Staying Employed After Welfare

Briefing Paper #128

Staying employed after welfare

Work supports and job quality vital to employment tenure and wage growth

by Heather Boushey

Executive summary
Real wage growth and long-term employment work together. People who remain consistently employed over time are more likely to experience real wage increases. But, causality runs both ways: those who start off at higher wages are more likely to stay employed. Remaining employed over time usually indicates higher starting wages and results in real wage increases. This relationship is especially important in the low-wage labor market.

Employment duration and wage growth are the most useful indicators for assessing the long-term success of welfare reform. Welfare recipients who are able to leave welfare and maintain employment have the best chance of experiencing real wage growth over the long term. Individuals who maintain employment will be on a path toward attaining a safe and decent standard of living for themselves and their families.

For many workers, and most welfare recipients, finding and keeping a job begins at home by finding a way to balance all of life’s responsibilities. For workers with family responsibilities, the capacity to maintain employment depends on whether they have access to safe and affordable child care. For working parents, and especially single mothers, using a center-based care provider and receiving help with child care are both associated with staying employed. In fact:

Single mothers of young children are twice as likely to still be employed after two years if they used center-based care.
Women who have a high school degree or less are almost three times more likely to still be employed after two years if they used center-based care.
Former welfare recipients with young children who used formal daycare are nearly three times as likely to still be employed after two years as those who do not.

Not all parents use center-based care settings, and for many families informal settings are a better child-care option. However, one-third of children are in center-based care settings, and the longer mothers stay in the labor force, the more likely they are to use center-based care.

Receiving assistance with child-care payments also increased employment durations:

Single mothers of young children are 40% more likely to still be employed after two years if they receive help paying for child care.
Mothers with a high school degree or less were just as likely as mothers with some college education to experience increased employment tenure when receiving child-care subsidies.
Former welfare recipients with young children are 60% more likely to still be employed after two years if they receive help paying for child care. Helping people stay employed also depends on the quality of the job. The better the first job is—whether it has health insurance and whether the starting wages are high—the more likely it is that the individual will remain employed over time. Workers who receive health insurance from their employer when starting their job are more likely to be employed two years later than those who do not:
Single women who receive employer-provided health insurance are twice as likely to still be employed after three years as those who do not.
Women with less education experienced a relatively greater increase in employment tenure if they began employment in a job with employer-provided health insurance than did women with more education.
Former welfare recipients who receive employer-provided health insurance are 2.6 times as likely to still be employed after three years as those who do not. Starting wages are also associated with maintaining employment:
Women who started their job earning in the top quintile of women earners are twice as likely to still be employed after two years as women who started in the bottom quintile.
Former welfare recipients who started their jobs earning in the second to bottom quintile are 63% more likely to still be employed after two years as those who started in the bottom quintile.

High-quality jobs offer higher starting wages and, critically, affordable health insurance. Placement in high-quality jobs creates a virtuous cycle by helping workers to maintain employment. Increased work experience leads to higher wages and develops on-the-job skills; the experience reinforces the advancement and the advancement reinforces the experience. For former welfare recipients, who start off at the very low end of the labor market in terms of wages and benefits, the quality of the first post-welfare job is especially critical to helping them move up the job ladder, maintaining their presence in the labor market, and supporting their families.

Maintaining employment has real implications for families. Staying employed is critical to enabling low-wage workers and former welfare recipients to support their families. Without stable employment, there is little hope that they will be able to do so, especially given that families can no longer rely on welfare as a source of income. Stable family income has long-term consequences for child development, and children whose parents are able to find and keep jobs appear to do better over the long-run.

Policy makers who want to help former welfare recipients and other low-wage workers move into the kind of employment that will enable them to support their families have a number of options. Since real wage growth and long-term employment work together, policy makers can either help women maintain jobs through work supports or help them find high-quality starting jobs. Helping former welfare recipients and other workers maintain employment means helping them balance their familial responsibilities so that they can devote attention to their jobs. This requires a significant expansion of affordable, safe, and enriching child care. Promoting better quality starting jobs means helping employers to offer health insurance through policies that make health insurance affordable for employers, and adopting policies that improve starting wages, such as raising the minimum wage and fostering the development of unions in low-wage sectors of the economy.

Programs such as WorkFirst, which emphasize moving welfare recipients into any job, may not have helped those workers to find the kind of high-quality jobs that would allow them to advance and maintain employment over the long term. In order to get former welfare recipients into the labor market and help them stay there, we need to ensure that they have access to the kinds of work supports that will help them balance work and family. Without these, it is difficult for them to remain employed. Without continued employment, the virtuous cycle will be broken.

Introduction
In 1996, by signing the Personal Responsibility and Work Opportunity Reconciliation Act, commonly known as “welfare reform,” President Clinton ushered in a new phase not only for American welfare policy, but for how Americans think about working mothers. The aim of the welfare program was refocused from one that only provided cash assistance to one that actively encouraged recipients to move as quickly as possible into the labor market and into “self-sufficiency.” Congress now requires that welfare mothers work outside the home in order to receive their welfare benefits, while limiting the amount of education and training a welfare recipient may receive. Increases in work supports, such as child-care subsidies, were implemented in tandem with the reforms. However, many states also implemented what are known as WorkFirst programs, which encourage caseworkers—often financially—to put w
elfare recipients into jobs as quickly as possible (Strawn, Greenberg, and Savner 2001). WorkFirst programs discourage education and training in favor of rapid entry into the labor market.

Some have heralded the welfare reforms as a success, because millions of welfare mothers moved off welfare and into employment. Since 1996, welfare caseloads have fallen dramatically: from their peak in March 1994 to June 2001, caseloads fell by 62% (Administration for Children and Families and Office of Planning Research and Evaluation 2000).¹ Most of those who left welfare—about 60%—found jobs (Loprest 1999), and among those who found a job, most worked at least 30 hours per week (Center for Law and Social Policy 2001).

The problem is that getting into the labor market and staying there are not the same thing. Historically, welfare recipients have cycled in and out of low-wage jobs; before the 1996 welfare reforms, more than 40% of former welfare recipients returned to welfare within one year. Many of those who had left welfare since 1996 have moved into low-wage, traditionally short-tenure jobs. Average wages are about $6 to $8 per hour, and only about one-quarter of the jobs provide health benefits. When surveyed, about two-thirds of those who left welfare were working, and about three-fourths had worked at some point in time (Strawn and Martinson 2000).

Staying in the labor market is one of the primary ways that workers move up the job ladder; wage growth and long-term employment work together, and therefore people who remain consistently employed over time are more likely to see real wage increases. But, causality runs both ways: those who start off at higher wages are more likely to stay employed over time. Maintaining employment is associated not only with higher starting wages, but also with more real wage growth over time. This relationship is especially important in the low-wage labor market.

Access to the kinds of jobs that allow workers to balance their familial and work responsibilities can make the difference between keeping a job and not. For working mothers—especially low-income mothers—access to child care is often the critical component. For all workers, job quality matters in whether they remain employed and experience real wage growth.

Length of employment and real wage growth also have real implications for families. Recent research has shown that the timing of poverty has implications for child development: family income is more important in shaping ability and achievement in young and middle-aged children than it is for adolescents (Duncan and Brooks-Gunn 1997). If low-wage workers with families are less likely to maintain employment than are low-wage workers without family responsibilities, there are implications not only for financial outcomes but also for the long-term success of the children.

We examine the role of work supports and job quality on employment durations and real wage growth among women workers over the 1990s. Using a longitudinal data set (Survey of Income and Program Participation) that covers both the early and late 1990s, we analyze the extent to which the labor market experiences of women with a history of welfare differ from those of other women. The first section addresses whether mothers who have access to child care are more likely to stay employed and whether this affects their wage growth. Next, we examine the role of starting job quality on employment durations and wage growth. We then explore the role of the economic boom of the 1990s on employment durations and real wage growth and look at its implications for families. Finally, policy implications are discussed.

Background
The labor market conditions for low-wage workers in the late 1990s were better than they had been in decades, and in this sense welfare reform was well-timed. While welfare reform pushed women off welfare, the strong economy pulled them into the labor market. Former welfare recipients, however, still face a number of specific barriers to employment. By definition, they are mostly single mothers, which means that any job they find must allow them to balance their familial responsibilities along with their job. Although work supports were a stated intention of welfare reform, they have been paltry at best. For example, the majority of children eligible for federal child-care subsidies under TANF have yet to receive anything. Also, former welfare recipients tend to end up in low-quality first jobs, which makes staying employed and advancing more difficult. The logic of WorkFirst—that getting even a low-paid job is better than no job at all—means that many former welfare recipients enter the labor market without the skills necessary to find and keep a job. As a result, many do not get “good” first jobs, and in the end, this may have a dampening effect on employment durations and real wage growth in the long term.

Work supports play an important role in helping workers balance work and family. For many workers, and most welfare recipients, finding and keeping a job begins at home by finding a way to balance all of life’s responsibilities. Especially for working mothers, negotiating the demands of child care (and often elder care) can make keeping a full-time job difficult. Work supports can play a major role in helping workers with family responsibilities maintain employment (Anderson and Levine 1999; Wadfogel 1998). Without work supports, a sick child can upset the fragile balance of work and family, causing a worker to quit or lose her job and thus lowering her chances of moving up the job ladder.

Work supports for working parents are critical because most mothers now work. Between 1975 and 2001, the labor force participation rates of mothers with children under three years increased from 34.3% to 61.3%, and among married-couple families with children, a majority (63.2 % in 2001) now have two incomes (Bureau of Labor Statistics 2002). The increase of mothers in the labor force has occurred both because more women are choosing to remain working and because of trends in male earnings. As male wages fell during the 1980s and early 1990s, families needed the incomes of mothers to maintain their standards of living (Mishel, Bernstein, and Schmitt 2000).

For parents to function successfully at their jobs, they need access to safe and affordable child care. Quality child care has been shown to affect both short- and long-term child outcomes in the areas of safety, school readiness, and problem behaviors (Giannarelli and Barsimantov 2000). Ideally, child care should be enriching and promote a child’s development, but there is debate as to what kind of child care is best for young children. Studies have found that the quality is generally low in both home-based and center-based care; although research has confirmed that home-based care in low-income neighborhoods is less well equipped and less stimulating than in middle-class settings (Fuller et al. 2001). Focusing on improving access to center-based care may be the best policy option, especially as 30% of children under six are in center-based care settings (Smith 2000). Furthermore, using center-based care, which has regular days and hours of operation, may reduce absenteeism due to unexpected child-care difficulties (Holzer and Stoll 2000).

Access to child care can be difficult for two reasons. First, in many communities, there are not enough available child-care slots in regulated or center-based care settings (Fuller et al. 2001; Rangarajan 1996). Second, child-care costs are high, especially for low-income mothers. Family budget studies find that child-care costs range from 13% to 29% of a basic family budget for a single parent family with one child (Boushey et al. 2001). Among low-income families, child care can eat up 35% of income (Committee on Ways and Means 2000). Even among the 27% of parents who use family for child care, payment may still be necessary (Zinsser 2002). The high cost of care a
ffects the employment of mothers, especially low-income, single mothers (Ribar 2000). Among former welfare recipients, recent research finds that increased funding for child-care subsidies significantly increases the probability that they will work (Connelly and Kimmel 2000; Lemke et al. 2000) and have increased earnings (Witte et al. 1999).

To evaluate access to child care as a work support, this study examines two aspects of child care. First, the analysis tests the effects of the availability of child-care subsidies on employment durations and real wage growth. Second, it examines the use of center-based care among pre-school children. This is not to presuppose that center-based care is appropriate for all families, but because subsidies are more often used for center-based care, it makes sense to examine the two in tandem (Cabrera, Hutchens, and Peters 2002).

Job quality is critical for upward mobility and longer employment tenure. Balancing work and family is not the only barrier that welfare mothers face when looking at the labor market. The WorkFirst ideology is that any job is better than no job, and caseworkers are asked to move welfare recipients into jobs as quickly as possible, rather than offering education, training, or even time to find a job that would be a good match and provide a decent standard of living. Long-term success in the labor market may be hindered if the first job is of low quality and does not offer upward mobility.

Prior research shows that job quality is critical for upward mobility and longer employment durations. Overall, there is a positive correlation between employment tenure and wage growth (Newman 2000). Workers who start jobs at low wages are less likely to stay employed; employment tenure in low-wage jobs is shorter than in more highly paid ones. Workers with longer work histories generally are paid higher wages, and increased work experience is positively correlated with at least moderate wage growth (Gladden and Taber 2000). However, early studies of former welfare recipients have found that those who left welfare have shorter job tenure than other low-income mothers. Studies of those who left welfare since welfare reform have found that about three-quarters of former recipients worked at their current employer for less than a year, and a third for less than six months, while about half of other low-income mothers have been on their job for more than a year (Center for Law and Social Policy 2001).

However, low-wage workers are less likely to see strong wage growth overall. Prior research on real wage growth among low-skill and/or low-wage workers is fairly consistent in finding that less-educated workers experience little or no real wage growth. Using data from the 1980s and early 1990s, most studies find that wages increase between 1.0% and 2.6% per year for low-skill workers (Burtless 1995; Card, Michalopoulos, and Robins 2001; Connolly and Gottschalk 2000; Moffitt and Rangarajan 1989). Gottschalk (2000) found that substantial proportions of workers experienced real declines in wages while working for the same employer or moving to a new employer between the mid-1980s and mid-1990s. Roughly 70% of jobs held by workers with less than a high school degree had negative real wage growth, and even among college graduates, 56% were in jobs with no real wage growth. Less-educated workers do not experience the average within-job wage growth; their real wage profiles within jobs are remarkably flat (Gottschalk 2000).

Job quality is not only indicated by higher wages, but also by whether the job offers health insurance. Low-wage jobs are less likely than high-wage jobs to offer health insurance: in 1998, only 30% of workers in the lowest wage quintile had employer-provided insurance, compared with 82% of workers in the highest quitile (Mishel, Bernstein, and Schmitt 2001). In addition, the types of jobs available to low-skill workers are less likely to include health benefits; for example, workers in the low-wage service or retail sector are 18-31% less likely to have employer-sponsored health care than are workers in the manufacturing sector (Wiatrowski 1995).

The quality of the initial job taken is critical in determining who will sustain a job over the long-term and who will not (Cancian and Meyer 2000). Historically, however, welfare recipients are less likely to get good initial jobs. Rather, they are likely to find jobs that are fairly unstable, provide low pay, few fringe benefits, and are associated with high turnover (Rangarajan, Schochet, and Chu 1998). Pavetti and Acs (1997) found that women who were on welfare, or who shared the demographics of welfare women, were less likely to move into a good job—one paying more than $8/hour and for more than

35 hours/week. Fewer than half of the young women studied who did not have a high school degree were in a good job at any time in their twenties, as opposed to almost three-fourths of all the young women and 59% of the mothers. Education is the key not only to finding a good job but also to keeping it: while 41% of all women work steadily in good jobs by ages 26 and 27, only 22% of mothers and 15% of women who have not completed high school do so.

We address the extent to which job quality matters in employment durations and real wage growth. Job quality is measured by starting pay and whether a worker receives health insurance from her employer when she starts her job. Jobs that pay higher initial wages are better jobs. What determines job quality is not whether an individual simply has health insurance but whether the employer provides it. Workers can receive health insurance from sources other than their employer (e.g., they may purchase it on the market or receive it through Medicare or an employed spouse). Jobs that offer health insurance typically offer other benefits, such as paid vacation and sick time, and tend to offer more stable employment and higher wages than jobs that do not. In this sense, employer-provided health insurance indicates whether the job is of high quality or not.

The strong economy made getting a job easier in the late 1990s. Undoubtedly, the late 1990s provided an excellent economy in which to implement welfare reform. The economy was close to full employment, and most workers who wanted to find a job could. The tight labor market boosted wages, especially for workers at the low end of the pay scale: after more than 15 years of declining wages for workers at the middle and bottom of the wage distribution, wages grew rapidly between 1995 and 1999. Among the bottom 10% of wage earners, inflation-adjusted wages rose by 9.3%. Strong wage growth at the bottom was accompanied by a slowing of the growth in wage inequality that had occurred over the past two decades. Strong labor demand was a key factor in moving families from welfare to work and increasing employment and earnings for low-wage workers.

As states implemented WorkFirst policies and caseloads rapidly declined, the question remained as to whether those who left welfare in the strong economy of the late 1990s would fare better than those who had left before. Early studies do not show benefits of WorkFirst programs in terms of employment durations or wages (Campbell, Maniha, and Rolston 2002; Hamilton et al. 2001). The important question for policy makers, then, is what promotes longer employment durations. What are the most effective policies for helping welfare recipients to move off welfare and into jobs with prospects for upward mobility? Are these strategies different than those that work for workers without a history of welfare use? We address four questions on the interaction of employment durations and wage mobility over the 1990s:

Are work supports critical for increasing employment durations, particularly among former welfare recipients?
Does having a high-quality job lead to longer employment durations and increased real wage growth? Does this differ for former welfare recipients?
Were former we
lfare recipients who left after welfare reform more likely to have longer employment durations and greater upward wage growth than those who left welfare before the reforms?
What was the role of the strong economy in employment durations and wage growth among individual workers, especially among disadvantaged workers?

Data and methodology
The Survey of Income and Program Participation (SIPP). In order to evaluate employment durations and real wage growth of workers over the 1990s, this analysis required a special kind of data set. The 1993 and 1996 panels of the Survey of Income and Program Participation (SIPP) provide a unique opportunity to evaluate workers’ experiences over the early 1990s economic recovery and the late 1990s economic boom. The panels also coincide with the implementation of welfare reform, allowing analysis of workers’ experiences pre- and post-welfare reform.²

The SIPP is a nationally representative, longitudinal panel data set. Each respondent is interviewed every four months—termed a wave—about their labor market and program participation experiences over the preceding four months. The 1993 panel of the SIPP has nine waves of data and covers the early 1990s (October 1992 to December 1995), while the 1996 panel has 12 waves of data and covers the late 1990s (December 1995 to February 2000). The sample of the SIPP used here only includes women age 18 to 64 who were employed in any month during the panel. Throughout this analysis, welfare refers to the federal programs Aid to Families with Dependent Children (AFDC) prior to 1996 and Temporary Assistance to Needy Families (TANF) after welfare reform in 1996. (See the Appendix for a complete description of the data.)

Measuring wage and income growth. This analysis measures both real wage growth and employment durations. Wage growth is measured across an individual’s first and last reported wage over the employment spell.³ Most of the analysis focuses on whether or not any wage growth was experienced among those who were employed for two or more waves. This does not take into account different spell lengths. Hourly wages are reported in 2000 dollars. (See the Appendix for more detail on the wage method and for a comparison of wages from the SIPP and Current Population Survey.)

The analysis also reports on the wage growth of women across starting wage quintiles. Workers are assigned to a wage quintile based on their starting wage. Then, the analysis shows whether they remain in that quintile or whether they move up or down the wage distribution.

Measuring employment durations. This report uses a relatively loose definition of employment, requiring that workers are at least marginally employed. An individual is marginally employed if he or she worked at least one week during the four-month wave. Employment is self-reported and can be either full time or part time and includes those who are self-employed. (See the Appendix for discussion of the implications of using this definition compared to alternative definitions.)

To calculate the proportion of workers who maintain consistent employment over time—that is, their employment duration—we estimate survival functions. A survival function calculates the percent of individuals remaining consistently employed after a given number of time periods, or is this instance a four-month wave.⁴

Child care is a necessary work support
Access to child care provides a critical work support for former welfare families and working mothers more generally. The child-care measures used here are: (1) whether the mother reported receiving assistance paying for child-care costs (the source could be the government, her employer, another parent, or another person), and (2) whether the mother reported using a daycare or family care center (i.e., center-based care) to care for any of her children under six.

Receiving a subsidy for child care promotes longer employment durations among women, regardless of marital status or educational attainment (Table 1). Unmarried women are a better comparison group to welfare mothers, because welfare mothers are highly likely to be unmarried. Both unmarried women and former welfare recipients are likely to face more child-care issues because of lower total family income and fewer adults in the family among whom to distribute the child-care responsibilities. Among unmarried women who received help with child-care payments, 52.5% were still employed after two years, while only 37.7% who received no help were still employed. The findings are similar among women with a history of welfare use (Table 2). Those who received assistance with child-care payments were more likely to maintain employment: 34.9% of former welfare recipients who received child-care assistance and 19.2% who did not were still employed after two years. Real wage growth is more prevalent among unmarried women and former welfare recipients who received help with child-care payments, relative to those who did not receive assistance.

Table 1

The findings do not hinge on the educational attainment of the mother, regardless of whether she was a former welfare recipient or not. Table 1 shows that the probability of remaining employed after two years is greater for women with more education but that the relative change associated with receiving child-care subsidies is the same for both educational groups. Women across the educational spectrum are 1.4 times as likely to stay employed after two years if they received child-care subsidies.

Women who have children under the age of six and use center-based care have much longer employment durations than women who do not, regardless of marital status or welfare use. Among unmarried women, 72.7% of those who used center-based care were still employed after two years as opposed to only 35.5% of those who did not. Among former welfare recipients with young children, using formal center care led to half (49.9%) remaining employed after two years, while only 18.7% of those who did not use center-based care remained employed for this long. Using center-based care may signal either that these women have a greater commitment to the labor market or that this kind of care allows them to maintain employment.

Less-educated women who used center-based care were 2.8 times more likely than those who did not to still be employed after two years. Among women with some college or more, those using center-based care were 1.9 times as likely to still be employed after two years.⁵ In terms of employment length, women with less education benefited more if they used center-based care relative to more educated women, even though overall they were less likely than higher educated women to stay employed.

A lower proportion of unmarried women and former welfare recipients who did not use center-based care and who received child-care subsidies experienced real wage growth over their employment spell, relative to those who received child-care assistance. This may either be because employment durations are shorter, and thus there is a shorter period over which to experience wage growth, or because having access to child-care assistance enables mothers to move up the job ladder more quickly.

In the end, access to child care matters. Receiving assistance for child-care payments increases the duration of employment among all mothers, regardless of welfare history, and these mothers are more likely to experience real wage growth. As mothers move to full-time employment, they are more likely to use center-based care (Fuller et al. 2001, 103). When welfare recipients first move off welfare and into the labor market, they are likely to use informal child-care arr
angements. However, among welfare recipients actually participating in welfare-to-work programs, use of center-based care rises (Fuller et al. 2001). Higher levels of maternal education and employment, higher levels of cognitive stimulation provided in the home, and residence not in public housing are all associated with use of center-based care (Zaslow et al. 1998). For both married and single women, mothers who use center-based care are more likely to work full time (Connelly and Kimmel 2000).

Using a daycare or family care center is not the only option available to working parents, and for many these are not the best options. Some families have relatives or friends who are willing—and able—to care for their children while they are employed. However, the findings point toward differences among women with young children who choose daycare over other options. These women are more likely to remain employed. This may be endogenous, as women who are more committed to the labor market are more likely to find center-based care; but providing the option of center-based care for more families would likely help to increase employment durations among women more generally.

School-age children also need care, especially in neighborhoods with high crime rates or rampant drug abuse. Often, mothers report that they are concerned about the safety of their school-age and pre-teen children, because they are concerned that they will fall victim to crime, drugs, or gangs. They also worry about child protective service agencies removing their school-age children from the home if they are found to be alone (Boushey forthcoming).

Child care is even more critical because the United States does not provide a great deal of flexibility to working parents. Many jobs do not provide substantial—or even minimal, for many workers—paid time off or workplace flexibility to address child-care needs,6 and mothers are less likely than other workers to be in jobs that offer flexibility (McCrate 2002). For low-wage workers, finding a job that provides paid time off to deal with a child-care emergency or a sick child may not be possible. The United States has a lower proportion of children in publicly supported care, relative to other advanced economies, and lower levels of child-care costs are covered by the government (Wadfogel 2001). Among low-wage employees, 37% had a job that offered paid leave for a sick child, whereas among high-wage employees, 61% did (Galinsky and Bond 2000).⁷ However, among firms that have family-friendly workplace policies for their employees, such as flexible schedules, on-site child care, or financial assistance with child care, there are positive impacts in terms of recruitment and retention, lower absenteeism, and increased job satisfaction (Friedman 2001).

Quality of first job matters
Job quality is measured by the woman’s starting wage and whether her employer provided health insurance when she began her job. We first place women into quintiles based on their starting wages and then follow whether women who started their job in a higher quintile stay employed longer or whether they experience a greater probability of wage growth relative to those who started in a lower quintile. The health care measure is whether an individual has access to employer-provided health insurance at the start of her employment. Among the sample population, 47% of those in the fifth wave of the 1996 panel had employer-provided health insurance when they began employment, and all individuals have data for that question.⁸ Single mothers are not able to access health insurance through a spouse, so employer-provided health insurance is even more important for them than for married mothers.

For women overall, higher starting wages are correlated with longer employment durations: of the women who began in the bottom 20% of the starting wage distribution, 35.5% were still employed after two years; while among those in the top 20%, 73.4% were still employed (Table 3). Thus, women in the top fifth were 2.1 times as likely to still be employed after two years, relative to women in the bottom fifth. The findings are similar for former welfare recipients. Starting off in a higher quintile is associated with a greater likelihood of remaining employed: among those in the bottom quintile, 20.0% were still employed after two years; and among those who began their employment in the middle quintile, 48.5% were still employed after two years (Table 4).⁹ Real wage growth was more likely among those who began in the lower quintiles among both women generally and those with a history of welfare receipt.¹⁰

Having employer-provided health insurance when starting a job leads to longer employment durations. If they had employer-provided health insurance, 73.7% of married women and 77.5% of unmarried women were still employed after two years. However, among those without employer-provided health insurance, only 44.4% of married women and 38.9% of unmarried women were still employed after two years. Thus, among unmarried women, those with employer-provided health insurance were twice as likely to still be employed after two years relative to those without. Real wage growth, however, was not much more likely among those with employer-provided health insurance relative to those without.

These results are not dependent on educational attainment. Women without a high school degree who had employer-provided health insurance were 2.3 times as likely to still be employed after two years relative to those without employer-provided health insurance. Among higher educated women, the increase in length of employment associated with having employer-provided health insurance is smaller; the likelihood of remaining employed, however, is still 1.4 among women with a college degree who have employer-provided health insurance.

The gap in employment durations between those with and those without employer-provided health insurance is even greater among former welfare recipients relative to those without a history of welfare use. Among former welfare recipients, 59.0% were still employed after two years if they had health insurance, while only 22.7% were still employed if they did not. Those with a high school degree or less were 2.5 times as likely to still be employed after two years if they had employer-provided health insurance. This indicates that, for former welfare mothers, getting health insurance may be even more critical to sustaining employment than for workers more generally, perhaps because these women are less likely than other women to have spouses who can put them on their health insurance.

Having an employer who provides health insurance increases employment durations among all women, but more so among former welfare recipients. This is obviously endogenous—jobs that have health insurance also tend to pay better, perhaps have better developed internal labor markets, and perhaps are better places to work. However, this does not change the conclusion that, to get individuals into sustained employment, finding jobs with health insurance is critical.

The role of the economic boom of the late 1990s
The long economic boom of the 1990s led to historic lows in unemployment and increased wages for workers across the income spectrum. In particular, low-wage workers made substantial wage gains. These aggregate phenomena should filter down to individuals such that workers should have experienced stronger wage growth and longer employment durations in the late 1990s relative to the early 1990s.

So, how did former welfare recipients fare in the strong economy? Were their labor market experiences different from those who left welfare earlier in the decade, prior to the welfare reforms and prior to the histo
rically low unemployment?

First, the late 1990s show little difference in the relationship between employment durations and starting wages. Figure A shows that women who began at higher starting wages were more likely to have longer employment durations if they had not just left welfare, but that there was little difference between this trend in the early and late 1990s. Among those who stayed employed for more than two years, the starting wages were higher than among those who stayed employed for less than two years. However, among former welfare recipients, the benefits of longer employment durations are not as clear. Although among women without a history of welfare, higher starting wages are positively correlated with two or more years of employment, among former welfare recipients, the relationship is relatively flat. Again, though, welfare recipients who began working in the late 1990s did not see more upward wage growth than those who began working in the early 1990s.

Second, employment durations were much longer among those who did not have a history of welfare use, but the gap between those who were on welfare and those who were not closed over the 1990s (Table 5). Over half (52.9%) of women who were not recent welfare recipients were still employed after two years in the early 1990s, and only slightly more (56.9%) were still employed after two years in the late 1990s. However, among former welfare recipients, only 16.5% and 25.3% were still employed after two years in the early and late 1990s, respectively. The increase in employment durations, however, was greater among former welfare recipients over the 1990s than among women more generally. In the early 1990s, those who had not been on welfare were 3.2 times as likely to still be employed after two years as those who had, but only 2.3 times as likely in the late 1990s. Thus, the strong economy of the late 1990s, is correlated with former welfare recipients in the late 1990s.

Former welfare recipients start their jobs at lower wages relative to women who were not on welfare prior to employment, but former welfare recipients saw a much larger jump in starting wages over the 1990s than did other women. Starting wages rose by 6.2% between the early and late 1990s among former welfare recipients, while rising by only 4.3% among nonwelfare workers (Table 5). Ending wages grew by twice as much for former welfare recipients (12.6%) relative to those who had not recently left welfare (6.2%). The probability of experiencing any real wage growth over the panel was lower among former welfare recipients than among those without a history of welfare use in both time periods.

Overall, many workers saw no real wage growth in the early and late 1990s. The proportion of all women with any real wage growth during an employment spell was 49% in the early 1990s and 56% in the late 1990s; among former welfare recipients, the proportion with real wage growth was 41% in the early 1990s and 51% in the late 1990s. Much of the real wage growth that individuals experienced was only minimal—their wages just barely outpacing inflation. Figure B shows that among all women and former welfare recipients and over both the early and late 1990s, the majority of workers saw little or no real wage growth during their employment spell. The majority of workers are clustered around zero real wage growth. In the late 1990s, the distribution of wage growth is skewed slightly more toward the right, meaning that women were more likely to experience growth in the late 1990s, but the pictures are not dramatically different over the two time periods.

Table 6 shows another measure of real wage growth: the proportion of women who remain in their starting wage quintile and the proportion who change quintiles. The changing of quintiles is often referred to as “wage mobility.” Over the late 1990s, women were more likely to move up a quintile than they had been in the early 1990s. The proportion of women who moved up a quintile rose among every quintile, except the top. Most women, however, remained in the same quintile that they were in when they began their employment spell. Among those in the top quintile, about 70% remained there in both the early and late 1990s.

Finally, Table 7 shows employment tenure for other groups of workers. Former welfare recipients are only one group of traditionally disadvantaged workers. Less-educated workers, non-U.S. citizens, women of color, and women who were unmarried with small children also all saw large increases in employment durations over the 1990s (Table 7).

Implications for families
Over the late 1990s, real family income grew along with wages. Comparing across women’s families, Table 8 shows that they were more likely to move up a quintile than to move down in the late 1990s relative to the early 1990s. Stronger wage growth led to an increase in family income so that families moved up. However, the majority of families that started off in the lowest quintile remained there across the 1990s. There was less mobility in terms of family income as compared to wage income, because individuals were more likely to experience wage increases that moved them up a quintile than were families overall (Table 6).

The economic boom of the late 1990s was beneficial for more disadvantaged workers. Former welfare recipients saw larger growth in starting and ending wages than women workers more generally, and demographic groups who are traditionally more disadvantaged in the labor market—Hispanic workers, foreign-born workers, and those with less education—increased their employment durations more so than other groups.

Overall, workers were more likely to experience real wage growth in the late 1990s, relative to the early 1990s. However, two out of every five workers experienced no real wage growth in the late 1990s. Among those with wage growth, many experienced growth that hovered around zero to slightly above zero.

Policy implications
The main implications for policy are that work supports truly matter and that the WorkFirst ideology—that any job is better than no job—may not be successful in the long-run. Access to child care is critical for women to stay employed and move up the job ladder. Getting a good starting job is also key.

Child care policies
The policy implications of this analysis for child care are simple: child care must be accessible and affordable to help mothers stay employed. There have been changes to the government’s role in child-care provision in the past few years, but they have been inadequate to address the pressing child-care needs. The government should expand the supply of safe and affordable child care.

Along with the implementation of welfare reform, funds available for child care increased substantially. Welfare reform made three changes to federal child-care funding: (1) existing child-care programs were combined into a single program, the Child Care Development Block Grant; (2) the federal legal guarantee of child-care support to current and former welfare recipients was eliminated; and (3) overall child-care funding was increased, by both increasing funds for child care directly and allowing TANF funds to be used for child care. Betw
een 1995 and 2000, child care funds available for welfare and poor working families increased from $2.5 billion to $8 billion (Fuller et al. 2001). However, since 1998, the TANF funds have grown much faster than funds from the Child Care Development Block Grant. Whereas in 1997, TANF funds comprised only 10% of total child care funds, by 2000 they provided just over half (Adams and Rohacek 2002). The increasing percentage of child-care funds from TANF is a problem because as caseloads rise, which they did over 2001, states may find that they need those funds instead for services for current welfare recipients or for cash assistance. Furthermore, this makes those funds vulnerable to cutbacks in the TANF block grants that may occur during reauthorization.

Even though child-care funding increased, the level of funds remains woefully inadequate. The demand for child care has increased by more than the availability of funding for child-care programs. By 1999, only 12% of eligible families received assistance through the Child Care and Development Fund (Layzer and Collins 2001; U.S. Department of Human Services 1999). Head Start serves less than half of the eligible children (Blank, Schulman, and Ewen 1999). Despite increased federal funding on child care over the past decade, wages for child-care workers stagnated, resulting in continued problems with recruiting and retaining qualified teachers (Whitebrook, Howes, and Phillips 1998).

Child-care funding needs to be increased and the availability of subsidies must be expanded. Furthermore, the supply of quality, affordable child care must be increased. Modest increases to current child-care funding streams will not improve quality or significantly improve access to child care. Current child-care initiatives focus on improving access through a variety of programs that parents are left to piece together. Most of these initiatives focus on expanding access to services with little emphasis on improving quality. To truly help families meet the basic need of quality, affordable child care, we need to develop a coordinated system of child-care programs, accompanied by a significant increase in funding.

Job quality policies
Welfare reform changed the rules within welfare offices around the country. The new model of casework is to increase work as quickly as possible for welfare recipients, rather than focus on education or training. This does a disservice to welfare recipients who move into low-paying, dead-end jobs that do not offer decent wages and health insurance. This disservice occurs on two levels: first, lower quality jobs mean shorter employment durations and lower real wage growth; and second, many low-wage workers go without health insurance.

When considering TANF reauthorization, rethinking the WorkFirst ideology is key to helping welfare recipients find higher quality jobs. Studies of welfare-to-work programs generally find little or no long-term effects on employment stability and wages, regardless of whether the program focuses on job search or education and training. However, there are programs out there that do help. For example, a program in Portland, Ore., led to large gains in stable employment and earnings growth through encouraging participants to take jobs at above-minimum wages, rather than take just any job (Campbell et al. 2000; Hamilton et al. 2001).

The government has a role in making it possible for employers to offer affordable health insurance to all of their workers. This past year, many health insurance companies raised premiums upwards of 20%, making health care unaffordable for many employers. If those costs are passed onto workers, many—especially those with lower incomes—will be unable to afford the premiums. Even when low-wage workers are offered employer-sponsored health insurance, their plans often include high premiums and deductibles and are more likely to require sizable employee contributions for family coverage premiums. According to the Agency for Health Care Policy and Research, in 1996 the average employee contribution for a health insurance premium for family coverage was $1,922 for employees in firms where more than 50% of the workforce worked at a low-wage level, compared to employee contributions of $1,484 in firms where less than 50% of the workforce was paid a low wage (Agency for Health Care Policy and Research 1997). Making it possible for employers to offer health insurance means working to keep premiums affordable and ensuring that the plans offered actually cover all the health services that workers need.

Health insurance is not only an indicator of a high-quality job but is critical for the health and well-being of families more generally, which has implications beyond employment tenure. Without health insurance, many families have trouble making ends meet, and they are more likely to experience a range of other family hardships (Boushey et al. 2001). Policy can also do more to ensure that families have access to health insurance even if their employer does not provide it. During the 1990s, Congress expanded the Medicaid program to cover more individuals and delink Medicaid from welfare. However, in the typical state, a family of three earning over $7,992 (59% of the poverty guideline) is not eligible for Medicaid coverage (Guyer and Mann 1999). Increasing coverage to more low-wage workers would help to address the health insurance issue.

In the end, however, creating high-quality jobs rests with firms. Recent research has pointed out ways that firms can help workers balance their familial and work responsibilities through policies that recognize these constraints as such. Providing opportunities for shorter work weeks, reduced hours, flexible schedules, and job sharing allows workers greater control over their time so that they can meet all their responsibilities (Appelbaum et al. 2001).

Conclusions
Although the strong economy of the 1990s lowered unemployment and led to real wage gains across the wage distribution, many workers still experienced no real wage growth. Even in the late 1990s, two out of every five workers saw no real wage growth, and a majority had wage growth that hovered around zero. Traditionally, more disadvantaged groups of workers benefited disproportionately from the long boom: women of color, immigrant workers, single mothers, and those with lower starting wages disproportionately benefited, relative to more advantaged workers, during the strong economy of the late 1990s. These groups of workers were more likely to see increases in their length of employment, and they saw larger increases in their probability of experiencing real wage growth.

The wage trend for former welfare recipients was not significantly different from that of women who had not been on welfare. The welfare reforms of the mid-1990s pushed more women off of welfare. If they did find employment, however—even occasional employment—the probability of remaining employed was greater in the late 1990s than it had been before welfare reform.

For many workers, and most welfare recipients, finding and keeping a job begins at home—by finding a way to balance all of life’s responsibilities. For workers with family responsibilities, the capacity to maintain employment depends on whether they have access to safe and affordable child care. Working parents, and especially single mothers, who use center-based care and receive help with child care are more likely to stay employed than those who do not. Among mothers of young children, the probability of remaining employed was greater among those who used center-based care, relative to mothers who did not: 72.7% of unmarried mothers of young children who used center-based care were still employed after two years, as opposed to only 35.5% of those who did not use center-based care. Among welfare recipients, 49.9% and 18.7% of those who did and did not use center-based care, respectively, were still employed after two years. Employment tenure was also longer amon
g those who received help paying for their child care.

The path to helping people stay employed also depends on the quality of their initial job. The better the first job is—whether it has health insurance and whether the starting wages are high—the more likely it is that the individual will remain employed over time. Women with health insurance are more likely to still be employed two years later than those without: among unmarried women, 77.5% who had employer-provided health insurance were still employed after two years. Only 38.9% of those without that benefit were still employed after two years. Among former welfare recipients, health insurance matters even more: 59.0% of those with employer-provided health insurance and 22.7% of those without were still employed after two years. Less-educated workers, who are overall less likely to receive employer-provided health insurance, disproportionately benefit from receiving health insurance when they begin employment. Starting off in a higher-paying job also increases length of employment overall.

Placement in high-quality jobs creates a virtuous cycle that helps workers to maintain employment. The high-quality job lets them reap rewards for their hard work, which leads to longer employment; the increased work experience and skill development leads again to higher wages, reinforcing the cycle. Especially for former welfare recipients who start off at the very low end of the labor market in terms of wages and benefits, placement in the first post-welfare job is critical to helping them move up the job ladder, maintaining their presence in the labor market, and supporting their families independently.

Policy makers who want to help former welfare recipients and other low-wage workers move into the kind of employment that will enable them to support their families have a number of options. Because real wage growth and long-term employment work together, policy makers can push to help people either maintain jobs or find the highest quality starting job. Helping former welfare recipients and other workers maintain employment means helping them balance their familial responsibilities so that they can devote their attention to their jobs. This requires a significant expansion of affordable, safe, and enriching child care. Promoting better quality starting jobs means helping employers to offer health insurance, through policies that make health insurance affordable for employers to offer, and adopting policies that improve starting wages, such as raising the minimum wage and fostering the development of unions in low-wage sectors of the economy.

Moving welfare recipients into any job through WorkFirst programs may not have helped those workers to find the kind of high-quality jobs that would have allowed them to move up the job ladder and maintain employment over the long-term. In order to get mothers, disadvantaged workers, and former welfare recipients into the labor market and help them stay there, we need to ensure that they have access to the kinds of work supports that will help them balance work and family. Without these, it is difficult for them to remain employed; without continued employment it will be difficult, if not impossible, for them to ever enter the virtuous cycle.

This project was generously funded by the Foundation for Child Development and the ILGWU Heritage Foundation.

This project benefited greatly from the contributions of my colleagues. Danielle Gao’s work to pull the SIPP data together facilitated the entire project and Brendan Hill’s efforts to put the results into a presentable format improved the final outcome. The comments and suggestions of Eileen Appelbaum, Chauna Brocht, Helen Connolly, the U.S. Census outreach department (in particular, Margaret Coleman, Carole Popoff, Jeffrey Sisson, and Martha Stinson), Ross Eisenbrey, Bethney Gundersen, John Schmitt, and Jeff Wenger appreciably improved the final product, and I owe them all a sincere debt of gratitude. The errors remain, of course, mine alone.

Appendix: The Survey of Income and Program Participation
This analysis makes use of the 1993 and 1996 panels of the Survey of Income and Program Participation (SIPP). The SIPP is a multipanel, longitudinal survey of the civilian, noninstitutional population in the United States, conducted by the U.S. Census. It is designed to examine issues related to participation in income maintenance programs, such as welfare and unemployment compensation, and contains extensive information on individuals’ backgrounds, employment and earnings, and access to services, including health insurance and child care. Unlike other available longitudinal data sets, such as the Panel Study of Income Dynamics or National Longitudinal Study of Youth, it covers all workers and has monthly, rather than annual, data on employment and earnings.

The SIPP is structured such that one-fourth of the sample is interviewed every month, and each four-month interval in which all sample members are interviewed is termed a wave. During each wave, respondents are asked a set of core questions for the preceding four months, which cover labor market participation, wages, and participation in income support programs, and questions from topical modules that change each wave. The first topical module, for example, includes employment and welfare history and asks questions that allow identification of a history of welfare use, as well as labor market experience prior to the panel. Other modules focus on child care, assets, training history, etc.

Editing the 1996 panel
The first step in setting up the SIPP for this analysis was to create a longitudinal panel for the 1996 SIPP. For each panel of the SIPP, the Census released the core and topical module files, along with a longitudinal file that has been edited for internal consistency and that provides panel weights. However, the 1996 longitudinal file was not available at the time of this analysis. Therefore, we took steps to ensure internal consistency of key variables, following Bureau of the Census (2000).

There were three sets of edits for the SIPP 1996 core data. First, all demographic variables (age, sex, race, and ethnic origin) were recoded so that they were consistent with answers given during the respondent’s first interview. Very small proportions of the demographic variables were fixed: 0.3% of sex, 0.4% of race, and 0.76% of ethnic origin.

Second, where possible, missing waves of data were imputed following the procedures outlined in Bureau of the Census (2000). If a missing wave was bounded on either side with nonmissing waves of data, the missing data were randomly imputed from the preceding or following wave’s data using a random carryover procedure. A value r was randomly assigned to each nonrespondent’s household for each missing wave, where r = 1, 2, 3, or 4. The first r reference months within the missing wave received their imputed values from the fourth month of the preceding wave, and the remaining 4-r reference months received their imputed amounts from the first month of the subsequent wave.

Third, within the SIPP there are problems with the consistency of the unique employer I.D. variable. The variable provides a unique employer I.D. for every job held, for a maximum of two jobs per month. If the respondent changes jobs, the employer I.D. changes (increases one step). If the respondent switches their first and second employer, the employer I.D. remains with the job. For example, assume a respondent’s first job was for a hospital as a nurse (employer

It has been found that there is a high degree of error associated with this variable and associated job characteristics. For the 1990-93 panels, employer I.D.’s are erroneous throughout the sample due to improper merges. The result is that the job characteristics for starting date and employer I.D. erroneously change values. Furthermore, if a r
espondent is not interviewed for a wave, then all employer I.D.’s change, even if the job does not.¹¹

Whether the respondent actually moved to a new job, however, can be determined with a high degree of certainty by examining the variables associated with the job, including starting date, industry, and occupation, at each unique employer I.D.’s first occurrence.¹² This is because the variables ESTLEMP (which indicates whether the respondent reported a change in job), industry, and occupation are not subject to the same errors as employer I.D. and starting date. To correct for erroneous employer I.D.’s, we determine whether job-related characteristics are consistent with their first appearance within the variable for employer I.D. number. Table A1 indicates the rules that we established for fixing the job-related characteristics.¹³ Again, the proportion of data with fixes is relatively small: 0.96% of employer I.D., 0.1% of industry, 0.96% of occupation, and 1.69% of starting dates.

Sample construction
The sample for this analysis was women who were between age 18 and 64. The 1993 panel of the SIPP covers from October 1992 through December 1995. The complete panel is nine waves of data and initially contained 21,823 households and 56,800 individuals. Data from the 1996 panel of the SIPP covers from December 1995 to February 2000. The entire panel is 13 waves of data and initially contained 40,188 households and 95,402 individuals (Westat 2001).¹⁴ Table A2 shows the means and standard deviations for the samples used in the analysis.

The SIPP data are restructured to provide information for each wave, rather than by each month. SIPP data are subject to a seaming problem: individuals often report similar states across the wave, and therefore between-wave transitions are more prevalent than within-wave transitions. For example, individuals are more likely to report a change of jobs or marital status between waves than within waves (Bureau of the Census 2000, 6-5). However, by using the data in “wavely,” rather than monthly increments, the problem may be attenuated in this sample.

Wages are adjusted using the CPI-RS. Wages are calculated from the monthly earnings and hours variables in the last month of the wave. If the respondent had missing earnings data for both their primary and secondary job, then wages were calculated from the hourly wage rate and usual hours per week. In the 1993 panel, 93% of earnings data came from the earnings per month variable, while in the 1996 panel, 90% did. (In the 1996 SIPP, about 1.5% of observations were mistakenly coded as $0.00 when earnings should have benn imputed.) The earnings variable was more likely than the hourly wage variable to contain information. Reported wages of less than $1.00/hour were recoded as missing (less than one-quarter of one percent of all person-waves in each panel).

Table A3 compares wages from the SIPP to the Current Population Survey Outgoing Rotation Group data (CPS-ORG). The annual data from the SIPP differs from the CPS data in that annual wages are comprised of the monthly wages for the individuals in the survey. In the CPS, an individual only appears at most twice given the rotation structure; whereas in the SIPP, the same individual will report their wage each month, providing 12 observations.

The SIPP and the CPS data show similar wages across the wage distribution. However, the SIPP data show falling real wages among men between 1995 and 1998, and the CPS shows increasing wages. The trends among women are less variant, as the SIPP shows positive wage growth except at the median where the CPS shows small positive wage growth. Comparing the SIPP and the CPS in 1995 and 1998, the differences are smaller in the 1996 panel than in the 1995 panel, although the SIPP shows higher wages than the CPS across the wage distribution. This is partially a result of the fact that the year 1995 is the last year of the 1993 panel, thus attrition has reduced the proportion of lower earners.

Employment definition
We define employment as being at least marginally employed. An individual is marginally employed if she worked at least one week during the four-month wave. Employment is self-reported and can be either full time or part time and includes those who are self-employed.

Other definitions of employment were tested, including strictly employed. Individuals are strictly employed if they worked all weeks over the four-month wave. If they did not work for even one week, it was considered a break in their employment spell. Findings do not differ substantively based on whether the strictly employed or at least marginally employed definition is used. Two other definitions were also tested, employed full time each wave during the month and employed the majority of weeks in a month (three if it was a four-week month and four if it was a five-week month). Again, these definitions provided results that were consistent with the two that are presented.

The use of the marginal employment definition does not substantially alter the findings in this report. Figure A-1 shows the survival functions for workers strictly employed and at least marginally employed in the early and late 1990s. The figure shows the percent of workers still employed by wave. It matters which employment definition is used in terms of the proportion of workers maintaining employment, because those at least marginally employed maintain employment longer than those who are strictly employed. Furthermore, the trends over time are not the same across the two definitions. Among workers strictly employed, there was no difference in employment durations over the 1990s. After six waves—two years—47% of workers were still employed in both the early and late 1990s. Among workers marginally employed, employment durations were greater in the late 1990s. After two years, 59.1% of workers were still employed in the early 1990s, and 63.1% were still employed in the late 1990s.

These findings on employment durations are consistent with prior research. Gottschalk and Moffitt (2000) use the SIPP to document job and employment stability in the 1990s. They find that employment exit rates among married men ages 20 to 62 hover around 20-22% per year between 1981 and 1995. This analysis finds that among all women age 18 to 64, 20-30% fall out of employment after one year.

Problems with the SIPP
The SIPP data have a number of known problems, including seaming, attrition and low response for some demographic groups, and other nonsampling errors.

The seaming problem stems from the fact that individuals often report similar states across the wave, and therefore between-wave transitions are more prevalent than within-wave transitions. For example, individuals are more likely to change jobs or marital status between waves than within waves (Bureau of the Census 2000, 6-5). However, since only one-quarter of respondents are interviewed in any month, the seams do not have large effects on cross section estimates. Analyses of spell data found that spell durations of multiples of four months were particularly common, indicating that this may bias our employment spell results. For this reason, and to reduce computation time, employment spells in this analysis are examined by wave, rather than by month.

Attrition and response rates are problems inherent to longitudinal data, as some panelist
s drop out of the survey over time and others enter. Between 15% and 25% of the SIPP was not interviewed in any month, and those months are missing from the sample, although individuals are not categorically deleted if they missed some months of the survey.¹⁵ Those who drop out are likely to be poorer than those who remain in the sample. This poses problems for our analysis because individuals with low employment rates and welfare recipients are less likely to be in the final waves of the SIPP data than are wealthier survey respondents.

Other nonsampling errors in the SIPP include under-coverage, nonresponse, and measurement errors. The effects of nonsampling error on survey estimates are that some demographic subgroups are underrepresented because of under-coverage and nonresponse. This includes young African American males, metropolitan residents, renters, people who changed addresses, and people who were divorced, separated, or widowed. Bureau of the Census (2000) analysis of these generally known problems with the SIPP concludes that:

estimates of income from government sources are generally low related to the amounts reported by administrative sources;
estimates of property income and assets, liabilities, and wealth are relatively poor;
estimates of poverty from SIPP panels prior to 1996 are lower than the CPS;
estimates of the working population differ from the CPS;
estimates of people without health insurance are much lower than the CPS estimates, and there are reasons to believe that the SIPP estimates are more reliable; and
birth statistics compare favorably with the CPS, but both are low relative to records from the National Center for Health Statistics.

A final issue is topcoding. Prior to 1996, there is a topcode of $33,332 for each month and wave income. In most cases, monthly income is topcoded at $8,333 ($33,332 divided by 4). However, if an individual only made a high amount of money in one month, then the topcode may not hold, although $33,332 will hold for the entire wave. Summary variables are not individually topcoded.

Endnotes
1. However, caseloads began to rise during the recession of 2001. Between October and December 2001, caseloads increased in 41 out of 54 states and territories, although the total number of welfare families nationally remained unchanged (Richer 2002).

2. A relatively small number of individuals in the SIPP data were on welfare. In both panels, 4% of individuals began the panel on welfare. Among those present in the first wave of their panel, 6.4% received welfare at some point during the 1993 panel, and 5.8% did so during the 1996 panel.

3. The analysis does not hinge on this specification, as the findings are not substantively different when conducted on continuous employment spells or over the entire panel.

4. We calculate the survival function using the Kaplan-Meier product-limit method and test differences using the log-rank test.

5. Sample sizes were too small to look at child-care subsidies or center-based care among former welfare recipients by educational attainment.

6. The federal government has required family leave and increased funding for child care over the 1990s. The Family and Medical Leave Act gives the right to 12 unpaid weeks off to workers who have worked at their firm for at least 1,250 hours in the past year—just over 24 hours per week—in firms that employ at least 50 people at their site. The time off can be used to care for a new child (including new adoptees), to care for a family member, or for the worker’s own serious health condition. However, this act only covers a little over half of American workers, and those with low incomes may be unable to afford unpaid time off.

7. A low wage is less than $7.71 an hour, and a high wage is more than $19.25 an hour.

8. Among workers overall in 1995, 59.1% had employer-provided health insurance (Mishel, Bernstein, and Schmitt 2000). This analysis only includes women and may explain the lower employer-provided health insurance figure, as women are less likely than men to receive insurance from their employer. Furthermore, the analysis only addresses whether a women had employer-provided health insurance when she became employed. For many workers, their health insurance may be delayed until they have completed a probation period so this may undercount the coverage of workers overall.

9 . Among former welfare recipients, the sample in the upper quintiles was too small to be of statistical value.

10 . The association of a lower starting wage quintile with a greater probability of real wage growth is partially a result of the way that the analysis was done: workers at the bottom have nowhere to go but up. This issue is addressed in the next section of the report.

11. This problem was based on conversations with Martha Stinson at the Census. She reported that, although many knew of the problem, there was no documentation at that time.

12. Conversations with Martha Stinson at the Census revealed that she had conducted analysis using the internal employer name variables and found that this is an appropriate fix for the data.

13. These rules only reflect the permutations of errors that we were able to identify, and thus they may not cover all errors found in the data.

14. In 1996, the U.S. Census Bureau changed the way it administered the SIPP to computer assisted survey information collection. In addition, a panel is now introduced every four years, rather than the previous overlapping design. The 1996 panel contains about 37,000 households compared to 20,000 for previous panels, and the 1996 panel oversamples the low income population. Census researchers have not found that the redesign affected outcomes (Hockd and Winters).

15. The U.S. Census imputes some values for those missing months.

References
Adams, Gina, and Monica Rohacek. 2002. Child Care and Welfare Reform. Washington, D.C.: The Brookings Institution.

Administration for Children and Families and Office of Planning Research and Evaluation. 2000. Temporary Assistance for Needy Families (TANF) Program: Third Annual Report to Congress. Washington, D.C.: U.S. Department of Health and Human Services. <http://www.acf.dhhs.gov/programs/opre/annual3.doc>.

Agency for Health Care Policy and Research. 1997. Trends in Personal Health Care Expenditures, Health Insurance and Payment Sources, Community Based Population, 1997-1995. Washington, D.C.: Agency for Health Care Policy and Research.

Anderson, Patricia M., and Phillip B. Levine. 1999. “Child Care and Mothers’ Employment Decisions.” Paper presented at Labor Markets and Less-Skilled Workers, Joint Center for Poverty Research, Washington, D.C.

Appelbaum, Eileen, Thomas Bailey, Peter Berg, and Arne L. Kalleberg. 2001. Shared Work, Valued Care: New Norms for Organizing Market Work and Unpaid Care. Washington, D.C.: Economic Policy Institute.

Blank, Helen, Karen Schulman, and Danielle Ewen. 1999. State Prekindergarten Initiatives, 1998-1999. Washington, D.C.: Children’s Defense Fund.

Boushey, Heather. Forthcoming. “‘This Country is Not Woman-Friendly or Child-Friendly’: Talking about the Challenge of Moving from Welfare-to-Work.” Journal of Poverty 6 (2).

Boushey, Heather, Bethney Gundersen, Chauna Brocht, and Jared Bernstein. 2001. Hardships in America: The Real Story of Working Families. Washington, D.C.: Economic Policy Institute.

Bureau of
Labor Statistics. 2002. Employment Characteristics of Families. <http://www.bls.gov/news.release/famee.toc.htm>.

Bureau of the Census. 2000. Survey of Income and Program Participation: User’s Guide. Vol. 1. Washington, D.C.: U.S. Department of Commerce, Economic Statistics Administration.

Burtless, Gary. 1995. “Employment Prospects of Welfare Recipients.” In D. S. Nightingale and R. H. Haveman, eds. The Work Alternative: Welfare Reform and the Realities of the Job Market. Washington, D.C.: The Urban Institute.

Cabrera, Natasha, Robert Hutchens, and H. Elizabeth Peters. 2002. “From Welfare to Child Care.” Research News 6 (2): 11-14.

Campbell, Nancy Duff, Judith Appelbaum, Karin Martinson, and Emily Martin. 2000. Be All That We Can Be: Lessons From the Military for Improving Our Nation’s Child Care System. Washington, D.C.: National Women’s Law Center.

Campbell, Nancye, John K. Maniha, and Howard Rolston. 2002. Job Retention and Advancement in Welfare Reform. Washington, D.C.: The Brookings Institute. No. 18.

Cancian, Marie, and Daniel R. Meyer. 2000. “Work after Welfare: Women’s Work Effort, Occupation, and Economic Well-being.” Social Work Research 24 (2): 69-86.

Card, David, Charles Michalopoulos, and Philip K. Robins. 2001. “The Limits to Wage Growth: Measuring the Growth Rate of Wages for Recent Welfare Leavers.” National Bureau of Economic Research, No. 8444. August. Cambridge, Mass.

Center for Law and Social Policy. 2001. Frequently Asked Questions About Working Welfare Leavers. Washington, D.C.: Center for Law and Social Policy. <http://www.clasp.org/pubs/TANF/leaversFAQ.pdf>.

Committee on Ways and Means. 2000. 2000 Green Book. 17th ed. Washington, D.C.: U.S. House of Representatives.

Connelly, Rachel, and Jean Kimmel. 2000. Marital Status and Full-time/Part-time Work Status in Child Care Choices. March.

Connolly, Helen, and Peter Gottschalk. 2000. “Returns to Tenure and Experience Revisited: Do Less Educated Workers Gain Less From Work Experience?” Boston, Mass.: Boston College.

Duncan, Greg J., and Jeanne Brooks-Gunn. 1997. “Income Effects Across the Life Span: Integration and Interpretation.” In G. J. Duncan and J. Brooks-Gunn, eds. The Consequences of Growing Up Poor. New York, N.Y.: Russell Sage Foundation.

Friedman, Dana E. 2001. “Employer Supports for Parents with Young Children.” Future of Children 11 (1): 63-77.

Fuller, Bruce, Sharon L. Kagan, Gretchen L. Caspary, and Christiane A. Gauthier. 2001. “Welfare Reform and Child Care Options for Low-Income Families.” Future of Children 11 (1): 97-119.

Galinsky, Ellen, and J. T. Bond. 2000. “Supporting Families as Primary Caregivers: The Role of the Workplace.” In D. Cryer and T. Harms, eds. Infants and Toddlers in Out-of-home Care. Baltimore, Md.: Paul Brookes Publishing.

Giannarelli, Linda, and James Barsimantov. 2000. Child Care Expenses of America’s Families. Occasional Paper Number 40. Washington, D.C.: Urban Institute.

Gladden, Tricia, and Christopher Taber. 2000. “Wage Progression Among Less Skilled Workers.” In D. Card and R. Blank, eds. Finding Jobs: Work and Welfare Reform.

Gottschalk, Peter. 2000. “Wage Mobility Within and Between Jobs: How Prevalent is Downward Mobility?” Paper presented at conference entitled Low-Wage Employment, Earnings Mobility, and the Eurpoean-American Employment Gap, University of Aberdeen. Nov. 17-18.

Gottschalk, Peter, and Robert A. Moffitt. 2000. “Job Instability and Insecurity for Males and Females in the 1980s and 1990s.” In D. Neumark, ed. On the Job: Is Long-term Employment a Thing of the Past? New York: Russell Sage Foundation.

Guyer, Jocelyn, and Cindy Mann. 1999. Employed but Not Insured. Washington, D.C.: Center on Budget and Policy Priorities.

Hockd, Elaine C., and Franklin Winters. “Comparing certain effects of redesign on data from the Survey of Income and Program Participation.” Census Bureau. No. 226.

Holzer, Harry, and Michael Stoll. 2000. “What Happens When Welfare Recipients are Hired?” Washington, D.C.:Georgetown Public Policy Institute.

Layzer, Jean I., and Ann Collins. 2001. Access to Child Care for Low-Income Working Families Report. Washington, D.C.: U.S. Department of Health and Human Services.

Lemke, Robert J., Ann Dryden Witte, Magaly Queralt, and Robert Witt. 2000. “Child Care and the Welfare to Work Transition.” National Bureau of Economic Research. No. 7583.

Loprest, Pamela. 1999. How Families that Left Welfare are Doing: A National Picture. Washington, D.C.: Urban Institute. No. B-1. <http://newfederalism.urban.org/html/series_b/anf_b1.html>.

McCrate, Elaine. 2002. Working Mothers in a Double Bind: Working Moms, Minorities have the Most Rigid Schedules, and are Paid Less for the Sacrifice. Washington, D.C.: Economic Policy Institute.

Mishel, Lawrence, Jared Bernstein, and John Schmitt. 2000. The State of Working America 2000-01. Ithaca, N.Y.: Cornell University Press.

Moffitt, Robert, and Anuradha Rangarajan. 1989. “The Effect of Transfer Programmes on Work Effort and Human Capital Formation: Evidence From the U.S.” In A. Dilnot and I. Walker, eds. The Economics of Social Security: Oxford, England: Oxford University Press.

Newman, Katherine S. 2000. “In the Long Run: Career Patterns and Cultural Values in the Low-Wage Labor Force.” Harvard Journal of African American Public Policy 6 (1).

Pavetti, LaDonna, and Gregory Acs. 1997. Moving Up, Moving Out or Going Nowhere? A Study of the Employment Patterns of Young Women. Washington, D.C.: The Urban Institute.

Rangarajan, Anu. 1996. Taking the First Steps: Helping Welfare Recipients Who Get Jobs Keep Them. Princeton, N.J.: Mathematica Policy Research, Inc.

Rangarajan, Anu, Peter Schochet, and Dexter Chu. 1998. Employment Experiences of Welfare Recipients Who Find Jobs: Is Targetting Possible? Princeton, N.J.: Mathematic Policy Research, Inc.

Ribar, David. 2000. “Transitions From Welfare and the Employment Prospects of Low-skill Workers.” September. Washington, D.C.: Georgetown University.

Richer, Elise. 2002. Welfare Caseloads Increased in Most States in the Last Quarter of 2001: Caseloads Up in Most States between September and December 2001 Caseloads Also Up Throughout 2001. Washington, D.C.: Center for Law and Social Policy.

Smith, Kristin. 2000. Who’s Minding the Kids? Child Care Arrangements. Washington, D.C.: U.S. Census Bureau. No. P70-70.

Strawn, Julie, Mark Greenberg, and Steve Savner. 2001. Improving Employment Outcomes Under TANF. Washington, D.C.: Center for Law and Social Policy.

Strawn, Julie, and Karin Martinson. 2000. Steady Work and Better Jobs: How to Help Low-Income Parents Sustain Employment and Advance in the Workforce. New York: Manpower Demonstration Research Corporation.

U.S. Department of Human Services. 1999. National Study of Child Care for Low-Income Families. Washington, D.C.: U.S. Department of Human Services.

Wadfogel, Jane. 1998. “Understanding the ‘Family Gap’ in Pay for Women with Children.” Journal of E
conomic Perspectives 12 (1): 137-56.

Wadfogel, Jane. 2001. “International Policies toward Parental Leave and Child Care.” Future of Our Children 11 (1). <http://www.futureofchildren.org>.

Westat. 2001. Survey of Income and Program Participation User’s Guide. Edited by I. Mathematica Policy Research. 3rd ed.Washington, D.C.: U.S. Census Bureau. Whitebrook, Marcy, Carolee Howes, and Deborah Phillips. 1998. Worthy Work, Unlivable Wages. National Center for the Early Childhood Workforce.

Wiatrowski, William. 1995. “Who Really Has Access to Employer-Provided Health Benefits?” Monthly Labor Review June.

Witte, Ann Dryden, Magaly Queralt, Tasneem Chipty, and Harriet Griesinger. 1999. “Unintended Consequences? The Welfare Reform ‘Package’ and the Earnings of Low-income Women.”June. Wellesley, Mass.:Wellesley College.

Zaslow, M. J., E. Oldham, K. A. Moore, and E. Magenheim. 1998. “Welfare Families Use of Early Child Care and Education Programs, and the Implications for their Children’s Development.” Early Childhood Research Quarterly 13 (4): 535-63.

Zinsser, Caroline. 2002. “Child Care within the Family.” Future of Children 11(1): 123-7.

Staying Employed After Welfare

Briefing Paper #128

Work supports and job quality vital to employment tenure and wage growth

Sign up to stay informed

Track EPI on Twitter