«DiscoVeriNg HomelessNess Volume 13, Number 1 • 2011 U.S. Department of Housing and Urban Development | Office of Policy Development and Research ...»
the effects of segregation on housing outcomes for Blacks through the following four hypotheses:
Hypothesis 1: As segregation increases, Blacks will be more likely than Whites to live in housing of inadequate quality.
Hypothesis 2: As segregation increases, Blacks will be more likely than Whites to live in housing that is crowded.
Hypothesis 3: As the affordable housing supply increases at the city level, housing inadequacy and crowding will decrease.
Hypothesis 4: As homeownership increases at the city level, Blacks living in more highly segregated areas will live in more inadequate and crowded housing than will Whites.
Controlling for affordable housing supply in testing hypothesis 3 and controlling for homeownership rates in testing hypothesis 4 provide an opportunity to evaluate how policies targeted at increasing affordable housing supply and homeownership may influence the relationship between segregation and housing quality for Blacks.
Data for Part I come from three sources: the 1997 AHS National Public Use File, the 1990 Decennial Census, and the 2000 Decennial Census. The AHS (formerly the Annual Housing Survey)
began collecting data on the nation’s housing in 1973. Since 1981, it has collected national data every odd-numbered year. The U.S. Census Bureau conducts the survey for HUD. It returns to the same housing units every other year until a new sample is selected.
Most of the data for Part I of the study come from the AHS, including information on the adequacy and crowding of housing units and information on the household and household head, who is referred to as the householder in the AHS. Data from the AHS were merged by standard metropolitan statistical area (SMSA), with a common segregation index—the index of dissimilarity, affordable housing measures, and measures of homeownership calculated from the 1990 and 2000 Decennial Censuses and linearly interpolated to 1997 values.
A series of nested logistic regression models were run to test the four hypotheses predicting the log odds that a householder is living in an inadequate housing unit or an overcrowded housing unit.
The dependent variables in the analyses are measures of housing inadequacy and overcrowding.
Housing Inadequacy. The housing inadequacy measure is constructed from the HUD housing inadequacy recode provided in the AHS Public Use File. A “1” on the housing inadequacy measure indicates that the housing unit is declared either severely inadequate or moderately inadequate by HUD standards and a “0” indicates that the housing unit is adequate. HUD defines a housing unit
as severely inadequate if any of the following conditions exist:
1. The unit lacks complete plumbing facilities.
2. Three or more heating equipment breakdowns occurred lasting 6 hours or more in the last 90 days.
3. The unit has no electricity.
4. The electrical wiring is not concealed, working wall outlets are not present in every room, and fuses and breakers blew three or more times in the last 90 days.
5. Five or more of the following exist: outside water leaks, inside water leaks, holes in the floor, cracks wider than a dime in the walls, areas of peeling paint or plaster larger than 8 ½ x 11 inches, rodents seen recently in the unit.
6. All of the following exist: no working light fixtures or no light fixtures at all in public hallways;
loose, broken or missing steps in common stairways; stair railing not firmly attached or no stair railings on stairs at all; three or more floors exist between the unit and the main entrance to the building and the building has no elevator.
A unit is moderately inadequate if it is not severely inadequate and any of the following conditions
1. The unit lacks kitchen facilities.
2. Three or more toilet breakdowns occurred, lasting 6 hours or more in the last 90 days.
3. An unvented room heater is the main heating equipment.
40 Discovering Homelessness From Exclusion to Destitution: Race, Affordable Housing, and Homelessness
4. Three or four of the following exist: outside water leaks, inside water leaks, holes in the floor, cracks wider than a dime in the walls, areas of peeling paint or plaster larger than 8 ½ x 11 inches, rodents seen recently in unit.
5. Three of the following exist: no working fixtures or no light fixtures at all in public hallways;
loose, broken, or missing steps in common stairways; stair railings not firmly attached; no stair railing on stairs at all.
6. Three or more floors exist between the unit and the main entrance to the building and the building has no elevator.
The unit is deemed adequate if it is neither severely nor moderately inadequate (ICF International, 1997).
Overcrowding. The overcrowding measure is a standard measure of housing density: the number of people per room.6 A unit is overcrowded if there is more than one person per room in the housing unit (Ringheim, 1990). In the analyses, a housing unit is classified as “1” if it is overcrowded and as “0” if it is not.
The following independent variables from the AHS were used in the analyses: central-city location;
rental status of the unit; public housing status; the race, age, and sex of the householder; the householder’s highest level of education; the household income; whether the household receives welfare income; and three region dummy variables (Northeast, Midwest, and South, with West serving as the reference group).
Residential segregation is measured by the index of dissimilarity at the metropolitan area level. In calculating the index of the dissimilarity, census tracts were used as proxies for neighborhoods.
Massey and Denton (1993) identify the index as the standard measure of segregation. The index of dissimilarity “captures the degree to which blacks and whites are evenly spread among neighborhoods in a city…[and]… gives the percentage of blacks who would have to move to achieve an ‘even’ residential pattern—one where every neighborhood replicates the racial composition of the city” (Massey and Denton, 1993: 20). Indices of dissimilarity were obtained from the 1990 and 2000 Decennial Censuses at www.census.gov. Dissimilarity values for 1997 were estimated by linear interpolation, using the 1990 and 2000 Decennial Census data. The index of dissimilarity for the 132 SMSAs in this study range from a low of.23 to a high of.86 with a mean value of.64. The index was split into quartiles (Dissimilarity1, Dissimilarity2, Dissimilarity3, and Dissimilarity4) with the first dissimilarity quartile (Dissimilarity1) serving as the reference group in the analyses. The dissimilarity quartiles were interacted with the Black householder dummy variable to create the main variables of interest in the analyses (Black*Dissimilarity2, Black*Dissimilarity3, and Black*Dissimilarity4, with the interaction of Black with the first dissimilarity quartile serving as the reference group).
The interaction terms represent the independent effect of Black headship compared with White headship within metropolitan areas with different levels of Black and White segregation.
Rooms include all finished rooms in the housing unit, including bedrooms, living rooms, dining rooms, kitchens, recreation rooms, permanently enclosed porches, lodgers’ rooms, and offices. Dining rooms must be separate to be counted.
Bathrooms, laundry rooms, utility rooms, pantries, and other unfinished rooms are not counted.
Two housing costs measures are included in the analyses. The first measure is an indicator of rent burdens: the proportion of renters in the metropolitan area making under $10,000 who pay more than 35 percent of their income on rent. Housing is considered affordable when no more than 30 percent of income is spent on housing costs. This measure estimates the extent to which the lowest income renters have high housing burdens in a given metropolitan area. The measure was interpolated for 1997 using data from the 1990 and 2000 Decennial Censuses. Although $10,000 was worth more in 1989 than it was in 1999, the $10,000 cutoff was used in both the 1990 and 2000 Decennial Censuses as the lowest income category for which rent-to-income ratios were calculated.
The second housing cost measure used in the analyses is the ratio of lowest rent units to lowest income households at the metropolitan area level. Such measures have been used in other research to indicate the extent of the affordable housing crunch (Jencks, 1994; Wright, 1989). With this measure, this study estimates the low-income-housing ratio for those most at risk of becoming homeless—those living below 50 percent of the poverty threshold. In 1990, affordable rents for a family of three living below 50 percent of the poverty threshold were approximately $150 a month or less. In 2000, affordable rents for a family of three living below 50 percent of the poverty threshold were approximately $200 a month or less. To estimate the number of households living below 50 percent of the poverty threshold, the number of individuals living below 50 percent of the poverty threshold was divided by 3. This approach is similar to that used by Wright (1989) to construct affordable housing ratios for households at the poverty line. To calculate the affordability measure, the number of lowest rent units was divided by the number of households living below 50 percent of the poverty threshold. Higher values on the measure indicate larger numbers of affordable units in relation to households below 50 percent of the poverty threshold, and lower values on the measure indicate fewer numbers of units in relation to households below 50 percent of the poverty threshold. The measure was calculated for both 1990 and 2000 and interpolated to estimate a value for 1997.
Homeownership was measured using the proportion of homeowners in each metropolitan area in
1997. The proportion was interpolated from proportions reported in the 1990 and 2000 Decennial Censuses.
Most of the independent variables in the analysis are dummy variables. The central city variable is coded “1” for households in the central city and “0” for those in suburbs or rural areas. The rental status variable is coded “1” for households who rent their units and “0” for households who own their units. The public housing variable is coded “1” if the housing unit is public housing and “0” if it is privately owned or rented. The Black headship variable is coded “1” if the householder (otherwise known as the household head) is Black and “0” if the householder is White. The female headship variable is coded “1” if the householder is female and “0” if the householder is male.
The welfare recipiency variable is coded “1” if the householder receives welfare and “0” if the householder does not receive welfare. The highest level of education attained by the householder is split into five dummy variables: 8th Grade or Less, 9th to 12th Grade, High School, Some College, College, and More than College (with More than College serving as the reference group). Region is split into four dummy variables: Northeast, Midwest, South, and West, with West serving as the reference group. Age is a continuous variable measured in years, and household income is a categorical variable with $125,000 or more serving as the reference category.
42 Discovering Homelessness From Exclusion to Destitution: Race, Affordable Housing, and Homelessness Part II: Analysis of Migration for Homeless Services Part II of the research tests the following hypothesis.
Hypothesis 5: Black homeless clients are less likely than White homeless clients to migrate for homeless services.
Data from the NSHAPC were used to test this hypothesis. The NSHAPC, conducted in 1996, was designed to be a nationally representative sample of both homeless programs and the clients who use them. Included in the NSHAPC were 76 primary sampling areas, including “the 28 largest metropolitan statistical areas in the United States; 24 small and medium-sized metropolitan statistical areas, selected at random to be representative of geographical regions (Northeast, South, Midwest, West) and size; and 24 rural areas (groups of counties)” (Burt et al., 1999: 3). The study collected information on programs within these sampling areas and sampled homeless clients within these programs. A homeless program had to have a focus on serving homeless people (although, not necessarily only homeless people), have direct service, and be within the geographical boundaries of the sampling area (Burt et al., 1999).
Homeless clients were sampled from within a sample of the homeless programs, taking into account program type and size (Burt et al., 1999). A client is defined as someone who uses a program and thus includes both homeless and nonhomeless clients. Between 6 and 8 clients were selected randomly at around 700 site visits, resulting in a total of 4,207 client interviews. Interviews were conducted by trained Census interviewers and, in most cases, the interview was held at the program location. Clients received $10 for participating in the study (Burt et al., 1999).
To assess the effect of differential access to homeless services, this study compares the migration patterns of the Blacks homeless clients to the patterns of White homeless clients. The NSHAPC contains data on migration patterns. If access to homeless services is more of a factor in Black homelessness, we should expect Black homeless people (especially within the inner city) to migrate less than White homeless people for homeless services, assuming equal preferences for the use of homeless services. Nested logistic regression models were run to test hypothesis 5. The dependent variable in the analysis is the log odds that a homeless client has migrated for homeless services.
Independent variables in the analysis include race, education, age, present mental health problems, present alcohol problems, present drug problems, incarceration at some point during lifetime, first-time homelessness, and central-city origin location.