Ensuring the stability of welfare-to-work exits: The importance of recipient knowledge about work incentives.

Anderson, Steven G.

Research on designing incentives to make low-wage work a more viable alternative to public assistance has played an important role in the development of state Temporary Assistance for Needy Families (TANF) programs (Bane & Ellwood, 1994; Edin & Lein, 1997; Jencks, 1992). This article considers a related issue that generally has been overlooked--whether welfare recipients clearly understand the incentives that have been developed to encourage work.

Respondent knowledge about work incentives has become increasingly important under TANF. With the imposition of five-year time limits and stricter sanctions for not fulfilling work requirements, failure to use available work supports can have serious effects on families. To explore how well recipients understand work-related benefits, I interviewed welfare recipients about a wide range of work incentives and support services in a pre-TANF waiver program in Michigan. The findings demonstrate that most respondents lacked knowledge about these incentives and thus often underestimated the benefits available from working. Furthermore, misunderstandings about incentives negatively influenced respondent perceptions about the welfare system, which may undercut future responsiveness to incentives.

This article discusses the implications of these findings for the structuring and implementation of TANF work incentive policies and makes the argument that benefit simplification could improve the effectiveness of work incentives and strategies for improving the dissemination of information about work incentives. Also recommended are the development of post-employment service strategies to increase access to public supports for those who leave TANF.

Literature Review

No research has analyzed welfare recipients' knowledge about the full range of work-related economic incentives available to them. However, studies have shown that welfare recipients lack knowledge about both transitional Medicaid and child care benefits (Hardina & Carley, 1997; Rofuth & Weiss 1991). In addition, Hardina and Carley and Sarri and colleagues (1984) found that recipients of Aid to Families with Dependent Children (AFDC) generally did not understand basic AFDC eligibility requirements. Other studies have shown low-income people more generally underuse and lack relevant information about the earned income tax credit (EITC), food stamps, state tax incentives, and other programs (Coe, 1983; Katz, Gutek, Kahn, & Barton 1975; Olson & Davis, 1994; Swanson, 1975).

Studies of the welfare system and bureaucratic relations offer possible reasons for knowledge deficits. Researchers have argued that the complexity and fragmentation of the welfare system confuses welfare recipients (Bane & Ellwood, 1994; Ellwood & Adams, 1990; Nightingale, 1989). Frequent program changes may exacerbate this complexity, as both recipients and social services personnel may not learn about relevant changes (Anderson, 1998; Bane & Ellwood). Furthermore, different programs may offer conflicting incentives and have varying eligibility rules, benefit determination criteria, and work requirements (Anderson).

Variations in interactions between caseworkers and clients also may lead to differential service access and knowledge about incentives (Brodkin, 1986, 1997; Lipsky, 1980). Such inconsistencies may have grown in recent years, as staffing cutbacks, caseworker deprofessionalization, and increasingly complex regulations have made thorough information dissemination difficult (Fabricant & Burghardt, 1992). Consequently, recipients often attribute their lack of knowledge about benefits to caseworker shortcomings in providing information (Coalition on Human Needs, 1988; Southport Institute for Policy Analysis, 1992).

Study Methods

To test how well welfare recipients understood work incentives, I conducted personal interviews with AFDC recipients in Lansing, Michigan, between September 1995 and April 1996. At that time, Michigan was implementing a welfare reform program often portrayed as a national model in policy debates preceding passage of the Personal Responsibility and Work Opportunity Reconciliation Act of 1996 (PRWORA) (P.L. 104193). Nearly all Michigan recipients were required either to be working, participating in schooling or a work training program, or volunteering in community services for at least 20 hours per week. In addition, the state had implemented simplified earnings disregards and support services designed to encourage work. Caseworkers were required to inform recipients about these changes in 1994.

All study respondents received services through the same local office of the Michigan Department of Social Services (MDSS). This was intended to minimize variations resulting from differing information dissemination and administrative practices. Although using one geographic area limits study generalizations, the study was exploratory and not intended to generate population estimates.

The sample consisted of 60 unmarried women who had received AFDC in the month before being interviewed. Thirty of these women were currently working or had worked within the past six months, whereas the remaining 30 had not worked in the past six months. This sampling choice was based on the expectation that those with recent work experiences would be more likely to have knowledge about work incentives than nonworking recipients would.

Because of concerns that MDSS involvement might inhibit the openness of response, two alternative recruiting methods were used. First, local social services agencies, education and training programs, and churches were asked to identify potential participants. Second, study advertisements were posted in grocery stores, check cashing exchanges, and second-hand stores in low-income areas. Brief telephone screening interviews ensured that people who called in response met the criteria for inclusion. In addition, women were screened out if they had learned of the study from someone already interviewed to minimize possible snowball effects.

Although the resulting sample is not random, the techniques used appeared to be successful in minimizing systematic biases. Respondents were referred from many different sources, including employment and training programs (25.0 percent), bulletin boards in private establishments (23.3 percent), social services agencies (20.0 percent), day care or Head Start programs (11.7 percent), adult education programs (8.3 percent), word of mouth from nonrespondents (8.3 percent), and church or neighborhood groups (3.3 percent). (Selected characteristics of the respondents are presented in Table 1).

I conducted interviews lasting from 1 1/2 to two hours. Most interviews took place in the homes of respondents, and all but seven were audio-taped. Respondents were paid $25 in cash at the beginning of the interview to build rapport and send a clear signal that payment was unrelated to any socially desirable response set.

Interview questions assessed how knowledgeable respondents were about work incentives available both to those who combined welfare with work and those who left welfare for work. Respondents were questioned about four types of income incentives:

1. Michigan disregarded the first $200 and 20 percent of additional earnings when calculating monthly AFDC benefits.

2. 20 percent of all monthly earnings were excluded in food stamps benefit calculations.

3. The federal earned income tax credit (EITC) provided refundable federal tax credits of up to $3,100 for low-income working people.

4. The state homestead property tax credit provided refundable tax credits which increased with earnings.

Questions also measured knowledge about support services. Full Medicaid coverage continued for AFDC recipients who worked part-time, and one-year transitional Medicaid coverage was provided for those leaving AFDC for work. Medicaid coverage for those exiting AFDC also was available indefinitely for most low-income children; poor adults with high medical expenses could qualify for a Medicaid spend-down program. State policies allowed up to $175 per child in child care to be disregarded in both AFDC and food stamps benefit calculations, and day care services were available indefinitely to those who left AFDC for work.

The development of questions on each work incentive followed a similar pattern. First, respondents were asked whether they were familiar with or knew the eligibility requirements for receiving an incentive. Respondents who expressed this minimal familiarity then were asked follow-up questions to determine how clearly they understood the incentive.

Analysis of whether respondents were knowledgeable about each incentive first involved developing criteria for what constituted a correct response. These criteria were developed from official program eligibility and benefit determination documents. Interview responses then were compared against these criteria to determine knowledge levels.

Because many of the incentives studied are complex and contain eligibility nuances, determining the level of understanding necessary for a respondent to be considered knowledgeable was an important concern. The general intent was to determine only if respondents knew about the availability of an incentive and understood basic eligibility criteria. Nonetheless, useful distinctions emerged during data analysis between this basic knowledge level and more detailed knowledge about incentive values or other incentive features. For example, a respondent may have known that food stamp benefits would be reduced if she worked, but be unaware of the level of these reductions. Although resulting coding distinctions between basic and detailed knowledge often are crude, they are helpful in clarifying the extent of respondent knowledge and so are included in the following discussions.

Qualitative techniques were used to analyze reasons for respondent confusion or misunderstanding about each incentive. Initial codes were assigned to these open-ended responses apriori based on reviews of interview tapes and transcripts (Grinnell, 1997), and similar codes then were aggregated into more general conceptual codes (Strauss & Corbin, 1990). The resulting coded information in turn was organized both within each incentive and also thematically across all incentives. Because respondent comments offered useful insights about incentives, quotes were transcribed verbatim and are presented as such.

Findings

Knowledge about Incentives

Knowledge varied considerably about different incentives. For example, 90 percent of the 60 respondents knew they could continue receiving Medicaid if they combined welfare with work, whereas fewer than one-half knew about the EITC, the homestead property tax credit, and child care available to welfare leavers (Table 2).

Although respondents frequently demonstrated basic knowledge about incentives, more detailed knowledge was lacking. Except for the $200 AFDC earnings disregard and the continuation of Medicaid eligibility if working part-time, fewer than one-third of the 60 respondents displayed detailed knowledge about each incentive. Medicaid coverage for welfare leavers provides a good example of this pattern of fairly high basic knowledge but lower detailed knowledge. Although 77 percent of respondents indicated that they could obtain some Medicaid coverage if they left welfare for work, only 8 percent understood the extent of coverage that was available.

Given these low levels of knowledge about specific incentives, it is not surprising that respondents rarely demonstrated even basic knowledge about the full range of available work incentives. Only 5 percent of respondents demonstrated basic knowledge about all of the six incentives available to those who combined welfare with work, and 3 percent knew about all incentives available for those who left welfare for work.

Because some incentives are considered particularly important, I also analyzed respondent knowledge about selected subsets or packages of key incentives. One subset for those who combined welfare with work included the $200 AFDC earnings disregard, the EITC, Medicaid, and child care assistance. Thirty percent of the 60 respondents displayed basic knowledge about all of these incentives, but only 2 percent demonstrated detailed knowledge. For those who left welfare for work, knowledge about a package including the EITC, transitional Medicaid, and assistance with child care expenses was examined. Twenty-three percent of respondents displayed basic knowledge about all three of these incentives, whereas 5 percent exhibited detailed knowledge.

Variations in Knowledge by Respondent Characteristics

One interpretation of these findings may be that respondent knowledge deficits resulted from a lack of interest in working (see, for example, Mead, 1992). However, 95 percent of the 60 respondents had some work experience, and 65 percent had worked in a job lasting at least six months during the preceding three years (Table 1). Furthermore, although the working stratum of respondents had higher levels of knowledge about each work incentive than nonworking respondents, such differences were small. Working respondents demonstrated statistically significant higher levels of knowledge only about child care incentives. More common, like their nonworking counterparts, those who worked demonstrated striking gaps in knowledge about basic incentives.

Knowledge comparisons also were made between high school graduates and nongraduates, those who had cycled on and off welfare ("recyclers"), and those who had not, long-term and short-term welfare recipients, and African American and white recipients. Differences again generally were inconsistent and not statistically significant. The largest differences occurred between white recipients and African American recipients, in that white people were significantly more knowledgeable about three key incentives--$200 disregard, EITC, and child care. These differences cannot be attributed to differences in formal education, as educational levels were comparable between these groups.

Because of the small and nonrandom nature of this sample, these between-group comparisons should be interpreted cautiously. The reasons for differences in knowledge between white and African American recipients are not apparent from this data and merit further research. Other differences based on recipient characteristics also may emerge with a larger sample. However, the lack of significant differences and low levels of knowledge across subgroups suggest that more general reasons may explain lack of knowledge about incentives.

Lack of Knowledge and Incentive Complexity

One alternative explanation for lack of knowledge is that incentives often are complex and may confuse recipients about the benefits to which they are entitled. Comments from respondents illustrate how confusion may result, and how this may undercut confidence in work incentives.

Seeing is Believing: How Calculation Complexities Affect Recipient Perceptions. Both the $200 and 20 percent earnings disregard and child care reimbursements for those who combine welfare with work operated through adjustments to the AFDC grant. Whereas such procedures were designed to enhance administrative efficiency, in practice they often confused respondents about the benefits actually received.

Table 3 presents a budgeting example for a typical work situation to demonstrate why application of the $200 and 20 percent disregard policy was confusing. The calculations under the first two headings show how this disregard actually was determined by caseworkers and how this affected grant calculations. The information presented under the third heading shows the effects of the disregards on recipients.

Although Table 3 demonstrates that recipients in this situation indeed received more income than if they did not work ($694.10 versus $489.00), even in this simple case the relationship between earnings and grants levels is not clear. Rather, as the lower portion of the table shows, recipients kept all after-tax earnings and also received reduced AFDC checks as a result of the effect of earnings. No information indicating how earnings affected the AFDC grant level was included with the AFDC check. Interpretations of how the $200 and 20 percent policy was applied were further complicated by a procedure known as retrospective budgeting, under which grant calculations for the current month were based on earnings from two months earlier.

Thus, although recipients were told that they would be allowed to keep the first $200 in earnings and 20 percent of earnings thereafter, they did not see how this was applied in determining their benefits. These elements of calculation complexity explain why even working recipients did not always understand the $200 earnings disregard. The reply of one working respondent to a question on how working would affect her AFDC grant is illustrative:

You know, I don't understand.... All I do is just fill out the forms [earnings reports] and send them in ... and then every month when they send me this [her monthly grant statement]. I just pray that it will be enough.... They never tell why it goes up or down.

These calculation complexities contributed to an unexpected finding: More than one-fourth of the respondents who knew about the $200 disregard did not believe that it was actually applied in their grant calculations. Such disbelief generally resulted from direct experiences in which the AFDC grant checks received while working did not have a clear relationship to the $200 disregard. Unfortunately, this sometimes appeared to lead to broader distrust of caseworkers and the welfare bureaucracy, as is reflected by the following respondent comment:

Well, they come in and they make it sound like... "It will benefit you, you'll see the benefits--it will benefit you more by going back to work than not working." And they lied. They sit down and break it down and they show you how the first $200 is yours--they don't deduct anything out of that, then 20 percent of the rest of your income is yours to keep. And you know, I never saw it.

If a recipient incurred child care expenses, grant calculations were further complicated. Although the basic child care disregard of $175 per child ($200 for child under two) again seems clear, the calculation process mirrored that presented in Table 3 for the $200 and 20 percent disregard. That is, documented day care expenses up to the limits were deducted from countable income when calculating grants, and details of how this calculation process worked were not provided to recipients. As a result, respondents did not understand and often questioned how the child care expenses they turned in to caseworkers affected their grants. As two respondents said:

DSS has a criteria you have to follow to get the [day care] payment.... And then they say they reimburse you--they put it back in your check. But sometimes you have to pay for your day care and sometimes they pay ... it gets so confusing ... they're trying to budget it but they're not budgeting it right.

Confusion Because of Multiple Coverage Options: Medicaid Coverage for Those Leaving Welfare. As discussed earlier, those who left welfare for work could receive continued Medicaid coverage through several different options, each of which provided different levels of coverage or co-payment requirements. This mix of Medicaid coverage options was reflected in respondent perceptions. Encouragingly, 77 percent of the 60 said that some Medicaid coverage would be available. However, respondents were confused about the different coverage options and consequently were uncertain about who in their family would be covered or whether they would face co-payments. Sixteen substantively different response categories emerged from an analysis of open-ended responses on perceived Medicaid coverage. These categories reflected a wide range of correct and incorrect information concerning the criteria for continued Medicaid eligibility. For example, two respondents knew that transitional Medicaid coverage would be available for on e year, but they incorrectly believed that the coverage was based on income.

This lack of accurate information about coverage options typically resulted in respondents underestimating the extent of available Medicaid coverage. Only 35 percent of the 60 respondents were aware of the transitional Medicaid coverage option, which offered the most complete coverage for those leaving welfare for work. At the same time, 60 percent of respondents indicated that losing Medicaid benefits would be a problem if they left welfare for work.

This suggests that, whereas respondents had assimilated information from various sources about medical coverage, the details often were vague. Unfortunately, such details are critical in the signals recipients receive about the value of working, and the complexity that resulted from having several different categorical medical programs confused these signals.

Benefits through Other Agencies: The Earned Income Tax Credit. Although the EITC is the most important income transfer system for poor working people, only two-thirds of study respondents had ever heard of it and only 30 percent could clearly describe the criteria for receiving it. This lack of knowledge derives partially from the fact that the EITC and other tax credits are not administered by human services agencies. Rather, low-income people must pursue credits on their own when they file income taxes. Furthermore, many low-income people do not have sufficient incomes to owe income taxes, so they may not file taxes and hence not learn about the availability of this refundable credit.

Because the tax system is complex, many recipients turn to private tax preparers or others to complete their taxes. Although the use of skilled preparers should increase the probability that welfare recipients receive the EITC, it paradoxically does not always result in recipients understanding the incentive. For example, more than one-fourth of the 60 respondents who had received the EITC did not understand the basis for having received it, and all of these respondents had their returns prepared by someone else. As one respondent who had her taxes prepared by a local agency said: I just do my taxes and they give me the money. I have no idea what I'm getting the money for.

Discussion

The most obvious implication of these findings is that large numbers of welfare recipients underestimate the benefits they can receive while working. At a time when public policies featuring work incentives, nonwork sanctions, and time limits on welfare receipt are being implemented nationwide, this may result in a failure to gain access to available work supports critical to the financial wellbeing of low-income families. It also may inhibit recipients from pursuing otherwise desirable work options, or cause them to leave jobs because of inadequate supports.

Because the incentives studied were objectively complex even to the informed policy analyst, respondents' lack of knowledge and confusion seemed reasonable. Therefore, one need not gravitate to negative assumptions about recipient motivations or capabilities to interpret lack of knowledge. Rather, complex incentives may be viewed as imposing high information costs on prospective beneficiaries. This not only suggests that considerable effort may be needed to understand various incentives, but also that errors in interpretation may be common.

A more subtle but important implication concerns the broader negative perceptions that may emerge when recipient experiences with incentives differ from expectations. Limited or inaccurate information sometimes creates unrealistic expectations that subsequently are frustrated by experience. The question arises as to whether recipients who experience such frustrations are less likely to pursue similar options in the future or develop more generalized negative opinions about the welfare system and welfare bureaucracies. Discussions with respondents indicated that this may be a problem, but further research in this regard is needed.

Study findings have clear implications for states as they develop and revise TANF programs. Perhaps the most basic lesson is that even improved incentives will often be misunderstood if not accompanied by careful implementation strategies. Although many approaches are consistent with this general proviso, I present three that seem promising.

First, efforts should be made to reduce the complexity associated both with calculating and obtaining incentives; this involves simplifying eligibility rules and benefit calculations, as well as minimizing confusing interaction effects between incentives. For example, the earnings disregard levels for the TANF and food stamp programs could be better integrated, as could allowable values for automobiles and other work-related assets. Welfare policies often are complex for good reasons, including the targeting of limited resources. However, the costs of policy variations in terms of potential misunderstanding and confusion also should be considered.

Second, more creative approaches are needed for presenting information to recipients. Recipients typically receive information on work incentives, but often it is so general that making a situation specific interpretation is difficult. One possibility would be to present recipients with comparisons of the net income and services available under various scenarios, such as welfare only, half-time work, and full-time work. Case managers could be trained to present this information to recipients in conjunction with the development of work activity and service plans and then to follow-up to ensure that available supports are obtained. Research could help determine the most effective timing, contexts, and presenters for conveying this information.

Finally, more attention should be given to developing postemployment strategies for recipients who obtain work. Public policy efforts have focused on getting recipients to leave welfare, but then have not followed up well with those who have left. Given that low-income job situations are tenuous and that welfare recycling is common (Bane & Ellwood, 1994; Harris, 1993, 1996), postemployment services may help reinforce the work situation (Herr, Wagner, & Halpern, 1996; Wagner, Herr, Chang, & Brooks, 1998). Specific strategies could include using exit interviews when recipients leave welfare to provide information on how to obtain work-related benefits. Follow-up calls or interviews soon after the person started work also would be helpful, as benefits then could be presented and discussed in a more specific context.

Because both caseworker time and agency resources are limited, targeting caseworker efforts on clients likely to have difficulty obtaining work benefits may be advisable. Because earlier research has shown that those with low education are likely to return to welfare (Harris, 1996), targeting such individuals is one possibility. Further research also could clarify if other characteristics appear to be important in TANF recycling and knowledge about benefits.

Nonetheless, none of these strategies appears overly costly, and all seem consistent with ongoing policy discussions regarding integrating welfare services, changing the culture of welfare offices, and creating a more active role for case managers in encouraging work. Costs should be justified if the strategies prove effective in reinforcing the value of work by ensuring that low-income workers receive any benefits to which they are entitled.

Table 1

Selected Characteristics of AFDC Recipients in Lansing, Michigan (N =
60)

Characteristic Median

Age 30.0
Highest grade of school completed 11.0
Years on AFDC as adult 5.3
Number of children living with 2.0
Characteristic %

Never married 63.3
Black 46.7
Hispanic 18.3
Received diploma or GED 53.3
Recyclers (ever stopped receiving AFDC) 48.3
Ever worked 95.0
Worked in past 3 years 65.0
Table 2

Knowledge about Incentives Available to AFDC Recipients Who Combine
Welfare with Work or Leave Welfare for Work (N = 60)

 Basic Detailed
Incentive n % n %

Welfare-work combination
 AFDC earnings disregard 48 80.0 -- --
 $200 -- -- 37 61.7
 20 percent -- -- 12 20.0
 Food stamp earnings disregard 45 75.0 0 0.0
 Earned income tax credit 24 40.0 18 30.0
 Homestead property tax credit 11 18.3 8 13.3
 Medicaid 54 90.0 54 90.0
 Child care 42 70.0 1 1.7
Leaving welfare for work
 Medicaid 46 76.7 5 8.3
 Child care 22 36.7 10 16.7
 Food Stamp continued 40 66.7 2 3.3
 eligibility

NOTE: -- = not applicable
Table 3

$200 and 20 Percent Calculations for AFDC Recipients with Two Children
Working Half-Time at the Minimum Wage (Assuming No Child Care Expenses):
1996


Budgetable or countable income
calculation

Gross earnings (20 x 4.3 x $4.25) $ 365.50
-$200 earnings disregard (200.00)
-20% earnings disregard (33.10
Total budgetable income $ 132.40

AFDC grant calculation

Grant if no other income $ 489.00
-Total budgetable income 132.40
AFDC grant received 356.60

Income actually received by recipients

Gross earnings $ 365.50
-Estimated taxes withheld (28.00)
Net earnings in paycheck 337.50
AFDC grant 356.60
Total income received 694.10

Original manuscript received March 16, 1999

Final revision received August 12, 1999

Accepted November 1, 1999

References

Anderson, S. G. (1998). The knowledge of welfare recipients about work incentives. Unpublished doctoral dissertation, University of Michigan, Ann Arbor.

Bane, M. J., & Ellwood, D. T. (1994). Welfare realities: from rhetoric to reform. Cambridge, MA: Harvard University Press.

Brodkin, E. Z. (1986). The false promise of administrative reform: implementing quality reform in welfare. Philadelphia: Temple University Press.

Brodkin, E. Z. (1997). Inside the welfare contract: Discretion and accountability in state welfare administration. Social Service Review 71, 1-33.

Coalition on Human Needs. (1988). How the poor would remedy poverty. Washington, DC: Author.

Coe, R. (1983). Participation in the Food Stamp program, 1979. In G. J. Duncan & J. N. Morgon (Eds.), Five thousand American families--Patterns of economic progress (Vol. 10, pp. 121-177). Ann Arbor: University of Michigan, Institute for Social Research.

Edin, K., & Lein, L. (1997). Making ends meet. New York: Russell Sage Foundation.

Ellwood, D. T., & Adams, E. K. (1990). Medicaid mysteries: Transitional benefits, Medicaid coverage, and welfare exits. Health Care Financing Review (1990 Annual Suppl.), 119-131.

Fabricant, M. B., & Burghardt, S. (1992). The welfare state crisis and the transformation of social service work. Armonk, NY: M. E. Sharpe.

Grinnell, R. M. (1997). Social work research & evaluation: Quantitative and qualitative approaches. Itasca, IL: F. E. Peacock Publishers.

Hardina, D., & Carley, M. (1997). The impact of increased allowable work hours on two-parent families receiving welfare. Social Work Research, 21, 101-109.

Harris, K. M. (1993). Work and welfare among single mothers in poverty. American Journal of Sociology, 99, 317-352.

Harris, K. M. (1996). Life after welfare: Women, work, and repeat dependency. American Sociological Review, 61, 407-426.

Herr, T. S., Wagner, S., & Halpern, R. (1996). Making the shoe fit: Creating a work-prep system for a large and diverse welfare population. Chicago: Erikson Institute.

Jencks, C. (1992). Rethinking social policy: Race, poverty, and the underclass. Cambridge, MA: Harvard University Press.

Katz, D., Gutek B. A., Kahn, R. L., & Barton, E. (1975). Bureaucratic encounters: A pilot study in the evaluation of government services. Ann Arbor, MI: University of Michigan, Institute for Social Research.

Lipsky, M. (1980). Street-level bureaucracy: Dilemmas of the individual in public services. New York: Russell Sage Foundation.

Mead, L. (1992). The new politics of poverty: The non-working poor in America. New York: Basic Books.

Nightingale, D. S. (1989). Institutional effects of reform. In P. H. Cottingham & D. T. Ellwood (Eds.), Welfare policy for the 1990s (pp. 199-218). Cambridge, MA: Harvard University Press.

Olson, L. M., & Davis, A. (1994). The earned income tax credit: View from the street level. Evanston, IL.: Northwestern University, Center for Urban Affairs and Policy Research.

Rofuth, T. W., & Weiss, H. (1991). Extending health care to AFDC recipients who obtain jobs: Results of a demonstration. Health & Social Work 16, 162-169.

Sarri, R. C., Beisel, N., Boulet, J., Butler, A., Churchill, S., Lambert S, Ray, A, Russell, C., & Weber, J. (1984). The impact of federal policy change on AFDC recipients and their families: An evaluation of 1981 OBRA policies in Michigan. Ann Arbor: University of Michigan, Institute for Social Research.

Southport Institute for Policy Analysis. (1992). It's not like they say: Welfare recipients talk about welfare, work, and education. Washington, DC: Author.

Strauss, A., & Corbin, J. (1990). Basics of qualitative research: Grounded theory procedures and techniques. Newbury Park, CA: Sage Publications.

Swanson, R. W. (1975). Utilization of the 1973 Michigan property tax credit by welfare recipients. Lansing: Michigan Department of Social Services.

Wagner, S., Herr, T., Chang, C., & Brooks, D. (1998). Five years of welfare: Too short? Too long? Chicago: Erikson Institute.

Steven G. Anderson, PhD, is assistant professor., School of Social Work, University of Illinois, Urbana-Champaign, 1207 West Oregon Street, Urbana, IL 61801; e-mail: sandersn@uiuc.edu. The author thanks Mary Corcoran and Sheldon Danziger for their advice as he developed this research, and Jill Doner Kagle and Anthony Halter for useful comments on an earlier draft of this article.