health outcomes. On the positive side, respondents felt significantly safer in their new neighborhoods: 44 percent felt safer, 22 percent felt less safe, and 34 percent were unchanged.
No data suggested a significant difference in employment rates. On the whole, these outcomes are largely negative. Only in their sense of safety did Harbor View families report an improvement after moving. The other five measures showed no change or showed negative outcomes.
Taken together, the findings in exhibits 3 and 4 mirror the outcomes seen in many studies of families involuntarily displaced by HOPE VI. Families in the Duluth HOPE VI project seem to have moved to better neighborhoods by the objective indicators available from the census (exhibit 3),
16 HOPE VI Better Neighborhoods, Better Outcomes? Explaining Relocation Outcomes in HOPE VI yet they reported little to no improvement on a range of subjective individual-level measures (exhibit 4).
The lack of benefits for the sample as a whole, however, may mask patterns of benefits to certain subpopulations. Some relocated people do report benefits, although the number doing so varies from measure to measure. If the same respondents are reporting benefits across different measures, it might be possible to identify subpopulations for which HOPE VI relocation works well. Bivariate correlations among the outcome measures indicate the degree to which respondents who report change (one way or the other) on one item are more likely to report similar change on other items.
A look at the correlation matrix for change in individual outcomes indicates that little overlap exists between these outcomes (exhibit 5). Change in economic security is positively correlated with change in employment but is statistically unrelated to all other changes measured. An increased sense of safety is correlated with a higher level of neighborhood satisfaction but is unrelated to changes in neighboring behaviors and employment. Changes in neighboring behaviors are not correlated with any other individual-level variables examined.
These patterns suggest that the individual changes reported by residents displaced from Harbor View are not cumulative. Those who report a positive change in one area, in general, are not more likely to report positive changes in other areas. Thus, it is not the case that some residents report uniformly rosier outcomes, while others consistently report worse outcomes. These findings suggest that models that explain one set of outcomes may not explain others.
Linking Better Neighborhoods and Better Individual Outcomes To examine more closely the link between neighborhood outcomes and individual outcomes, the analysis tested the hypothesis that the degree of neighborhood change is related to the degree of individual-level change. Bivariate correlations were calculated for each of the six individual outcome variables and each of nine measures of neighborhood change described earlier. Of the 54 bivariate correlations produced, only 4 achieved statistical significance (data not shown), and all 4 were related to changes in the racial characteristics of the neighborhood (both an increase in non-White population and an increase in African-American population were correlated with decreases in economic security and employment). At the bivariate level, it seems, changes in the objective conditions of the neighborhoods were largely unrelated to the changes that people relocated by HOPE VI reported in their own lives.
It is possible, however, that when multiple dimensions of neighborhood change are considered, better outcomes might occur. Thus, an index of neighborhood change was created, using changes in poverty, racial composition, and housing market value. Displaced people were then divided into two groups, with those who experienced the greatest change on all three dimensions put into one group and everyone else put in the other group. Respondents reporting a reduction of more than 20 percentage points in the non-White population of their neighborhood (41 percent of the sample) and a reduction of more than 30 percentage points in poverty (40 percent of the sample) and an increase of more than $10,000 in median housing value (30 percent of the sample) were classified as having had significant change in neighborhood. This categorization classified 21 respondents (19 percent of the sample) as having experienced the greatest neighborhood change on all three dimensions. These 21 people reported individual outcomes that were not statistically different than the rest of the sample for all five outcome measures examined (data not shown). Thus, even a combination of different types of neighborhood change is unrelated to individual outcomes.
If neighborhood change is not related to individual outcomes, what is? The literature suggests a range of individual-level attributes may influence the relocation experience. Senior citizens may be more adversely affected by being forced to move away from their long-time community, and residents for whom English is not a first language may also experience more difficulties in a relocation process (Kleit and Manzo, 2006). Other characteristics, such as household size, gender, marital status, the presence of small children in the family, education level, and, of course, race, may have important effects on how HOPE VI families fare during relocation.
Attachment to the original neighborhood (and thus a person’s willingness to move) may color a person’s perceptions of the new neighborhood. Respondents who felt a close attachment to the old neighborhood may resent being forced to move. These respondents may report worse outcomes than those for whom HOPE VI provided the opportunity to leave a neighborhood they wanted to escape.
The following multivariate analysis tests each of these propositions. Regression models were run for each of the six individual change variables. Equation 1, which is estimated using a linear ordinary least squares (OLS) model, was rerun for three additional dependent variables: changes in sense of safety, neighboring behaviors, and economic security.
where Y equals the respondent’s change in neighborhood satisfaction, NBHDCH is a vector of neighborhood change measures, IND is a vector of individual attributes, and ATTACH is the respondent’s lack of desire to have moved from Harbor View.
Equation 2 is estimated as a binary logistic model. The independent individual-level variables are described in exhibit 6.
where EMPLOY3 equals the respondent’s employment status at the time of the year-3 survey (a binary variable taking the value of 1 if the respondent is employed and 0 otherwise), and EMPLOY1 is the employment status at the time of the intake interview (coded in the same manner as the previous variable).
18 HOPE VI Better Neighborhoods, Better Outcomes? Explaining Relocation Outcomes in HOPE VI
The mix of individual-level variables for any given dependent variable was determined so as to maximize the explanatory power of the equation (that is, to produce the highest adjusted r-squared). In some cases, the analysis uses alternative measures of the same concept. For example, with education, the data were collected in ordinal categories. Two alternative dummy variables were created: one variable differentiated those with at least a high school diploma from those without and a second differentiated those with any education beyond high school from those without.