Housing Instability Among Current and Former Welfare Recipients

Abstract

Objectives. We examined correlates of eviction and homelessness among current and former welfare recipients from 1997 to 2003 in an urban Michigan community.

Methods. Longitudinal cohort data were drawn from the Women’s Employment Study, a representative panel study of mothers who were receiving cash welfare in February 1997. We used logistic regression analysis to identify risk factors for both eviction and homelessness over the survey period.

Results. Twenty percent (95% confidence interval [CI]=16%, 23%) of respondents were evicted and 12% (95% CI=10%, 15%) experienced homelessness at least once between fall 1997 and fall 2003. Multivariate analyses indicated 2 consistent risk factors: having less than a high school education and having used illicit drugs other than marijuana. Mental and physical health problems were significantly associated with homelessness but not evictions. A multivariate screening algorithm achieved 75% sensitivity and 67% specificity in identifying individuals at risk for homelessness. A corresponding algorithm for eviction achieved 75% sensitivity and 50% specificity.

Conclusions. The high prevalence of housing instability among our respondents suggests the need to better target housing assistance and other social services to current and former welfare recipients with identifiable personal problems.

The Personal Responsibility and Work Opportunity Reconciliation Act of 1996 ended the federal government’s guarantee of cash assistance to poor families by replacing the 60-year-old entitlement program Aid to Families with Dependent Children with the transition-to-work program Temporary Assistance for Needy Families (TANF). TANF imposed a cumulative 60-month time limit on receipt of federally funded cash benefits for most recipients, tied welfare receipt directly to work activity, and devolved a great amount of program authority to the states.1

Many studies have explored the impact of the Personal Responsibility and Work Opportunity Reconciliation Act on the work behavior and welfare status of low-income women.2^,3 Large declines in welfare case-loads, increased work activity among single mothers, and reductions in official child poverty rates are indicators of the success of welfare reform. However, related research on the material well-being of those who no longer receive welfare (welfare leavers) has documented high rates of hardships, including lack of health insurance coverage, food insecurity, and housing problems.4^–8

Housing problems are particularly acute among low-income families with children. Wood and Rangarajan, using 2003 survey data, reported that 16% of unsubsidized current and former welfare recipients in New Jersey had experienced eviction, homelessness, doubling up with friends or relatives, or frequent moves in the previous year.9 In a review of welfare-leaver studies across many states, Acs and Loprest found that between 25% and 50% of welfare leavers reported falling behind on housing payments, and between 6% and 26% reported moving as a result of high housing costs. Smaller percentages experienced eviction (4% to 7%) or homelessness (1% to 3%). On average, rates of housing problems were somewhat lower for welfare leavers than for welfare stayers, although some of the studies reviewed by Acs and Loprest showed that more than 20% of welfare leavers continued to experience housing problems.3

Despite high rates of housing problems among current and former welfare recipients, the characteristics of individuals and families that are associated with heightened risks of unstable housing conditions have not received much attention. Given the probable relationships between housing-related hardships, material well-being, and labor market outcomes, the lack of knowledge about these characteristics represents a significant gap in the literature.

The studies published to date have tended to use observational data from samples of homeless or otherwise disadvantaged individuals. Few studies have compared the characteristics and circumstances of low-income women experiencing housing problems with those of women in similar socioeconomic situations who are not experiencing such problems.

Observational studies of severely disadvantaged individuals may be especially misleading in light of evidence indicating a high prevalence of key risk factors among the broader population of low-income individuals. For example, previous research involving the panel data used in our study showed that more than one third of all current and former welfare recipients satisfied diagnostic screening criteria for at least 1 psychiatric disorder in a given year and that about two thirds experienced at least 1 disorder over the 6-year study period.10^,11 Such mental and physical health problems are likely to contribute to housing instability among low-income families.

PREDICTORS OF HOUSING INSTABILITY

The homeless have been shown to have high levels of physical and mental illness.12^,13 Physical health problems may foster housing instability by depleting economic resources or interfering with an individual’s ability to work steadily.14^,15 Psychiatric disorders may estrange individuals from family and friends, leaving them with fewer social and material resources to draw upon in times of need. Interpersonal problems may place people with mental illness at greater risk for eviction or homelessness if they are less able to negotiate conflicts with landlords or to address contentious living situations.16^,17

Substance use and related disorders are also risk factors for homelessness,18^,19 because they may deplete social and material resources. It has been shown that drug use, particularly use of crack, heroin, and cocaine, is more pronounced in homeless populations.13^,19^–22 Criminal offenses related to the possession, use, and distribution of illicit substances are explicit criteria for eviction from public housing and loss of housing voucher aid.23 These offenses also constitute grounds for eviction from private housing in many jurisdictions.

Demographic attributes are related to housing instability as well. On average, homeless women are disproportionately young and non-White.12^,13^,22 Young adults may be vulnerable to housing instability because they have not developed the economic and social resources to help them obtain and retain housing.24 Discrimination may amplify housing instability among members of minority groups.20^,25^,26 Racial differences with respect to wealth may place these individuals at greater risk for housing instability, because wealth or savings can help people maintain stable housing during periods of financial hardship.27

Marriage and cohabitation may be protective factors against eviction and homelessness. In comparison with families headed by married couples, single-mother families are much more likely to be poor and to have fewer economic resources upon which to call in times of need. Similarly, cohabitation may contribute to housing stability by increasing household economic resources. Marriage and cohabitation may also reflect otherwise unobserved individual characteristics such as interpersonal skills.22 In contrast, cohabitation is less stable than marriage.28^,29 Thus, women who cohabit may be more at risk for housing problems than are single women were a relationship to end and they experience difficulty finding a new residence.

Domestic violence may increase the risk of eviction and homelessness, because women experiencing abuse often leave their primary residence to escape the abuser and may seek refuge in emergency shelters.13 Battered women are also vulnerable to eviction that arises from partners’ abusive behavior directed toward other tenants, landlords, or the physical property itself.30

Women who have been under criminal justice supervision may be more likely to experience housing difficulties. Employers are less likely to hire individuals with criminal records31; jail time may erode human capital or social networks and exacerbate physical or mental illness.32 Diminished social networks resulting from criminal conviction may also reduce the chance that individuals can rely on friends or family for social or material support in times of need, thus heightening the risk of housing instability. Criminal offending may foster social contacts that heighten specific risks such as domestic violence. It may also signal other individual characteristics and behaviors that may interfere with an individual’s ability to maintain secure and stable housing.24 In addition, individuals convicted of drug-related felonies are no longer eligible for some forms of housing assistance and public aid.23^,33

Finally, human capital—work experience, work skills, and education—affects an individual’s ability to find and retain employment,15^,34 to navigate the housing assistance system or complicated evictions proceedings, or to secure affordable housing through housing searches. Human capital deficiencies may foster housing instability if those who are jobless or working in low-paying jobs have difficulty making monthly rental payments.

METHODS

We analyzed data from the Women’s Employment Study (WES), a longitudinal survey of single mothers who received cash welfare in one Michigan urban county in February 1997. Trained interviewers conducted in-person interviews with these women in the fall of 1997, 1998, 1999, 2001, and 2003; interviews averaged about 1 hour at the first wave and about 1.5 hours at the final wave. Respondents were selected with equal probability from all women who received TANF benefits in February 1997 and were White or African American, single, heads of households, and between 18 and 54 years of age. Response rates at the 5 waves were 86%, 92%, 91%, 91%, and 93%, respectively. We used data from the 536 WES respondents who completed all 5 interviews. Because there was little evidence that attrition from the sample was nonrandom, we did not use sample weights.35

Given the panel design, data were available on whether a woman had been evicted or experienced a period of homelessness between each survey wave. All independent variables, with the exception of criminal conviction, were measured as baseline characteristics reported at wave 1 in fall 1997. The dependent variables were based on the respondents’ experiences after the baseline characteristics had been measured. We analyzed correlates associated with a respondent’s report that (1) she was evicted from her residence at some time in the period starting after the fall 1997 interview and ending with the fall 2003 interview and (2) she experienced a spell of homelessness over the same 6-year period.

We used multiple logistic regression analysis to examine the baseline characteristics associated with each outcome. We included 7 employment barriers using definitions developed in earlier research involving this data set15: low educational attainment (did not graduate from high school), low human capital (low levels of work experience or low levels of use of specific skills on previous jobs), and whether a respondent had a criminal conviction, met diagnostic screening criteria for 1 of 3 measured psychiatric disorders (major depression, generalized anxiety disorder, and posttraumatic stress disorder), had a physical health problem, had used “hard” drugs (defined as stimulants, cocaine, crack, heroin, hallucinogens, or inhalants) at any time in her life before the fall 1997 interview, or had experienced severe domestic abuse at any time up to the fall 1997 interview.

We also examined sociodemographic characteristics, including whether a respondent was married or cohabiting, her race and age, the number of children residing with her, and the percentage of years she had spent on welfare from 18 years of age to the 1997 interview. All of these variables other than criminal conviction were measured at the first interview. Table 1 ▶ includes definitions of all dependent and independent variables.

TABLE 1—

Variable Definitions and Prevalence of Characteristics Among Sample Respondents: Women’s Employment Study, 1997–2003

	Sample
Dependent variables
Reported eviction in at least 1 wave after 1997 interview, %	19.8
Reported episode of homelessness in at least 1 wave after 1997 interview, %	12.3
Independent variables
No high school degree at 1997 interview, %	29.9
Human capital barrier (worked less than 20% of the time between age 18 y and 1997 interview or had performed 4 or fewer of 9 specific skills on a job before 1997 interview), %	26.4
Criminal conviction before 1999 interview, %	4.5
Mental health problem in 12 months before 1997 interview (met diagnostic screening criteria using the CIDI short-form for 1 or more of the following disorders: posttraumatic stress disorder, depression, and generalized anxiety disorder), %	36.0
Physical health problem in 12 months before 1997 interview (had age-specific physical limitation or self-reported fair or poor health), %	54.4
Domestic violence (experienced severe abuse at any time before 1997 interview), %	52.8
Hard drug use (use of cocaine, crack, stimulants, heroin, or other hard drugs at any time before 1997 interview), %	17.4
Married at 1997 interview, %	10.3
Cohabiting at 1997 interview, %	14.7
African American (coded as 1 vs 0 for White), %	54.7
Aged 18–24 y at 1997 interview, %	25.0
No. of children in household at 1997 interview, mean	2.24
Percentage of years on welfare from age 18 y until 1997 interview	59.3

Note. CIDI = Composite International Diagnostic Interview.

RESULTS

Table 2 ▶ shows descriptive statistics for the entire sample as well as for women classified according to their experiences of eviction and homelessness. Twenty-four percent of respondents (n = 130) had unstable housing situations at some time between 1997 and 2003 (95% CI = 20%, 28%); 19.8% (n = 106) had been evicted at least once (95% CI = 16%, 23%), and 12.3% (n = 66) had been homeless at least once (95% CI = 10%, 15%).

TABLE 2—

Employment Barriers, Sociodemographic Characteristics, and Experiences of Eviction and Homelessness: Women’s Employment Study, 1997–2003

		Evicted During 1997–2003		Homeless During 1997–2003
	Total (N = 536)	No (n = 430)	Yes (n = 106)	No (n = 470)	Yes (n = 66)
Employment barriers, %	100.0	80.2	19.8	87.7	12.3
No high school degree	29.9	25.6	47.2***	27.0	50.0***
Human capital barrier (low work experience or skills)	26.4	26.2	27.4	25.4	33.9
Criminal conviction	4.5	3.5	8.5**	3.4	12.1***
Mental health problem	36.0	33.6	46.2**	33.4	54.6***
Physical health problem	54.4	51.4	66.7***	50.6	81.5***
Domestic violence	52.8	50.0	64.2***	49.8	74.2***
Hard drug use	17.4	15.7	24.8**	16.0	27.7**
Sociodemographic characteristics
Married, %	10.3	11.2	6.6	10.6	7.6
Cohabitating, %	14.7	15.1	13.2	14.5	16.7
African American, %	54.7	55.6	50.9	53.8	60.6
Age 18–24 y, %	25.0	24.7	26.4	23.8	33.3*
No. of children in household, mean	2.24	2.22	2.32	2.25	2.15
Years on welfare as an adult, %	59.3	58.2	63.8**	58.3	66.2**

Note. Domestic violence and hard drug use were assessed over a respondent’s lifetime, prior to the first interview in fall 1997. Criminal conviction was assessed prior to 1999. As a result of missing data, values in some columns do not reflect the total sample size.

*P < .10; **P < .05; ***P < .01.

Bivariate comparisons indicated that women who experienced 1 of these housing problems were more disadvantaged than other respondents. Among those evicted or homeless between 1997 and 2003, approximately one half had not completed high school, as compared with only about one quarter of those who had not been evicted or experienced homelessness. Mental and physical health barriers and episodes of domestic violence were also significantly associated with evictions and homelessness. As of the 1997 survey, about one quarter of women with housing problems reported that they had used hard drugs at some point in their lives, in comparison with about one sixth of those not experiencing these problems. Those experiencing housing problems were more than twice as likely as those not experiencing housing problems to have been convicted of a crime.

At the same time, many respondents who reported no housing difficulties also reported barriers traditionally associated with housing risks. More than 30% reported mental health barriers at the first wave, and approximately half reported physical health barriers. Also, half had experienced domestic violence at some time in their life prior to fall 1997, and nearly 20% reported some history of hard drug use in 1997 or preceding years.

The results of our multivariate analysis are shown in Table 3 ▶. (In a related analysis, we used 4 survey waves and estimated fixed effects and conditional logistic regression models for eviction and homelessness over the 6-year survey period. In that analysis, we used the contemporaneous value of each variable rather than the value at the first interview. Findings are available on request.) Women who had less than a high school education and those with a history of hard drug use were more likely to have been evicted at some time during the study period.

TABLE 3—

Adjusted Odds Ratios (AOR; With 95% Confidence Intervals [CIs]) for Eviction and Homelessness (n = 523): Women’s Employment Study, 1997–2003

	AOR for Eviction (95% CI)	AOR for Homelessness (95% CI)
No high school degree	2.57*** (1.56, 4.24)	2.21** (1.19, 4.10)
Human capital barrier (low work experience or skills)	0.70 (0.41, 1.21)	1.15 (0.60, 2.20)
Criminal conviction	1.78 (0.71, 4.51)	2.79* (0.99, 7.89)
Mental health problem	1.39 (0.86, 2.25)	1.69* (0.93, 3.07)
Physical health problem	1.39 (0.84, 2.29)	3.19*** (1.54, 6.60)
Domestic violence	1.44 (0.87, 2.36)	2.23** (1.16, 4.29)
Hard drug use	1.59* (0.90, 2.81)	1.83* (0.90, 3.72)
Married	0.49 (0.19, 1.25)	1.18 (0.39, 3.55)
Cohabiting	0.79 (0.39, 1.58)	1.82 (0.80, 4.17)
African American	0.78 (0.48, 1.29)	1.83* (0.97, 3.47)
Aged 18–24 y	1.07 (0.63, 1.86)	1.83* (0.94, 3.56)
No. of children in household	1.10 (0.92, 1.32)	0.87 (0.68, 1.11)
Percentage of years on welfare as an adult	1.53 (0.55, 4.27)	2.38 (0.60, 9.38)
χ2	42.6***	64.6***

Note. Although 13 respondents had missing data on 1 or more of the independent variables, decreasing the total number of respondents included in the logistic regression analysis to 523, we conducted a sensitivity analysis and confirmed that the missing data did not appear to bias the regression results.

*P < .10; **P < .05; ***P < .01.

A greater number of variables were significant in the logistic regression examining the correlates of homelessness than in the logistic regression examining evictions. Having less than a high school education, having a criminal conviction, experiencing a mental or physical health problem, experiencing domestic violence, using hard drugs, being African American, and being between the ages of 18 and 24 years in 1997 were all associated with being homeless at least once after fall 1997. The relationship between physical health problems and homelessness was particularly strong and statistically significant (adjusted odds ratio [OR] = 3.19, P < .01).

To determine the relative magnitude of each of the independent variables, we computed the probability of eviction or homelessness during the survey period for a representative respondent with no health problems or barriers to work activity and median demographic characteristics (African American, not married or cohabiting, older than 24 years, caring for 2 children, and 59% of years on welfare as an adult). With these characteristics, the baseline predicted probability that a representative respondent would be evicted at some point after the 1997 interview was 8.3%; the probability of homelessness was 1.6%.

We calculated the predicted probability of a respondent experiencing an eviction or homelessness if she had one of the characteristics that was significant in the regression but was otherwise identical to the representative respondent relative to the baseline prediction (8.3% for eviction, 1.6% for homelessness). Varying the independent variables one at a time allowed us to determine the extent to which the probability of eviction or homelessness changed in response to changes in the characteristics of the representative respondent. Her probability of experiencing an eviction increased to 12.6% if she had used hard drugs and to 18.9% if she had not completed high school.

Having less than a high school education or having a criminal conviction increased the probability of homelessness to 3.6% and 4.5%, respectively. In addition, the probability increased to 5.1% if she had physical health problems, to 2.7% if she had mental health problems, and to 3.0% if she had used hard drugs. If she had experienced domestic violence, her probability of homelessness increased to 3.6%, and it fell to 0.9% if she was White. The probability increased to 3.0% if she was between the ages of 18 and 24 years.

We calculated sensitivity and specificity curves for both homelessness and eviction to examine how well our model correctly distinguished between respondents with and without housing problems. For homelessness, the area under the receiver operating characteristic curve of 0.79 indicated a rather tight model fit (an area of 1.0 indicates a perfectly discriminatory model). For eviction, the area under the receiver operating characteristic curve was 0.69. Screening individuals with a predicted probability of homelessness exceeding 0.11 would result in a sensitivity of 75% and a specificity of 67%. This same cutoff led to a positive predictive value of 24% and a negative predictive value of 95%. In the case of eviction, screening individuals with a predicted probability of eviction exceeding 0.15 would result in a sensitivity of 75% but a much lower specificity, 50%. This cutoff led to a positive predictive value of 27% and a negative predictive value of 89%.

Our results reveal that many factors associated with work outcomes are also associated with housing outcomes among current and former welfare recipients, although the relative importance of these factors differs for each of the various outcomes. In other analyses (data not shown but available from the authors on request), we found no consistent pattern between employment and eviction or homelessness. That is, respondents who experienced one of these negative housing events were not necessarily less likely to be working, to have experienced job loss, or to have lower household incomes. Work provides incomplete protection against housing instability.

DISCUSSION

Other studies, including those involving the same data used here,8^,36 have documented improvements in the average socioeconomic status of single mothers after welfare reform. However, these studies have not focused on the substantial minority of these women who experience housing instability. We found that about one fifth of low-income WES mothers had been evicted and about one eighth had been homeless at some point between 1997 and 2003. Risks of these housing problems were much higher within specific subgroups: 50% of those with criminal convictions, 38.8% of high school dropouts, 34.4% of hard drug users, and 30.4% of those exposed to domestic violence had experienced at least 1 of the 2 housing insecurity problems assessed.

Although our 6.5-year panel investigation does not permit extensive causal analysis, our regression analyses demonstrated that several baseline characteristics were associated with subsequent episodes of housing instability and are thus potentially important targets of practice and policy interventions. Our results revealed a strong association between 4 personal problems—health limitations, psychiatric disorders, hard drug use, and domestic violence—and periods of homelessness. Many studies of homelessness and health have been unable to disentangle the causal relationship between these variables. In our analysis, these problems were measured in fall 1997 and thus were a precursor to the spells of homelessness assessed over the subsequent 6 years. Among the 4 problems, only hard drug use was associated with future evictions, suggesting that different factors contribute to different unstable housing outcomes.

Study Limitations

This study involved 2 principal limitations. First, our data regarding drug use, psychiatric disorders (assessed using the Composite International Diagnostic Interview screening batteries for the 12-month prevalence of psychiatric disorders), and health status were based on self-reported information collected by trained survey interviewers. These measures were not based on clinical or diagnostic examinations. We did use standard epidemiological measures validated in other surveys.37

Second, our sample was drawn from non-Hispanic White and African American mothers 18 to 54 years of age who received cash welfare in fall 1997 in 1 urban Michigan county. Although the WES represents well the population from which it was drawn, it is not nationally representative or representative of a population of welfare recipients in 2005. In particular, because of its urban Michigan location and its cohort design, the WES does not provide representative data on the youngest subgroups of current TANF recipients, immigrant or noncitizen-headed households, or households in rural areas.

Other data sets, such as the National Survey of America’s Families, the National Longitudinal Survey of Youth, and the National Survey of Drug Use and Health, are more nationally representative. However, the first 2 are (repeated) cross-sectional data sets, and the third does not provide detailed information on mental health, health, drug use, experiences of domestic violence, and some of the other correlates of housing instability that we analyzed in our study. The numbers of current and former welfare recipients in the WES compare favorably to the numbers in these national samples. Also, descriptive statistics from previous WES investigations of substance use and psychiatric disorders suggest close agreement between the WES and national data when overlapping data items are available.10

Policy Implications

Our results indicate a number of risk factors that should be given more attention by policymakers and practitioners. For example, it is likely that screening welfare recipients for health, mental health, drug use, and domestic violence problems could identify many of the women facing the greatest risks of homelessness. Almost half of our respondents who reported eviction or homelessness lacked a high school degree, and nearly 30% had low skill levels or low levels of work experience. Welfare caseworkers might investigate housing concerns among welfare applicants with a history of exposure to domestic violence (74% of the homeless individuals in our sample had such a history). Exploring housing problems among those with physical or mental health problems may be particularly important.

Current federal housing programs are not entitlements; relatively few low-income households receive housing assistance. A few states use TANF dollars to provide supplemental housing assistance to families who have left welfare, although under current rules receipt of such assistance for more than 4 months would count against the 60-month federal time limit.38 A New Jersey study showed that rates of eviction and homelessness were markedly lower among current and former welfare recipients who received a housing subsidy or lived in public housing than among those living in unsubsidized housing.9 In addition, the positive effect on employment rates and earnings was greater among welfare recipients in the Minnesota Family Investment Program who received housing assistance in combination with other forms of assistance than among those who did not receive housing assistance.39

Our findings have identified a set of risk factors that can help service providers target social services and housing assistance to those at high risk for eviction and homelessness. Some of the risk factors that might be used for such targeting are similar to those used to allocate assistance with employment. However, our findings also indicate differences between determinants of housing instability and those related to work. Some current and former welfare recipients who work still face significant risks of housing instability and may thus require assessment and services. Such findings underscore that wages and hours worked are not the only measures of well-being to consider when examining the transition from welfare to work. Increased attention to housing as an outcome can improve the well-being of welfare recipients.