The welfare myth: disentangling the long-term effects of poverty and welfare receipt for young single mothers.

McNamara, Justine M.

This study investigates the effects of receiving welfare as a young woman on long-term economic and marital outcomes. Specifically, we examine if there are differences between young, single mothers who receive welfare and young, single mothers who are poor but do not receive welfare. Using the 1968-1997 Panel Study of Income Dynamics, our findings suggest those who receive welfare for an extended period as young adults have the same pre-transfer income over a 10 to 20 year period as those who are poor but do not receive welfare as young adults. While we found some differences between the two groups in income levels and the likelihood of having relatively low income when control variables were not included in our models, once appropriate controls were used, these differences became statistically insignificant. The only statistically significant difference found between the two groups in our 10, 15, and 20 year models was the likelihood of being married in year 15. Our results indicate that it is income level as a young adult, as well as such factors as the unemployment rate in the area of residence, but not welfare receipt, that affect long-term income and marital outcomes.

Key words: AFDC, TANF, PRWORA, Poverty, Dependency

**********

The welfare reform legislation in 1996 was motivated in part by the belief that welfare recipients had become too dependent on welfare, especially those who began using welfare at a relatively young age and who continued to rely on it for many years. Many believe that policies intended to alleviate poverty instead intensified economic problems for the poor by making them less self-reliant. In particular, arguments have been made that welfare receipt reduces earnings and decreases the likelihood of marriage (Herrnstein and Murray, 1994; Horn, 2002; Mead, 1986, 1998; Murray, 1984, 2001).

In response, the Personal Responsibility and Work Opportunity Reconciliation Act of 1996 (PRWORA) replaced the Aid to Families with Dependent Children (AFDC) program with Temporary Assistance for Needy Families (TANF). This law included provisions for limiting welfare receipt by young mothers, including requiring teen mothers to live at home with their parents. Since the passage of this Act, there has been a drastic reduction in the number of recipients of welfare. Nationally, there has been a 50 percent decline in the number of welfare recipients from 1996 to 2002 (Committee on Ways and Means, 2004). While PRWORA gives states greater authority over their welfare populations, it also imposes strong federal rules designed to increase work and decrease the length of time for welfare receipt (Albelda and Tilly, 1997). The welfare rolls may have also decreased because of the strong U.S. economy in the latter half of the 1990s, although we saw no subsequent increase in rolls when the economic conditions worsened after this period.

While government policies reflect concerns about the potential negative effects of welfare, there has been relatively little research on the actual nature of these effects in the long-term. In particular, we know little about how relatively young recipients of welfare fare economically over a 10 to 20 year period compared with young single mothers who are poor but do not receive welfare. Do young welfare recipients fare worse than those young women who have children, are poor but who do not receive welfare? And, how do those who receive welfare or are poor fare economically relative to those single mothers who are not poor at a relatively young age? This study examines these issues in order to determine if making welfare less available to relatively young women appears likely to change their probability of being married or their level of economic well-being later in life. We focus particularly on young women because it has been hypothesized or inferred that the negative effects of welfare may be more severe for young recipients (Mead, 1986; Murray, 1984; Tanner, 1996).

Our study examines income and marital status at 10, 15, and 20 years after initially receiving welfare, including the effects of the length of welfare receipt on these outcomes. While welfare proponents suggest that long-term welfare receipt has negative overall outcomes for single mothers, others have suggested that it is the impact of poverty that has deleterious effects on long-term income and marital status. We thus also examine how relatively short-run or longer-run stays in poverty or near-poverty affect economic outcomes 10 to 20 years after initially falling into poverty and leading a household as a young single mother. In particular, we focus on the differences between poor women who spend a considerable amount of young adulthood without receiving welfare and those women who spend a relatively long period receiving AFDC in young adulthood.

Previous Research

Despite public policy concerns about the effects of welfare, there has been little research that directly examines the effects of receiving welfare at a young age on an individual's economic outcomes later in life. A considerable amount of existing research about the possible negative effects of welfare has focused on welfare-dependence issues such as the long-term use of welfare, and the intergenerational transfer of welfare dependence. Research examining the intergenerational use of welfare has generally found that adults whose parents used welfare may be more likely to themselves use welfare (An et al., 1993; Dolinsky, Caputo & O'Kane, 1989; Gottschalk, McLanahan, & Sandefure, 1994; Hill and Ponza, 1989; McLanahan, 1988; Rank and Cheng, 1995). Gottschalk (1992) and Vartanian (1999) found that for blacks, but not whites, parental welfare receipt strongly predicted welfare receipt of the daughter, although other childhood or adolescent factors such as income level and education of the head of household also contributed to higher likelihoods of welfare receipt. However, these studies only follow young parents through early adulthood.

Research on the length of AFDC receipt has generally found that characteristics such as non-completion of high school, little work experience and having young children tend to be associated with relatively lengthy receipt of AFDC (Bane and Ellwood, 1994; Blank, 1989; Ellwood, 1986; Fitzgerald, 1991; O'Neill, Bassi, and Wolf, 1987; Vartanian, 1997). Other studies on the likelihood of leaving AFDC have found that residential location makes a difference in AFDC exit probabilities (Fitzgerald, 1995; Gleason, Rangarajan, and Schochet, 1998; Vartanian, 1997). Long periods on AFDC may include either a long single spell on AFDC or several spells (Bane and Ellwood, 1994; Harris, 1996). Harris (1996) found, for example, that 36 percent of women who leave AFDC return within 18 months after first leaving and that 57 percent return within seven years after first leaving.

It has been hypothesized that women who receive welfare for an extended period of time may be less likely to marry and leave welfare relative to those who stay on welfare for a short time (Blank, 1989). One reason for this lower likelihood may be because women develop a "taste" for welfare after receiving aid for an extended period rather than a reliance on spouses or their own labor income for self-sufficiency (Blank, 1989). Generally, studies that have examined the effects of welfare benefits have found little or no association between benefit levels and marriage disincentives (Bane and Ellwood, 1994; Gottschalk et al., 1994; Moffitt, 1992; Wilson & Neckerman, 1986). It has also been hypothesized that the use of AFDC may change household composition by increasing family size (to receive higher benefits), and thereby potentially decreasing the likelihood of marriage in the future. Blank and others (Bane and Ellwood, 1994; O'Neil, Bassi and Wolf, 1987) find relatively low rates of exits from AFDC by means of marriage but also find little evidence that the likelihood of exit from AFDC by means of marriage decreases over the length of an AFDC spell. Part of the reason for the low likelihood of leaving welfare by means of marriage (or for poor women being unmarried) may be due to the low numbers of "marriageable" men, especially for African Americans (Coontz and Folbre, 2002; Lichter, McLaughlin, LeClere, Kephart, and Landry, 1992; Wilson, 1987). Edin and Lein, (1997) find that poor single mothers often do not marry because the risks of marriage, such as the threat of domestic violence, loss of control, and potential for sexual abuse of their children, often outweigh the potential gains from entering into marriage.

Welfare use may also decrease future labor force participation by decreasing labor market experience (Blank, 1989). Lower levels of work experience due to welfare receipt may mean lower wages and thus lower income in the future. Blank (1989) and Bane and Ellwood (1994) found evidence that the longer the length of time on welfare, the lower the likelihood of exiting welfare from earnings, although Vartanian (1997) did not find evidence of such a relationship.

The concern reflected in the 1996 welfare reform law that the combination of teen childbearing and welfare use may have particularly negative effects on economic outcomes for young women has prompted research focused on the interaction between young childbearing, welfare use and later economic outcomes. Duncan and Hoffman (1990), for example, use the Panel Study of Income Dynamics (PSID) to examine whether African American teens who have children outside of marriage and who use welfare have lower economic outcomes when they reach age 26 than those who do not have children as teenagers or use welfare. After controlling for many background variables, they found that those teens who used welfare within two years of the birth of their child, whether married or not, fared far worse than those who did not. While these results suggest that the receipt of welfare is the critical factor in determining later-adult economic outcomes, sister-pair research reaches somewhat different conclusions. Corcoran and Kunz (1997) also used the PSID to examine the effects of the interaction of teen childbearing and welfare use on economic outcomes at age 26 or slightly beyond, using sister pairs in order to decrease the level of heterogeneity of the comparison groups. They found that once family background differences are controlled, the effects of having a child as a teenager and receiving welfare were less than shown in previous research. For example, the negative effects on family income for teen mothers dropped by more than 50 percent when controlling for the effects of family background in a sister-pair model as compared to a model that did not control for such factors. Sample sizes in this sister-pair study were small, however: there were 31 sister pairs where one received AFDC as a teen mother and the other sister did not. Also, the time frame for examining outcomes is relatively short in the study--soon after age 25.

Studies examining the economic situation of broader groups of former welfare recipients have generally found high poverty rates among these women (Cancian and Meyer, 2000; Meyer and Cancian, 1998; Vartanian and Gleason, 1999; Vartanian and McNamara, 2000). Using data from the Survey of Income and Program Participation, Vartanian and Gleason (1999) examined economic status in the initial period after leaving welfare and found that over 50 percent of former recipients were poor 18 months after leaving welfare, while Vartanian and McNamara (2000), using data from the PSID, found 38 percent poverty rates four years after leaving welfare. Meyer and Cancian (1998) found somewhat higher (50%) poverty rates five years post-welfare among women in the National Longitudinal Survey of Youth, although in subsequent work these authors noted an improvement in women's economic status over time, with poverty rates lower in the second and subsequent years than in the first year after welfare (Cancian and Meyer, 2000). These studies, however, generally failed to find strong links between the length of time women spend on welfare and the likelihood of poverty after leaving welfare, finding instead that factors such as education and the employment status of both the recipient and her spouse had powerful effects on the likelihood of poverty after welfare (Cancian and Meyer, 2000; Meyer and Cancian, 1998; Vartanian and Gleason, 1999; Vartanian and McNamara, 2000).

The studies on the economic effects of welfare described above have a number of limitations: they usually examine outcomes only a few years following the receipt of welfare, they generally do not compare the outcomes of welfare recipients with poor nonrecipients, and a number of studies are limited by small sample size and insufficient control of background variables.

Using the 1993 PSID, Dunifon (1999) examined longer-term outcomes for AFDC recipients by comparing women who received any AFDC income between 1968 and 1972 with women who were heads of households, had children and were between 21 and 39 in those years but did not receive welfare during that period. She found that 20 years following this initial period, women who had received welfare had similar income-to-needs ratios and number of hours of work as the non-welfare group, although differences in hours of work did emerge in some of the periods examined prior to the twenty-year period. Dunifon's study was not designed to examine the effects of AFDC receipt by young single mothers on long-term outcomes, but instead she determined the effects of receiving any AFDC relative to no cash assistance for single mothers at varying ages. Also, she examined only post-transfer income later in life (that is, total income including all transfer payments such as General Assistance and AFDC). Pre-transfer income excludes all transfer payments, including government assistance. Examining pre-transfer income later in life allows for a better assessment of the effects of welfare on single mothers' self-sufficiency. Our study addresses these issues by examining marriage rates and pre-transfer income of a relatively large sample of young mothers to assess the effects of substantial (at least 10 percent of total income) and early welfare receipt.

Methodology

Sample

This study uses data from the 1997 Panel Study of Income Dynamics (PSID). The PSID is a longitudinal data set that dates back to 1968. In 1968, there were approximately 5,000 families in the sample and 18,000 individuals. By 1997, the data contained over 6,000 families and 19,000 individuals. The longitudinal nature of the PSID allows for the examination of economic and other outcomes across a 30-year span, by selecting data on women at an early period in the survey, and then selecting data for these same women in periods that are approximately 10, 15, and 20 years after we initially examine their characteristics. When appropriately weighted, the PSID is representative of the non-immigrant, United States population. We used the individual weights in the PSID when examining our sample.

We chose three samples of women that would allow us to examine their characteristics at a relatively early age and then examine these same women again at 10, 15 and 20 years after our initial examination. First, we chose a sample of women who started in the PSID anytime from 1968 to 1987 in order to get their characteristics after 10 years (our final sample year was 1997) for our 10 year sample. Next, we chose a sample of women from 1968 to 1982 and looked to see how they were doing 15 years later. Last, we chose a sample of women from 1968 to 1977 to see how they fared 20 years hence.

We then further limited our samples by choosing women who were age 24 and under, who were heads of household when they first entered the sample, and had one of the following three characteristics: AFDC income made up at least 10 percent of their total income; (1) total income was at or below 150% of the poverty line and AFDC income made up less than 10 percent of their total income during this period; or, AFDC income made up less than 10 percent of total income and total income was above 150 percent of the poverty line for this period. (2) For example, a woman who was a head of household who first received a substantial amount of AFDC at age 20 would have her initial period begin at age 20. A woman who was first a head of household and had income at or below 150 percent of the poverty line at age 22 and had no substantial AFDC income at or before age 24 would have her initial period begin at age 22. A woman who was first a head of household at age 17 and had income above 150 percent of the poverty line for all years while she was a head of household until age 24 and never received a substantial amount of AFDC would have her initial period begin at age 17. Using this design allowed us to determine those female heads of household who ever received a substantial amount of AFDC at a young age, those who were ever poor or near-poor during this period without substantial AFDC receipt, and those who avoided both substantial AFDC receipt and low income during their early adult years.

We chose only single women who were mothers in order to compare similar young women who had children, as well as to focus our study on the effects of early AFDC use, rather than the effects of early childbearing. We chose women who were heads of household because the PSID contains specific income, education and other information on these women that is not available for those who are children or have some other relationship to the head of household. Also, by choosing only single mothers, we felt that our comparison groups were more comparable than by also allowing women who were married into our sample at the beginning of the period.

From these samples we wished to determine if the use of AFDC as a young woman helped to predict long-term outcomes for a number of dependent variables. Critics of welfare hypothesize that it is early welfare use that causes women to be dependent on government aid, to be poor, and to be less likely to be married (Mead, 1986, 1998; Herrnstein and Murray, 1994). According to this hypothesis, relatively long-term use of AFDC as a young woman should have a more detrimental effect on outcomes than shorter-term AFDC use because women grow more dependent on AFDC the longer they use it, and because labor market skills deteriorate the longer a person is out of the labor market. In order to examine this hypothesis, we chose to compare women who received AFDC, either for a relatively short or long period, with women who were poor or near poor during young adulthood. That is, we wanted to determine whether the receipt of AFDC, and the poverty that generally accompanies AFDC, affects outcomes relative to the non-receipt of AFDC and poverty. We also included in our sample non-poor women in order to determine the relative impact of having higher levels of income at an early age on long-term outcomes. We therefore examined whether the long-term or short-term receipt of welfare at an early age directly affected long-term economic and marital outcomes, with controls for family and personal circumstances and other factors at this early age. Our goal was not to determine the possible effects of factors after this early period of a woman's life on these long-term outcomes, but instead to be able to predict from a given set of circumstances early in adulthood, outcomes later in life.

Our primary interest in this study is the comparison between single mothers who receive AFDC for a long time during early adulthood and single mothers who were poor for an extended period of time but did not receive AFDC. These two groups were of primary interest because it is concern about long-term dependency on welfare that underlies welfare reform. In the descriptive results we present below, we found that the long-term poor group and the long-term AFDC group were of similar ages and had similar income levels when we initially examined them. We also found that the two groups spend almost the same amount of time with income levels below 150 percent of the poverty line at the beginning of the sampling period. However, most of the income for the AFDC group during the initial four year period was from transfer payments, with a median level of 67 percent (with a median of 65 percent for welfare payments), while the median level of transfers for the long-term poor group was 16 percent (with a median of 0 percent for welfare payments). Thus, the long-term poor group appear to represent those single mothers who remain in or near poverty for an extended period in young adulthood but rely far less on financial assistance than the long-term AFDC group.

Independent Variables

Once our sample was determined, we examined a four year period during and after the initial year that the woman entered the sample. During this four year period, we examined whether AFDC recipients stayed on AFDC for one or two years, or three or four years in this four year period after they initially became heads and started receiving AFDC. Those AFDC recipients receiving AFDC for one or two years were labeled the short-term AFDC group. Those who stayed on AFDC for three or four years were labeled the long-term AFDC group. The low-income, non-AFDC sample, were grouped in a similar fashion, whereby those who had incomes at or below 150 percent of the poverty line for one or two years were labeled as the short-term poverty group and those who had low incomes for three or four years were labeled as the long-term poverty group. We use 150 percent of the poverty line instead of the poverty line because many critics of the current measure of poverty claim that the poverty line is too low, and 150 percent of the poverty line better captures a realistic measure of what is necessary to survive (Edin and Lein, 1997; Smeeding, 1992). These poverty or near poverty groups are simply indicators of low income without the receipt of AFDC. We named those who neither received a substantial amount of AFDC nor had incomes at or below 150 percent of the poverty line the "non-poor group". Each of the five groups--the short-term and long-term AFDC and poverty or near-poverty groups, and the non-poor group--was mutually exclusive. (3)

We then created dummy variables for each of these five groups. In the regression models that we ran, we chose four of these groups as included variables within our regression analyses--the short-term and long-term AFDC groups, the short-term poverty group, and the non-poor group. The long-term poverty group was the excluded category in the regression analyses. We excluded the long-term poverty group in order to easily determine if receiving AFDC, either short-term or long-term, was a key factor in explaining future income and other outcomes or if AFDC recipients fared similarly to those who were poor for an extended period of time. We also ran models where the short-term poverty group was the excluded group in our regression analyses. Our main results were little changed from what we present here when we compared the short-term poor group and the long-term or short-term AFDC groups. We also ran models that excluded from our samples women who were in the non-poor group. We discuss these results below.

We then created other variables, including a variable that indicated whether the family had one child (the included group in our regressions) or more than a single child. We also control for the effects of having very young children when first entering the sample by using a variable in our models for whether the woman has a child who is below age 3. We did this because some women who enter our sample, especially those in their twenties, may have children who are older. We also control for the age of the woman by creating three dummy variables for age: one variable that has a value of one for those under the age of 18, another dummy variable that takes a value of one for those who are 18 to 21, and another (the excluded group) for those who are over the age of 21. We use this set of dummy variables in our models because of the possible non-linear effects of age--or the potentially highly negative effects of being very young and being a head of household--on our outcomes.

In addition, we control for the amount of money that individuals receive from relatives over a four year period, as an indication of family support. Family money support could either have negative or positive effects on long-run economic outcomes. If women become dependent on this support and work less because of it, family income support could have negative effects on future income. If women who use this family income to help them increase training or help them find work, the effects may be positive on future income.

In order to determine the effects of childhood circumstances, we control for the effects of growing up poor relative to not growing up poor for the head of the woman's household. Many of the problems associated with adult outcomes may stem from low income during adolescence or childhood. While we did not have enough years of data to directly examine the level of income for the woman during childhood, the PSID provides a variable that indicates the level of income of the head of household while growing up. To further examine the woman's childhood circumstances, we controlled for the education level of the head of household and the occupation of the head of household while the woman was growing up.

We also control for a number of other variables when the woman first enters our sample in our models, including the county unemployment rate, the region of residence, education, race, city size, and the year in which the individual first entered the sample. We also control for whether the woman had any work disabilities at the beginning of the period we examined.

Our models also include a continuous variable for family income-to-needs in the initial four year period. One reason to use such a variable is because of the potential for some families to be in the long-term poverty group but be just below 150 percent of the poverty line, or be in the non-poor group and be just above the poverty line. Such a variable would also give us a way of comparing similar AFDC recipients who may have relatively high income levels if they live in a more generously paying welfare state or who may have relatively low levels of income. We in fact found a good amount of variation of income levels within groups. We found that our standard errors changed little, and often decreased, when we included beginning income in our regression models relative to when we did not include it, indicating that the collinearity among our AFDC and poverty group variables and income was not causing the significance levels to decrease because of potentially high standard errors.

We included several variables at the end of the 10 to 20 year examination periods, including the unemployment rate of the state or county of residence, the maximum welfare payment available in the state of residence for a family of four (each states sets their own level of cash assistance available to those eligible for AFDC, and these levels differ by factors such as income and family size), and whether work was limited by disabilities. These measures control for economic conditions and physical/mental factors associated with work.

There are a number of variables that we contemplated using in our analyses but did not because of their collinearity with AFDC receipt. These variables included hours of work, AFDC maximum payments available in the state of residence, and marital status, all over the initial four year period. When we did include these variables in our analyses, our results changed little from the main results reported here.

Dependent Variables

We chose a number of dependent variables to examine the effects of early adulthood AFDC receipt. As we have stated above, critics of the welfare system hypothesize that young welfare recipients are less likely to marry because they are in less need of additional income that being married may provide them. We test this hypothesis by examining the effects of receiving AFDC, either long-term or short-term, on the likelihood of being married later in life. We examined this hypothesis with the set of independent variables we described above.

Critics of the AFDC system also hypothesize that those who received welfare at an early age would be more likely to be poor or have low income later in life, claiming that being on welfare teaches young women not to work or develop the skills necessary for high earnings. We test this hypothesis by examining the woman's pre-transfer family income-to-needs ratios and the likelihood of having pre-transfer income below 150 percent of the poverty line later in life. We use pre-transfer income-to-needs instead of earnings to get a better measure of overall well-being of the woman without government cash assistance. We do this, in part, because women are sometimes the secondary earners in the family and may devote more time to child care than the primary earners. We use 150 percent of the poverty line because the poverty line is an extremely low measure of economic adequacy (as briefly described above).

We measured our set of dependent variables at three different points after first examining our sample of young women. We measured our income variables over a two year period in order to smooth out large variations in any single year. We also examined marital status over two year periods to see if the woman was married for either of the two years. Each dependent variable was measured after 9 and 10 years, 14 and 15 years, and 19 and 20 years after we first examined respondents' characteristics as young women.

In results not shown, we also tested to determine if particular groups were more or less likely to drop out of the sample before the end of our first testing period. We only examined those women who had completed the four years of the initial period because we wanted to determine which of our groups they belonged to--the long-term AFDC or poverty groups, the short-term AFDC or poverty groups, or the non-poor sample. One hundred and nineteen women did not make it to the 10 year ending period. We did not find that the AFDC groups were any more or less likely to drop out of the sample relative to the high poverty group, both with a full set of controls for beginning circumstances and using bivariate regressions. In fact, none of the groups showed any higher likelihood of dropping out of the sample relative to other groups. Factors such as disability status and growing up poor showed positive effects for dropping out of the sample.

Statistical Methods

We used logistic regression analysis to determine if the either of the AFDC groups were more likely to drop out of the sample than the long-term poverty group before the 10 year period. We use ordinary least squares regression analysis in our models to examine pre-transfer family income-to-needs. The likelihood of marriage and the likelihood of being at or below 150 percent of the poverty line later in life use logistic regression models because of the binary response dependent variables. In our statistical models, we use ten percent significance levels (or better), instead of five percent levels, to better see if any differences are found among the groups we examined.

Results

Descriptive Statistics

Table 1 presents mean values and standard deviations for women when they first started in the sample and who lasted in the sample for at least 10 years. There is some similarity between the short-term poverty and short-term AFDC groups as well as the long-term poverty and long-term AFDC groups in their level of mean income over the first four years they are in the sample. Both the short-term AFDC and poverty groups have mean income levels that are above the poverty line. The long-term AFDC group has income that is 9 percent below the poverty line while the long-term poverty group has income that is 5 percent above the poverty line for their average levels of income. Thus, these long-term AFDC and poverty groups begin their sampling period at roughly equal levels of income. The non-poor group have income levels at the beginning of their sampling periods that are nearly 3 times the poverty line.

These initial comparisons examine post-transfer income in the initial period. Part of this income for the AFDC group is from government assistance, specifically from AFDC. Table 1 shows that 11.09 percent of total income for the short-term AFDC group comes from AFDC income while AFDC makes up 56.44 percent of income for the long-term AFDC group. The table shows that the poor groups and the non-poor groups receive almost none of their income from AFDC.

The poor and the AFDC groups have similar numbers of children while the non-poor have only 1.21 children, on average. Not surprisingly, we find that the long-term AFDC group is far less likely to be married than the other groups. We also find that the proportion of whites in both the long-term AFDC and long-term poor groups to be far smaller than for the other income/welfare groups.

The level of education for the AFDC groups is lower than for the poor groups, when examined by short-term and long-term status. For example, 48.70 percent of the short-term AFDC group never finished high school, while 36.05 percent of the short-term poor group never finished high school. More than 68 percent of the long-term AFDC group never finished high school, as compared to 53.02 percent of the long-term poverty group and 21.15 percent of the non-poor group.

Table 2 show the means and standard deviations or percentages at different periods of time after the initial examination of the different groups of women, as well as whether there are significant differences (without control variables) between each of the four groups (both AFDC groups, the short-term poverty group and the non-poverty group) and the long-term poverty group, for each period examined. The top portion of Table 2 shows that the proportion of women who are married is significantly lower for the long-term AFDC group relative to the long-term poor group. However, by year 20, only 31.0 percent of the long-term poor group are married, as compared to 37.9 percent of the long-term AFDC group. (4) There is a gradual increase in the rates of marriage for the short-term AFDC group, while the short-term poverty group and the non-poor group see marriage rates increase in year 15 but fall by year 20.

The middle portion of Table 2 also shows that there is some upward movement in pre-transfer income levels from 10 to 20 years after the sample was initially examined for most of the groups. (5) For both AFDC groups, income levels are higher 10 years after the beginning of their sampling periods, as well as 15 years and 20 years after, but for the long-term AFDC group, income is only at 30 percent above the poverty line 20 years after initially receiving AFDC as a head of household while pre-transfer income is 120 percent above the poverty line for the short-term AFDC group 20 years later. For the poor groups, there is a similar rise in income levels at the 10 year period, but the long-term poverty group has pre-transfer income that is only 40 to 60 percent above the poverty line up to 20 years after they were initially examined. Only in year 10 is pre-transfer income significantly higher for the long-term poor group relative to the long-term AFDC group. For the non-poor group, pre-transfer income is 180 percent to 270 percent above the poverty line at 10, 15 and 20 years after being first examined, somewhat higher than when the non-poor were first examined.

The bottom portion of Table 2 shows the likelihood of having pre-transfer income at or below 150 percent of the poverty line 10 to 20 years after initially entering the sample. For the long-term AFDC and long-term poor groups, well over half of their members are below this income level at each of the different periods. The levels decrease somewhat for the long-term AFDC group but rise in year 15 for the long-term poor group from year 10, then fall slightly in year 20. The only significant difference between the two groups is in year 10, where the long-term AFDC group has a significantly greater proportion of families having pre-transfer income below 150 percent of the poverty line relative to the long-term poor group.

Without controls for characteristics when first becoming a head or first receiving AFDC, these initial results indicate that those who start off non-poor have higher income levels, lower likelihoods of pre-transfer poverty or near poverty, and higher likelihoods of marriage later in life than the long-term poverty group. The long-term AFDC group generally shows lower pre-transfer income levels than the long-term poor group but these statistically significant differences disappear by year 15. The short-term poverty and short-term AFDC groups appear to do somewhat better than the long-term poverty group over time, but not nearly as well as the non-poor group. Without the use of controls for other factors which may contribute to later life outcomes, however, it is impossible to determine whether the differences observed between the groups of women are due to the effects of welfare, the effects of poverty or the effects of background variables not included in these models. In the full models presented below, many of these effects are controlled for, allowing us to test the hypotheses about the effects of welfare receipt we presented earlier.

Full Regression Models

The Likelihood of Marriage

Table 3 shows the logistic regression results for the likelihood of marriage at different points in time. These results indicate that there is a significant difference at the .05 level for the likelihood of being married between the long-term AFDC and the long-term poverty groups in year 15, but no significant difference in years 10 and 20. The significant difference for year 10 between the long-term AFDC group and the long-term poverty group (at the.05 level) as shown in Table 2 are no longer significant once controls are used in the models. There is no significant difference in marriage likelihoods between the non-poor group, or the short-term poverty group and the long-term poverty group. In year 20, there is a statistically significant difference between the short-term AFDC group and the long-term poverty group in marriage likelihoods, with the short-term AFDC group more likely to be married in year 20. These results show no consistent pattern for marriage differences between the long-term AFDC and poverty groups. In models where we excluded women who were in the non-poor group from our samples (results not shown), we found few differences in our main results. Interestingly, in year 20 we found that the long-term AFDC group was more likely to be married relative to the long-term poor group (p < .05). We also found that our adjusted [R.sup.2] values and --2 log likelihood values decreased somewhat in these models.

Other factors that negatively affect the likelihood of marriage in year 20 include being non-white, being a high school dropout, and growing up poor. State welfare maximum payments have no significant effect on marriage in years 10, then go from positive to negative in years 15 and 20. Unemployment rates at the end of the period have no effect on marriage likelihoods. We find that having higher levels of income at the beginning period has a significant effect on marriage likelihoods only in year 10.

Income Models

Table 4 shows the OLS regression results for pre-transfer family income-to-needs, when controlling for the effects of the women's condition and characteristics when they first entered the sample. Unlike our initial models without controls (Table 2), these results indicate that women who receive AFDC for a relatively long period at an early age have pre-transfer incomes that are no lower than those women who are poor for a relatively long time but do not receive AFDC in that early period. In fact, none of the models show that the long-term AFDC group is significantly different from the long-term poor group even at the .20 level of significance. Our results indicate that none of the groups have significantly higher pre-transfer family income-to-needs in the later periods relative to the long-term poor group, once we control for beginning family income-to-needs.

Factors affecting later pre-transfer family income-to-needs in various years include family income-to-needs at the beginning (in all years), race (in year 10), education level (in all years), work limitations (in years 15 and 20), growing up poor (in year 20), and state welfare maximums (showing positive effects in years 10 and 20). Age has little effect on our outcomes, while having more than one child relative to those who have only one child affects outcomes (in a positive direction) only in year 10. The unemployment rate in the area of residence at the end of the period has significant negative effects in each of the three models, increasing in strength through time.

Table 5 shows similar effects to those shown in Table 4 when examining the likelihood of having pre-transfer income at or below 150 percent of the poverty line. The long-term AFDC group is no more likely to be at or below 150 percent of the poverty line relative to those in the long-term poverty group.

Tables 4 and 5 indicate that work limitations at the end of the initial period have a strong negative effect on income in years 15 and 20. Other factors strongly affecting pre-transfer income and the likelihood of having pre-transfer income below 150 percent of the poverty line include race, being a high school dropout, and to a lesser degree, the amount of money received from relatives. Being a high school dropout has a strong negative effect on income throughout the 20 year period. Interestingly, being very young at the beginning of the period, under 18 years of age, has no effect on pre-transfer income later in life relative to those who are over 21. Growing up poor shows negative effects on pre-transfer income in years 10 and 20 in both Table 4 and 5.

The evidence from our analyses indicates that with or without control variables in our models, young single mothers receiving AFDC for an extended period of time do not have significantly lower income levels relative to those young single mothers who are poor but do not receive welfare in years 15 and 20, and there are no differences in year 10 when controls are used. Income at the beginning of the period, however, has strong effects on income later in life. Growing up poor has some negative effects in both our income models.

Conclusions

We have tested the question of whether welfare use itself results in economic or other hardships, or whether simply being poor, or the combination of poverty and other factors, produce particular outcomes for single mothers. Overall, we found that long-term welfare receipt in young adulthood is seldom associated with outcomes that are any more negative than those associated with the experience of long-term poverty or near poverty in young adulthood for these women.

First, we examined the hypothesis that those who receive welfare at a young age are less likely to marry than those who do not receive welfare early. We found only weak support for this hypothesis. We found that women who had used AFDC for a relatively long period in young adulthood were no less likely to be married 10 or 20 years after their initial receipt than women who were poor for a long period as young adults, but we did find them significantly less likely to be married at 15 years after initial receipt. Short-term AFDC recipients were never any less likely to marry than women who experienced substantial time in poverty in early adulthood, and by year 20 were in fact significantly more likely to be married.

Second, we examined the hypothesis that early receipt of AFDC is associated with lower pre-transfer income-to-needs later in life. Here, our results provide no support for this hypothesis. No significant differences were found between the two long-term groups in pre-transfer family income-to-needs in any of the years examined when appropriate controls were used. Women in the short-term AFDC group are no different from the long-term poverty group in terms of pre-transfer income-to-needs in any of the outcome periods. We found, however, that factors such as the unemployment rate, growing up poor, level of education, and initial family income-to-needs had strong effects on our outcomes.

Finally, we examined the hypothesis that early AFDC use is a cause of long-term poverty and low income. Our findings offered no support for this hypothesis. There were no differences between the long-term AFDC group and the long-term poverty group at either 10, 15 or 20 years in the likelihood of having pre-transfer income at or below 150% of the poverty line. The long-term economic well-being of young AFDC recipients appears to be no worse (or better) than non-recipients who started out with low income for an extended period in the sample. Indeed, those who use welfare for a relatively short period of time have a significantly lower likelihood of having pre-transfer income at or below 150% of the poverty line by year 20 than women who were in the long-term poor group in early adulthood. In contrast to the long-term AFDC recipients, women who were poor for only a short time in early adulthood are significantly less likely to have low pre-transfer income than women who were poor for longer periods in early adulthood, perhaps suggesting that it is the persistence of poverty in early adulthood (whether or not it is associated with AFDC receipt) rather than welfare receipt itself that best predicts the likelihood of later life poverty or near poverty.

These results support the notion that it is income (or unmeasured factors associated with income) rather than welfare itself that affects the economic well-being of young women. Single mothers who are poor for a substantial period in early adulthood are just as likely to find themselves in or near poverty in later life as single mothers who receive AFDC for a substantial period at the same time of their lives. In addition, we found that these two groups have similar marriage likelihoods. The somewhat different picture of the relationship of early welfare receipt to later-life economic outcomes presented in the models where control variables were not used (Table 2) illustrates the importance of teasing out the effects of welfare itself from the effects of those background characteristics which young welfare recipients share with other young women with low incomes: a lack of education, the experience of growing up poor and the effects of institutional racism.

Our findings have clear public policy implications. The current emphasis on the reduction of welfare use rather than the reduction of poverty is unlikely, based on the results in this paper, to positively affect these women's long-term economic outcomes. Just pushing young single mothers into low-wage work, which will not necessarily lift them out of poverty, will not promote long-term economic well-being. Rather, attention to retaining at-risk young women in the education system, and providing them with further job training may be a more effective long-term strategy for reducing economic hardship among this vulnerable group. In addition, it is important that policy address the issue of the care of the children of these young, poor, single mothers, whether or not the mothers receive welfare. Growing up poor, for the women in our study, often had negative effects on long-term economic outcomes, and there is every reason to suppose that these women's own children will suffer similar negative effects.

If we hope to reduce poverty among women with children, denying these women public assistance seems unlikely to result in any long-term change in their economic well-being. Instead of focusing on getting women off welfare in order to improve their economic chances, policies need to instead focus on lifting young single women with children out of poverty in early adulthood.

Appendix Table 1

Means and Standard Deviations For those Who Stayed in the Sample At
Least 10 Years

 AFDC Sample

Variable Short-Term Long-Term

Money from relatives (1999 $s) 121.0 (238.6) 127.1 (304.3)
 Over 4 year period
Percent with money from relatives 43.0 (47.5) 36.5 (45.6)
 over 4 year period
State welfare maximum for a 739.6 (352.8) 868.2 (363.8)
 family of 4 at the beginning of
 the period
Percentage under age 18 7.7 13.4
Percentage aged 18-21 52.7 51.1
Percentage age 22-25 39.5 35.5
Percentage growing up poor 47.3 41.3
Percentage whose father was a H.S. 57.6 49.7
 dropout
Percentage whose father was a H.S. 36.1
 grad
County unemployment rate at 6.8 (2.0) 7.3 (2.1)
 beginning
County unemployment rate at end 6.5 (2.3) 7.4 (2.7)
Percentage in South 37.8 26.8
Percentage in West 18.9 17.8
Percentage in North Central 23.0 41.9
Percentage in Northeast 20.3 13.4
Percentage with disabled head of 6.3 16.0
 household at the beginning
Percentage with disabled head at 6.6 16.0
 the end
Monthly state welfare maximum for 606.6 (252.8) 719.1 (296.9)
 a family of 4 at the end (1999
 $s)
Largest city in county of 27.9 35.9
 residence has a pop. of
 500,000 or more: SMSA
Largest city in county of 24.6 35.3
 residence has a pop. of
 100,000-499,999:SMSA
Largest city in county of 8.1 8.2
 residence has a pop. of
 50,000-99,999:SMSA
Largest city in county of 9.9 7.3
 residence has a pop. of 25,000-
 49,999:non-SMSA
Largest city in county of 29.6 13.2
 residence has a pop. of 1-
 24,999:non-SMSA
Year entered the sample: 1968-1971 10.5 8.4
Year entered the sample: 1972-1975 14.5 22.7
Year entered the sample: 1976-1979 27.7 23.2
Year entered the sample: 1980-1984 27.8 31.4
Year entered the sample: 1985-1987 19.6 14.2
Sample size 160 250

 Poor or Near-Poor Sample

Variable Short-Term Long-Term

Money from relatives (1999 $s) 218.8 (933.0) 201.5 (355.2)
 Over 4 year period
Percent with money from relatives 32.1 (52.9) 46.9 (46.3)
 over 4 year period
State welfare maximum for a 742.5 (343.7) 623.9 (303.0)
 family of 4 at the beginning of
 the period
Percentage under age 18 3.4 16.6
Percentage aged 18-21 50.1 49.0
Percentage age 22-25 46.5 34.3
Percentage growing up poor 44.4 45.7
Percentage whose father was a H.S. 48.3 57.9
 dropout
Percentage whose father was a H.S.
 grad
County unemployment rate at 6.3 (2.1) 6.7 (1.8)
 beginning
County unemployment rate at end 6.1 (2.0) 6.0 (2.0)
Percentage in South 41.4 55.0
Percentage in West 20.9 3.5
Percentage in North Central 28.5 10.8
Percentage in Northeast 9.1 30.7
Percentage with disabled head of 4.4 15.2
 household at the beginning
Percentage with disabled head at 8.9 14.6
 the end
Monthly state welfare maximum for 663.8 (329.1) 521.8 (208.8)
 a family of 4 at the end (1999
 $s)
Largest city in county of 32.2 11.4
 residence has a pop. of
 500,000 or more: SMSA
Largest city in county of 18.4 24.4
 residence has a pop. Of
 100,000-499,999:SMSA
Largest city in county of 18.4 17.8
 residence has a pop. Of
 50,000-99,999:SMSA
Largest city in county of 4.2 20.5
 residence has a pop. of 25,000-
 49,999:non-SMSA
Largest city in county of 26.8 26.0
 residence has a pop. of 1-
 24,999:non-SMSA
Year entered the sample: 1968-1971 22.4 11.0
Year entered the sample: 1972-1975 12.1 7.4
Year entered the sample: 1976-1979 22.0 19.8
Year entered the sample: 1980-1984 27.1 34.5
Year entered the sample: 1985-1987 16.3 27.4
Sample size 109 124

 Non-Poor Sample

Variable All

Money from relatives (1999 $s) 40.6 (137.4)
 Over 4 year period
Percent with money from relatives 26.5
 over 4 year period
State welfare maximum for a 848.9 (339.5)
 family of 4 at the beginning of
 the period
Percentage under age 18 0
Percentage aged 18-21 29.3
Percentage age 22-25 70.7
Percentage growing up poor 43.6
Percentage whose father was a H.S. 40.8
 dropout
Percentage whose father was a H.S.
 grad
County unemployment rate at 6.1 (2.2)
 beginning
County unemployment rate at end 6.1 (2.4)
Percentage in South 29.8
Percentage in West 20.5
Percentage in North Central 33.3
Percentage in Northeast 16.4
Percentage with disabled head of 2.6
 household at the beginning
Percentage with disabled head at 2.0
 the end
Monthly state welfare maximum for 863.9 (360.8)
 a family of 4 at the end (1999
 $s)
Largest city in county of 28.2
 residence has a pop. of
 500,000 or more: SMSA
Largest city in county of 36.9
 residence has a pop. Of
 100,000-499,999:SMSA
Largest city in county of 12.4
 residence has a pop. Of
 50,000-99,999:SMSA
Largest city in county of 2.6
 residence has a pop. of 25,000-
 49,999:non-SMSA
Largest city in county of 19.9
 residence has a pop. of 1-
 24,999:non-SMSA
Year entered the sample: 1968-1971 20.4
Year entered the sample: 1972-1975 18.9
Year entered the sample: 1976-1979 24.2
Year entered the sample: 1980-1984 19.2
Year entered the sample: 1985-1987 17.4
Sample size 75

Table 1

Percentages, Means and Standard Deviations At the Beginning of
Their Sampling Period For Women Who Stay in the Sample for At
Least 10 Years

 AFDC Sample

Variable Short-Term Short-Term Long-Term

Sample size 160 250
Income and work measures
Family income-to-needs, average
 over 4 year period 1.28 (.57) .91 (.41)
Years with income below 150% of
 the pov. line in the 4 year
 period 2.71 (1.21) 3.36 (1.01)
AFDC income as a % of total
 income in the 4 year period 11.09 (11.00) 56.44 (26.72)
Avg. annual hours of work 861.86 (600.20) 380.81 (433.90)
Age, marriage and children
 measures
Age at the beginning 20.65 (2.21) 20.58 (2.16)
Number of children 1.50 (.76) 1.62 (.86)
Percentage married in 4 year
 period 48.70 13.84
Years married in the 4 year
 period 1.09 (1.20) .22 (.58)
Race
Percentage white 60.19 41.77
Percentage African American 33.95 53.96
Highest level of education
Percentage of HS dropouts 48.70 68.07
Percentage of HS grads 19.49 20.46
Percentage some college 31.82 11.22
Percentage college grads 0.00 0.01
Total sample size: 718

 Poor or Near-Poor Sample

Variable Short-Term Short-Term Long-Term

Sample size 109 124
Income and work measures
Family income-to-needs, average
 over 4 year period 1.92 (.56) 1.05 (.29)
Years with income below 150% of
 the pov. line in the 4 year
 period 1.46 (.56) 3.55 (.46)
AFDC income as a % of total
 income in the 4 year period 0.72 (2.71) 0.99 (4.81)
Avg. annual hours of work 1293.31 (663.13) 1020.65 (602.50)
Age, marriage and children
 measures
Age at the beginning 21.23 (2.57) 20.28 (2.21)
Number of children 1.25 (.78) 1.43 (.78)
Percentage married in 4 year
 period 53.37 35.31 (44.38)
Years married in the 4 year
 period 1.31 (1.51) .66 (.94)
Race
Percentage white 66.67 48.14
Percentage African American 28.62 50.28
Highest level of education
Percentage of HS dropouts 36.05 53.02
Percentage of HS grads 20.25 14.97
Percentage some college 37.80 28.33
Percentage college grads 5.90 3.68
Total sample size: 718

 Non-Poor Sample

Variable Short-Term All

Sample size 75
Income and work measures
Family income-to-needs, average
 over 4 year period 2.98 (1.65)
Years with income below 150% of
 the pov. line in the 4 year
 period 0.00 (.00)
AFDC income as a % of total
 income in the 4 year period 0.03 (.20)
Avg. annual hours of work 1543.83 (782.79)
Age, marriage and children
 measures
Age at the beginning 22.14 (1.99)
Number of children 1.21 (.67)
Percentage married in 4 year
 period 56.60 (58.54)
Years married in the 4 year
 period 1.19 (1.44)
Race
Percentage white 66.82
Percentage African American 22.65
Highest level of education
Percentage of HS dropouts 21.15
Percentage of HS grads 25.35
Percentage some college 52.03
Percentage college grads 1.48
Total sample size: 718

Table 2

Percent Married, Family Pre-Transfer Family Income-to-Needs, and
Percent with Pre-Transfer Family Income-to-Needs at or Below 150
Percent of the Poverty Line At the End of the Periods

 Poor or Near-
 AFDC Sample Poor Sample

Variable Short-Term Long-Term Short-Term

Percent married
10 years later 48.1 36.7 (b) 60.3
15 years later 59.8 (d) 29.1 (b) 56.4 (d)
20 years later 67.9 (b) 37.9 52.4
Pre-transfer family income-to-needs
10 years later 1.5 (l.9) 0.9 (0.9) (a) 2.2 (1.6) (b)
15 years later 1.6 (l.5) 1.0 (1.0) 2.1 (1.6) (b)
20 years later 2.2 (1.2) (d) 1.3 (l.2) 2.6 (1.5) (b)
Percentage with pre-transfer family income at or below 150% of the
poverty line
10 years later 56.3 68.5 (b) 24.7 (a)
15 years later 58.2 69.2 33.3 (a)
20 years later 21.6 (a) 53.4 21.1 (a)

 Poor or Near-
 Poor Sample Non-Poor Sample

 Long-Term
Variable (excluded group) All

Percent married
10 years later 53.4 66.1 (d)
15 years later 52.6 78.0 (b)
20 years later 31.0 67.7 (c)
Pre-transfer family income-to-needs
10 years later 1.4 (l.1) 2.6 (2.1) (a)
15 years later 1.4 (l.0) 2.8 (2.4) (a)
20 years later 1.6 (1.0) 3.7 (2.8) (a)
Percentage with pre-transfer family income at or below 150% of the
poverty line
10 years later 54.8 24.2 (a)
15 years later 65.4 29.8 (a)
20 years later 56.2 22.4 (b)

Note: a: p<.001; b: p<.01; c: p<.05; d: p<.10

All significance levels are relative to the long-term poor or near-poor
group for the given year. Significance levels were determined by
running regression models with the long-term poverty group as the
excluded group. N=718 for the 10 year period; N=462 for the 15 year
period; N=252 for the 20 year period.

Table 3

Coefficient Estimates (Standard Errors) for the Logistic Regression
Models for the Likelihood of Being Married in Subsequent Years After
Initially Becoming a Head of Household or Receiving Welfare, for Single
Mothers Age 24 and Under

 Coefficient Estimates
 (Standard Errors)

Variable 10 Years After

Personal and family background variables
Having one child .309 (.233)
Child under the age of 3 .093 (.222)
Race--white .940 (.223) ****
Age under 18 at beginning .520 (.370)
Age 18 to 21 at beginning .610 (.227) ***
Never married -.025 (.233)
Head was a professional while growing up .793 (.441) *
Head was a HS graduate while growing up -.929 (.227) ****
Head was a college graduate while growing up -.895 (.449) **
Grew up poor -.293 (.205)
Income, work and education
Money from relatives 1.050 (.325) ***
High school dropout -1.689 (.262) ****
High school graduate -.063 (.279)
Any work limitations at the end .251 (.326)
Any work limitations at the beginning .365 (.350)
County unemployment rate at the beginning .093 (.058)
County unemployment rate at the end .051 (.050)
Welfare and income measures
Long-term AFDC group .020 (.327)
Short-term AFDC group -.457 (.332)
Short-term poverty group -.673 (.371) *
Non-poor group -1.240 (.478) ***
Family income-to-needs at the beginning 1.043 (.214) ****
State welfare maximum for AFDC at end -.405 (.521)
Intercept -1.138 (.812)
-2 log likelihood 741.145
N 718

 Coefficient Estimates
 (Standard Errors)

Variable 15 Years After

Personal and family background variables
Having one child .260 (.315)
Child under the age of 3 .339 (.297)
Race--white 1.286 (.298) ****
Age under 18 at beginning .738 (.501)
Age 18 to 21 at beginning .317 (.313)
Never married -.121 (.309)
Head was a professional while growing up .531 (.684)
Head was a HS graduate while growing up -.174 (.312)
Head was a college graduate while growing up -.360 (.625)
Grew up poor .355 (.280)
Income, work and education
Money from relatives .568 (.398)
High school dropout -1.848 (.342) ****
High school graduate -.435 (.370)
Any work limitations at the end .677 (.398) *
Any work limitations at the beginning -.095 (.432)
County unemployment rate at the beginning .075 (.073)
County unemployment rate at the end .090 (.068)
Welfare and income measures
Long-term AFDC group -.948 (.457) **
Short-term AFDC group .359 (.443)
Short-term poverty group .030 (.497)
Non-poor group -.264 (.625)
Family income-to-needs at the beginning .410 (.284)
State welfare maximum for AFDC at end 1.487 (.821) *
Intercept -2.275 (1.322) *
-2 log likelihood 435.778
N 462

 Coefficient Estimates
 (Standard Errors)

Variable 20 Years After

Personal and family background variables
Having one child -.317 (.409)
Child under the age of 3 -.335 (.400)
Race--white .590 (.404)
Age under 18 at beginning 1.207 (.759)
Age 18 to 21 at beginning 1.389 (.420) ****
Never married .392 (.386)
Head was a professional while growing up -.500 (.818)
Head was a HS graduate while growing up -.155 (.416)
Head was a college graduate while growing up -1.259 (.937)
Grew up poor -.823 (.405) **
Income, work and education
Money from relatives .119 (.242)
High school dropout -1.376 (.491) ***
High school graduate -1.183 (.490) **
Any work limitations at the end -.774 (.480)
Any work limitations at the beginning -.064 (.512)
County unemployment rate at the beginning .279 (.136) **
County unemployment rate at the end -.011 (.133)
Welfare and income measures
Long-term AFDC group .479 (.682)
Short-term AFDC group 1.095 (.666) *
Short-term poverty group -.100 (.786)
Non-poor group .865 (.915)
Family income-to-needs at the beginning .252 (.349)
State welfare maximum for AFDC at end -3.014 (1.143) ***
Intercept .560 (1.541)
-2 log likelihood 251.165
N 252

*: p<.10; **: p<.05; ***: p<.01; ****: p<.001.

Note: Each model is significant at the .001 level. Each model also
included variables for region of residence, year entering the
sample, and whether living in a rural or more urban area.

Table 4

Coefficient Estimates (Standard Errors) for the Pre-Transfer Family
Income-to-Needs Ratios in Subsequent Years After Initially Becoming a
Head of Household or First Receiving Welfare, for Single Mothers Age 24
and Under

 Coefficient Estimates
 (Standard Errors)

Variable 10 Years After

Personal variables
Having one child (more than 1 child is excluded) .315 (.127) ***
Child under the age of 3 -.200 (.119) *
Race--white .550 (.122) ****
Age under 18 at beginning .108 (.211)
Age 18 to 21 at beginning -.010 (.121)
Never married .125 (.124)
Head was a professional while growing up .106 (.220)
Head was a HS graduate while growing up -.236 (.120) **
Head was a college graduate while growing up -.026 (.229)
Grew up poor -.257 (.110) **
Income, work and education
Money from relatives .105 (.120)
High school dropout -.439 (.140) *
High school graduate -.167 (.156)
Any work limitations at the end -.256 (.183)
Any work limitations at the beginning -.142 (.189)
County unemployment rate at the beginning -.004 (.032)
County unemployment rate at the end -.052 (.027)
Welfare and income measures
Long-term AFDC group -.264 (.180)
Short-term AFDC group -.207 (.182)
Short-term poverty group -.274 (.198)
Non-poor group -.325 (.249)
Family income-to-needs at the beginning .641 (.079) ****
State welfare maximum for AFDC at end .642 (.283) **
Intercept 1.199 (.434) *
Adjusted [R.sup.2] .324
N 718

 Coefficient Estimates
 (Standard Errors)

Variable 15 Years After

Personal variables
Having one child (more than 1 child is excluded) -.065 (.145)
Child under the age of 3 -.176 (.138)
Race--white .216 (.138)
Age under 18 at beginning .099 (.239)
Age 18 to 21 at beginning -.290 (.144) **
Never married .116 (.140)
Head was a professional while growing up .553 (.293) *
Head was a HS graduate while growing up .049 (.142)
Head was a college graduate while growing up -.038 (.269)
Grew up poor -.091 (.128)
Income, work and education
Money from relatives .207 (.115) *
High school dropout -.891 (.159) ****
High school graduate -.595 (.174) ****
Any work limitations at the end -.580 (.184) ***
Any work limitations at the beginning -.222 (.207)
County unemployment rate at the beginning .033 (.035)
County unemployment rate at the end -.101 (.031) ***
Welfare and income measures
Long-term AFDC group -.051 (.218)
Short-term AFDC group -.017 (.213)
Short-term poverty group -.143 (.239)
Non-poor group .022 (.303)
Family income-to-needs at the beginning .830 (.129) ****
State welfare maximum for AFDC at end .878 (.371) **
Intercept .958 (.593)
Adjusted [R.sup.2] .436
N 462

 Coefficient Estimates
 (Standard Errors)

Variable 20 Years After

Personal variables
Having one child (more than 1 child is excluded) -.019 (.225)
Child under the age of 3 -.389 (.225) *
Race--white .188 (.228)
Age under 18 at beginning .074 (.415)
Age 18 to 21 at beginning .050 (.228)
Never married -.033 (.212)
Head was a professional while growing up .619 (.469)
Head was a HS graduate while growing up -.196 (227)
Head was a college graduate while growing up .289 (.517)
Grew up poor -.581 (.220) ***
Income, work and education
Money from relatives .096 (.149)
High school dropout -.527 (.265) **
High school graduate -.509 (.272) *
Any work limitations at the end -.862 (.264) ***
Any work limitations at the beginning -.782 (.297) **
County unemployment rate at the beginning .137 (.072) *
County unemployment rate at the end -.223 (.072) **
Welfare and income measures
Long-term AFDC group -.045 (.383)
Short-term AFDC group .144 (.364)
Short-term poverty group .313 (.429)
Non-poor group .386 (.512)
Family income-to-needs at the beginning .484 (.193) ***
State welfare maximum for AFDC at end .484 (.628)
Intercept 3.251 (.841) ****
Adjusted [R.sup.2] .436
N 252

*: p<.10; **: p<.05; ***: p<.01; ****: p<.001.

Note: Each model is significant at the .001 level. Each model also
included variables for region of residence, year entering the
sample, and whether living in a rural or more urban area.

Table 5

Coefficient Estimates (Standard Errors) for the Logistic Regression
Models for the Likelihood of Having Pre-Transfer Income At or Below
150 Percent of the Poverty Line in Subsequent Years After Initially
Becoming a Head or Receiving AFDC for Single Mothers Age 24 and Under

 Coefficient Estimates (Standard Errors)

Variable 10 Years After 15 Years After

Personal variables
Having one child (more
 than 1 is excluded) -.531 (.237) ** -.221 (.306)
Child under the age of 3 .284 (.226) .281 (.293)
Race--white -.345 (.223) .082 (.228)
Age under 18 at begin-
 ning .177 (.339) -.076 (.547)
Age 18 to 21 at begin-
 ning -.341 (.227) -.138 (.314)
Never married .079 (.236) -.382 (.320)
Head was a professional
 while growing up .070 (.460) -.790 (.655)
Head was a HS graduate
 while growing up .455 (.222) ** -.503 (.296) *
Head was a college
 graduate while growing
 up .245 (.441) .175 (.606)
Grew up poor .367 (.207) * -.027 (.275)
Income, work and
 education
Money From relatives -.323 (.307) -.366 (.416)
High school dropout 1.027 (.252)**** 1.280 (.328) ****
High school graduate .362 (.278) .837 (.356) **
Any work limitations at
 the end .305 (.339) 1.865 (.525) ****
Any work limitations at
 the beginning .250 (.377) -.140 (.469)
County unemployment rate
 at the beginning -.032 (.062) .020 (.076)
County unemployment rate
 at the end .182 (.055) **** .109 (.067)

Welfare and poverty
 measures
Long-term AFDC group -.356 (.330) .049 (.494)
Short-term AFDC group .266 (.331) -.245 (.466)
Short-term poverty group .287 (.362) -.370 (.521)
Non-poor group .782 (.479) -.040 (.648)
Family income-to-needs -1.656 (.241) **** -1.087 (.284) ****
 at the beginning
State welfare maximum -.058 (.544) -1.758 (.790) **
 for AFDC at end
Intercept .542 (.834) 1.654 (1.252)
-2 log likelihood 709.963 440.166
N 718 462.000

 Coefficient
 Estimates (Standard
 Errors)

Variable 20 Years After

Personal variables
Having one child (more
 than 1 is excluded) -.277 (.513)
Child under the age of 3 .005 (.534)
Race--white .141 (.560)
Age under 18 at begin-
 ning 1.575 (1.085)
Age 18 to 21 at begin-
 ning -.569 (.537)
Never married -.596 (.557)
Head was a professional
 while growing up -1.483 (1.187)
Head was a HS graduate
 while growing up .771 (.522)
Head was a college
 graduate while growing
 up -1.188 (1.552)
Grew up poor 2.443 (.545) ***
Income, work and
 education
Money From relatives -2.858 (1.000) ***
High school dropout 1.227 (.634) **
High school graduate 1.530 (.648) **
Any work limitations at
 the end 1.286 (.622) **
Any work limitations at
 the beginning 2.996 (.876) ***
County unemployment rate
 at the beginning -.266 (.177)
County unemployment rate
 at the end .332 (.169) **

Welfare and poverty
 measures
Long-term AFDC group -1.324 (.954)
Short-term AFDC group -2.522 (.964) ***
Short-term poverty group -1.610 (1.014)
Non-poor group .560 (1.168)
Family income-to-needs -1.616 (.531) **
 at the beginning
State welfare maximum 2.535 (1.685)
 for AFDC at end
Intercept -1.382 (2.206)
-2 log likelihood 169.516
N 252

*: p<.10; **: p<.05; ***: p<.01; ****: p<.001.

Note: Each model is significant at the .001 level. Each model also
included variables for region of residence, year entering the sample,
and whether living in a rural or more urban area.

Notes

(1.) We chose 10 percent of income as a cutoff for AFDC so we only included those with substantial AFDC income in the AFDC groups we define later in the paper. We also examined different cutoffs for AFDC, including 20 percent of income coming from AFDC without substantial changes to our main results.

(2.) It is possible for women in the AFDC group to have been poor before they received AFDC, but were put into the AFDC sample because they received AFDC later.

(3.) We also ran tests examining a 6 year period after initially entering the sample. To be in either the long-term AFDC group, the woman had to have "substantial" AFDC income for at least 5 out of the 6 years. To be in the long-term poverty group, the woman had to have income at or below 150% of the poverty line for at least 5 out of the 6 years without any substantial AFDC income in this period. The shorter term groups were determined by being in either poverty or receiving AFDC for 1 to 4 years. Our results changed little when using these measures relative to the measures presented in this article. If anything, our coefficient estimates (and significance levels) examining the differences between the long-term AFDC and long-term poverty groups decreased when using this alternative measure. We chose to present the results for the 4 year period because this time frame allowed us to better focus on early adulthood experiences and characteristics relative to the 6 year period.

(4.) We did an examination of why we got such a large reduction in the percentage of the long-term poor group who were married between years 15 and 20. We lost 20 women who were married and 24 women who were not married between year 15 and year 20. Without the PSID weights, the year 15 and 20 marriage rates would have been far closer--40 percent in year 15 and 36 percent in year 20.

(5.) Only post-transfer income levels were shown in Table 1. However, in results not shown, pre-transfer income has increased for all of the groups.

References

Albelda, Randy and Chris Tilly. 1997. Glass Ceilings and Bottomless Pits: Women's Work, Women's Poverty. Boston, MA: South End Press.

An, C., Haveman, R., & Wolfe, B. (1993). Teen out-of-wedlock births and welfare receipt: The role of childhood events and economic circumstances. Review of Economics and Statistics, 72, 195-208.

Bane, M. J., & Ellwood, D. (1994). Welfare Realities: From Rhetoric to Reform. Cambridge, MA: Harvard University Press.

Blank, R. (1989). Analyzing the length of welfare spells. Journal of Public Economics, 39, 245-273.

Cancian, M. and Meyer, D.R. (2000). Work after welfare: Women's work effort, occupation, and economic well-being. Social Work Research, 24: 69-86.

Committee on Ways and Means Committee, U.S. House of Representatives (2004). 2004 Green Book, http://waysandmeans.house.gov/Documents.asp ?section=813

Coontz, S. and Folbre, N. (2002). "Marriage, Poverty, and Public Policy", Poverty Research News, The Joint Center for Poverty Research, Vol. 6, No. 3, 9-11.

Corcoran, M. E. and Kunz, J.P. (1997). Do unmarried births among African-American teens lead to adult poverty? Social Service Review, 71: 274-287.

Dolinsky, A. L., Caputo, R. K., & O'Kane, 12 (1989). Competing effects of culture and situation. Social Service Review, 63,359-371.

Duncan, G. J. and Hoffman, S.D. (1990). Welfare benefits, economic opportunity, and out-of-wedlock births among black teenage girls. Demography, 27: 519-535.

Dunifon, R. (1999). Short- and long-term effects of AFDC Receipt on Families. Joint Center for Poverty Research Working Paper, no. 24.

Edin, K. & Lein, L. (1997). Making Ends Meet: How Single Mothers Survive Welfare and Low-Wage Work, Russell Sage Foundation, New York.

Ellwood, D. (1986). Targeting 'Would-Be' Long-Term Recipients of AFDC. Princeton, N J: Mathematica Policy Research.

Fitzgerald, J. (1991). The Effects of the Marriage Market and AFDC Benefits on Exit Rates From AFDC. Journal of Human Resources, Summer.

Fitzgerald, J. (1995). Local labor markets and local area effects on welfare durations. Journal of Policy Analysis and Management, 14, 43-67.

Gleason, 12, Rangarajan, A., and Schochet, P. (1998). The dynamics of receipt of Aid to Families with Dependent Children among teenage parents in inner cities, The Journal of Human Resources, 33: 988-1002.

Gottschalk, P. (1992). The intergenerational transmission of welfare participation: Facts and possible causes. Journal of Policy Analysis and Management, 11,254-272.

Gottschalk, P., McLanahan, S., & Sandefur, G. D. (1994). The dynamics and intergenerational transmission of poverty and welfare participation. In S. H. Danziger & D. H. Weinberg (Eds.), Confronting poverty (pp. 85-108). Cambridge, MA: Harvard University Press.

Harris, K. M. (1996). Life After Welfare: Women, work, and repeat dependency. American Sociological Review, 61: 407-26.

Herrnstein, R. J., & Murray, C. (1994). The bell curve: Intelligence and class structure in American life. New York: The Free Press.

Hill, M. S., & Ponza, M. (1989, October). Intergenerational transmission of welfare dependence. Paper presented at the annual meetings of the Population Association of America, Washington, DC.

Horn, Wade (2002). "Welfare Reform Reauthorization: Promoting self-sufficiency, protecting children, and strengthening marriage," Poverty Research News, The Joint Center for Poverty Research, Vol. 6, No. 3, 3-5.

Lichter, D. T., McLaughlin, D., LeClere, F., Kephart, G. and Landry, D. (1992). "Race and the retreat from marriage: A shortage of marriageable men?", American Sociological Review, 57, 781-99.

McLanahan, S. (1988). Family structure and dependence: Early transitions to female household headship. Demography, 25, 1-16.

Mead, L. (1986). Beyond Entitlement: The Social Obligations of Citizenship. New York, Free Press.

Mead, L. (1998). Telling the Poor What to Do. The Public Interest, 132:97-112.

Meyer, D. R., and Cancian, M. (1998). Economic Well-Being Following an Exit From Aid to Families with Dependent Children. Journal of Marriage and the Family, 60: 479-92.

Moffitt, R. (1992). Incentive Effects in the U.S. Welfare System: A Review. Journal of Economic Literature, 30, 1-61.

Murray, C. (1984). Losing ground: American social policy 1950-1980. New York: Basic Books.

Murray, C. (2001). "Family Formation", in The New World of Welfare, edited by Rebecca Blank and Ron Haskins, Brookings Institute Press, Washington D.C.

O'Neill, J. A., Bassi, L. J., & Wolf, D. A. (1987). The Duration of Welfare Spells. Review of Economics and Statistics, 69, 241-248.

Rank, M. R., & Cheng, L.C. (1995). Welfare use across generations: How important are the ties that bind? Journal of Marriage and the Family, 57, 673-84.

Smeeding, Timothy (1992). "Why the U.S. antipoverty system doesn't work very well," Challenge, January-February.

Tanner, M. (1996). The End of Welfare: Fighting Poverty in the Civil Society. Washington, D.C.: Cato Institute.

Vartanian, T. P. (1997). Neighborhood Effects on AFDC Exits: Examining the Social Isolation, Relative Deprivation, and Epidemic Theories, Social Service Review, 71: 548-573.

Vartanian, T. P. (1999). Childhood Conditions and Adult Welfare Use: Examining Neighborhood Factors. Journal of Marriage and the Family, 61:225-237.

Vartanian, T. P., & Gleason, P. (1999). Income and Job Market Outcomes After Welfare: 1990-1995, Joint Center for Poverty Research, Working Paper No. 92. http://www.jcpr.org/wp/WPprofile.cfm?ID=92.0

Vartanian, T. P. and McNamara, J. M. (2000). Work and economic outcomes after welfare. Journal of Sociology and Social Welfare, 27: 41-78.

Wilson, William Julius (1987), The Truly Disadvantaged, Chicago: The University of Chicago Press.

Wilson, W. J. and Neckerman, K. (1986). "Poverty and Family Structure: The Widening Gap Between Evidence and Public Policy Issues," in Fighting Poverty, Eds.: S.H. Danziger and D.H. Weinberg. Cambridge: Harvard University Press, pp. 232-59.

U.S. Department of Health & Human Services Administration for Children and Families, 2000 http://www.acf.dhhs.gov/news/stats/caseload.htm.

THOMAS P. VARTANIAN

JUSTINE M. MCNAMARA

Bryn Mawr College

Graduate School of Social Work and Social Research