National   Poverty   Center
Gerald   R.
Ford   School   of   Public   Policy,   University   of   Michigan   www.npc.umich.edu
This   paper   was   delivered   at   a   National   Poverty   Center   conference.
Any   opinions,   findings,   conclusions,   or   recommendations   expressed   in   this   material   are   those   of   the   author(s)   and   do   not   necessarily   reflect   the   view   of   the   National   Poverty   Center   or   any   sponsoring   agency.

Welfare Participation and Marriage
Julien O. Teitler*
Nancy E. Reichman
Lenna Nepomnyaschy
Irwin Garfinkel
Word count: 6638
* Address correspondence to Julien Teitler, Columbia University School of Social Work, 1255
Amsterdam Avenue, New York, NY 10027, or by email at jot8@columbia.edu
.
This research was supported by the Department of Health and Human Services’ Administration for
Children and Families and Office of the Assistant Secretary for Planning and Evaluation
(90PA00007/01). An earlier version of this paper was presented at the National Poverty Center conference: "Marriage and Family Formation Among Low-Income Couples," Washington DC,
September 4-5, 2003.

Welfare Participation and Marriage
Abstract:
Despite recent interest in the potential of the welfare system as a tool to affect marriage behaviors among low-income women, little is known about how welfare participation, as opposed to welfare policy, affects decisions to marry. We employ an event history approach to examine transitions to marriage over a three-year period among mothers who have had a non-marital birth.
We find that welfare participation under the new Temporary Assistance to Needy Families program (TANF) reduces the likelihood of transitioning to marriage (hazard ratio is .60, p < .01) but only while the mother is receiving welfare. Once the mother leaves TANF, past receipt has little effect on marriage. We project that over an 18-year period, TANF participation results in a 4 to 6 percentage point reduction in marriage and delays marriage by up to 31 months. We infer that the negative association between TANF participation and marriage reflects economic disincentives or stigma rather than differences in participants’ and non-participants’ orientations toward marriage.

INTRODUCTION
The Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) of
1996 was designed to reduce reliance on welfare, make fathers more accountable, and increase marriage. The last goal has been prominently featured in recent debates over welfare reauthorization. The belief that there is a causal link from welfare to family structure is rooted in two distinct perspectives. One is that the welfare system can affect marriage behaviors through economic incentives. Another is that reliance on public assistance and single parenthood are both part of a “tangle of pathologies” (to borrow from Moynihan 1965) afflicting poor women. The latter perspective is reflected in current welfare policy debates. For example, Rector (2004) refers to “behavioral poverty” (in contrast to material poverty), in which welfare participation and single parenthood are two key elements of a self-perpetuating culture of poverty.
Despite widespread speculation about the links between the welfare system and marriage behaviors of low-income women, almost all of the research in this area has focused on the effects of welfare policies on marriage. Very little is known about how welfare participation affects decisions to marry. In the only previous study on this topic, Fitzgerald and Ribar (2004) found significant effects of welfare participation on transitions into and out of female headship.
However, that study focused exclusively on concurrent effects (how current welfare participation is related to transitions). No study has systematically explored the associations between past versus current welfare participation and marriage.
In this paper, we address the previously unexplored question of whether associations between Temporary Assistance to Needy Families (TANF) participation and marriage extend beyond the recipiency period and how the effects of past welfare participation compare with those of current participation. The answer to this question is critical for understanding how the
1

present welfare system affects family structure and whether single mothers who will participate in welfare differ from those who will not in terms of orientation towards marriage. If the effects of past and current welfare participation are similar, it would suggest that welfare participation confers a permanently debilitating, or “scarring” effect on women that leads to long-term reductions in the likelihood of marriage or that associations between welfare participation and marriage reflect differences in characteristics of individuals who do and do not have experience with the welfare system. If, however, the effects pertain to current participation only, it would suggest that the association is due to financial incentives to remain on welfare or short-term stigma rather than scarring or selection effects.
We employ an event history approach to examine the effects of TANF participation on entry into marriage among mothers who were unmarried when their child was born. We include measures of current and past welfare in the models in order to disentangle associations between
TANF participation and marriage that are short-term (i.e., during the recipiency period only) and those that are long-term. We use the results from those analyses to project the cumulative impact of welfare participation on marriage over the life course.
Background and significance
There has been much interest among researchers and politicians, dating back at least 25 years, in how the welfare system affects family structure. Most studies in this area have focused on the effects of welfare policies on female headship, marriage, and cohabitation. Some studies have investigated the effects of policies on non-marital fertility, which affects eligibility for welfare. The evidence from studies conducted after 1990 1 suggests that, on balance, welfare discourages marriage and encourages fertility. However, the effects tend to be small and often are
1 These tend to be more methodologically rigorous than the earlier studies.
2

significant only for whites (see Blank 2002; Moffitt 1998; Peters et al. 2001; Ratcliffe,
McKernan, and Rosenberg 2002). For example, Moffitt found that a $100 increase in the monthly benefit level reduces the probability of marriage by 2.5 to 5 percent (Moffitt, Reville, and Winkler 1998). A review of experimental studies indicate no effects on marriage (Gennetian and Knox 2003) and a another review of recent studies indicates that, overall, changes in welfare rules under PRWORA that replaced Aid to Families with Dependent Children (AFDC) with time limited TANF are unlikely to result in large changes in marriage behavior (Peters, Plotnick, and
Jeong 2001).
Finding relatively small policy effects does not mean that welfare has little impact on marriage, as policies represent only one aspect of the welfare system. It is possible that the experience of participating in the welfare system has profound effects on individuals’ marriage behaviors. It may reduce or delay marriage by altering the value participants place on marriage, by affecting women’s mental health, by making participants less attractive as potential spouses to marriageable men, or by providing economic or social disincentives to marriage. A number of different mechanisms could lead to participation effects that are long-term (persisting after welfare participation ends), or short-term (occurring only while participants receive welfare).
Long-term:
Welfare participation may have long-term scarring consequences on mothers’ attitudes toward marriage, perceptions of themselves, or others’ perceptions of them. However, there could be underlying differences between participants and non-participants that are related to both welfare participation and long-term marriage behaviors. The potential long-term mechanisms and empirical evidence in support of them are discussed below, as is the potential selection effect.
3

(a) Values: Welfare participation may change individuals’ attitudes toward work and marriage. The notion that reliance on public assistance alters work and family values is one that has existed since at least the 1960s. This argument was formalized by Oscar Lewis (1968), who coined the termed “culture of poverty,” but it has since been espoused in various forms among conservatives and liberals and persists in current debates over welfare reform under a variety of different labels. Despite the preponderance of this belief, very little research has investigated the extent to which welfare participation changes individuals’ values. Studies that have investigated the effects of having received welfare as a child on non-marital fertility provide indirect empirical evidence that welfare changes individuals’ marriage attitudes. However, this research is fraught with methodological problems (see 1995 review article by Corcoran). Whether welfare participation changes family values remains very much an open question.
(b) Self-perception and stress: Welfare participation may have psychological effects. It may depress self-esteem or increase stress, leading to difficulty in finding partners and maintaining relationships (Goodban 1985; Jarrett 1996; Nichols-Casebolt 1986; Jayakody and
Stauffer 2000; Rainwater 1982). There is recent evidence that mental health is strongly associated with marital status among mothers of young children (DeKlyen, McLanahan, and
Brooks-Gunn forthcoming ), although the directionality between mental health and marital status remains unclear. As far as we know, there have been no research studies on the extent to which a mother’s psychological state mediates the relation between welfare and marriage.
(c) Scarlet lettering: Having participated in welfare may affect marriage prospects by devaluing participants’ attractiveness to potential marriage partners. Prior research indicates that employers’ views of welfare participants makes them less willing to hire former recipients and more likely to pay them lower wages (Noonan and Heflin forthcoming ), but little is known about how past welfare participation relates to marriage directly.
4

(d) Selection: Associations between participation and marriage may reflect selection; that is, differences in marriage rates of participants and non-participants may be due to unobserved characteristics of women associated both with welfare participation and marriage. Many have argued that behaviors of poor mothers that have been attributed to welfare actually reflect poverty more generally (see, for example, Schram 2000). Thus, poverty may be the underlying source of associations between welfare participation and marriage behavior. Recent qualitative research suggests that low-income couples view economic self-sufficiency as a prerequisite for marriage and therefore tend to delay marriage until they have obtained a desired financial status
(Edin and Kafalas 2005; Gibson, Edin, and McLanahan forthcoming ). Poverty may also be associated with a unique set of values that results in an increased likelihood of welfare participation (versus work) and a decreased likelihood of marriage. That is, living in poverty (not welfare participation per se) may contribute to a culture of poverty.
Short-term:
(e) Economic disincentives: According to the classic economic theory of utility maximization, in particular Becker’s theory of marriage (Becker 1973, 1974), welfare eligibility rules counting the earnings of spouses provide a disincentive to marriage. The disincentive is likely to be strongest during welfare spells and thus have a short-term impact. Welfare participants face an imminent loss of benefits if they marry, whereas non-recipients will discount the value of a loss of benefits associated with marriage because they may not need welfare assistance in the future.
(f) Stigma: While on welfare, mothers may be stigmatized by others, which may make them less marriageable. Prior research has documented negative perceptions of welfare participants, generally (e.g., Klugel and Smith 1986; Rainwater 1982). However, there is no
5

direct evidence that this translates into reduced marriage prospects. Being on welfare may also affect how women view themselves. As discussed earlier, low-income couples strive to achieve self-sufficiency before entering marital unions (Edin and Kafalas 2005; Gibson, Edin and
McLanahan forthcoming ). Mothers on TANF may postpone marriage because welfare participation is an incontrovertible and highly visible marker (to themselves and to others) of not having achieved that goal. In other words, it is proof that they have not met an economic
“marriage bar.”
Evidence of participation effects
As indicated earlier, there has been very little empirical work investigating the effects of welfare participation on marriage. Only one study has examined the link directly. Using the
Survey of Income and Program Participation, Fitzgerald and Ribar (2005) found significant negative effects of current welfare participation on exits from female headship (the most common pathway being through marriage), after controlling for individual and policy measures.
By also estimating simultaneous models of welfare participation and headship, they found that unobserved factors associated with both participation and headship inflated their single equation estimates by about 10 percent. We extend their analysis by asking whether the effects of TANF participation on marriage persist beyond the recipiency period and by investigating potential mechanisms. We use post-1996 data and focus on an urban population at high risk of welfare dependency.
6

Determinants of welfare participation and marriage
As discussed earlier, welfare participation, poverty, and being an unmarried parent are highly related. Single women, those who have low levels of education, those who have more children to care for, and those who are members of minority groups all are more likely than women without these characteristics to be poor and use welfare (Pavetti 1997). Immigrant mothers are less likely to rely on welfare than native-born mothers due to eligibility restrictions and issues of legal status, but they are more likely to be poor and therefore financially eligible for welfare (Tumlin and Zimmerman 2003).
Zedlewski (2002), using data from the 1997 and 1999 National Survey of America’s
Families, imputed TANF eligibility status for families and compared characteristics of participants and non-participants. Among eligible families, participants had more children, were less likely to live with other adults (including partners), had higher rates of physical and mental health problems, were less likely to have worked in the past three years, were younger, and were more likely to be black and less likely to be Hispanic than non-participants. Reichman et al.
(2004) conducted a similar analysis of TANF use among urban families with one-year-old children and found very similar results. We control for most of these characteristics in our analyses of the effects of TANF participation on marriage.
Local labor markets are associated with welfare participation (Hoynes 2000; Fitzgerald
1995; Ribar 2005) and marriage (Blau, Kahn, and Waldfogel 2000). State policies also have been associated with both welfare participation and marriage. For example, studies have found positive associations between state welfare generosity and welfare caseloads (Mead 2000, 2003).
They also have found that stronger child support enforcement decreases welfare participation
(Meyer 1993; Huang, Garfinkel, and Waldfogel 2004) and that state Medicaid generosity decreases welfare participation (Yelowitz 1995) and increases marriage (Yelowitz 1998). We
7

include city fixed effects in our analyses of the effects of TANF participation on marriage to control for local economies and state policies that may be associated with both TANF participation and marriage.
DATA AND MEASURES
The Fragile Families and Child Wellbeing Study (hereafter, FF) follows a cohort of parents and their newborn children in 20 U.S. cities (located in 15 states). Mothers were interviewed in the hospital at the time of their child’s birth (baseline) and over the telephone one and three years later. Baseline interviews were conducted with a probability sample of 3,712 unmarried mothers and a comparison group of 1,196 married mothers from 1998 to 2000 (see
Reichman et al. 2001 for details of the research design). Response rates of unmarried mothers were 87 percent at baseline, 90 percent (of baseline mother interviews) at the one year follow-up, and 87 percent (of baseline mother interviews) at the three year follow-up wave.
Of the 3,293 mothers who were unmarried at baseline and completed follow-up interviews at one year, 15 gave inconsistent reports of marital status at either baseline or followup and were excluded from the analyses. Another 324 were excluded because of missing data on
TANF spells (220 cases) or covariates (104 cases). The remaining 2,954 cases form the sample for the analyses that follow. Our outcome of interest is marriage, either to the baby’s father or to someone else. We used information from the one and three year follow-up interviews to determine marital status of the mother at each month after her baseline interview until her final interview date or until she got married. For the 14 percent of cases for which we did not have exact marriage dates, we imputed them based on the mother’s report of her marital status at each wave.
8

We focused on whether TANF participation affects entry into marriage among mothers who were unmarried when their child was born. We used retrospective reports of TANF participation dates to construct monthly welfare histories from 1997 until the focal child was three years old (2001 to 2003). Observations for which there is no three year interview were right censored at the first follow-up interview, when the child was one. For some observations, the exact dates of the beginning or ending of TANF spells were missing. In these cases, we used available data on TANF participation status at each interview to impute dates. The TANF participation dates were used to construct three time-varying measures of TANF participation to estimate short and long- term effects. The first is a measure of current TANF participation, which was coded as 1 for months in which respondents were on TANF and as 0 for months in which they were not. The second is a measure of whether the mother was ever on TANF (since 1997), which was coded as 0 until the first month that the respondent received TANF (the baseline interview if they received TANF prior to that date) and as 1 thereafter throughout the observed period. The third is a measure of past TANF, which was coded as 1 for any given month if the respondent had been on TANF at any time since 1997 and was not currently on TANF, and coded as 0 otherwise.
2
We incorporated the following baseline measures that may be associated both with TANF participation and transitions to marriage: mother’s race/ethnicity (non-Hispanic white, non-
Hispanic black, Hispanic, and other), mother’s educational attainment (less than high school, high school or equivalent, or more than high school), mother’s nativity (U.S.-born vs. foreignborn), whether the mother was cohabiting with the baby’s father, parity (whether the child was the mother’s first), mother’s age (whether she was at least 20 years old), the difference in age
2 By considering only welfare participation since 1997, we are excluding previous AFDC participation from our measure of past participation. It is therefore possible that a mother who relied on AFDC but not on TANF would be coded as not having relied on TANF in the past.
9

between the mother and father (in years), whether the birth was covered by Medicaid, whether the mother lived with both of her biological parents at age 15, the mother’s health (excellent, very good or good, compared to fair or poor), and whether she attended religious services at least several times per month. We also included city fixed effects to control for state policies and characteristics that may impact both TANF participation and marriage.
METHODS
We employed an event history approach to model the effect of TANF participation on the likelihood and timing of marriage. Specifically, we estimated Cox proportional hazard models in which duration was measured in months from the child’s birth. As indicated earlier, all baseline unmarried mothers who completed one-year follow-up interviews were included, whether or not they completed 3-year follow-up interviews. Individuals who did not marry during the period of time they were observed were right-censored at the time of their last interview. Breslow’s method was used for handling ties. The Schoenfeld residuals method confirmed that we were not violating the proportionality assumption. We tested the sensitivity of our results to various sample restrictions.
Using event history models has several advantages: First, we did not have to choose an arbitrary time point at which to assess marital status and could determine the extent to which
TANF participation is associated with delays in marriage; second, we could be confident that our results were not driven by the effect of marriage on TANF participation (reverse causality).
By including measures of current and past TANF in our analyses, we were able to disentangle associations between participation and marriage that are short-term (i.e., they exist during the recipiency period only) and those that are long-term. Doing so also allowed us to estimate the expected cumulative impact of participation on marriage over the life course and to
10

examine the underlying mechanisms. While we could not test specific hypotheses, we could explore whether associations between TANF participation and marriage reflect short- versus long-term mechanisms. In other words, if we found evidence of short- but not long-term effects, we could rule out the values, self-perception and stress, scarlet lettering, and selection explanations. In that case, the only plausible underlying mechanisms would be economic disincentives or short-term stigma. If we found evidence of long-term effects, then any of the hypothesized mechanisms or explanations described above could play a role.
Data limitations prevented us from modeling unobserved heterogeneity using simultaneous hazards procedures as did Fitzgerald and Ribar (we only observed one transition to marriage). However, we controlled for an extensive set of individual-level covariates and included city fixed effects that remove potential confounding effects of state policies and citylevel characteristics that may affect both TANF participation and marriage. As indicated earlier,
Fitzgerald and Ribar’s analysis of selection effects revealed that unobserved factors inflated estimates of the effect of welfare participation on exits from headship by about 10 percent. Given that finding and our extensive set of controls, our estimates of the effect of TANF participation on marriage are unlikely to be inflated by more than 10 percent. Furthermore, as mentioned earlier, the magnitude of past TANF participation effects can provide information about selection effects.
DESCRIPTIVE ANALYSIS
Characteristics of mothers by whether they ever participated in TANF between 1997 and their last interview are presented in Table 1. Overall, the urban population of unmarried mothers is poor or near-poor (40 percent of mothers had less than a high school education, and 75 percent had births covered by Medicaid). However, there are notable differences between participants
11

and non-participants: Participants are less likely than non-participants to be non-Hispanic white, to have high educational attainment levels, to be immigrants, and to be cohabiting with the baby’s father at the time of the birth. They are also more likely to have more than one child, to have relied on Medicaid to pay for the birth, and to have at least spent some time growing up in a household without both parents.
TANF participation rates in this sample are high, but the average duration of spells on
TANF is relatively short. Fifty four percent of the sample (1,589 out of 2,954 mothers) relied on
TANF at any time between 1997 and when they were last interviewed, which was between 2001 and 2003 for most mothers in the study.
3 For this group, the average length of first TANF spell that occurred between the focal child’s birth and the mother’s last interview was 11 months.
4
As shown in the survival curve in Figure 1, marriage rates are relatively low and decline slightly over the observed period. Within the first 12 months after the birth of the child, approximately 9 percent of mothers married. During the second 12-month period an additional 5 percent married, and during the third 12-month period just over 4 percent married. Marriage rates during this period vary considerably by TANF participation, at least in relative terms. Of those who were ever on TANF, 12 percent married compared to 24 percent of those who were never on
TANF (not shown in figure).
5 This difference translates into a gap of 4 percentage points per year in the proportion of mothers marrying. The extent to which these observed differences in marriage rates are due to TANF participation is the question we sought to answer. As explained
3 The mothers who did not complete three-year interviews were last interviewed between 1999 and 2001.
4 Six percent of participants were still on their first TANF spell when they were last interviewed so the actual average duration is somewhat longer.
5 The figures for percent marrying by TANF participation status are based on the three year snapshot, without consideration of whether or not TANF participation pre-dated marriage. These figures are therefore not directly comparable to the survival data for the full sample, referred to above, which are based on life-table estimates.
12

earlier, differences could reflect marriage delays associated with current TANF participation, marriage delays due to having been on TANF in the past, or characteristics (observed and unobserved) of mothers associated both with TANF participation and marriage behavior.
MULTIVARIATE ANALYSIS
We estimated three sets of Cox proportional hazard models. The first set is the most comparable to the analyses of exits from headship by Fitzgerald and Ribar. Specifically, it estimates the effect of currently being on TANF on the likelihood of transitioning into marriage.
The second set of models estimates the effect of ever having been on TANF on marriage, and the third set estimates the separate effects of current and past TANF participation. For each set, we show results from an unadjusted model, a model that adds only city fixed effects, and a model that also includes an extensive set of sociodemographic characteristics, as well as our measures of family structure as a child, maternal health, and religiosity.
The effects of current TANF participation on marriage are shown in Table 2. When controlling for the full set of covariates (Model 3), the hazard ratio associated with currently being on TANF is .61 and highly significant. This compares to a ratio of .65 from the model of
Fitzgerald and Ribar that controls for unobserved heterogeneity. The effects of covariates are consistent in direction with those obtained in prior studies of marriage and reduce the association between TANF participation and marriage. They do so by considerably more than the city fixed effects, which change the hazard ratio only from .42 to .45.
Our estimates of the effect of ever having been on TANF are shown in Table 3. These effects are smaller than those of currently being on TANF, across all models. The hazard ratio in the fully controlled model (Model 3) is .83, indicating that ever participating in TANF is not strongly associated with marriage and is of marginal statistical significance ( p = .084).
13

The effect of ever having been on TANF captures both the effect of currently being on
TANF and the association between marriage and having been on TANF in the past. In order to differentiate between the two effects, we estimated models that include both current and past
TANF participation. The latter is similar to the measure of ever on TANF used for Table 3, but coded as 0 for months in which respondents received TANF. These results are shown in Table 4.
A comparison of the estimates for current and past TANF participation indicates the relative magnitude of immediate versus longer-term associations between TANF participation and marriage, and allows us to evaluate the plausibility of hypothesized mechanisms.
In Model 4, the effect of past receipt of TANF is close to zero and highly insignificant (p
= .85). The effect of current TANF participation is virtually identical to the corresponding estimate in Table 2. It is large and highly significant. These results strongly suggest that the association between TANF participation and marriage does not extend past the recipiency period.
Once off of TANF, the likelihood of marriage reverts to that of mothers who have never been on
TANF.
We conducted a number of supplemental analyses to assess the robustness of these results. Specifically, we ran analyses corresponding to Model 4 in Tables 2, 3, and 4, restricting the sample to three economically homogeneous populations: mothers who were eligible for
TANF during the year after their child’s birth, 6 had births paid for by Medicaid, and had at most a high school education. We also estimated models that restricted the sample to mothers who were born in the U.S., as welfare participation of foreign-born mothers has been dramatically restricted since PRWORA. Finally, we estimated models that restricted the sample to cases for which we had complete information on marriage dates to insure that the results were not sensitive to the missing data imputations. The results of these analyses are shown in Table 5. The hazard ratios
6 For details on the TANF eligibility imputation method, see Reichman et al. 2004 (inclusive method).
14

for currently being on TANF (first column) are similar in magnitude to those in Table 4, and are all significant at p < .10, despite the smaller sample sizes. Substantively, the findings based on the sub-samples are similar to those from the full sample: Effects of ever having been on TANF are somewhat smaller than those of currently being on TANF (second column) and effects of having been on TANF in the past are close to 0 (likelihood ratios are close to 1) and not significant for any sample.
PROJECTED CUMULATIVE EFFECTS
In the last step of our analysis, we used the results from Table 4 to project the effects of
TANF participation on the probability of marriage and on the average delay in marriage over an
18-year period.
7 Specifically, we assumed that the effect of TANF participation on marriage over the life course would be equal to the expected number of years on TANF multiplied by the participation effect.
First we computed an expected marriage rate for our sample population over 18 years by weighting race/ethnic-specific marriage rates of women with non-marital births (computed by
Graefe and Lichter (2002) using the 1995 National Survey of Family Growth) to the racial/ethnic composition of the urban FF sample.
8 This results in an estimated marriage rate of 62 percent over the 18 years, or an average of 3.5 percent per year.
We then estimated the percentage of unmarried mothers who would be on TANF in any given year over the 18-year period. Using data from the National Longitudinal Survey of Youth from 1979 to 1996, Moffitt (2002) found that welfare recipients received AFDC for an average of
7 We focused on an 18-year period, which corresponds to the period of time before which the focal child would reach majority age. It also would, on average, correspond to the end of TANF eligibility for that mother.
8 The Graefe and Lichter rates are 82 percent for whites, 62 percent for Hispanics, and 59 percent for blacks. Our sample is 15 percent white, 28 percent Hispanic, and 54 percent black.
15

39 months over a 10-year period. We assumed 3 years (36 months) as a lower bound and 4 years
(48 months) as an upper bound average for TANF over an 18-year period.
9 Using the proportion of baseline unmarried mothers in our sample who were ever on TANF by the three year followup interview (54 percent) as a guide, we assumed 60 percent as a lower bound estimate of the percentage that will ever be on TANF over an 18 year period and 75 percent as an upper bound estimate. Based on these figures, 10 to 17 percent of baseline unmarried mothers would be on
TANF in any given year.
10 Using the coefficient of current TANF from Model 4 in Table 2 (.61) as the likelihood of marriage among TANF participants relative to that of non-participants, we calculated the average annual marriage rate for non-participants to be 3.65 and that of TANF participants as .61 times that, or 2.23.
11 Based on our finding that the effects TANF participation do not extend past the recipiency period, we project that TANF participation would decrease the proportion of mothers who marry by at most 4 to 6 percentage points over 18 years.
12 That is, approximately 60 percent of mothers who had spent any time on TANF would be married within
18 years of having their child, compared to 65% of those who had never been on TANF. While this difference is not substantial, the association of TANF participation with marriage delay is considerably larger. We estimate that the average delay in marriage among mothers who
9 The average amount of time on TANF is likely to be lower than what it was on AFDC because of the time limits and other restrictions under PRWORA and a stronger labor market. However, Moffitt’s figures cover a 10-year, rather than an 18-year, period. For these reasons, we use figures that frame Moffitt’s figure of 39 months.
10 Lower bound estimate: (3 years / 18 years)* .60 = .10; upper bound estimate: (4 years / 18 years)* .75 = .17.
11 The annual marriage rate, divided by: the proportion on TANF in a year multiplied by the proportion who will ever participate in TANF, plus the proportion not on TANF in a year = 3.5/(.10*.6 + .9) = 3.65 and 3.5/(.17*.75 + .
83) = 3.66. As these figures are virtually identical, we use 3.65.
12 (3.65-2.23) * 3 years total time on TANF = 4.26 and (3.65-2.23) * 4 years total time on TANF = 5.68.
16

participate in TANF compared to those who do not ranges from 23 to 31 months.
13 These projections are insensitive to the assumptions we made.
DISCUSSION
We examined the extent to which TANF participation is associated with the likelihood and timing of marriage of mothers who had non-marital births. We did not address the muchstudied question of whether welfare policies affect marriage (and if so, by how much); rather, we focused on the little explored question of whether participation in TANF is associated with decreased likelihood of marriage. This is the first and only study to assess whether there are welfare participation effects on marriage beyond the immediate recipiency period. This is an important question, as longer-term effects would suggest deeper-rooted and permanent changes in behavior and would support scarring or selection arguments.
We found evidence that TANF participation has a negative effect on the likelihood of marriage, but that the effect is for the most part confined to the period of participation. Our estimated short-term effects are similar in magnitude to those found by Fitzgerald and Ribar, despite different data, populations, and time periods studied. We projected that the total effects would translate into modest differences in marriage rates over an 18-year period, but that TANF participation may result in marriage delays of 2 to 2.5 years over this period. The projections do not hinge on our assumed average length of TANF spell, which is a pre-welfare-reform figure and therefore likely to be somewhat inflated given the new restrictive welfare environment.
Since marriage is a potentially important route out of poverty for unwed mothers (Lichter,
Graefe, and Brown 2003), delays in marriage associated with TANF participation may have
13 4.26 percent fewer mothers marrying, divided by the annual marriage rate of 2.23, times 12 months = 22.9
months; (5.68 / 2.23) * 12 = 30.6 months.
17

detrimental effects on mothers’ and children’s economic well-being. On the other hand, delays in marriage could have favorable effects on family stability by leading to more selective searches for mates, which could result in higher quality or longer-term relationships.
We did not directly model selection into TANF. It is therefore possible that our estimates of the effects of TANF participation on marriage are biased. Given what is known about characteristics associated with both marriage and welfare participation, we would expect that unobserved characteristics lead to overestimates of the effects of TANF participation on marriage. That is, TANF participants have characteristics that are negatively associated with marriage. This argument is consistent with the finding of Fitzgerald and Ribar that unobserved heterogeneity leads to upward biases in estimates of the effects of welfare participation on marriage. With even more control variables than Fitzgerald and Ribar included in their models and city fixed effects, our estimates of the effect of current TANF participation on marriage are likely to be biased upward only slightly. Given that our estimates of past TANF participation are close to zero and not significant, we do not think that selection is of much concern here.
Earlier, we discussed potential mechanisms by which TANF participation could affect marriage. Our findings help to adjudicate between some of the hypothesized mechanisms.
The effects of current welfare participation appear to be either due to direct financial repercussions of getting married (at least to a man who has earnings), which could lead to immediate ineligibility for further benefits, or to short-term stigma. In terms of the latter, the association of TANF with financial insecurity may label mothers as not marriageable or place them below a self-perceived marriage bar, as suggested by Edin and Kafalas (2005) and Gibson,
Edin and McLanahan (forthcoming). Our analyses do not speak to the relative importance of these two classes of short-term mechanisms, as the results are consistent with both of them.
18

Our finding of minimal associations between TANF participation and marriage beyond the recipiency period does not support the other (long-term) arguments. While we do not directly examine mothers’ attitudes toward marriage, the fact that marriage behaviors of participants and non-participants are so similar after TANF spells suggests that welfare participation does not have scarring effects (through values, self-perception, or scarlet lettering). The absence of longterm effects also contraindicates the selection argument. It is important to note that our analyses do not pertain to the effect of poverty on marriage, nor do they test mechanisms by which poverty affects marriage. Any of the long-term mechanisms discussed earlier could be important pathways by which poverty is associated with family structure.
Welfare debates in Washington have increasingly focused on marriage promotion. While conservatives and liberals may not agree on why marriage is important – some view marriage as a moral issue, others believe marriage can help break the cycle of poverty and dependence on public assistance, and still others are most concerned about its impact on child well-being – the two sides generally agree that marriage should be encouraged, or at least not be discouraged, by the welfare system. Our paper indicates the extent to which one aspect of the welfare system – participation – affects marriage behaviors among low-income women. We find that TANF participation does not discourage marriage over the long run but that it may delay marriage by up to 2.5 years. Further, and perhaps most importantly, these findings provide indirect evidence that participants and non-participants are not distinguishable in terms of cultural or personal orientations toward marriage. This suggests that marriage behaviors are shaped by welfare participation only through short-term economic incentives or stigma.
This study is subject to some limitations. It is based on an exclusively urban sample. It is possible that TANF participation effects are different in rural, small town, or suburban communities. For example, if short term stigma or longer term scarring associated with TANF
19

participation is greater in non-urban settings, where TANF participation is less prevalent than in urban settings (Leonard and Kennedy 2001), TANF participation effects on marriage may be larger than those we estimated using the FF sample. Also, we extrapolated cumulative effects using only three years of data. Effects of current participation on marriage may not be uniform over the child’s first 18 years. It is also possible that there are longer-term “sleeper effects” that we are unable to observe in the three year time frame. Finally, our estimates of welfare participation effects are based solely on the post-1996 period. It is possible that scarring effects of participation existed under the former AFDC program.
20

REFERENCES
Becker, G. 1973. A theory of marriage: Part I.
Journal of Political Economy , 81:813-846
Becker, G. 1974. A theory of marriage: Part II.
Journal of Political Economy , 82:511-526
Blank, R. 2002. Evaluating welfare reform in the United States.
Journal of Economic Literature
40(4): 1105-1166.
Blau, F., Kahn, L., and Waldfogel, J. 2004. The impact of welfare benefits on single motherhood and headship of young women - Evidence from the census.
Journal of Human Resources.
39(2): 382-404.
Corcoran, M. 1995. Rags to rags: poverty and mobility in the United States.
Annual Review of
Sociology 21: 237-267.
DeKlyen, M., Brooks-Gunn, J., McLanahan, S., and Knab, J. Forthcoming. The mental health of parents with infants: do marriage, cohabitation, and romantic status matter?
American Journal of Public Health .
Edin, K. and Kafalas, M. 2005.
Promises I can keep: Why poor women put motherhood before marriage.
University of California Press.
Fitzgerald, J. 1995. Local labor markets and local area effects on welfare duration.
Journal of
Policy Analysis and Management 14(1): 43-67.
Fitzgerald, J. and Ribar, D. 2004. Transitions in welfare participation and female headship.
Population Research and Policy Review.
23: 641-670.
Gennetian, L. A. and Knox, V. 2003.
Staying Single: The Effects of Welfare Reform Policies on
Marriage and Cohabitation . New York, MDRC.
Gibson, C., Edin, K., and McLanahan, S. Forthcoming. High hopes, but even higher expectations: The retreat from marriage among low-income couples.
Journal of
Marriage and Family.
Goodban, N. 1985. The psychological impact of being on welfare.
Social Service Review . 59:
403-422.
Graefe, D. and D. Lichter 2002. Marriage among unwed mothers: Whites, blacks, and Hispanics compared. Perspectives on Sexual and Reproductive Health 34(6): 286-293.
Hoynes, H. 2000. Local labor markets and welfare spells: Do demand conditions matter?
The
Review of Economics and Statistics 82(3): 351-368.
Huang, C.-C., I. Garfinkel, and J. Waldfogel. 2004. Child support enforcement and welfare caseloads.
The Journal of Human Resources 39(1):108-134.
Jarrett, R.L. 1996. Welfare stigma among low-income, African American single mothers.
Family Relations . 45: 368-374.
Jayakody, R. and Stauffer, D. 2000. Mental health problems among single mothers:
Implications for work and welfare reform.
Journal of Social Issues . 56(4): 617-634.
Klugel, R. and Smith, E, 1986.
Beliefs about inequality: Americans’ views of what is and ought to be . New York: Aldine De Gruyter.
Leonard, P. and Kennedy, M. 2001. “What Cities Need from Welfare Reform Authorization.”
Center on Urban and Metropolitan Policy, The Brookings Institution.
Discussion Paper
[http://www.brook.edu/dybdocroot/es/urban/publications/leonkencitieswelfare.pdf]
Lewis, O. 1968. The culture of poverty, in D. Moynihan (ed.) On Understanding Poverty . New
York: Basic Books.
Lichter, D., Graefe, D., and Brown, J. 2003. Is marriage a panacea? Union formation among economically disadvantaged unwed mothers. Social Problems, 50(1):60-86.
21

Mead, L. 2000. Caseload change: An exploratory study.
Journal of Policy Analysis and
Management 19(3): 465-472.
Mead, L. 2003. Caseload change: An alternative approach.
Policy Studies Journal 31(2): 163-
185.
Meyer, D. 1993. Child support and welfare dynamics: evidence from Wisconsin.
Demography
30(1): 45-62.
Moffitt, R. 1998. “The effect of welfare on marriage and fertility,” pp. 50-97 in R. Moffitt (ed.),
Welfare, the family, and reproductive behavior . Washington, DC: National Academy
Press.
Moffitt, R. 2002. “Experienced-based measures of heterogeneity in the welfare caseload,” in
Ver Ploeg, M., Moffitt, R., and Citro, C. (eds.) Studies of Welfare Populations: Data
Collection and Research Issues.
Panel on Data and Methods for Measuring the Effects of
Changes in Social Welfare Programs. Committee on National Statistics. Division of
Behavioral and Social Sciences and Education. National Research Council.
http://aspe.hhs.gov/hsp/welf-res-data-issues02
Moffitt, R. Reville, R., and Winkler, A. 1998. Beyond single mothers: cohabitation and marriage in the AFDC program.
Demography , 35(3): 259-278.
Moynihan, D. 1965.
The Negro Family:The Case for National Action . Washington DC: The U.S.
Department of Labor Office of Policy, Planning, and Research.
Nichols-Casebolt, A. 1986. The psychological effects of income testing income-support benefits.
Social Service Review . 60: 287-302.
Noonan, M. and Heflin, C. Forthcoming. Does welfare participation affect women's wages?
Social Science Quarterly.
86(5).
Pavetti, L. 1997. “How much more can they work? Setting realistic expectations for welfare mothers.” Washington, DC: Urban Institute.
Peters, H., Plotnick, R., and Jeong, S. 2001. “How will welfare reform affect childbearing and family structure decisions?” Institute for Research on Poverty No. 1239-01. University of
Wisconsin, Madison, WI: IRP Publications.
Rainwater, L. 1982. “Stigma in income-tested programs,” in I. Garfinkel (ed.), Income tested programs: The case for and against . New York: Academic Press.
Ratcliffe, C., McKernan, S., and Rosenberg, E. 2002. “Welfare reform, living arrangements, and economic well-being: A synthesis of literature.” Washington, DC: The Urban Institute.
Rector, R. and Johnson, K. 2004. Understanding poverty in America. Heritage Foundation
Backgrounder, no 1713.
Reichman, N., Teitler, J., Garfinkel, I. and Garcia, S. 2004. Variations in maternal and child wellbeing among financially eligible mothers By TANF participation status.” Eastern
Economic Journal 30 (1): 101-118 .
Reichman, N., Teitler, J., Garfinkel, I., and McLanahan, S. 2001. Fragile families: sample and design.
Children and Youth Services Review.
23(4-5): 303-326.
Ribar, D. 2005. Transitions from welfare and the employment prospects of low-skill workers.
Southern Economic Journal 71(3): 514-33.
Schram, S. 2000. In the clinic: the medicalization of welfare.
Social Text 18(1): 81-107.
Tumlin, K. and W. Zimmerman. 2003. “Immigrants and TANF: A look at immigrant welfare recipients in three cities.” The Urban Institute
[http://www.urban.org/UploadedPDF/310874_OP69.pdf].
Yelowitz, A.1995. The Medicaid notch, labor supply, and welfare participation: Evidence from eligibility expansions.
The Quarterly Journal of Economics 110(4): 909- 39.
22

Yelowitz, A. 1998. Will extending Medicaid to two-parent families encourage marriage?
Journal of Human Resources 18(1): 65-81.
Zedlewski, S. 2002. “Left behind or staying away? Eligible parents who remain off TANF.”
The Urban Institute. Policy Brief. Series B, No. B-51.
23

Table 1: Sample Characteristics By TANF Participation Status (Since 1997)
Married to father or partner
Non-Hispanic white
Non-Hispanic black
Hispanic
Other race/ethnicity
Less than high school
High school graduate
More than high school
Born in U.S.
Cohabiting with father at baseline
First birth
Age >=20
Difference in parents' ages (years)
Medicaid birth
Lived with both biological parents at age 15
Good, very good, or excellent health at baseline
Attends religious services several times/month
Length of first TANF spell (months)
N
Ever on
TANF
2.8
84
29
90
41
31
76
34
19
3
47
94
12
10
66
22
31
11
1589
37
1365
Never on
TANF
2.7
64
46
93
58
53
78
32
36
3
31
79
24
21
41
35
All
Mothers
2.8
75
37
91
49
41
77
33
27
3
40
87
18
15
54
28
34
11
2954
24

Table 2: Effect of Current TANF Participation on Hazard of Marriage
Model 1 Model 2 Model 3 Model 4
Currently on TANF
Non-Hispanic black
Hispanic
Other race/ethnicity
High school graduate
More than high school
Born in U.S.
Cohabiting with father at baseline
First birth
Age >=20
Difference in parents' ages
Medicaid birth
Lived with both biological parents at age 15
Good, very good, or excellent health at baseline
Attends religious services several times/month
City fixed effects
N (2954)
0.42
(.000)
No
91058
0.45
(.000)
Yes
91058
Yes
91058
1.65
(.000)
0.68
(.007)
2.34
(.000)
0.83
(.058)
1.00
(.989)
1.00
(.903)
0.92
(.443)
0.99
(.930)
0.61
(.002)
0.48
(.000)
0.75
(.049)
0.73
(.226)
1.24
(.060)
0.99
(.949)
1.20
(.053)
Yes
91058
1.67
(.000)
0.64
(.002)
2.31
(.000)
0.82
(.046)
1.00
(.988)
1.00
(.910)
0.92
(.394)
0.61
(.002)
0.50
(.000)
0.77
(.071)
0.73
(.239)
1.26
(.046)
Figures are proportional hazard ratios and p -values
25

Table 3: Effect of Ever Having Received TANF on Hazard of Marriage
Model 1 Model 2 Model 3 Model 4
Ever received TANF 0.58
(.000)
0.63
(.000)
Non-Hispanic black
Hispanic
Other race/ethnicity
High school graduate
More than high school
Born in U.S.
Cohabiting with father at baseline
First birth
Age >=20
Difference in parents' ages
Medicaid birth
Lived with both biological parents at age 15
(.002)
2.31
(.000)
0.80
(.028)
1.00
(.975)
1.00
(.942)
0.92
(.393)
0.82
(.069)
0.49
(.000)
0.77
(.074)
0.74
(.241)
1.28
(.034)
1.71
(.000)
0.64
Good, very good, or excellent health at baseline
Attends religious services several times/month
City fixed effects
N (2954)
No
91058
Figures are proportional hazard ratios and p -values
Yes
91058
Yes
91058
(.933)
0.92
(.435)
0.99
(.923)
1.00
(.998)
1.20
(.057)
(.007)
2.35
(.000)
0.81
(.037)
1.01
(.950)
1.00
0.83
(.084)
0.47
(.000)
0.75
(.051)
0.73
(.224)
1.26
(.045)
1.69
(.000)
0.67
Yes
91058
26

Table 4: Effects of Past and Current TANF Receipt on Hazard of Marriage
Model 1 Model 2 Model 3
Received TANF in past
Currently on TANF
Non-Hispanic black
Hispanic
Other race/ethnicity
High school graduate
More than high school
Born in U.S.
Cohabiting with father at baseline
First birth
Age >=20
Difference in parents' ages
Medicaid birth
0.72
(.003)
0.39
(.000)
0.78
(.025)
0.42
(.000)
(.000)
0.65
(.002)
2.30
(.000)
0.82
(.044)
1.00
(.994)
1.00
(.913)
0.92
(.418)
0.97
(.783)
0.60
(.002)
0.50
(.000)
0.77
(.073)
0.74
(.247)
1.26
(.047)
1.67
Lived with both biological parents at age 15
Good, very good, or excellent health at baseline
Attends religious services several times/month
City fixed effects
N (2954)
No
91058
Figures are proportional hazard ratios and p -values
Yes
91058
Yes
91058
Model 4
(.985)
1.00
(.905)
0.93
(.460)
0.99
(.925)
0.99
(.000)
0.68
(.008)
2.34
(.000)
0.83
(.058)
1.00
(.949)
1.20
(.055)
0.98
(.853)
0.60
(.002)
0.48
(.000)
0.75
(.051)
0.73
(.231)
1.24
(.061)
1.65
Yes
91058
27

Table 5: Effects of TANF Participation on Hazard of Marriage (Sub-Samples)
Sample:
Currently on TANF Ever on TANF Past TANF
Corresponds to
Table 2, Model 4
Corresponds to
Table 3, Model 4
Current
TANF
Corresponds to Table 4, Model 4
Eligible for TANF
N = 1187
0.69
(.095)
0.97
(.870)
1.16
(.458)
0.73
(.207)
Medicaid births
N = 2220
High school education or less
N = 2154
0.63
(.007)
0.60
(.004)
0.87
(.264)
0.73
(.013)
1.04
(.290)
0.85
(.265)
0.64
(.012)
0.57
(.002)
U.S.-born
N = 2568
0.62
(.004)
Non-imputed marriage dates
N = 2891
0.55
(.001)
Figures are proportional hazard ratios and p -values
0.80
(.056)
0.79
(.038)
0.93
(.544)
0.94
(.632)
0.60
(.003)
0.53
(.001)

1.0
0
Figure 1: Kaplan-Meier Unmarried Survival Estimates
0.7
5
0.5
0
0.2
5
0.0
0
0 10 20
Analysis Time: Months
30 40
29