«AmericAn neighborhoods: inclusion And exclusion Volume 16, Number 3 • 2014 U.S. Department of Housing and Urban Development | Office of Policy ...»
Housing quality measured by self-report for any of four maintenance deficiencies: (1) lack of heat for 6 hours or more during f past winter, (2) lack of hot water for 3 hours or more in past year, (3) the number of cockroaches seen in the home on a typical day (none is considered no maintenance defiency here), and (4) seeing any signs of mice or rats in the building in the past 90 days.
Study Site The affordable housing units in our study site were targeted to low-income households, as defined by HUD, with earnings at the time of initial qualification for housing ranging between 69 and 80 percent of HUD Income Limits, depending on household size and the unit type for which they qualified.12 Of the participating households, 21 applied with a Section 8 voucher, which enabled them to meet eligibility guidelines with a lower household income than would otherwise be required.
The study site includes two midrise buildings containing affordable rental units on either side of a single block that terminates in a large public waterfront park. Each building is next to a luxury condominium tower that sits between the affordable-housing building and the park. Both affordablehousing buildings have elevators, and no stairs are required to enter the buildings or reach any of the units. Each building has a single point of entry that opens into a small lobby area where mailboxes for all residents are located. The street frontage of the buildings is a substantial portion of the block face, with the entrance doors midblock on either side of the street. People come and go at most times of the day and evening, and residents regularly pass one another at the entrance or immediately in front of the buildings.
Qualifying incomes ranged from a minimum of $37,370 for a household of one (studio unit) to a maximum of $61,450 for a family of four (two-bedroom unit).
The two buildings with affordable housing units are typical in design and include no features that would specifically encourage neighborly interaction—that is, no community room or outdoor space shared by residents and no seating in the lobby areas; however, several respondents mentioned during the course of the interviews a common laundry room in the basement of each building as a place where they frequently see neighbors. The complex allows pets, which is not typical for newly constructed affordable housing in New York City. Of study participants, 21 percent reported having a dog, which increases foot traffic to and from the building and also provides an opportunity for residents to see one another on the street when they walk their dogs. Dog owners reported walking their pet an average of 2.5 times per day.
All study participants moved to the study site approximately 3 1/2 years before being interviewed, although some had moved to another unit in the complex (N = 14), including 6 who moved between the two affordable-housing buildings and 8 who moved within the same building. When they applied for housing, 42 percent of respondents lived in the community district13 where the study site is located. For these households, the average length of residence in the community was 14.7 years at the time of the interview compared with 3.8 years for those households that moved from another neighborhood in New York City.
Residents reported improvements in both neighborhood safety and housing quality relative to where they lived when they applied for housing. The vast majority of all respondents—91 percent—rated the streets at night in the study site neighborhood as either “very safe” or “safe.” At baseline, 76 percent of respondents reported their neighborhood as “very safe” or “safe” at night. At the time of the followup interview, 75 percent of residents reported no maintenance deficiencies in their affordable housing unit (no instances of heating breakdown, loss of hot water, signs of rodents in the building, or cockroaches in the home). At baseline, only 27 percent reported no maintenance deficiencies in their home.
Social Network Measures and Analytic Strategy We captured data on three types of networks: the overall network of the respondent (“ego”) regardless of geographic proximity, relationships with individuals who lived within the same neighborhood (as defined by the respondent), and ties to neighbors within the same building. Exhibit 2 shows the overall structure and flow of the interview modules.
Six name generators enabled respondents to nominate a maximum of 18 individuals. Up to 3 names were captured14 for each of the following: (1) people with whom the respondent discussed an important personal matter in the last 6 months, (2) people the respondent asked for small favors in the last 2 months, and (3) people the respondent asked for advice or information in the last 12 months. Individuals named in these three generators are considered the respondent’s core network A community district is an administrative boundary used by the City of New York to allocate municipal resources and define local political representation. These boundaries roughly correspond to Public Use Microdata Areas; New York City contains 59 community districts. Affordable housing that is allocated through a lottery process, such as the units studied in this article, include a 50 percent set-aside for qualified applicants who live in the community district where the study site is located.
For the first three generators (the core network), respondents were able to name as many individuals as they chose, but only the first three were captured for each generator. For the final three generators (those limited to people on the same block), the respondent was specifically asked to name up to three individuals for each generator.
and were not limited to a specific geography (that is, they could name anyone regardless of where s/he lived). Because we were particularly interested in neighbor ties, we also asked the respondent to nominate up to 3 individuals who live on the same block for each of the following: (4) neighbors with whom the respondent interacted most frequently, (5) any other neighbors not already named from whom the respondent sought advice or information, and (6) any other neighbors not already named to whom the respondent provided advice or information. Respondents were able to nominate the same individual more than once; however, the final network comprised only unique individuals (“alters,” or “ties”) named in one or more of the six generators.
Basic information was collected for each unique individual who was named, including the tie’s relationship to the respondent, whether the tie was the same race or ethnicity as the respondent, the gender of the tie, and whether the tie was foreign born. We also asked whether the tie had one or more children younger than 18 years old, whether the respondent thought the tie was generally “better off, worse off, or about the same” as the respondent, and the frequency of interaction between the respondent and the tie. The question about interaction included visiting face to face, talking on the phone, e-mailing, and texting. Frequency was measured using a six-item categorical variable coded to estimate the total number of interactions per year, with “every day” coded as
365.25 interactions (to account for leap years), “a few times a week” coded as 156, “once a week” coded as 52, “once a month” coded as 12, “a few times a year” coded as 5, and “less than once a year” coded as 1 interaction.
For each unique individual, we asked geographic proximity (for example, in the same household, neighborhood, or building). Any tie who lived with the respondent was treated as part of the overall social network but was excluded from calculations of building and neighborhood networks.15 More detailed information was collected for each individual who lived in the same neighborhood as the respondent (local tie). The density of the local network was derived from information gathered on which of the local ties interacted regularly with other individuals in the respondent’s network;
all answers were treated as symmetrical and assumed to be undirected—that is, if the respondent indicated that one tie interacted regularly with another person, the data were coded so that the other person also interacted regularly with the tie. We define density as the proportion of ties who interact regularly with one another, ranging from 0 (none of the ties interact) to 1 (all the ties interact regularly).
Content and activation were measured using 18 true-or-false statements such as “I have loaned money to ______” and “_______ has loaned money to me.” Each interaction was coded as falling into one of six categories: expressive, instrumental, or informational support and the direction of the interaction—provided or received by the respondent. Two additional measures were coded based on whether the respondent had named the local tie in one of the core generators that corresponded to the true-or-false statement for the provision of that type of support. If the local tie was named for that generator, it was coded the same way as if the respondent had indicated “true.” Exhibit 3 lists each of these items and their corresponding category.
The interview data were used to generate two complementary datasets: (1) a respondent-level dataset of 120 individuals and their overall network characteristics (for example, composition, homophily, Coresident family represents a minimal proportion of all nominated individuals. See exhibit 3.
range, geographic proximity ties, and activation for specific types of support) and (2) a dataset comprising the 282 building ties named by the 120 respondents.16 For this second dataset, we employed logistic regression models to estimate the odds ratios of the respondent receiving or providing each of the three main types of support: expressive, instrumental, and informational. These models used robust standard errors to account for multiple ties within a given respondent’s network. For these analyses, frequency of interaction with the individual was group centered based on the mean frequency of interaction within the given network. Outcomes were coded based on the items listed in exhibit 3, with binary values indicating whether one or more of the items were coded as “true” for each type and direction. All data are unweighted.
The social network data analyzed here are cross-sectional. As such, the present study does not attempt to draw any conclusions about changes in social networks or the effect of moving to affordable Of all respondents, 15 did not nominate any within-building ties, including 8 who did not name any local ties and 4 who did not nominate any ties at all. These cases are therefore included in all descriptive analyses but excluded from the tie-level dataset and corresponding statistical models.
housing on social connectedness. The nature of the affordable-housing selection process makes it very unlikely that any two residents knew each other before moving to the study site; however, we cannot assess net changes in social networks or whether these relationships may have formed even in the absence of moving to this particular housing complex. We focus on describing the social context of this low-income population, including the characteristics of individual networks and the use value of interactions for particular ends.
Findings Social Networks of Low-Income Working Households Exhibit 4 shows descriptive statistics for the average network composition, including all ties regardless of geographic proximity and separately for all local ties. We also parse local ties into those who live in the same building as the respondent (“same building”) or in the neighborhood but not in the same building as the respondent (“elsewhere in neighborhood”).
Coresident family members are excluded from all local ties, same building, and elsewhere in neighborhood calculations.
b May not add to 100 percent because of item nonresponse.
Overall, respondents living in affordable housing reported an average network size of 5.9 unique people, including 3.8 people in the core network. The average annual frequency of contact with all ties was 149 interactions, equivalent to between two and three times per week. On average, onethird of unique ties are kin, comprising mainly family members from outside the respondent’s household. The demographics of nominated individuals show that residents of affordable housing interact with similar individuals. On average, 64 percent of ties are identified as being the same race or ethnicity as the respondent and 66 percent are the same gender—most of whom are female.
The average network comprises mostly people whom the respondent indicated as being generally “about the same”; 30 percent of the ties in the average network are “better off” and 13 percent are “worse off.” Neighbor Networks Local ties—unique individuals who live in the same neighborhood as the respondent—represent 32 percent of the average core network and 54 percent of the average overall network. Respondents interact less frequently with local ties than with the overall network, with an average frequency of 97 interactions per year, or between one and two times per week. A lesser, but still substantial, portion of local ties are kin. The average local network shows a lesser proportion that is the same race or ethnicity and a greater proportion that is the same gender than the overall network but remains consistent with the pattern that people interact primarily with similar individuals. Most local ties are doing “about the same” as the respondent. On average, local networks show a relatively low level of density;17 31 percent of local ties interact regularly with one another.