«DiscoVeriNg HomelessNess Volume 13, Number 1 • 2011 U.S. Department of Housing and Urban Development | Office of Policy Development and Research ...»
$5,000–$9,999 –.135***.874.274*** 1.315.288*** 1.334.283*** 1.327.311*** 1.365.305*** 1.357.325*** 1.384 (.007) (.013) (.013) (.013) (.013) (.013) (.013) $10,000–$14,999.213*** 1.237.942*** 2.565.939*** 2.557.931*** 2.537.954*** 2.596.949*** 2.583.966*** 2.627 (.007) (.013) (.012) (.012) (.012) (.012) (.012) $15,000–$19,999.322*** 1.380 1.288*** 3.626 1.276*** 3.582 1.270*** 3.561 1.306*** 3.691 1.280*** 3.597 1.306*** 3.691 (.007) (.012) (.012) (.012) (.012) (.012) (.012) $20,000–$29,999.525*** 1.690 1.361*** 3.900 1.362*** 3.904 1.356*** 3.881 1.387*** 4.003 1.392*** 4.023 1.415*** 4.116 (.007) (.012) (.012) (.012) (.012) (.012) (.012) $30,000–$39,999.630*** 1.878 1.390*** 4.015 1.394*** 4.031 1.397*** 4.043 1.417*** 4.125 1.405*** 4.076 1.420*** 4.137 (.007) (.012) (.012) (.012) (.012) (.012) (.012) $40,000–$49,999.447*** 1.564 1.307*** 3.695 1.314*** 3.721 1.303*** 3.680 1.334*** 3.796 1.299*** 3.666 1.323*** 3.755 (.007) (.012) (.012) (.012) (.012) (.012) (.012) $50,000–$74,999.561*** 1.752 1.171*** 3.225 1.167*** 3.212 1.163*** 3.200 1.183*** 3.264 1.168*** 3.216 1.183*** 3.264 (.007) (.012) (.012) (.012) (.012) (.012) (.012) $75,000–$99,999.357*** 1.429.995*** 2.705.992*** 2.697.977*** 2.656 1.001*** 2.721.975*** 2.651.993*** 2.699 (.008) (.013) (.013) (.013) (.013) (.013) (.013) $100,000–$124,999 –.229***.795.663*** 1.941.623*** 1.865.608*** 1.837.628*** 1.874.582*** 1.790.596*** 1.815 (.009) (.014) (.014) (.014) (.014) (.014) (.014)
a crowded unit. The lower likelihood of crowding in public housing is possibly a function of occupancy restrictions placed on public housing units. Model 2, in which the dissimilarity index dummy variables are introduced into the model, indicates that segregation increases the likelihood of living in a crowded unit. Units in the highest segregation quartile were 3.071 times more likely to be crowded compared with units in the lowest segregation quartile.
In model 3, hypothesis 2 is tested. In exhibit 5, coefficients from model 3 were used to graph the predicted probability of living in a crowded unit for Blacks and Whites. At low levels of segregation, Blacks and Whites are equally likely to live in a crowded unit. At medium levels of segregation, Whites are slightly more likely to live in a crowded unit. In the highest segregation quartile, Blacks are almost two times as likely as Whites to live in a crowded unit. This finding partially supports hypothesis 2, because the highest levels of segregation have more detrimental effects on crowding in Black households than in White households. This relationship is not linear, because it does not increase crowding among Blacks more than Whites living in less segregated areas. As with the findings for the housing inadequacy models, only segregation rates in the highest quartile result in more detrimental effects for Black households.
In addition, those living in the central city and those living on welfare had higher odds of living in overcrowded housing compared with those who did not live in the central city and those not living on welfare. Householders who rented their units had higher odds than those who owned their units of living in an overcrowded dwelling, whereas older householders were less likely than younger householders to live in an overcrowded dwelling. The odds of living in an overcrowded unit decreased as education levels increased. As in the housing inadequacy analyses, income had a nonlinear effect on crowding, which is perhaps due to the combination of households that live Exhibit 5
Black White Note: In calculating the predicted probabilities, means were used where possible. Modal values were used for dichotomous control variables. This method likely produces conservative estimates, because Blacks are more likely than Whites to rent, live in central cities, and have lower incomes, characteristics that put them at greater risk of living in crowded housing.
56 Discovering Homelessness From Exclusion to Destitution: Race, Affordable Housing, and Homelessness in smaller units by necessity and households that live in smaller units by choice. Households with very low incomes must live in smaller housing units out of necessity, but households with high incomes may choose smaller units because they are located in areas that are close to valued amenities. Little difference exists in the odds of living in a crowded dwelling in the three regions.
Female-headed households were less likely than male-headed households to live in crowded units.
This finding may be attributed to female headship serving as a proxy for single-parent families, thus decreasing family size and the probability of adult crowding (Rosenbaum, 1996).
Models 4 and 5 examine the effects of the affordability measures on overcrowding, testing the second part of hypothesis 3. As expected, housing units in metropolitan areas with higher proportions of low-income renters with high rent burdens are more likely to be crowded. Higher numbers of lowest rent units in relation to households below 50 percent of the poverty threshold reduce the likelihood that a householder is living in a crowded unit. Both affordability measures decrease, but do not erase, the effects of segregation on crowding for Black householders.
In models 6 and 7, the second part of hypothesis 4 is tested. Higher homeownership rates at the metropolitan level decrease the likelihood of householders living in crowded units and reverses the relationship between segregation and crowding, but higher homeownership rates do not erase the relationship between Black headship and crowding. Adding homeownership rates reduces the odds ratio for Black-headed housing units from 1.29 to 1.14. This finding suggests that increasing homeownership rates will decrease home crowding overall but that this effect does not remove the effects of segregation for Blacks in general and especially for Blacks living in the most highly segregated areas. As in models 4 and 5, increasing affordable housing decreases crowding, but it does not erase the effects of segregation on crowding for Black householders living at the highest segregation levels.
NSHAPC Descriptive Statistics Before the 1980s, the homeless population was primarily composed of White middle-aged males.
After 1980, Blacks became overrepresented in the service-using homeless population with respect to their share of the national population (12.8 percent) and their proportion of the poverty population (28.4 percent of individuals and 26.1 percent of families).7 In the 1996 NSHAPC, 40.1 percent of homeless clients were Black non-Hispanic and 40.9 percent were White non-Hispanic.8 Geographic Location of the Black Homeless Population Both homeless clients in general and homeless clients who are Black were found at greater rates within central-city areas (exhibit 6). A much higher percentage of Black homeless clients surveyed in the NSHAPC were found in central-city areas (81 percent) than White homeless clients (62 percent). Black homeless clients were concentrated in large central-city areas (63.1 percent), with smaller percentages (17.8 percent) experiencing their homelessness in less dense, mid-size centralcity areas. In both suburban and rural areas, White homeless clients are more prevalent than Black March 1997 Current Population Survey.
This research excludes Hispanic homeless clients, because some evidence indicates that the determinants of homelessness for Hispanic clients are likely different than the determinants for both Black and White non-Hispanic clients. See Baker (1994) for more information about the “Latino paradox,” the underrepresentation of Hispanics in the homeless population.
homeless clients. When this study examined the geographic location of previous residence (exhibit 6), Black homeless clients were also more likely than White homeless clients to have lived previously in central-city locations (76.2 percent), with 57.8 percent living in large central-city areas and 18.4 percent living in mid-size central-city areas. Slightly less than one-half of White homeless clients lived in central-city locations before their current homeless episode.
As Black poverty has become concentrated in center cities, so has Black homelessness. These findings suggest that geographic explanations of Black overrepresentation should focus on conditions in central-city areas. If housing and neighborhoods are related to Black homelessness they will be housing and neighborhoods located in central-city areas, in particular, large central-city areas. Research by Burt et al. (2001) on program data from the NSHAPC found that large central-city areas had more service availability than smaller areas surveyed, although not necessarily higher levels of services in relation to population and poor population size. Given greater service availability in the areas in which they become homeless, it is expected that Black homeless clients would be less likely than White homeless clients to migrate for homeless services. This hypothesis is tested in the next section.
Migration Patterns This study now turns to the pull factor of access to shelter and homeless services. Overall, 44 percent of clients surveyed in the NSHAPC moved from the place where they became homeless to the service location at which they were interviewed (exhibit 7). As Baker (1994) and Lee and Farrell (2004) found, homeless services are more likely to be sited in minority communities than 58 Discovering Homelessness From Exclusion to Destitution: Race, Affordable Housing, and Homelessness
Exhibit 9 examines the moves made by homeless clients who migrated for services. As noted by Burt et al. (2001), most moves are made to locations of larger size, such as from suburban locations to central cities. The highest percentage of Black movers (32.3 percent) moved from one large central city to another large central city, suggesting a segregation of Black homelessness within large central cities. More than one-half (51.4 percent) of all Black movers moved to large central cities compared with a little more than one-fourth (26.4 percent) of White movers. In addition, Black movers were more likely than White movers to move to the same type of location as the location of their last regular residence (that is, from large central to large central, from mid-central
to mid-central, and so on) (53.7 versus 34.7 percent). White homeless clients who moved were more likely to be sampled in an emergency shelter than White homeless clients who did not move (34.5 versus 19.6 percent). Thus, White movers are (1) more likely to move, (2) more likely to move to larger locations, and (3) more likely to be sampled in emergency shelters than Black movers.
Black movers are (1) less likely to move, (2) more likely to move from one large central-city location to another large central-city location, and (3) more likely to move to a place similar in size to the location they left.
Duration, Transiency, and Alternative Explanations of Homelessness After becoming homeless, Black homeless clients have longer mean homeless spells than White homeless clients have (an average of 3 versus 2.4 years). Using the 1996 NSHAPC, Allgood and Warren (2003) found that White homeless clients had shorter homeless spells than non-White homeless clients. Homeless spells are longer for Blacks in central cities than in other areas and longest for Whites in rural areas. Rural findings should be viewed with caution, however, due to low sample sizes in these areas. Around 50 percent of both Black and White homeless clients were experiencing their first homeless spell at the time of the survey. The frequency of homeless spells was similar across the racial categories examined.
Exhibit 10 presents variation in transiency by race. Transiency is measured as the number of towns or cities that a homeless client stayed in for 2 or more days while homeless. This study found the experience of White homelessness to be more transient. More than two times the percentage (29 versus 13 percent) of White homeless clients stayed in three or more towns or cities during their current homeless spell. Greater White transiency could be due to lack of homeless services or 60 Discovering Homelessness From Exclusion to Destitution: Race, Affordable Housing, and Homelessness
greater freedom of Whites to move from town to town. Although living in more cities is correlated with length of current homeless spell for White homeless clients (r =.307***9), no correlation exists for Black homeless clients.
NSHAPC Multivariate Analyses Exhibit 11 presents the results of logistic regression models that predict the likelihood that a homeless client has moved to receive homeless services. Model 1 includes controls for race, education, age, mental health problems, alcohol problems, drug problems, incarceration history, and first-time homelessness. Jencks (1994) and Hopper (2003) suggest that the crack epidemic10 played a role in increasing homelessness in the 1980s and 1990s, in particular among Blacks.
Model 2 adds central-city origin location to the model. Hypothesis 5 receives some support in both models, because Black homeless clients are less likely than White homeless clients to have migrated for homeless services after controlling for other factors. In model 1, Black homeless clients are.365 times as likely as White homeless clients to have moved. In model 2, we see that homeless clients who become homeless in central cities are less likely than clients who become homeless outside of central cities to migrate for homeless services. The addition of central city to the model explains part of difference between Black and White homeless clients in model 1. Although taking a central-city location into account explains part of the difference in migration, in model 2, Black homeless clients are still less than one-half as likely as White homeless clients to have moved. In both models, male homeless clients are almost two times as likely as female homeless clients to have moved.
Golub and Johnson (1997) suggested that crack cocaine use had declined or at least remained stable in the late 1990s.
Thus, it might not be less of a factor for current homelessness among Blacks. More recent studies may find more White rural homelessness due to increases in methamphetamine abuse.