On the Declining Health Status of Welfare Caseloads: Emerging Dilemmas for Serving the Poor

Abstract

An emerging concern for public policy is welfare reform’s potential to inadvertently affect caseload composition by increasing the proportion of recipients with health-related barriers to employment. We examine this using data from the Welfare Client Longitudinal Study, an in-depth case study of a large California county. Through quantitative analyses, we examine the extent of change in health-related problems since welfare reform and their potential to progressively impact overall composition of the caseload. We augment this with qualitative data on how local welfare providers are responding to the health-related needs of aid recipients. Results suggest that the burden of health-related problems is growing and that welfare providers may be poorly equipped to respond effectively on their own. The changing composition of welfare caseloads may foster several new policy dilemmas that demand broader attention: states and localities may face difficulties meeting federal workforce participation requirements, may need to restructure welfare-to-work programs to serve a more functionally impaired population, and take steps to better integrate health and welfare services at the local level.

An emerging concern for public policy is welfare reform’s potential to inadvertently affect caseload composition by gradually increasing the proportion of recipients with health-related barriers to employment. Welfare reform has helped many recipients make successful transitions from welfare to work (Blank and Haskins, 2001; Blank, 2002). Yet there are growing concerns that many of those left behind on the rolls are there due to health-related problems that pose significant barriers to self-sufficiency. Over time, the gradual “silting up” in the caseload of recipients with mental health problems, addictions, physical disabilities, and the closely-related social problems of violent victimization, criminality and homelessness, could make increasing proportions of caseload “hard to employ” (Blank, 2002; Pavetti and Bloom, 2001; Zedlewski and Loprest, 2001). If recipients’ health care needs are indeed steadily growing over time, then policymakers may face a new constellation of interests and issues in financing and organizing public services for the poor.

There is limited research on secular change in the health status of populations on aid. Still, state and local welfare providers widely perceive growth in the proportion of hard-to-employ recipients since welfare reform (Moffitt and Stevens, 2001). In a series of site visits to state welfare agencies, the U.S. General Accountability Office found many providers preoccupied by concerns about responding to health-related barriers to work in their caseloads (United States General Accounting Office, 2001; Weil, 2002; Weil and Finegold, 2002). Investigators from the Urban Institute similarly found that welfare systems in 17 cities were shifting more attention to health issues due to concerns about a “growing disconnect between the needs of some recipients and their ‘welfare-to-work program capabilities’” (Holcomb and Martinson, 2002, p2).

Such concerns underscore the need for an empirically based understanding of how the composition of welfare caseloads may be changing and the capacity of welfare providers to meet the health-related needs of their clientele. We address this with an analysis of the Welfare Client Longitudinal Study (WCLS), an in-depth case study of welfare recipients and providers in a large California county. We present results from quantitative analyses of representative surveys of the countywide caseload to examine changes over time in the health-related needs of aid recipients and whether they could progressively impact caseload composition. We augment this with an analysis of qualitative data on local welfare providers throughout the study site. This allows us to examine how welfare workers are responding to the health care needs of recipients, and the policy-related issues they face in meeting those needs.

Background: The Policy Debate

The 1996 welfare reform law, the Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA, P.L. 104–193), brought about a broad shift from the traditional model of welfare entitlement to one that promotes self-sufficiency through mandatory work requirements and time limits on aid. The first few years of reform generated considerable optimism due to evidence that many aid recipients were making successful transitions from welfare to work (Acs and Loprest, 2001; Blank and Schmidt, 2001; Brauner and Loprest, 1999; Karoly, 2001; Loprest and Wissoker, 2002). Research further documented that solid gains could be made through job search programs using the “work first” approach, without requiring long-term public investments in the education and vocational rehabilitation of aid recipients (Bloom and Michalopoulos, 2001; Hamilton, 2002).

Despite all this, by the first debate over welfare reform reauthorization in 2001–2, policymakers, academics and advocates had begun to voice new concerns about welfare reform’s potential to inadvertently increase the proportion of aid recipients with health-related barriers to work. In the Washington Post, Joe Califano (2002), former Secretary of Health, Education and Welfare, argued that while the overall size of caseloads had declined significantly due to reform, those left behind on the rolls were the most disadvantaged due to significant health-related barriers to work. Helping these recipients enter the labor force would require more sustained and costly assistance for domestic violence, physical disabilities, mental health problems, criminal involvement and substance abuse (for related discussions, see: Allen and Kirby, 2000; Blank and Haskins, 2001; Kuttner, 2002; Loprest and Zedlewski, 2002; Weil and Finegold, 2002). If the most disadvantaged recipients were to represent a greater proportion of the aid population over time, then policymakers would need to develop long-term solutions to meet their health and welfare needs (United States General Accounting Office, 2001). Most states, having implemented a “work first” approach, focused on moving individuals rapidly into the workforce (Blank and Haskins, 2001). Yet this emphasis on short-term transitioning clients into work left most poorly equipped to provide remedial assistance to those facing significant health problems.

Despite these concerns, there were few data that documented a growing burden of hard-to-employ welfare recipients (United States General Accounting Office, 2001). What data there were suggested mixed results. Moffitt and Stevens (2001), using the Current Population Survey (CPS), found that the composition of the caseload had changed little over the course of reform, at least in terms of what they could measure, which was confined to indicators of labor market potential such as education and recent work experience. In another analysis of CPS data, Moffitt (2002) found conflicting evidence in the form of an increased tendency for more educated women to successfully leave aid, although the effect appeared to be modest. One other available study compared barriers to employment in the aid population over time, finding little change in characteristics that hinder employment. Although the researchers found slight declines in health and mental health status, they were not statistically significant (Zedlewski and Loprest, 2001; Zedlewski and Alderson, 2001).

Practically speaking, much of the 2001–2 reauthorization debate over the hard-to-employ centered on whether more recipients should be exempted from welfare reform’s mandatory work requirements and time limits. Federal guidelines already included a “hardship exemption” that allowed welfare agencies to exempt up to 20 percent of their caseloads; beyond that proportion, states were required to cover the costs of exempted recipients on their own (Blank and Haskins, 2001; Loprest and Zedlewski, 2002). Pointing to the lack of research evidence, White House officials argued against raising the exemption threshold. To them, claims of growing disadvantage in the welfare caseload seemed largely a perception rather than a reality. They argued that the perception of declining health status was due to the fact that, under reform, welfare workers were more directly “engaging the caseload,” which meant they were simply more likely to notice “hidden barriers” to employment than before welfare reform’s push to move recipients into jobs (Kaufman, 2002).

At the end of the day, the reauthorization debate produced no meaningful changes in policies affecting the hard-to-employ. And since this time, the thrust of welfare policymaking has been towards making even fewer accommodations for recipients with health-related barriers to employment. Thus, in changes to the welfare reform law during 2006, Congress increased work requirements and tightened restrictions on the extent to which participation in mental health and substance abuse treatment could count towards work requirements (Goldstein, 2006).

METHODS

The analysis brings together quantitative and qualitative data from the Welfare Client Longitudinal Study (WCLS). The focus of interest is recipients and providers of Temporary Aid to Needy Families (TANF, formerly Aid to Families with Dependent Children or AFDC). Figure 1 shows the timing of WCLS data collection in relation to relevant welfare policy changes. The WCLS includes two representative surveys of the study site’s AFDC/TANF caseload in 1989 and 2001, collected at similar intervals before and after passage of the PRWORA. Participants in the 2001 survey were later followed over a three-year period to track changes in a wide range of health, mental health and related social problems. As Figure 1 shows, the WCLS also included a three-year qualitative study of welfare reform implementation throughout TANF agencies in the study site. Through unobtrusive observation and in-depth interviews with TANF workers, we captured day-to-day workplace practices with a focus on how workers managed health-related problems in the caseload.

Figure 1.

Timeline

Study Site

The study site is a California county of nearly one million residents. The site was selected for its large size and wide variation in ethnic and income groups, including more affluent communities as well as pockets of urban ghetto poverty and economically depressed rural areas. In 1997, the county welfare system began adopting and implementing California’s version of the federal welfare reform guidelines. Like most states, California adopted the federal 5-year time limit, work requirements and the “work first” model. “Work first” focuses on quickly moving recipients into jobs rather than providing education and vocational training; access to health services occurs only after a recipient proves unsuccessful at finding employment. Compared to other states, California differs in that it offers relatively generous cash benefits, removes only adults (not children) when time limits are applied and has moderate earned-income disregards.

Between the 1989 and 2001 AFDC/TANF caseload surveys, the county population grew in size by approximately 18 percent, and grew older and more ethnically diverse (for more detailed descriptions, see: Schmidt et al., 2006; Zivot and Jacobs, 2004). These demographic changes in the countywide general population are mirrored in the caseload surveys. Table 1 shows the distributions of welfare caseload characteristics in 1989 and 2001, about 6 years before and after passage of the PRWORA. In keeping with broader trends in the study site, welfare recipients were significantly older and more ethnically diverse in 2001 than in 1989. Single parents were more common in the 2001 welfare caseload than in 1989, while the gender, marital status, family characteristics, education and prior history of aid receipt appear unchanged. It is notable that, after welfare reform in 2001, recipients report higher levels of recent workforce engagement than they did prior to reform, in 1989. This is an important change in the caseload composition that we take up in more detail below.

TABLE 1.

Characteristics of Welfare Recipients, Before and After Welfare Reform (in percents)

Demographic and Human Capital Characteristics	AFDC Before reform 1989	TANF After reform 2001	p-valuea
Age			.00
18–24	35	33
25–34	48	39
35+	17	28
Gender			.55
Male	6	7
Female	94	93
Race/ethnicity			.00
White	46	30
African American	36	38
Other minority	19	32
Married or cohabitating	19	22	.23
Single parent	61	75	.00
Child under age 4 at home	63	60	.40
Prior history of NAFDC/TANF	60	63	.36
High school education	69	67	.59
Employed during the prior year	61	75	.00
Unweighted N	362	691

Note. Results are weighted for sample design and non-response.

Percents may not sum to 100 due to rounding.

^ap-values are for two-tailed tests.

Data Sources and Procedures

Quantitative Data on the AFDC/TANF Caseloads

The 1989 and 2001 welfare caseload surveys include representative samples of the population applying for, and receiving AFDC/TANF throughout the study site. In the 1989 survey, we successfully interviewed 465 AFDC applicants with a response rate of 92 percent. In 2001, we returned to the same local agencies and drew a new sample of 691 TANF applicants in an identical manner, with a response rate of 85 percent. Both surveys used identical instruments and procedures to ensure comparability and to facilitate comparisons of the caseload before and after welfare reform. Subjects were systematically sampled, taking every “nth” applicant from the daily intake rosters at all four of the study site’s AFDC/TANF offices. In-person interviews in English and Spanish took place before or after the welfare intake interview in private locations, and always preceded the final determination of acceptance onto aid. Applicants were tracked in county records to determine the outcome of their aid application.

In addition to the caseload surveys, we randomly selected 455 TANF recipients from the 2001 study to track and re-interview on a yearly basis over three years. At each wave of follow-up, participants were interviewed by telephone if possible, but those without phones or who could not otherwise be reached were interviewed in-person in a variety of private locations, including their homes, prisons and homeless shelters. Intensive tracking efforts allowed us to successfully locate and re-interview 82 percent of the baseline sample at the third year of follow-up, with no evidence of biased attrition on key measures in this analysis (see: Odierna, 2006; in press). All study participants gave informed consent and participants in the 2001 study are protected by a federal Certificate of Confidentiality. Throughout, we use sample weights to equalize uneven probabilities of selection due to survey design, non-response and attrition. For more details on methodology, see prior WCLS publications (e.g., Lown et al., 2006; Schmidt et al., 2007; Schmidt et al., 2002; Schmidt et al., 2006).

Qualitative Data on Welfare Providers

The qualitative portion of this study took place during 2002–2005 at all four TANF offices in the study site. A staff of five researchers, including two PhD-level ethnographers, conducted the fieldwork. The team logged over 150 hours directly observing day-to-day work routines and program activities in TANF offices, such as intake processing and welfare-to-work programs. They also conducted 36 in-depth interviews with TANF intake and caseworkers to document changes in work routines since welfare reform, particularly those involving the provision of health-related services. Welfare provider interviews were open-ended and semi-structured, lasting 60–90 minutes. The interview guide was loosely centered on topics rather than specific questions, which included: workers’ primary tasks and responsibilities, their views of the recipient population, the management of recipients with health-related problems, and in appropriate cases, their experiences working with welfare clients before and after welfare reform. For other analyses of these data and further details on data collection, see: Dohan et al., 2005; Henderson et al., 2006; Schmidt et al., 2006.

Measures

Quantitative analyses compare changes over time in the prevalence of seven different health-related problems in the AFDC and TANF caseloads, in addition to demographic, family and employment characteristics. Psychiatric distress is evaluated using the Brief Symptom Inventory (BSI). This 53-item scale assesses the global severity of psychiatric symptoms using Derogatis’ standardized protocol for case definition (Derogatis, 1992). Assessments of substance abuse are based on standard measures widely used in prior epidemiological studies (Clark and Hilton, 1991; Weisner and Schmidt, 1993; Weisner and Schmidt, 1995; Wilsnack et al., 1991). A recipient is classified as a problem drinker if he or she satisfies two of the following criteria during the year prior to interview: (1) consumption of five or more drinks at one sitting on a monthly basis or more often; (2) at least one of five alcohol dependence symptoms; and (3) at least one of five alcohol-related social consequences. Heavy drug use is defined as the unprescribed use of at least one psychoactive substance on a weekly basis or more often during the prior year, including: cocaine or crack, amphetamines or crank, sedatives, heroin, other opiates, marijuana or hashish, and psychedelics (see: Weisner and Schmidt, 1993). Health status is self-assessed by recipients using the RAND Corporation’s standard measure, which has been used in prior research on diverse populations (e.g., Phillips et al., 2000)). Those who assessed their health as “poor” are classified as such.

We also compare the AFDC and TANF caseloads on the prevalence of three related problems that significantly impact health status: violent victimization, criminal involvement and homelessness. Recipients are classified as victims of violence if they report being physically or sexually assaulted during the prior year using a standard measure from the interpersonal violence literature (see: Lown et al., 2006; Salomon et al., 2002; Tolman et al., 2001)). Literal homelessness is defined as having been without a regular home at some time during the prior year, excluding episodes of being “doubled up” with relatives or friends at the time. Criminal involvement is designated for recipients who reported any time in jail, prison or on probation during the past year.

Data Analysis

The quantitative analysis begins with cross-tabulations and the Mantel-Haenszel chi-square tests to evaluate changes in the AFDC/TANF caseload surveys, comparing rates of health-related problems and their co-occurrence in 1989 versus 2001. The 1989 and 2001 caseloads are further compared using a multinomial logistic regression analysis that predicts any one and any two or more health-related problems (versus no problems) while controlling for changes over time in the demographic, employment and family characteristics of the welfare caseload. We complete the quantitative analysis by investigating whether health-related problems could cumulatively impact the composition of the caseload over the course of welfare reform implementation. Here, we draw upon the three-year prospective data on TANF recipients to examine whether health-related problems increase the likelihood of remaining on aid over time, thereby giving rise to the gradual “silting up” of more disabled recipients in the caseload over time. Here, we examine the month-by-month cumulative probability of aid receipt for those with 2 more problems (versus 0 or 1 problem) using a linear-by-linear association to assess statistical differences. Because there were few differences in time on aid for individuals with no problems and one health-related problem, they are grouped in this analysis.

We position this quantitative analysis within the context of a qualitative data on how TANF workers manage the health-related needs of recipients on a day-to-day basis in the post-welfare era. Data collection, coding and analysis for this portion of the study were performed in an inductive and iterative manner. Interviews were taped and professionally transcribed. Transcripts and field notes were then coded and analyzed using the qualitative software program NVivo. The initial coding was broad, producing over 150 codes. Through discussions between the authors, the most prominent codes were identified and, with additional analysis and discussion, codes were grouped into themes. Once themes were identified, the authors returned to the raw interview data looking for confirmation, modification, or refutation of these themes, and identified relevant direct quotations from respondents.

Methodological Limitations

While our case study approach provides an in-depth view of changes unfolding on the local level, there is a more limited capacity to generalize findings to other welfare systems throughout the country. Under the PRWORA’s policy of devolving federal authority to lower levels of government, there is growing diversity in local welfare systems, making local case studies both more compelling and more difficult to generalize from. To maximize generalizability, we selected a large study site that captured the demographic diversity of the U.S. population, and a site where the welfare policies most relevant to this study are similar to those implemented elsewhere.

Another methodological limitation pertains to our ability to attribute changes over time in the composition of the welfare caseload to welfare reform policies, as opposed to secular changes in the population and economic environment. For example, the strong economy that emerged between our 1989 and 2001 surveys could have affected observed changes in the caseload by affecting job opportunities. To make strong causal claims about the role of welfare policy, one would at minimum require control groups of people in the study site who were eligible for public aid but did not apply. In the absence of such data, our approach is to examine our surveys for evidence of patterns that are consistent with the hypothesized impacts of welfare reform, and to use multivariate techniques to control for potentially confounding changes in demographic and employment characteristics over time.

RESULTS

Quantitative Analysis: Health-Related Problems in the Welfare Caseload

Analyses of the WCLS caseload surveys suggest that prevalences of some, but not all, health-related problems have increased in the aftermath of welfare reform. Table 2 compares the prevalence of seven health-related problems in the study site’s 1989 AFDC caseload and 2001 TANF caseload. In the 2001 sample collected after reform, recipients are much more likely to report psychiatric distress than in 1989. In the 2001 sample, 32 percent of TANF recipients met BSI criteria for psychiatric distress compared to 18 percent in the 1989 sample – an 83 percent increase. There are also significant increases in homelessness, with 21 percent of recipients in 2001 having reported that they had lived on the streets at some time during the prior year, compared to 16 percent in 1989. Rates of drug abuse appear to have also increased at a marginal level of statistical significance, while rates of problem drinking remain virtually the same. The only health-related problem that is less common in the 2001 sample is violent victimization.

TABLE 2.

Prevalence of Problems Among Welfare Recipients, Before and After Welfare Reform (in percents)

Health and Social Problems	AFDC Before reform 1989	TANF After reform 2001	p-valuea
Problem drinking	11.3	11.3	.53
Heavy drug use	15.3	19.2	.07
Psychiatric distress	17.5	32.1	.00
Violent victimization	17.3	12.8	.03
Literal homelessness	16.4	21.0	.04
Criminal justice encounter	12.1	10.7	.28
Poor health	5.8	4.7	.27
Most Common Co-Occurring Problems
Psychiatric distress & literal homelessness	3.5	8.7	.00
Psychiatric distress & heavy drug use	4.0	8.6	.00
Psychiatric distress & violent victimization	6.3	8.1	.17
Heavy drug use & literal homelessness	4.3	6.7	.06
Heavy drug use & problem drinking	5.0	6.0	.32
Number of problems			.06
None	47.9	41.6
One	27.2	27.0
Any two or more	24.9	31.5
Unweighted N	362	691

Note. Results are weighted for sample design and non-response.

^ap-values are for one-tailed tests, except for number of problems which is two-tailed.

As Table 2 further shows, there were no changes between 1989 and 2001 in the proportion of recipients reporting one of the seven health-related problems we examined. However, there were significant changes among those reporting more than one, or “co-occurring,” problems. Among recipients interviewed after welfare reform, nearly one-third (32 percent) report co-occurring problems, compared to one-quarter (25 percent) of those interviewed prior to reform. Psychiatric distress figures prominently among those co-occurring problems on the rise in the welfare caseload, particularly in combination with homelessness and heavy drug use, where rates more than doubled between 1989 and 2001.

We also conducted a multivariate analysis to test whether these changes in health-related problems remain apparent after controlling for changes in the demographic and employment characteristics of aid recipients between 1989 and 2001. Table 3 presents the results of a multinomial logistic regression analysis where the three-way dependent variable reflects one health-related problem (versus none), and two or more co-occurring problems (versus none). The results suggest that there have been significant increases over time in aid recipients with two or more problems after controlling for other changes in caseload demography. This result is shown by the relative risk ratios (RRRs) for survey year. Other factors being equal, there is no statistically significant change between 1989 and 2001 in recipients with only one health-related problem, but a significant increase of those with co-occurring problems. The regression results also provide some insights into the characteristics of aid recipients who are most prone to co-occurring problems. Men, unmarried recipients, those with a previous history of aid receipt, and recipients with no recent work history appear to be the most likely to experience 2 or more problems.

TABLE 3.

Multinomial Logistic Regression Results Predicting Number of Problems

Demographic characteristics	Number of Problems (vs no problems)
	Any One		Any Two or More
	OR	p-value	OR	p-value
Year
1989	–		–
2001	1.16	.35	1.60	.00
Age	1.00	.98	1.00	.78
Gender (Male)	3.03	.00	4.73	.00
Race/ethnicity
White	–		–
Black	.74	.11	.74	.10
Other minority	.76	.15	.76	.17
Married or cohabitating	.93	.68	.54	.00
Previous history of AFDC or TANF	1.31	.10	1.77	.00
Human capital characteristics
Less than high school education	1.27	.16	1.29	.14
No recent work history	.98	.90	1.41	.04
Model X2, df	59.495, 18 **

Note. Results are weighted for sample design and non-response.

It is possible that this evidence of an increased burden of co-occurring health problems is the result of a “blinking light” phenomenon whereby, after welfare reform, disabled recipients would be more likely to apply for aid precisely because agencies are making health-related more available. To rule out this possibility, we examined changes in the number and types of services sought by applicants in our pre- and post-reform samples. On average, both AFDC (in 1989) and TANF (in 2001) applicants hoped to obtain two kinds of services in addition to receiving cash aid. This suggests that expectations of the services available in welfare agencies did not significantly change under reform.

Finally, we draw on the 3-year follow-up of our 2001 TANF sample to examine evidence of a progressive increase in recipients with health-related problems over the course of welfare reform implementation, from 2001–4. If recipients with health-related problems take longer to leave welfare than other recipients, the likely consequence would be a disproportionate rise in the burden of health-related problems in the TANF caseload over time. Figure 2 examines the extent to which health-related problems are associated with the time recipients remain on welfare between 2001 and 2004. The figure shows cumulative distributions of total months on aid among recipients who entered TANF with 2 or more health-related problems at baseline, compared to recipients with one or no problems. Statistically significant differences in the slope of the curves suggest that, over a 3-year period, aid recipients with two or more health-related problems accrued significantly more months on aid than other recipients.

FIGURE 2.

Months on Aid for TANF Recipients With and Without

Qualitative Analysis: TANF Worker’s Response to Health-Related Needs

Examination of the qualitative data on TANF workers generally supported findings from our quantitative analysis. A majority of workers (80 percent) had worked in the county welfare administration for long enough to witness the implementation of welfare reform firsthand. Particularly among these workers, there was a strong perception that, over time, the composition of the caseload was changing to produce a growing burden of health-related barriers to work. Some workers thus compared the older breed of “unmotivated” recipients from before welfare reform with a newer crop of “dysfunctional” recipients after reform. As one specialty caseworker put it, “These people are not lazy, they just can’t do it.” The problem today, it seemed, was no longer dealing with recipients who chose not to work, but rather, with those who were not able to work. One worker in the county’s job search program thus reflected on the need to adapt to progressive changes in caseload composition:

When welfare reform first started we got all the gung-ho people who had been on assistance – [those] who wanted to work but weren’t given the supportive services that enabled them to work. So, for the first probably year or two, we had the cream of the crop, those people who were ready and willing and really didn’t have a lot of life issues. Just something happened to them…. In the last couple of years, we’ve really had to switch how we’re doing things…. We’ve had to change with the population change that we’re seeing now.

TANF workers frequently commented on the growing presence of psychiatric problems in the welfare caseload. However, on an almost uniform basis, they were quick to add that the mental health problems seldom occurred on their own. Many workers spoke of “hard to employ” recipients with multiple, inter-related health and social problems, including substance abuse, domestic violence, physical disabilities and learning disabilities. As one caseworker described these recipients:

Well, they’re dysfunctional usually because they have, like I said, childhood trauma, sexual molest[ation] is very common, rape in their battered woman relationship, severe learning disabilities – all of these things combined – and [it] has led them into substance abuse and now depression and other issues, physical injuries, etc.

The growing burden of health-related problems posed a variety of concerns and dilemmas for welfare workers. The most common was that this change in caseload composition compromised their ability to meet federal workforce participation requirements. Some workers expressed this in terms of concerns for the recipients themselves:

[T]hey’re never really going to be self-sufficient, that little group, which is the majority – a huge proportion – of our recipients, the ones who are on aid right now, you know, are at the bottom of the barrel, I worry about those people. They’re very, very marginal….

Others seemed more concerned about unrealistically high expectations on the workers themselves, who lacked adequate resources for moving adequate numbers of their clients into jobs:

Well, there aren’t the resources out there for the number of people that are severely disabled. There just aren’t…. So that’s my gripe. That’s my issue is we don’t have the resources to help these people and the county is giving lip service to ‘Yes, we’re helping these folks. Yes, we’re giving them a chance at a better life.’

Even if adequate resources were available, most TANF workers seemed dubious about whether their more disabled recipients would be able to take advantage of them. Workers at all levels of the system expressed frustrations with “compliance” by recipients, who “don’t always cooperate or they’re not even functional enough to follow through.” Simple tasks, such as making appointments or keeping the proper forms on file, proved impossible for a growing number of recipients. In the end, some seemed too dysfunctional to keep up with the simple administrative demands for staying on aid.

There’s frustration with some of the simplest things. I mean, I think probably, oh gee, I can’t even tell you how many recipients will call and they’ll say they have an appointment. ‘I have an appointment, I don’t know with who and I don’t know what time.’ I said, ‘Well, do you have the…’ (It’s like give me something to work [with]!) ‘Do you have the appointment slip?’ ‘No, I lost it.’ It’s like that. ‘I’ve lost it, I’ve lost the slip, I lost this, I lost….’ And it’s like, when you think of it, it’s like, well, how many times do we lose our ATM card?

Workers in the county’s job search programs expressed particular concerns about some recipients’ capacity to comply with and benefit from their programs. The job search process simply seemed unrealistic for many recipients. In response, some workers described how they had adjusted program curricula to better meet the declining functional capacities of their clientele. This meant downplaying job-search skills, such as resume writing and interviewing, and emphasizing “life skills,” that is, training in simple daily tasks such as time management, self-esteem building, healthy eating and anger control. As one worker explained,

In the last couple of years we’ve really had to switch how we’re doing things, as far as teaching life skills to the class, versus just [the] curriculum on how to get a job. So we’ve had to change with the population change that we’re seeing now.

Workers also described difficulties brokering health services in the wider community to meet the increased needs of recipients. Under the “work first” model, such services were not usually made available until a recipient had failed to obtain a job, often multiple times. At this point, the case was usually transferred to the Assessment and Intensive Services Divisions where specialty workers had three months to develop a welfare-to-work plan that could include “intensive services,” such as mental health care, addiction treatment or services for domestic violence. Workers in these divisions, however, described a variety of administrative barriers to successfully engaging recipients with health services in the community. Most notable was the fact that the referral process was extremely time-consuming and required considerable follow-through by multiple workers as well as the recipient. Workers first needed to identify the recipient’s needs – a process that seldom met with high rates of success (for further discussion of this problem, see: Henderson et al., 2006). Recipients were then referred assessment specialists from the county health department. Assuming the recipient was deemed eligible for services, a formal referral required authorizations from higher-ups in the welfare department, a process that could take several more weeks to complete. This step was followed by standard diagnostic tests required for final determination of the type of services needed. Recipients approved for intensive services were finally provided with a referral to an outside agency for services, which assumed that there were treatment slots available.

Given the complexity of these procedures, it is not surprising that TANF workers described a variety of ways that this process could run amok. Recipients often failed to complete the needed paperwork or dropped out of the application process before the necessary steps had been completed. Alternatively, after completing all necessary steps to secure a formal referral, programs in the community might be full. After being wait-listed, many recipients lost interest in pursuing services, failed to attend services once the slot came open or were later rejected for failure to comply with program rules.

For those recipients who failed to qualify for needed services, or who failed to comply with treatment or respond to rehabilitation, there were three ultimate outcomes. If the recipient obtained intensive services and complied with treatment yet still failed to obtain stable work, it became possible to move the case off the TANF rolls and onto permanent disability in the federal Supplemental Security Income (SSI) program. Obtaining documentation for SSI eligibility was, however, a lengthy and demanding process fraught with many of the same problems as those just described. A second possibility was permanent exemption from welfare time limits. This alternative was discouraged by welfare department managers and therefore infrequently used in our study site, because it meant that the state government might have to assume the long-term costs of supporting the recipient. The third and final possibility involved the recipient’s permanent departure from aid after reaching welfare reform’s mandatory 5-year time limit on aid receipt. The growing frequency of this outcome concerned some staff, given that their more disabled recipients had rather dim prospects for economic self-sufficiency:

I am starting to see some of those people timing out on welfare and they’re never really going to be self-sufficient…. Even if I can get those people jobs, they can’t keep them because they’re so disorganized. They just can’t organize their lives well enough to keep a job…. Those people are falling through the cracks.

DISCUSSION

Welfare reform has benefited many aid recipients who have left aid for stable jobs. Even so, our data lend credence to the concern that a growing proportion of those left behind face significant health-related barriers to work. We approached this issue using several different sources of data. And all of these data proved consistent with claims that the overall health status of the population on aid is declining over time. Comparing caseloads before and after welfare reform, we found significant increases in most, but not all, health-related problems. This included an 83-percent increase in the rate of psychiatric distress, a large increase in the rate of homelessness, a marginally higher rate of illicit drug abuse, and a pronounced increase in rates of multiple, or co-occurring, problems. Our prospective data on TANF recipients further showed that the presence of co-occurring health problems significantly increases the amount of time that recipients spend on aid. This is consistent with the claim that the composition of the caseload is changing because, in the current system, recipients with health-related problems have difficulty moving from welfare for work, thereby leading to a gradual “silting up” of the most disabled people in the caseload over time.

Paradoxically, the increases we observed in health-related problems appear to have taken place within a context where the aid recipient population has otherwise grown more work-ready, according to standard labor market indicators. In the caseloads we observed, the overall burden of health-related problems declined while at the same time, levels of recent work experience increased. As noted at the outset, some researchers and policymakers have relied on economic indicators, such as recent work experience, to show that the welfare population is not necessarily growing more disadvantaged over time (e.g., Moffitt, 2003). Indeed, data from many local and state systems also suggest that employment indicators have improved over the course of implementing welfare reform (Holcomb and Martinson, 2002; Moffitt, 2003; Zedlewski, 2002; Zedlewski and Alderson, 2001; Zedlewski et al., 2003). The results of this case study suggest, however, that it is possible for employment indicators to improve even as indicators of health impairments and social disadvantages worsen. This suggests that the health impairments of aid recipients and their work readiness are not necessarily one in the same. Researchers should not view these indicators as interchangeable, and policymakers may find that these problems call for rather different solutions.

Results from our qualitative research on welfare providers were consistent with the quantitative results, and also shed light on some emerging policy dilemmas that may arise from the changing composition of welfare caseloads. As noted above, some policymakers have argued that claims of a shift in caseload composition may be simply be a perception among welfare providers. They point out that welfare reform has led providers to become more closely “engaged” with aid recipients’ struggles to find work. Consequently, workers may be more likely to notice any dysfunctions that stand in the way of employment, even in the absence of real changes in overall caseload composition. Our data, however, suggest that providers’ perceptions of declining health status in their caseloads can be independently confirmed by survey data on the recipients they serve, at least in the county welfare system we studied. We also observed a high degree of engagement and awareness of health-related problems on the part of many welfare workers. But the preoccupation seemed more to grow out of the day-to-day complexities and frustrations inherent in trying to move these recipients into the workforce and to broker health care services on their behalf.

Welfare providers throughout the study site faced a common set of problems in attempts to respond to growing health-related needs in their caseloads. On the one hand, these problems stemmed from what workers perceived to be the unrealistic expectation of policymakers, namely, that most recipients could benefit from job search programs and become stably employed. On the other hand, these problems stemmed from the expectations that workers had of their clients, namely, that most should or could comply with complex administrative requirements for performing in work programs and obtaining other services.

These problems seemed to point to several ways that the “work first” system fell short of meeting the needs of its more disabled clientele. Workers found that existing job search programs were too demanding for many recipients, and some took it upon themselves to rescale expectations by teaching basic “life skills” to those who had little chance of finding jobs. Workers also pointed to the fact that, under “work first,” health care needs were seldom addressed until after the recipient had “proven” him or herself truly disabled after multiple failures to find work. However, once workers were in a position to broker health services for their recipients, a new set of problems emerged that made it difficult to succeed. Here, what we observed was a mismatch between the drawn-out, complex administrative procedures required to determine eligibility and link the client with health services in the community, and the recipient’s capacity to comply with these procedures and wait for paperwork completion and an open slot in services. Bureaucratic hurdles of this nature also hampered workers’ capacity to find permanent solutions by moving disabled recipients onto the federal SSI program or by obtaining a permanent “hardship” exemption. These administrative dysfunctions left some workers concerned that growing numbers of disabled recipients were permanently leaving aid due to time limits with limited capacity for self-sufficiency.

POLICY IMPLICATIONS

If other welfare systems are like the one we studied, then the declining health status of recipients is likely to pose several new dilemmas with which policymakers will need to grapple. Changes in caseload composition are, for one thing, likely to make it more and more difficult for states to meet federal workforce participation requirements. Unfortunately, the drift in welfare policy since passage of the PRWORA has been towards increasing, rather than decreasing, the proportion of recipients that states are required to move into labor force. At the broadest level, then, federal policy may be poorly attuned to changes on the ground in welfare caseloads. Despite these political realities, there may be a growing need for federal policymakers to rethink workforce participation requirements, to increase the federal threshold for exemption beyond 20 percent, and to provide more allowances for treatment in-lieu-of work requirements so as to better accommodate the changing health status of aid recipients.

With respects to policymaking at the state and local levels, our findings point to need to consider the possibility that the “work first” approach may not be working equally well for all recipients. As demonstrated by our data, job search programs may place unrealistic expectations on aid recipients with co-occurring problems and may unnecessarily postpone the time it takes to identify health care needs and link recipients with services. States and localities that routinely screen all applicants for disabilities may, to some extent, circumvent these problems and delays (Henderson, et al., 2006). Additionally, a few innovative state and local systems have experimented with specialized “welfare-to-work” programs focused on post-employment services (Allen and Kirby, 2000), intensive case management and other approaches that bundle health and social services (Morgenstern et al., 2001). While often costly on a per-case basis, such efforts have proven effective (United States General Accounting Office, 2001; Zedlewski and Loprest, 2001).

More fluid, efficient systems for linking aid recipients with health care services may require formal policy interventions that merge or integrate welfare and health care systems at the local level. Admittedly, the barriers to integrating local health and welfare services are substantial and demand cooperation between providers with differing financial incentives, work practices, and professional ideologies. The growing prevalence of co-occurring conditions may, however, prove to be a common ground for welfare and behavioral health service providers. In mental health and substance abuse circles, there has been growing attention to the high prevalence of co-occurring conditions, and much careful study of how providers can resolve the dilemmas that arise, such as the lack of insurance coverage, the lack of treatment supply, fragmentation in the organization and delivery of health services, and philosophical disagreements among providers about how best to treat specific conditions (Grant et al., 2004; Hasin et al., 1990; Kessler et al., 1996; Kessler et al., 1995; D’Aunno et al., 1991; Ridgely et al., 1990; Rosenheck and Lam, 1997; Rosenheck et al., 2001; Schmidt, 1991). To the extent that welfare providers will increasingly find themselves in the position of serving recipients with co-occurring conditions, they are likely to confront challenges similar to those faced by behavioral health providers in their communities, giving impetus to collective efforts to better integrate services.