How Much More Can They Work?

The nonpartisan Urban Institute publishes studies, reports, and books on timely topics worthy of public consideration. The views expressed are those of the authors and should not be attributed to the Urban Institute, its trustees, or its funders.

The author would like to thank Alison Earle, David Ellwood, Robert Lerman, and Demetra Nightingale for their thoughtful comments and Amy-Ellen Duke, Daniel Dowhan, Stacy Poulos and Robert Solorazano for help with computer programming and research assistance. The author would also like to thank Greg Acs whose ideas on how to examine the transition from bad to good jobs helped inform the analysis for this paper. Any opinions expressed in this paper are those of the author and do not necessarily reflect the views of the Urban Institute. Funding for this project was provided by the Annie E. Casey Foundation.

Executive Summary

The social safety net for low-income families is currently undergoing a radical transformation. For the last 61 years, families with children with limited income and assets were entitled to ongoing cash assistance from the Aid to Families with Dependent Children (AFDC) program. Now, in a rapidly growing number of states, families with limited income or assets can only receive cash assistance if they agree to look for work or work in exchange for the receipt of government assistance. This transformation of the social safety net for low-income families with children began with the implementation of numerous state welfare reform demonstration projects. ¹ The shift was codified into federal law with the passage of the Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) of 1996. PRWORA eliminated the AFDC program and replaced it with a block grant to states to establish a Temporary Assistance for Needy Families (TANF) program. Although PRWORA provides states with considerable flexibility to decide what support they will provide to families in need of assistance, TANF is clearly intended to emphasize short-term, employment-related assistance.

Widespread dissatisfaction with the old AFDC program and the dramatic increase in employment among all women and among mothers with children have generated broad-based support for requiring welfare recipients to work or look for work in exchange for receiving government assistance. In March 1996, 66 percent of all mothers with children--including 55 percent of mothers with children under the age of three--were employed, most of them full-time. Rules regarding the treatment of earned income have made most welfare recipients who find employment ineligible for benefits, making it difficult to compare how the work experiences of women who ever use welfare actually compare to those of women who never turn to the welfare system for support.

To examine how the work experiences of welfare recipients compare to non-recipients, I use data from the National Longitudinal Survey of Youth (NLSY) to construct complete employment and welfare histories for a cohort of women over a ten-year period, from the time they turn age 18 through the year in which they turn 27. To estimate how much more women who ever receive welfare would have worked had they followed the same employment paths as women with similar characteristics who never turned to welfare, I develop a dynamic microsimulation model that predicts women's movement in and out of the labor market over the ten-year period.

This analysis shows that between the ages of 18 and 27, women who ever receive welfare work only about 60 percent as much as women who never receive welfare. Though the majority of welfare recipients work at some point, a substantial fraction spend minimal time in the labor market. About a third of recipients work for less than 25 percent of the ten-year period; a third work between 25 and 50 percent of the time, and the remainder work for more than half of the ten-year period. Welfare recipients who spend the least amount of time in the labor market are primarily recipients with low education and skill levels. Lower rates of employment in any given year and shorter periods of employment when they are employed contribute to welfare recipients spending less time in the labor force than non-recipients.

I estimate that if welfare recipients followed the same employment paths as non-recipients with similar characteristics, they would work 30 percent more than they currently work. The greatest employment gains would be realized by women with low levels of education, low basic skills, and more children, since they currently spend the least amount of time working. Even if welfare recipients were to follow the same employment paths as non-recipients, they would continue to experience substantial periods of joblessness. I estimate that, by age 27, fewer than half of recipients who had not completed high school would be employed steadily. These results suggest that a substantial fraction of welfare recipients are likely to continue to need a safety net to support them during periods when they are unable to find employment.

I. Introduction

The social safety net for low-income families is currently undergoing a radical transformation. For the last 61 years, families with children with limited income or assets were entitled to ongoing cash assistance from the Aid to Families with Dependent Children (AFDC) program. Now, in a rapidly growing number of states, families with limited income or assets can only receive cash assistance if they agree to look for work or work in exchange for the receipt of government assistance (Pavetti, Holcomb, and Duke 1995; General Accounting Office 1997).

This transformation of the social safety net for low-income families with children began with the implementation of numerous state welfare reform demonstration projects. The shift was codified into federal law with the passage of the Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) of 1996. PRWORA eliminated the AFDC program and replaced it with a block grant to states to establish a Temporary Assistance for Needy Families (TANF) program.

Although PRWORA provides states with considerable flexibility to decide what support they will provide to families in need of assistance, TANF is clearly intended to emphasize short-term, employment-related assistance. Families are eligible to receive TANF assistance for only 60 months in their lifetime, increasing the urgency for families to find employment or alternative means of support quickly. TANF recipients are also required to perform community service after receiving assistance for two months and to work once they are determined to be job ready or after receiving assistance for 24 months. States can, however, opt out of the community service requirement by writing a letter to the Secretary of the U.S. Department of Health and Human Services indicating that they intend to do so. To ensure that state TANF programs emphasize work, PRWORA requires states to meet steadily increasing work participation rates to receive their full TANF allocation.

Widespread dissatisfaction with the old AFDC program and the dramatic increase in employment among all women and among mothers with children have generated broad-based support for requiring welfare recipients to work or look for work in exchange for receiving government assistance. In March 1996, 66 percent of all mothers with children--including 55 percent of mothers with children under the age of three--were employed, most of them full-time. The rules regarding welfare receipt make it difficult to come up with comparable employment figures for welfare mothers. Almost by definition, most welfare recipients are not employed, at least at the time they are receiving welfare benefits. Until recent changes in a number of states, after four months of employment recipients who were employed lost welfare benefits dollar for dollar, making it difficult for mothers to work and receive assistance. Administrative data indicate that in any given month, only about 10 percent of the AFDC caseload report income from employment (Zedlewski and Giannarelli 1997). However, when one takes into account the dynamic nature of the AFDC caseload, it becomes clear that work is more common among welfare recipients than this information suggests. Research indicates that between 50 and 70 percent of recipients who first receive welfare leave the welfare rolls within a year, and that between half and two-thirds of these exits occur when a recipient finds employment or otherwise works her way off welfare (Gritz and MaCurdy 1991; Harris 1993; Pavetti 1993).

Given that work is already common among at least some portion of the welfare caseload, it is not obvious how much more welfare recipients would need to work to mirror the work experiences of women in similar circumstances and with similar characteristics. Since a woman's welfare and employment status can change on a monthly basis, to obtain such an estimate one needs data on women's employment and welfare experiences over time. Fortunately, such data are available through several nationally representative longitudinal surveys. To examine how the work experiences of welfare recipients compare to non-recipients, I use data from the National Longitudinal Survey of Youth (NLSY) to construct complete employment and welfare histories for a cohort of women over a ten-year period, from the time they turn 18 through the year in which they turn 27. To estimate how much more women who ever receive welfare would have worked had they followed the same employment paths as women with similar characteristics who never turned to welfare, I develop a dynamic microsimulation model that predicts women's movement in and out of the labor market over the ten-year period.

This analysis shows that between the ages of 18 and 27, women who ever receive welfare work only about 60 percent as much as women who never receive welfare. Though the majority of welfare recipients work at some point in time, a substantial fraction spend minimal time in the labor market. About a third of recipients work for less than 25 percent of the ten-year period; a third work between 25 and 50 percent of the time, and the remainder work for more than half of the ten-year period. Welfare recipients who spend the least amount of time in the labor market are primarily recipients with low education and skill levels. Lower rates of employment in any given year and shorter periods of employment when they are employed contribute to welfare recipients spending less time in the labor force than non-recipients.

I estimate that if welfare recipients followed the same employment paths as non-recipients with similar characteristics, they would work 30 percent more than they currently work. The greatest employment gains would be realized by women with low levels of education, low basic skills, and more children, since they currently spend the least amount of time working. Even if welfare recipients were to follow the same employment paths as non-recipients, they would continue to experience substantial periods of joblessness. I estimate that by age 27, fewer than half of recipients who had not completed high school would be employed steadily. These results suggest that a substantial fraction of welfare recipients are likely to continue to need a safety net to support them during periods when they are unable to find employment.

This paper is organized as follows. I begin by describing the level and type of work that is expected of welfare recipients under TANF, what we already know about the work experiences of women on welfare, and the data and methodology used for this analysis. In subsequent sections, I explain how women on welfare differ from other women, highlighting differences that are likely to affect their employment prospects; describe how the work experiences of welfare mothers compare to non-recipients and examine the characteristics of welfare recipients based on their level of labor force attachment; and estimate how much more we can expect recipients to work given their characteristics and circumstances--assuming recipients follow the same employment paths as women who never receive welfare. I conclude with a discussion of the findings and their implications for transforming the current welfare system into a more work-oriented program.

II. Work Expectations Under TANF

PRWORA provides states with considerable flexibility in determining how they will use their TANF funds, but it is quite prescriptive in the area of work participation. Parents or caretakers receiving assistance under TANF are required to engage in work once the state determines that the individual is ready to do so or has received TANF assistance for more than 24 months. States can set shorter time limits for work participation and many have already done so (U.S. Department of Health and Human Services 1997). If an individual refuses to engage in work, TANF requires states to reduce the family's TANF grant, but states can also completely terminate assistance for families who do not comply with the work requirement. Medicaid coverage can also be terminated for an adult who refuses to work, although Medicaid coverage for children must remain intact.

The legislation sets forth explicit work participation rates that states must meet, including the minimum number of hours recipients must participate in a work-oriented activity (table 1) and the activities that can count toward the participation requirement. In 1997, the first year of full implementation, the law requires states to have 25 percent of their mandatory TANF caseload participating in work activities for at least 20 hours per week. The percentage of the mandatory TANF caseload required to participate in program activities increases by 5 percentage points each year, reaching a high of 50 percent in FY 2002. The level of participation required by each participant increases to 25 hours in FY 1999 and reaches the maximum level of 30 hours per week in FY 2000. With a few exceptions, these participation requirements apply to all TANF household heads, including teen mothers. ² States have the option to exempt single-parent families with a child under the age of one from the work requirement, and may only require single-parent families with a child under the age of six to work 20 hours per week. The TANF work participation rates are substantially higher than those required under the Job Opportunities and Basic Skills (JOBS) training program, the welfare-to-work program that states were required to operate prior to TANF. Due to numerous exemptions from program participation, only about 10 percent of the AFDC caseload was required to participate in JOBS activities (Pavetti, Holcomb, and Duke 1995).

What counts as participation is also substantially different under TANF than under the JOBS program. Under TANF, activities that count toward the participation rate focus primarily on employment; under JOBS they focused more on employment preparation activities such as assessment or education and training. States are not prohibited from placing recipients in these activities under TANF, but only recipients participating in those activities specified in the law can count toward a state's work participation rate. Allowable activities under TANF include: (1) unsubsidized or subsidized private or public sector employment; (2) on-the-job training; (3) work experience; (4) job search and job readiness assistance for up to six weeks; (5) community service programs; (6) provision of child care services to an individual participating in a community service program, and (7) vocational educational training (limited to 12 months for any individual and 30 percent of all work participants). Once recipients are required to work more than 20 hours per week they must participate in one of these activities for the first 20 hours, but can meet the additional required hours by participating in job skills training, education directly related to employment, or receipt of a GED or job search and job readiness beyond the six-week limit.

The TANF work participation rates are structured so that states can meet them either by engaging recipients in allowable program activities while they are receiving assistance or by reducing assistance caseloads below their 1995 level. The caseload reduction provision included in PROWRA allows states to reduce their required work participation rate by 1 percentage point for every percentage point reduction in their assistance caseloads since 1995. The caseload reduction credit was included in PROWRA to reward states for getting recipients into private sector jobs that would make them ineligible for program benefits. However, because the legislation does not require states to show that their caseload reductions have resulted from increased employment, advocates have expressed concern that the easiest and cheapest way for states to meet the work participation rates is for states to make it more difficult for families to qualify for assistance. In the absence of bureaucratic hurdles that may discourage recipients from applying for assistance, states could reduce their TANF caseloads by increasing the work effort of recipients in several different ways, including (1) helping recipients who would have gone to work anyway enter the labor market sooner than they might have done absent strict work requirements; (2) getting recipients into the labor market who would not have done so on their own, and (3) reducing the rate of return to the welfare system by helping recipients stay employed longer than they do currently.

III. The Work Experiences of Welfare Recipients

Most of our knowledge about the work experiences of welfare recipients comes from studies that examine movement on and off the welfare rolls. The seminal work in this area (Bane and Ellwood 1983 and 1994) found that although it was quite common for women to move on and off the welfare rolls, exits due to an increase in women's earnings only accounted for 21 percent of all exits from welfare.

Subsequent research in this area shows that Bane and Ellwood underestimated exits from welfare to work because the annual data they used for their analysis failed to capture work exits that lasted for relatively short periods of time. Thus, studies using monthly rather than annual data find substantially higher rates of women leaving welfare for work, ranging from 33 to 69 percent. Using monthly data from the Seattle/Denver Income Maintenance Experiment control group, Blank (1986) found that 33 percent of all completed welfare spells ended with an increase in the household head's labor market income. Using data from the NLSY, Gritz and MaCurdy (1991) and Pavetti (1993) estimated that about half of all welfare exits could be attributed to work. In a study using monthly data for 1984-1986 from the Panel Study of Income Dynamics (PSID), Harris (1993) found that 69 percent of all women who left welfare left for work. In an analysis of self-reported reasons for welfare exits in Washington state, Weeks (1991) found that 54 percent of all welfare recipients who left the welfare system reported leaving welfare to enter the labor market.

Although studies using monthly data find much higher rates of women leaving welfare for work, they also find extremely high rates of return to the welfare system among women who leave welfare for work. Pavetti (1993) found that 40 percent of all women who leave welfare for work return to welfare within 12 months and two-thirds return within five years. Harris (1996) finds somewhat lower rates of return, with 41 percent of women who leave welfare for work returning to the welfare rolls within three years.

In a comprehensive analysis of women's labor market involvement over a two and-a-half year period, Spalter-Roth et al. (1995) find that 43 percent of the single mothers who ever received welfare during the period of their analysis were engaged in substantial hours of paid employment. Welfare recipients who were employed exhibited two different patterns of work: one group cycled back and forth between welfare and work, and the second group combined work and welfare. Studies designed to evaluate the impacts of welfare-to-work programs show similar levels of employment among welfare recipients, including those who did not receive program services. For example, Riccio, Friedlander, and Freedman. (1994) show that 57 percent of recipients who participated in California's Greater Avenues to Independence (GAIN) program and 51 percent of recipients who did not participate in the program worked over a three-year period. Other welfare-to-work evaluation studies show employment rates ranging from 31 to 78 percent (Nightingale and Holcomb 1997).

Recent research based on interviews with single mothers to ascertain how they make ends meet shows that work among welfare recipients is not restricted to time when they are off the welfare rolls. Edin (1991) found that just over half of a group of 50 welfare mothers she interviewed in Chicago engaged in part-time or full-time work that they did not report to the welfare department: 7 of the 50 welfare mothers worked at regular jobs under false Social Security numbers; 22 worked in regular or odd jobs and were paid in cash off the books, and 10 worked in the underground economy. In a larger similar study of mothers in four cities, Edin and Lein (1997) found similar results: 46 percent of welfare mothers across the four cities reported earnings from reported work, unreported work, or work in the underground economy. Income from these sources accounted for an average of 15 percent of welfare mothers' total monthly income.

In sum, this research suggests that although work is common among welfare recipients, much of it is short-term and relatively unreliable. In addition, few studies show more than half of the AFDC caseload working, even over an extended period of time. Thus, the available evidence suggests that there is likely to be substantial room to increase work among welfare recipients, both by increasing the fraction of recipients who work and by increasing the duration of their employment.

IV. The Data

The data used for this analysis come from the National Longitudinal Survey of Youth (NLSY), an annual survey of a nationally representative sample of young people who were between the ages of 14 and 21 as of January 1, 1979. The original sample of 12,686 members includes a cross-sectional sample, a supplemental sample designed to oversample Hispanic, African-American, and economically disadvantaged non-Hispanic, non-African-American youth, and a military subsample. About half (6,283) of the original survey members were young women. Due to funding constraints, the military subsample was dropped in 1984 and the economically disadvantaged supplemental sample was dropped in 1991. Because this analysis focuses on examining women's work experiences over a fixed period of time, only women who are interviewed in every survey year (1979-1993) and whose labor market experiences are observed from ages 18 to 27 are included. Women who report receiving welfare for at least one month between the ages of 18 and 27 are classified as ever having received welfare and all others are classified as never receiving welfare. This selection criteria yielded a total sample of 2,044 women, comprising 511 women who ever received welfare and 1,533 women who did not.

To examine the work experiences of women over time, I initially constructed job histories for each of the women in my sample. This job file, created by linking jobs across all fifteen years of the survey, includes 13,665 jobs; 2,860 of those jobs were held by women who ever received welfare and 10,805 were held by women who did not. For ease of presentation, I mostly examine women's work and welfare experiences on a quarterly basis. When the data are collapsed in this manner, a woman who receives welfare in one month during the quarter is classified as receiving welfare during the quarter. Similarly, a woman who works at all during the quarter is classified as being employed during the quarter.

For each of the 40 sample quarters, I also distinguish whether a woman works in a "bad" or a "good" job using the same criteria used by Pavetti and Acs (1997) in their analysis of transitions from bad to good jobs. Under this classification scheme, a quarter is classified as a good job quarter if a woman usually works at least 35 hours a week and earns at least $8 per hour (in real 1993 CPIX dollars). To qualify as a good job quarter, a woman has to work at least 70 hours in the quarter. If a woman works at all during the quarter but does not meet the good job criteria, the quarter is classified as a bad job quarter.

V. Conceptual Framework and Methods

There are many ways to examine how much more work we realistically can expect from welfare recipients. For example, one could construct a behavioral economic model that attempts to capture the work incentives and disincentives facing recipients under a reformed welfare system. Alternatively, one could construct estimates based on the results of previous welfare-to-work demonstration projects. However, given that current efforts to reform the welfare system differ dramatically from previous initiatives, both of these approaches would require numerous assumptions about how recipients are likely to respond to a new set of incentives and penalties that may bear little resemblance to those they have faced in the past.

Alternatively, one could use the work experiences of women who have never used welfare as a starting point for estimating how much more work we might expect from welfare recipients. If applied within a dynamic context, this approach allows us to answer the question, "How much more work can we expect from welfare recipients if they follow the same employment paths as women who never receive welfare?"

To implement such an approach, I develop a relatively simple dynamic microsimulation model to estimate women's movement in and out of jobs over a ten-year period, from the time they turn 18 to the time they turn 27. This period is a critical one in most women's lives. It is the time when most young women make their initial entry into the labor market. It is also a time when many women make important decisions about how much schooling they will pursue and whether and when they will marry and/or have children.

To capture the process of moving in and out of employment for young women, I use a multivariate framework known as a competing risk model, where the competing risks are one of three employment states: (1) joblessness; (2) employment in a bad job, or (3) employment in a good job. I estimate a separate competing risk model to capture transitions from each of the three employment states. The models are based only on the experiences of women who never receive welfare during the ten-year observation period. I then use the estimates from these models and the observed characteristics of women who ever turn to the welfare system to simulate the quarter to quarter employment transitions for welfare recipients. ³

The three competing risk models that form the heart of the microsimulation model provide a framework for estimating how various characteristics--including demographics, parental status, education and mastery of basic skills, local labor market conditions, previous work experience (in good and bad jobs), and length of time in one's current employment state--affect the probability of making the transition from one employment state to another, while holding other characteristics constant. These models allow us to ask questions such as: holding all else the same, what is the impact of education or marital status or the presence of children or the condition of the local labor market on the likelihood that a woman who is jobless will become employed in a bad or a good job in the next quarter? The actual coefficients from the model are difficult to interpret and are, therefore, included in the appendix. However, the general pattern of the results provides important insights into how the characteristics and life circumstances of women who ever turn to the welfare system are likely to influence their employment outcomes.

Not surprisingly, a woman's transition between employment states is significantly affected by the amount of education she has and whether or not she is in school. While women are in school, their employment status is relatively unstable. However, once young women are within one to two quarters of completing their education, the likelihood that they will move from joblessness to a bad job or from a bad job to a good job increases substantially. Once a woman has completed her education, more schooling is associated with better employment outcomes. Women who have not completed high school are more likely to lose jobs once they become employed, while women who have completed college have a significantly better chance of making the transition to a good job, either directly from a state of joblessness or from a bad job.

The presence of young children affects women's transitions in and out of the labor market in several important ways. Having a child under the age of five significantly reduces the likelihood that a woman who is jobless will make the transition to employment. Having a child under the age of one significantly reduces the likelihood that a woman will move from a bad to a good job and increases the likelihood that a woman employed in a good or bad job will make the transition to a state of joblessness.

The structure and condition of the local labor market affect the likelihood that a woman will find employment if she is jobless and will make the transition from a bad to a good job once she is employed. Women living in areas of lower unemployment are more likely to find employment if they are jobless and to move from a bad to a good job than women living in areas with higher levels of unemployment. Women living in the Midwest and the South are less likely than women in the Northeast to make the transition from bad to good jobs. Good jobs in the Midwest and the South also tend to be less stable than in the Northeast. Employment outcomes are also somewhat better for women living in urban areas. Women living in or near cities are significantly more likely to find a job if they are without one and to move from a bad job to a good job than are women in non-urban areas.

A woman's transition from her current employment state is also significantly affected by the length of time she has been in her current state and her previous work history. A woman's previous work experience increases the stability of her current employment and increases the likelihood of moving from a bad to a good job. Time spent working in a bad rather than good job reduces a woman's employment stability by increasing the likelihood that she will lose her employment altogether or that she will move from a good to a bad job. In general, transitions from one employment state to another occur quite rapidly. The longer a woman stays in a particular employment state, the less likely she is to leave it.

I now turn to a description of how the characteristics and work experiences of women who ever receive welfare compare to those of women who have never received welfare. I then present estimates of how much more we can expect welfare recipients to work if they follow the same employment paths as women who never receive welfare.

VI. Who Turns to the Welfare System for Support?

Although stereotypes of welfare mothers abound, welfare mothers are, in fact, a diverse group of women facing a broad range of life circumstances that lead them to the welfare system for support. Mothers who have completed high school or even some college may find themselves seated in a welfare office or participating in a job search workshop alongside women who have less than a tenth-grade education. Mothers who have never been married, mothers whose marriages have failed, and even a small number of mothers who are currently married all avail themselves of support provided by the welfare system.

Even though there is considerable diversity within the welfare population, as a group, welfare mothers differ substantially from women who never turn to welfare (Table 2). On average, women who ever turn to the welfare system have substantially lower levels of education and basic skills, more children to care for, and are more likely to be a member of a racial or ethnic minority than women who never receive welfare. Given these characteristics, one would expect welfare recipients to have a far more difficult time finding and keeping employment than women in general.

The low education levels and skill deficits of welfare recipients are quite striking as shown in table 2. Forty-five percent of welfare recipients had not completed high school by age 27, compared to just 10 percent of women who never received welfare. While more than half of non-recipients had pursued some education beyond high school, only 15 percent of welfare recipients had done so. Furthermore, more than half of welfare recipients' scores on the Armed Forces Qualifying Test (AFQT), an achievement test that has a strong relationship to higher levels of employment and earnings, fell in the bottom quartile of the AFQT distribution for all women; one out of every three recipients' scores were so low that they fell in the bottom decile of the AFQT distribution.

Early childbearing also distinguishes welfare recipients from other women. Sixty-three percent of welfare mothers had their first child by the time they turned 20 , including 30.8 percent who gave birth to their first child before they turned 18. In contrast, only 10 percent of non-recipient mothers gave birth to a child when they were a teenager and 54.2 percent did not give birth to their first child until after their 22nd birthday. Because many early births among welfare recipients are followed by subsequent births, by the time they reached 27, almost one-third of recipient mothers had three or more children, compared to just 11 percent of non-recipient mothers.

Although the majority of welfare recipients are white, blacks and other minority women are over-represented among welfare mothers. Black, non-Hispanic women account for just 15 percent of this sample of women, but account for more than one-third of welfare recipients. The over-representation among other minority women is also evident, but far less dramatic: Hispanic and other minority women account for about 6 percent of the overall sample, but 9 percent of women who ever use welfare.

VII. How Do the Work Experiences of Recipients Differ from Non-Recipients?

Nearly all women spent some time in the labor market over this ten-year period, regardless of whether they ever received welfare. As shown in table 3, about 95 percent of recipients and virtually all non-recipients reported working during this ten-year period. Recipients entered the labor market at a somewhat slower pace than non-recipients, possibly because their lower education and skill levels made it more difficult to find employment or because they were caring for young children. Nonetheless, most of those who would eventually enter the labor market had done so by age 20, with the majority ( 65 percent of recipients and 86 percent of non-recipients) entering by age 18.

As the data in table 3 show, even though recipients were almost as likely as non-recipients to work during this ten-year period, on average, recipients spent far less time working than non- recipients. While non-recipient mothers spent about 70 percent (369.2 weeks) of the total ten-year period working, non-recipients worked for only about 38 percent (195.0 weeks) of the time period. This gap in the total time spent in the labor force results from shorter periods of employment and longer periods of joblessness among recipients compared to non-recipients. On average, recipients' periods of joblessness were longer than their periods of employment, while the reverse was true for non-recipients. On average, recipients were employed in 4.5 consecutive quarters and jobless for 5.4 quarters while non-recipients were employed in 7.1 consecutive quarters and jobless for just 3.5 quarters.

Figure 1 provides additional insight into how the work experience of welfare recipients compares to that of other mothers. Because of recipients' long lags between spells of employment and their somewhat lower rates of overall labor force participation, in any given year, they were far less likely to work than non-recipient mothers. At age 18, welfare mothers were substantially less likely to work than non-recipient mothers, including those who had not completed high school. Over time, the employment rates of welfare recipients increased while those of non-recipient mothers stayed relatively constant, resulting in a narrower gap in employment rates by age 27. In fact, by age 27, the employment rates of recipients were almost identical to those of non-recipient mothers who had not completed high school, but were still substantially lower than non-recipient mothers in general.

Welfare recipients were not only less likely to work in any given year, but when they did work, did so substantially less than other mothers, including those who have not completed high school. At age 18, welfare mothers who were employed during the year worked an average of 25 weeks, compared to 33 weeks for all non-recipient mothers and 28 weeks for non-recipient mothers who had not completed high school. By age 27, non-recipient mothers who did not complete high school worked almost the same number of weeks (40) as all non-recipient mothers (42), but welfare mothers still lagged behind, working an average of 35 weeks.

The lower rates of employment at a given age and the shorter periods of employment contribute equally to the gap between the employment of recipients and non-recipient mothers.

If welfare recipients worked the same number of weeks as non-recipient mothers, but the fraction of recipients working remained the same, the average weeks worked over the ten-year period would increase by 30 percent. Similarly, if recipients worked the same number of weeks as they do now, but the proportion working increased to the level of non-recipient mothers, the average weeks worked over the ten-year period would also increase by about 30 percent. These results suggest that a welfare-to-work strategy that focuses both on getting more recipients into the labor force and helping recipients stay employed for longer periods of time will go much further toward reducing the employment gap between recipient and non-recipient mothers than either strategy by itself.

Looking at the experience of the "average" welfare recipient is useful in some respects, but masks the fact that some women who ever use welfare do reasonably well in the labor market while others do poorly. Table 4 shows that welfare recipients can be divided into three groups, roughly equal in size, based on the time they spend working over the ten-year period: 35.9 percent spend 25 percent or less of the period working; 31.1 percent spend between 26 and 50 percent working and 33 percent spend more than half of the period working, but only 8.3 percent work for 75 percent or more of the period.

In stark contrast, more than half of women who never received welfare worked for more than 75 percent of the ten-year period and an additional 31 percent worked between 51 and 75 percent of that time.

As the data in table 5 show, recipients who spent the least amount of time in the labor force are overwhelmingly women who one would expect to have the least favorable economic prospects or for whom work may simply not pay. Almost two-thirds of recipients with minimal involvement in the labor market had not completed high school and 59 percent had AFQT scores that fell in the bottom decile of the AFQT distribution. Almost three-quarters gave birth to their first child before they turned 20 and almost half also had three or more children. Minority women also accounted for the majority of recipients with minimal employment. Recipients with moderate levels of employment also had low levels of education and low basic skills although not to the extent of recipients with minimal employment. Almost one-third also had three or more children. Women with minimal employment received welfare in an average of 19 out of the 40 quarters compared to 9.7 quarters for recipients with moderate employment and 7.1 quarters recipients with high levels of employment.

VIII. How Much More Work Can We Expect?

The tabulations presented in table 6 show that if women on welfare were to follow the same employment paths as women with similar characteristics who never used welfare, their employment would increase by 30 percent over the ten-year period. On average, I estimate that welfare recipients would work in 24.5 of the 40 quarters, up from the 18.9 quarters they actually worked. While this increase brings welfare recipients' work experiences substantially closer to the experiences of non-recipients, because women who fare poorly in the labor market are over-represented among welfare recipients, a substantial gap remains. In order to achieve complete parity with all non-recipients, welfare recipients as a group would have to work 20 percent more, on average, than non-recipients with similar characteristics.

Recipients with the lowest levels of employment are predicted to experience the greatest gains in employment, primarily because they have the biggest gaps to close. Recipients who have not completed high school could potentially increase their employment by 41 percent, from 15.3 to 21.6 quarters. The projected relative gains are even greater for women with the lowest basic skills. If women who score in the bottom decile of the AFQT distribution were to follow the same employment paths as their non-recipient counterparts, I predict they would increase their employment by 48 percent, from 12.5 to 18.5 quarters. Recipients with three or more children would experience a somewhat larger gain of 48.6 percent, increasing their quarters with employment from 13.8 to 20.5 quarters. Not surprisingly, welfare recipients who spent only a minimal amount of time in the labor force could potentially experience the greatest gains, almost tripling the quarters they are employed, from an average of just 7 quarters to 20.4 quarters. Given that these recipients use welfare for far longer periods of time than other recipients, employment gains for women with a minimal attachment to the labor force could potentially have a substantial impact on a state's ability to meet the TANF work participation requirements, especially in later years.

These estimates suggest that there is considerable room to increase the amount of time welfare recipients spend working. It is important to note, however, that even if recipients were to follow the same employment paths as non-recipients with similar characteristics facing similar circumstances, they would spend a substantial amount of time without employment. As shown in the second column of table 6; on average, recipients are predicted to spend just 24 of the 40 quarters working (61 percent of the ten-year period). Recipients who have not completed high school are predicted to work in just 54 percent of the quarters and women with the lowest basic skills are predicted to work in just 46 percent of the quarters. The tripling of the labor force involvement among recipients who had only minimal involvement in the labor market is only sufficient to have them employed in half of the ten-year period.

Some of the time spent jobless occurs while women are young and just making the transition to steady employment, but not all of it does. By the time they reach 26 and 27, only about 61 percent of all recipients are predicted to be working steadily, though there is substantial variation among recipients with different characteristics. At least 70 percent of women who have completed high school, have just one child to care for, and have higher skills are predicted to be working steadily, compared to fewer than half of women who have not completed high school, have the lowest basic skills, or have three or more children to care for.

These estimates assume that welfare recipients' work experiences will mirror those of non-recipients. To the extent that unobserved characteristics associated with a young woman's reliance on welfare reduce the likelihood that she will find and keep employment, these estimates will overstate the employment outcomes for welfare recipients. Alternatively, if, when faced with no other alternatives, welfare recipients work more than non-recipients, these estimates will understate their levels of employment.

IX. Discussion and Conclusion

This analysis shows that even though the majority of welfare recipients spend some time working, they work substantially less than women who never turn to the welfare system for support. The lower levels of employment among welfare recipients result from shorter spells of employment and lower employment rates in any given year. If welfare recipients were to follow the same employment paths as non-recipients with similar characteristics and similar family responsibilities, we could expect them to work 30 percent more than they do currently. This represents a substantial increase over their current levels of employment, but recipients would still spend substantial periods of time jobless. The lower rates of employment among welfare recipients result, at least in part, from the fact that women with the least favorable employment prospects are over-represented among the welfare population. In order to achieve complete parity with all non-recipients, welfare recipients as a group would have to work 20 percent more, on average, than non-recipients with the same characteristics.

To meet the work expectations set forth under TANF, states will be faced with developing strategies for increasing employment among a diverse group of recipients. Some recipients already have substantial labor market experience and will probably have little difficulty increasing their levels of employment if required to do so. However, a large group of recipients has only a limited attachment to the labor force and is likely to have considerable difficulty finding employment. Forty-five percent of the sample of women who ever received welfare and 62 percent of recipients with limited labor force involvement had not completed high school by the time they turned 27. Even in the relatively robust current economic environment, high school dropouts face an unemployment rate that is about four times that of college graduates. In March 1996, when women faced an unemployment rate of 5.1 percent, never-married mothers with children faced an unemployment rate of 18.5 percent or more than three times that of all women. For never married mothers with children under the age of three, the unemployment rate was even higher, at 21.8 percent.

It is hard to imagine a scenario where states could meet the work participation rates required under TANF without increasing employment among the most disadvantaged recipients and/or by increasing the length of employment spells among recipients who have already managed to find employment on their own. Unfortunately, states have had limited experience and/or limited success with accomplishing either of these tasks. Previous welfare-to-work programs have had moderate success at increasing employment among recipients with moderate barriers to employment, but more limited success with increasing employment among the most disadvantaged (Gueron and Pauly 1991). Program strategies that are focused on increasing job retention are in their infancy and have only been implemented in a small number of programs across the country (Hershey and Pavetti 1997).

The results presented here suggest that, even if states are successful at getting the most disadvantaged recipients into the labor market, these recipients are likely to continue to experience substantial periods of joblessness. Even if recipients with the lowest skill levels were to follow the same employment paths as non-recipients, fewer than half of them would be steadily employed by the time they reach their late twenties, suggesting that during periods of joblessness these families may continue to need access to a safety net over an extended period of time.

It is clear from these data that a woman's education and her skill levels make a substantial difference in her employment outcomes. However, efforts to increase recipients' skill levels have proven largely unsuccessful (Martinson and Friedlander 1994; Heckman et al. 1997). Over the long term, the best prospects for improving women's employment outcomes may be to intervene when they are young, before they have failed at many of life's important endeavors that will shape the quality of their adult lives.

There is clearly room to increase employment among welfare recipients. These data suggest, however, that welfare recipients are already working 70 percent of what we would expect if they were to follow the same employment paths as women with similar characteristics who never turned to the welfare system. It is also clear that women with the lowest skills often turn to welfare and remain there for extended periods of time. What remains to be seen is whether these women will be able to find work if they are required to do so, especially in a labor market that increasingly puts a premium on higher level skills.

* indicates significance at the 90% level
** indicates significance at the 95% level

* indicates significance at the 90% level
** indicates significance at the 95% level

* indicates significance at the 90% level
** indicates significance at the 95% level

Usage and reprints: Most publications may be downloaded free of charge from the web site and may be used and copies made for research, academic, policy or other non-commercial purposes. Proper attribution is required. Posting UI research papers on other websites is permitted subject to prior approval from the Urban Institute—contact [email protected].

If you are unable to access or print the PDF document please contact us or call the Publications Office at (202) 261-5687.

Disclaimer: The nonpartisan Urban Institute publishes studies, reports, and books on timely topics worthy of public consideration. The views expressed are those of the authors and should not be attributed to the Urban Institute, its trustees, or its funders. Copyright of the written materials contained within the Urban Institute website is owned or controlled by the Urban Institute.