Introduction

Although adolescent parenthood has shed some of its stigma in recent years, teenagers who have children can still suffer adverse consequences, both for the children and for themselves (Aparicio et al. 2016; Barr and Simons 2012; Chumbler et al. 2014; Ellis-Sloan 2014; Smithbattle 2007; Yardley 2008). Teenage mothers tend to attain lower levels of education and skill formation, largely due to disadvantageous family and personal characteristics (Kalil and Kunz 1999; Lichter and Qian 2008; Miller and Moore 1990; Schoen et al. 2007). Teen mothers also are likely to continue having children, and to further delay the education and training needed for the labor market (Beutel 2000; Boden et al. 2008; Davis 2002; de Fátima Rato Padin et al. 2009; Furstenberg 1976; Hoffman et al. 1993; Hofferth et al. 2001; McDermott and Graham 2005; Teti and Lamb 1989; Thrane and Chen 2012). Furthermore, teen mothers have an increased likelihood of exhibiting behavioral problems (Ketterlinus et al. 1992; Manlove 1997) and experiencing disruptions in their psychological well-being (Booth et al. 2008; Nayak and Kehily 2014).

Similarly, with respect to teenage fathers, the precursor and consequences of teenage fatherhood have remained relatively stable. Teen fathers tend to have lower educational achievement, lower socio-economic status, and a greater likelihood of involvement in a variety of risky behaviors than their non-father peers (Card and Wise 1978; Fletcher and Wolfe 2012; Glikman 2004; Marsiglio 1988; Pirog-Good 1988, 1995, 1996; Robbins et al. 1985; Stouthamer-Loeber and Wei 1998; Thornberry et al. 1997; Vinnerljung et al. 2007; Wilkinson et al. 2009). As a result of their lower job skills and risky or delinquent behaviors, teen fathers are less likely to fulfill their child support obligations and less likely to share the domestic burden of caring for children (even those with whom they cohabit), both of which negatively affect children’s wellbeing (Cancian et al. 2011, 2013; Nepomnyaschy and Garfinkel 2010; Pirog-Good and Good 1995).

A vast number of previous studies identify consistent patterns of economically disadvantaged backgrounds, educational deficits, and relatively weak labor market outcomes of teen parents (Card and Wise 1978; Fergusson and Woodward 2000; Geronimus and Korenman 1993; Hofferth and Moore 1979; Upchurch and McCarthy 1990; Waite and Moore 1978). However, scholars have questioned whether these disadvantages are the causes or consequences of early parenting or both. For example, some scholars argue that the time spent child rearing and the stress associated with parenting are likely to hinder teens’ educational attainment and/or employment opportunities (Barber et al. 1999; Becker 1981). Others, however, suggest that the observed lower levels of educational attainment and poorer labor market outcomes could be attributable to pre-existing socioeconomic difficulties, such as teen parents’ family environments. These pre-existing deficits are understood to be predictive of teen parenting (Geronimus and Korenman 1993; Hotz et al. 2005). In general, scholars accept both causal interpretations: Disadvantages in childhood and early adolescence are predictive of becoming a teen parent and having become a teen parent is predictive of subsequent disadvantages. In this paper, we set aside a discussion of the precursors of teen parenting and look instead at how the life experiences of teen parents have changed in recent decades. It is hypothesized that the national trends in cohabitation and out-of-wedlock parenting may have reduced the stigma associated with teen parenting and that this might be reflected in somewhat better outcomes for teen parents.

In this study, we focused on two decades of unique periods that witnessed distinct changes in sexual behavior among youth. Since the beginning of our study period (1979), there has been sizeable decline in teen birth rates (Santelli and Melnikas 2010). The proportion of sexually experienced male and female teenagers aged 15–19 has consistently declined across selected years: 56% in 1988, 37.9% in 1995, and 30.6% in 2006–2010 (Abma and Sonenstein 2001; Martinez et al. 2011). During the period, contraceptive utilization among sexually active teenagers has increased from 70.7% in 1995 to 85.6% in 2006–2010 (Abma and Sonenstein 2001; Kearney and Levine 2015; Martinez et al. 2011). A combination of these two factors at least partially explains changes in teenager fertility over our study periods.

At the same time, state-level public policies attributed to the unique trends in our study period. States initiated policies on limited access to contraceptives and abortion services in varying degrees, disproportionately affecting teens. For example, two states (Texas and Utah) require parental consent for teen’s contraceptive services; twenty-one states and the District of Columbia explicitly allow minors to obtain contraceptive services without a parent’s involvement; another twenty-five states determine access to contraceptives as teen’s privacy rights (Frost 2013). An implicit purpose of these laws is to prevent abortions among teens, either by increasing the cost of abortions or by inducing substitution from abortion to contraception. Researchers have found some association between the parental involvement law and the reduction in the abortion rate, but the law’s causal impact on teens’ birthrate showed mixed results (Dennis et al. 2009).

Also, some states enforced mandatory counseling and waiting period laws on abortion. It requires that women should receive counseling before an abortion is performed and then must wait a specified time period between the counseling and the procedure. As of 2009, 36 states require a mandatory counseling before an abortion is performed; 27 of these states requires a woman seeking an abortion to wait a mandatory period, most often 24 h (Joyce et al. 2009). Overall, studies found that the law had no impact on abortion rates or birthrates (Bitler and Zavodny 2001; Meier et al. 1996; Medoff 2007). However, in Mississippi, there was a decline in the abortion rate in Mississippi, the state with the most restrictive laws regarding abortion (Althaus and Henshaw 1994; Joyce et al. 1997; Joyce and Kaestner 2001).

More recently, in Texas, the legislature passed a law that imposes strict requirements on abortion providers. It requires abortion facilities to meet the standards of Ambulatory Surgery Centers (ASCs) and mandates physicians performing abortions to have admitting privileges at a nearby hospital. Studies found that even though the supreme court struck down these two provisions as unconstitutional, the abortion rate among the teens in Texas decreased substantially (Colman and Joyce 2011; Grossman et al. 2014). Research also found that women continued to seek abortion services across state lines; however, abortions obtained out of state did not offset striking declines in Texas (Colman and Joyce 2011).

Additionally, the federal government’s policy to promote marriage and two-parent families could affect marriage among teens (Graefe and Lichter 2008; Hao and Cherlin 2004; Kaestner et al. 2003; Lopoo and DeLeire 2006). After the 1996 Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) was enacted, the welfare program encouraged marriage and the formation and maintenance of two-parent families. Under the PRWORA, welfare policies make it harder to qualify for receiving benefits for single mothers and imposed family caps on welfare benefits (Acs and Nelson 2004). Also, PRWORA mandated that states develop systems to establish voluntary paternity acknowledgement, provide information about the rights and responsibilities of minor parents, and required most teenage mothers to live with a parent or another approved adult (Rozie-Battle 2003). Studies attribute that these PRWORA provisions have discouraged nonmarital births, including teenage parenthood (Hao and Cherlin 2004; Kaestner et al. 2003; Lopoo and DeLeire 2006). At the same time, by 1998, more coordinated state and federal child support enforcement was implemented through the Uniform Interstate Family Support Act (UIFSA), while child support enforcement responsibility moved from the courts to administrative agencies (Huang 2010; Pirog and Ziol-Guest 2006). These changes in the child support enforcement provided teen mothers with greater incentives to inform fathers of their paternity, which likely contributed to increase reported rates of teen fatherhood (Case et al. 2003; Cassetty and Hutson 2005; Huang 2010; Nepomnyaschy and Garfinkel 2010).

In addition to changes in public policy, shifting norms related to cohabitation and parenting outside of marriage may influence the number and composition of teenagers who choose to become parents. Over past decades, there has been an increase in cohabitation among teen parents. In 2002, about one-third (35.1%) of unmarried teenage mothers were cohabitating with their partners; however, cohabitation of unmarried teens increased to nearly half (45.6%) between 2006 and 2010 (Chandra et al. 2005; Martinez et al. 2012). At the same time, a majority of teen mothers give birth while cohabiting and outside of marriage (Martinez et al. 2012). After the birth of the child, only about half of teen mothers are living with or married to the father of their child within a year (Eshbaugh 2008). During past decades, cohabitation has been the most common family formation activities among teenagers, followed by teen childbearing as a close second (Manning and Cohen 2015).

The growth in cohabitation among teen parents has raised concerns over the potential consequences for well-being of children. Research consistently show worse outcomes for children raised by a teen parents, including academic performance, social, emotional and behavioral problems, and delinquency issues (e.g., Cooksey 1997; Hardy et al. 1997; Hofferth and Reid 2002). Teenage cohabitation is also tied to subsequent life course outcomes among teen parents. Their relationship is less enduring after the birth of the child, and divorce is common among teen parents who choose marriage (Eshbaugh 2008; Furstenberg et al. 1987).

These normative and policy shifts could, in turn, influence teens’ subsequent choices and outcomes related to living arrangements, educational attainment, and labor market outcomes. In this study, we examine these longer-term consequences not only to better understand the changing profiles of teenage mothers and fathers and their life experiences after a child’s birth, but also to inform relevant public policies intended to prevent unplanned teenage parenthood and develop public programs that help teenage parents form healthy and nurturing family environments for their children.

Thus, the purpose of this paper is to update previous studies by examining the changing social characteristics of teenage mothers and fathers in recent decades. Specifically, we compare the characteristics of teen parents have changed across two nationally representative birth cohorts in the National Longitudinal Survey of Youth (NLSY) 1979 and 1997, report on differences in adult cohabitation rates, and discuss the choices teen parents are making about living with their children. We also examine the risk factors associated with becoming a teen parent and track teen parents’ educational and labor market outcomes until the age of 29 to examine whether the outcomes associated with become teen parents have changed in recent decades.

We find the NLSY 1979 and 1997 data to be ideal for our study. The NLSY 1979 birth cohort had passed their teenage years by the time PRWORA was enacted in 1996. Respondents in the 1997 NLSY were still in teenagers in 1996, and would therefore would have been subject to PRWORA provisions as teen parents. Therefore, the new PRWORA provisions on welfare benefits are likely to have impacted teen parents in the NLSY 1997 cohort, but would not have affected the earlier cohort at any point of interest to our study. In addition to PRWORA, the de-stigmatization of non-marital cohabitation over the years and greater enforcement of child support laws have potential consequences for living arrangements, educational attainment, labor earnings, and public policies targeting these groups.

Method

Data and Sample

This study relies on two sources of data: the 1979 National Longitudinal Survey of Youth (NLSY79) and the 1997 National Longitudinal Survey of Youth (NLSY97), provided by the U.S. Bureau of Labor Statistics (BLS). The NLSY79 is a sample of 12,686 young men and women who were born between January 1, 1957 and December 31, 1964. At the time of their first interview at 1979, their ages ranged from 14 to 22. After the first survey, follow-up interviews were conducted annually until 1994 and biannually after that year. The NLSY97 consists of a nationally representative sample of approximately 9000 youths, born between January 1, 1980 and December 31, 1984. At the time of the first interview, respondents’ ages ranged from 12 to 18, and they were interviewed on an annual basis. During both NLSY79 and NLSY97 interview period, an interviewer visited randomly selected households to identify all survey eligible youths. If an eligible youth lived in the household, the interviewers also interviewed one of the youth’s parents (including non-biological parents) or parent-type figure (including guardian or foster parents).

One advantage of the NLSY is that the BLS provides weights for each observation in each panel year to allow us to generate nationally representative profiles of teen parents and youth who do not become teen parents. The BLS also provides explicit instructions on how to construct a single weight per observation when either pulling observations from different panel years or pooling data across years. Our descriptive and multivariate analyses all utilize either the set of sampling weight provided by the BLS or customized longitudinal weights using BLS’s guidelines, which adjusts the sampling design and the use of data for more than one round of surveying.

There is considerable variation in the age of the survey respondents in the baseline years of the NLSY79 and NLSY97 cohorts; youth in the NLSY79 cohort were 14–22 years old and those in the NLSY97 cohort were 12–17 years old. Thus, simply using the two cohorts as given does not provide a valid comparison of teen parents across the two cohorts. For example, because of differences in maturity and life experiences, it would be inappropriate to compare 18–22 year olds in the baseline year of the NLSY79 cohort to 11–12 year olds in the baseline year of the NLSY97 cohort. Additionally, some NLSY respondents retrospectively reported their child’s births up to three years prior to their survey year. For example, 10.52 and 11% of births occurred before the first survey year among teen fathers of our sample in NLSY79 and NLSY97. For teen mothers of our sample, it is 25.22 and 22.74% in NLSY79 and NLSY97. As a result, differences in ages at the baseline might impact the recollection of birth dates of children and introduce considerable measurement error, particularly for men who less frequently live with their children. Therefore, we restrict our sample to youth aged 15–17 in the baseline years of both panels. In other words, respondents born in 1962, 1963, or 1964 in the NLSY79 and those born in 1980, 1981, or 1982 in the NLSY97 are included here in order to create comparably aged birth cohort data from both panels.

There are many identical questions in both panels which facilitate comparisons across the two panels on such topics as fertility, family backgrounds, and economic well-being. Further, the NLSY panels follow respondents for a long time. To be specific, this allows us to follow the birth cohorts (ages 15–17 at the baselines) for fifteen years when respondents are in their early thirties. When we restrict the two panels to comparably aged birth cohorts, we have 4415 individuals whom we follow from the NLSY79 and 5406 individuals from the NLSY97.

Measurement

This study has three key dependent variables. First, we create an indicator variable for teenage parenthood, based on the reported month and year of the birth of the respondent’s first child. Births and birthdates are reported retrospectively in the NLSY panels, allowing us to capture births that occurred even if a respondent was not interviewed in every panel year. Consistent with the National Center for Health Statistics, we defined teenage parenthood as being age 19 or younger at the birth of their first child, based on the definition by National Center for Health Statistics, where calculate teen birth rate as births per 1000 females aged 15–19. Indeed, previous studies on teenage pregnancy using NLSY 1979 and 1997 followed this definition (Argys and Peters 2003; Chafel 1994; Furstenberg et al. 1990; Geronimus and Korenman 1993; Hofferth and Reid 2002; Hoffman et al. 1993; Levine et al. 2001; Su et al. 2015) The second dependent variable is an indicator variable equal to one if the respondent received a college degree by age 29, and zero if the respondent did not. The third dependent variable is the respondents’ labor earnings at age 29. The independent variables include the educational attainment of the respondents’ parents, family structure, family income in constant 2017 dollars, and the respondents’ ethnic background.

Among independent variables, family structure is constructed using questions about respondents’ childhood household composition. The surveys asked respondents with whom they lived at age 14 (NLSY79) and at age 12 (NLSY97), and these data provide information on the household structure during the respondent’s teenage years. Family structure is measured with five variables: living with both biological parents; living with either a stepfather or stepmother; living with a biological mother only or mother’s marital status is unknown; living with a biological father only or father’s marital status is unknown, and living with other family arrangements such as adoptive or foster parents.

It is well documented that childhood family structure is closely linked to the life course of children, and structural disadvantages associated with single parent households influence social and economic hardship for children (Amato 2005; Aquilino 1996; Bumpass and Lu 2000; Cherlin 2010; Musick 2002; Seltzer 2000; Teachman 2003; Wu and Martinson 1993). Previous literature mainly focused on childhood living arrangements with single mothers and their children’s behavioral, cognitive, or educational outcomes (Brown 2004, 2006; Carlson and Corcoran 2001). Since 1990, scholars have increasingly recognized that changing demography of single fatherhood and its influence on children needs to be studied (Carlson 2006; Marsiglio 1994; Marsiglio et al. 2000). However, single father families are still less common than single mother households (Brown et al. 2015; Hofferth 2006; Hofferth and Anderson 2003), and despite the relative richness of the NLSY for analyzing single father families, the data on this type of household is too limited for us to analyze once we restrict the sample to 15–17 year olds at the baseline. For example, less than 2 and 4% of our sample in NLSY79 and NLSY97 are composed of single father-head family (see more details in Table 2). For this reason, we chose to examine how family structure relates to the risk of teen parenthood, particularly focusing on single mother-head families. In our regression, the reference group (i.e., omitted category) of family structure is living with both biological parents. We also control for race and include Hispanic and Black indicator variables in our model, leaving all other races (predominately Whites) as the reference group.

In terms of family income, Both NLSY79 and NLSY97 have contain repeated measures of family income. Beginning 1979, NLSY79 annually collected ‘Total Net Family Income’, measuring sources of family members’ income in the household. NLSY97 also annually asks gross family income. We used the family income information recorded at the first year of survey, 1979 for NLSY79 and 1997 for NLSY97 and they are adjusted in 2017 dollars. Details on variables for our analyses are described in Table 1. Also, standard summary statistics are included in Table 2.

Table 1 Descriptions of variables
Full size table
Table 2 Summary statistics
Full size table

Also, we constructed a cohabitation variable to see changes of marriage and cohabitation among teenagers. To do that, we built a series of living arrangement variables using the NLSY’s household relationship records. In every year of the NLSY79 panel, household members are enumerated and categorized into 66 different relationships relative to the youth respondent, such as respondent him(her)self, married spouse, unmarried partner, father or mother (both biological and step), etc. (For more information, please find codebook of NLSY79 household record: http://nlsinfo.org/content/cohorts/nlsy79/topical-guide/household/household-composition). Similarly, in every year of the NLSY97 panel, the household roster used 87 categorizations of relationships to the respondent for every person living in the household. (For more information please find codebook of NLSY97 household record: https://www.nlsinfo.org/content/cohorts/nlsy97/topical-guide/household/household-composition/page/0/1). Even though NLSY97 panel has more detailed categories of relationship among household members, we found that household relationships in both panel can be collapsed into a few broader categories in the same manner.

Based on household relationship data, we built the same yearly living arrangements of respondents for members of both cohort using the following categories: living with married spouse, unmarried partners, biological mother (and father), step mother (and father), grandparents, brothers and sisters, older relatives (aunts or uncles), and other relatives. This yearly living arrangement data allows us to investigate cohabitation relationship histories using comparable data for respondents in both the NLSY79 and NLSY97 panels.

Data Analyses

This study examines the relationships between teens’ socio-economic backgrounds and their status as teen parents, using logit estimation with marginal effects of an independent variable, holding other variables at their observed values. We also examine the absolute changes in educational attainment and labor earnings between the NLSY79 and NLSY97 cohorts. Then, we examine the relative changes of these outcomes between teen parents and non-teen parents (i.e., those who did not become parents in their teenage years). To examine absolute and relative changes, we first estimate changes in educational attainment and labor earnings among individuals who were childless in their teen years between the NLSY79 and NLSY97 cohort. The same comparisons are then made for teen parents in the two cohorts.

In the analysis of the educational attainment and labor earnings of teen and non-teen parents, the selection into teen parenthood is the focal concern. Confounding factors, such as personal and family backgrounds, may influence the likelihood of becoming teen parenthood, as well as educational attainment and labor earnings. To address this selection issue, we use the difference-in-differences approach in our analyses (Angrist and Pischke 2008). The difference-in-difference estimates compare the differences in mean outcomes between NLSY79 and NLSY97 cohorts for teen parents, as well as for non-teen parents. It allows us to identify the effects of becoming teen parents on their educational attainment and labor earnings, so long as both teen and non-teen parents are subject to the same time trends over the years of the NLSY79 and NLSY97 panels. In our analysis, we assume that the pre-existing differences between teen parents and non-teen parents were similar over time. Based on this assumption, the difference-in-difference estimates can control confounding unobserved time-constant factors, such as trends among the teen and non-teen parents’ personal and family backgrounds, that may influence both the likelihood of becoming a teen parent and later educational and employment outcomes.

Although we are using the difference-in-difference method to limit time-constant confounding factors, we acknowledge that unobserved time-varying confounding factors could still bias our estimates. For example, the PRWORA provisions enacted in 1996 may discourage teen parenthood in the NLSY97 cohort. If this is the case, the difference-in-difference estimates would be biased upwards.

Findings

Changes in Parenting Patterns

Cumulative rates of teenage paternity and maternity for all teenage parents are shown in the top panel of Fig. 1. The cumulative rates of teenage parenting for unmarried teenagers are shown in the lower section of Fig. 1. First, we find that teen fatherhood is more common in the NLSY97 cohort (about 8.8% at age 19) compared to the NLSY79 cohort (about 7% at age 19). The differences are statistically significant throughout the teen years, as indicated along the horizontal axis. Therefore, a higher proportion of male teenagers reported having children in the NLSY97 cohort, compared to male teenagers in the NLSY79 cohort. For female teenagers, the two cohorts show very similar cumulative rates of parenting over all ages. In both cohorts, female teenagers were much more likely than male teenagers to report becoming parents.

Fig. 1
figure 1

Cumulative rate of early parenthood by age (N = 5016 males and 4805 females). Notes All data are weighted by the customized longitudinal weights for the two cohorts. Men and women in our analysis sample from NLSY79 and NLSY97 were aged 15–17 at the first year of survey (N = 5016 for Men and N = 4805 for Women). ***p < 0.01; **p < 0.05; *p < 0.1

Full size image

As seen in the lower panel of Fig. 1, we find that the cumulative rate of births to unmarried male teenagers is consistently higher in the 1997 cohort than in the 1979 cohort. A similar pattern holds for unmarried female teenagers, but the difference in the cumulative birth rates becomes larger at older ages. Using the cumulative rate at age 19, we also calculate the non-marital birth rate among teen mothers, dividing all unmarried teen mothers by all teen mothers. The same calculation method was done for teen fathers. In the 1979 cohort, 77.2% of births to male teenagers and 66.5% of births to female teenagers were non-marital. These percentages increased to 84.0% for teen fathers and 77.2% for teen mothers in the NLSY97 cohort. This increase suggests that our data accurately reflect national trends regarding birthrates outside of marriage, which have increased in the United States (Guzzo and Furstenberg 2007; Smith et al. 1996; Ventura and Bachrach 2000; Wu 2008).

One possible issue in our analysis is the possibility of under-reporting of teen parenting during the survey. In particular, the issue of under-reporting of teenage paternity is longstanding and has been discussed in previous literature (Cherlin et al. 1983; Joyner et al. 2012; Stykes et al. 2013). Some scholars suggest that under-reporting of teenage paternity is due to the lack of representation of incarcerated fathers in household surveys (Marsiglio et al. 2000; Hernandez and Brandon 2002). However, under-reporting due to incarceration is likely to be less problematic in the NLSY79 and NLSY97 data, because both surveys consist of representative samples of non-institutionalized youth who are followed annually. Even if fathers are at some point incarcerated and are absent from a number of NLSY panels, retrospective questions on the ages of their children allow us to capture early parenting better than simple cross-sectional data. Joyner et al. (2012) found that the NLSY teenage paternity data was superior to that found in the National Survey of Family Growth (NSFG). The researchers attributed the difference to the fact that NSFG respondents’ first opportunity to report having a child might occur up to a decade after the birth. In contrast, NLSY respondents were queried about children and their birthdates in all survey years. Both of the NLSY panels are nationally representative and have been used extensively to study young parents (Geronimus and Korenman 1993; Guzzo and Furstenberg 2007; Hynes et al. 2008; Manlove et al. 2013). Additionally, between the two birth cohorts, there was serious improvements in genetic testing and paternity establishment rates. Better knowledge of paternity is likely to have reduced under-reporting in the 1997 birth cohort. Despite these possible issues, the NLSY are arguably the best available data on teenage paternity.

Changes in Marriage and Cohabitation

Figures 2 shows a pronounced shift from marriage to non-marital cohabitation between the two NLSY panels. In Fig. 2, marital and non-marital partners are not restricted to the biological parents of the child. The NLSY data do not provide this level of detail. Nonetheless, we can clearly see the shift from marriage to non-marital cohabitation that has been previously described in the literature (Hynes et al. 2008; Ventura and Bachrach 2000; Wu 2008). Also, because the NLSY follows individuals for many years, we are additionally able to identify shifts in cohabitation before and after the birth of children. In Fig. 2, we displayed the percentage of teen parents living with married or unmarried partners, relative to the birth year of their first child. We create the relative year to the birth of the first child by calculating the differences between the birth of the first child and the interview date of survey. For example, information at the relative year zero indicates what percentage of teen fathers/mothers lived with married/unmarried partners or not, at the time of child birth. Figure 2 shows our finding that a very small percentage (less than 10%) of teen mothers and fathers in both cohorts lived with a partner a year before the birth of the first child, irrespective of marital status. In the NLSY79 birth cohorts, during the pregnancy year and afterwards, there were sharp increases in rates of marital cohabitation seen among both teenage fathers and mothers. These increases were much more subdued in the NLSY97 birth cohorts. The magnitudes of these differences are striking, given that only 18 years elapsed between the two cohorts. Cohabitation with a non-marital partner among teenage fathers and mothers was much more common in the NLSY97 cohort than in the NLSY79 cohort. Non-marital teenage cohabitation, 2 years after the birth of the first child, rose roughly 30% points between the two cohorts. Across the two cohorts, the proportion of (married or unmarried) parents living together is roughly the same, about 50–60%, but there has been a sharp shift from marriage to non-marital cohabitation between the two cohorts. We also find that teen mothers remained more likely than teen fathers to live with their biological children.

Fig. 2
figure 2

Percentage of teen parents living with married and unmarried partners (N = 5016 males and 4805 females). Notes Data are weighted by yearly sample weights. These weights provided us with information to reflect U.S. population as a whole. Relative year to the birth of first child is the year difference between the birth of first child and the time of survey in the data. ***p < 0.01; **p < 0.05; *p < 0.1

Full size image

Predicting for Teen Pregnancy

In Table 3, we examine how family structure relates to the risk of teen parenthood. Previous research has shown that adolescents who live in non-traditional families without two married parents have higher chances of becoming teen parents because of the structural disadvantages associated with single parent and step-parent families (Bumpass and Lu 2000; Cherlin 2010; Teachman 2003; Wu and Martinson 1993). To see how the influence of family structure on teen parenthood has changed between NLSY79 and NLSY97, we created an interaction term between family structure and the NLSY97 cohort indicator variable. This interaction term allows us to investigate whether or not family structure has a stronger or weaker effect on the probability of becoming a teen parent over time. This question is salient as single parent households and other non-traditional families became increasingly common over our study period.

Table 3 Prediction for teen parenthood, with interaction (sample size = 5016 men and 4805 women)
Full size table

Table 3 shows that teen fatherhood is more common in the NLSY97 cohort than the NLSY79 cohort. The estimated coefficient of the NLSY97 dummy indicates that the NLSY97 cohort has more teen fathers by 7.9% points compared to the NLSY79 cohort, controlling for family structure, ethnic background, teens’ parents’ level of education, and family income. The difference between cohorts is noticeable: a higher proportion of male teenagers reported having children in the more recent cohort, compared to teenage males born in the NLSY79 cohort. It also appears that teen motherhood is less common in the NLSY97 cohort and this difference is statistically significant.

Table 3 reveals an apparent anomaly in our findings. Between the two cohorts, rates of teenage paternity increased while rates of teenage maternity decreased. One possible explanation for this is that teenage girls are more likely to partner with young men closer to their own ages in the NLSY 1997 panel. Age differences between teen mothers and their older partners was a topic of concern among some researchers and policymakers (Landry and Forrest 1995; Males and Chew 1996; Taylor et al. 1999) in the 1980s and 1990s. Concerns about age differences have included issues of meaningful consent in sexual relationships and behavioral problems of teen mothers (Boyer and Fine 1992; Larson et al. 1996; Lamb et al. 1986; Moore et al. 1989). However, in spite of the common assumption that adult men play a significant role in teen pregnancy, studies found that only a small percentage of births to teen mothers involved older partners (Lindberg et al. 1997). It is also worth noting that despite the concerns raised, older male partners are normative in American society, regardless of mother’s age, and older fathers are likely to have more economic stability than teen fathers (Hardy et al. 1989).

Another possibility is that more male teenagers were partnering with somewhat older women in the NLSY97. The popularization of older, female “cougars” in television and the popular press may make it more acceptable for younger males to partner with older females. Unfortunately, data on the ages of partners are not collected in the NLSY panels, so we cannot verify either of these possibilities.

A third possible explanation for our finding is that young men may more accurately report their paternity in the NLSY97. This change might occur if the stigma associated with early or non-marital parenting declined over time. And finally, it may be that young fathers in the later cohort are better informed about their paternity, a likely outcome of greater child support enforcement. We suspect that all of these factors are at play in explaining the increased rate of teenage paternity and a constant rate of teenage maternity.

We should also stress here that the panel nature of the NLSY surveys in both panels does allow us to capture the birthdates of children even many years after they are born (in subsequent surveys) and hence to obtain more precise measures of adolescent paternity. The issue of under-reporting of teenage paternity is longstanding and has been discussed in previous literature (Cherlin et al. 1983; Garfinkel et al. 1998; Joyner et al. 2012; Sorensen 1997; Stykes et al. 2013). Some scholars suggest that under-reporting of teenage paternity is due to the lack of representation of incarcerated fathers in household surveys (Marsiglio et al. 2000; Hernandez and Brandon 2002). However, under-reporting due to incarceration is likely to be less problematic in the NLSY79 and NLSY97 data, because both surveys consist of representative samples of non-institutionalized youth who are followed annually. Even if fathers are at some point incarcerated and are absent from a number of NLSY panels, retrospective questions on the ages of their children allow us to capture early parenting better than simple cross-sectional data. Joyner et al. (2012) found that the NLSY teenage paternity data was superior to that found in the National Survey of Family Growth (NSFG). The researchers attributed the difference to the fact that NSFG respondents’ first opportunity to report having a child might occur up to a decade after the birth. In contrast, NLSY respondents were queried about children and their birthdates in all survey years. Also, we use the reporting of children in all survey years to retrospectively calculate the parent’s age at first birth and thus generate the most accurate possible teenage paternity rate given the data available. Although we acknowledge that the NLSY data may not be without bias, we do apply comprehensive coding to retrospectively capture late reports of births to obtain the best available estimates.

Table 3 also shows that the influence of family structure on teenage motherhood is greater in the NLSY97 cohort than in the previous cohort. For both NLSY79 and NLSY97 cohorts, living with a single mother and other family members predicts higher risk of teen motherhood compared to those living with both biological parents. However, this risk was much greater for the NLSY97 cohort. The risk of becoming a teenage mother increased by 11% points for female teenagers living with a single motherhood family across the two cohorts. There were no significant patterns of family structure risk for male teenagers across the two cohorts. This contrast suggests that disadvantaged family structure has become more strongly associated with risk of teen motherhood, but not of teen fatherhood.

Educational Outcomes by Age 29

We also investigate whether the intervening period between the two cohorts mitigated some of the educational deficits of teen fathers and mothers. First, we examine the absolute changes of educational outcomes, comparing teen parents between NLSY79 and NLSY97. Second, we examine the relative changes of educational outcomes, comparing changes of teen parents between NLSY79 and NLSY97 to changes between those who did not become teen parents.

The estimated coefficient of the NLSY97 dummy in Table 4 indicates that teen parents made absolute progress in their educational attainments over the years, in terms of earning a college degree by age 29. The progress of teen mothers is particularly impressive. Teen mothers’ who completed a college degree by age 29 in the NLSY97 cohort is higher by 17.3% points, compared with teen mothers in the NLSY79 cohort (statistically significant at p < 0.01).

Table 4 Predicted absolute and relative change in education attainment and labor earnings (sample size = 5016 men and 4805 women)
Full size table

Despite this progress, the educational attainment of teen parents by age 29 has regressed compared to individuals who did not parent in their teen years. In Table 4, the estimated coefficient of the difference-in-difference indicator (the interaction term between teen fathers and NLSY97) indicates that the difference in earning a college degree for teen fathers versus male respondents who were childless in their teens was lower by 4.4% points in the NLSY97 compared to the NLSY79. While this pattern is similar for females, in neither case do these differences reach traditional levels of statistical significance.

Labor Earnings at Age 29

We turn to our attention to teen parents’ labor earnings at age 29, though this type of data can often be limited. Young men and women who leave school early and enter the labor force are likely to out-earn their counterparts who remain in school, but only for a few years in their early 20s. As those counterparts complete high school or college and enter the labor force, they are likely to increase their earnings quickly and later exceed those with less education (Becker 1981; Card 1995, 1999, 2001; Goldin and Katz 2009; Pirog-Good 1996). Thus, more educated groups in our analysis may show faster earnings growth in the later years of our study period.

In terms of absolute change in earnings, teen parents in the NLSY97 cohort are better off than those of NLSY79 cohort. The estimated coefficient of NLSY97 in Table 3 indicates that the earnings of NLSY97 teen fathers at age 29 are $9972 per year higher than those of NLSY79 teen fathers at age 29 (in constant 2007 dollars). The earnings of teen mothers in NLSY97 cohort are also higher than those of the NLSY79 cohort, by $8565. Both coefficients are statistically significant at p < 0.01.

However, even though teen parents’ labor earnings improved over the years, they still earned less than their peers who delayed parenting. In Table 3, dummy variables for teen fathers indicate that teen fathers earned $2924 per year less than other male respondents in the NLSY97 cohort. The difference is much greater for teen mothers, by $4141 per year. Findings for both groups are statistically significant. This result implies that early parenthood has suppressed women’s earnings more than men’s earnings, at least in the early stages of their careers. The reduced earnings of teen mothers in their early 20s may be explained by their role as primary caregivers for their children, as well as the number of children teen mothers has during teenage years (Beutel 2000; Boden et al. 2008; Davis 2002; de Fátima Rato Padin et al. 2009; Furstenberg 1976; Hoffman et al. 1993; Hofferth et al. 2001; McDermott and Graham 2005; Teti and Lamb 1989; Thrane and Chen 2012). Also, our finding above implies that teen mothers became more likely to obtain a college degree and may be spending more time in school rather than working.

Although the signs of the difference-in-difference indicator (i.e., the interaction terms between the teen parent and the NLSY97 variables) suggest that the earning gaps at age 29 have grown between those who become parents in their teenage years and those who delayed parenting, these coefficients do not meet traditional thresholds for statistical significance.

Discussion and Conclusion

The two decades from 1980 to 2000 saw many changes in social phenomena including fertility rates, marriage rates, divorce rates, non-marital birthrates, cohabitation rates, and educational attainment (Ellwood and Jencks 2004; Isen and Stevenson 2011; Musick et al. 2012; Smock and Greenland 2010). Furthermore, employment-oriented welfare reform and more coercive and coordinated child support enforcement altered incentives and consequences for becoming teenage mothers and fathers (Huang and Han 2012; Nepomnyaschy and Garfinkel 2010; Plotnick et al. 2007). These rapid changes motivated our study of successive generations of teen parents in recent decades using matched panel data cohorts in NLSY79 and NLSY97 to assess whether changes had occurred in teen parenting outcomes.

Findings from this study have meaningful implications for public policies and programs related to teenage fathers and mothers and their children. First, the reported birth rates for teenage motherhood remained stable across the two cohorts, but the reported rates of teenage fatherhood increased. At first glance, these two findings seem somewhat contradictory. This finding can be explained by a combination of reasons.

First, teenage males may be more likely to partner with older women in the NLSY 1997 panel than in prior decades. Recently, the popular press and television has been popularizing the partnering of older women with younger men as exemplified by shows like The Cougar, Extreme Cougar Wives, Cougar Town, dating websites like CougarLife.com and a long series of articles on female cougars in the Huffington Post. From a more academic perspective, partnering between young males and older females is less common than is partnering between young females and older males. However, when it does occur, the age gap tends to be larger. According to the 2002 National Survey of Family Growth (NSFG 2002), more than one-quarter of teen males whose first sexual intercourse occurred before age 16 partnered with females who were three or more years females (Manlove et al. 2006). Even though data limitation in the NLSY surveys prevent further researches to calculate age difference between teen parents, younger males partnering with older female could play a certain role in the increased rate of teenage fatherhood.

Second, it is also possible that teenage girls are selecting male partners who are closer to their own ages. While Lindberg et al. (1997) found that only a small percentage of births to teen mothers involved older partners, any further shift towards men still in their teens could also contribute to our overall finding.

Third, the increase in teenage paternity could also be partially attributed to the overall increase in states’ paternity establishment rates over the relevant years. Researchers have found that advances in genetic testing, voluntary in-hospital paternity establishment mandated by Congress, and more coordinated child support enforcement have been major reasons of increase in paternity establishment (Miller and Garfinkel 1999; Sorensen and Hill 2004). In the 1995 Annual Report to Congress, the Office of Child Support Enforcement (OCSE) noted a 2658% increase in recorded paternities between 1976 and 1995 (OCSE 1995). Also, in 1998, the Child Support Performance Incentive Act (CSPIA) modernized the child support enforcement incentive, including paternity establishment rates as part of the formula for distributing funding to states (Solomon-Fears 2013). Previous research indicates that CSPIA improved child support outcomes (Gerrish 2017; Huang and Edwards 2009; Pirog and Gerrish 2015), and changes in the child support enforcement could also contribute to increase paternity establishment and reported rates of teen fatherhood.

Finally, the stigma associated with nonmarital parenting has diminished between the NLSY 1979 and the NLSY 1997. Both better knowledge of the parenthood via paternity establish procedures as well as greater acceptance of unmarried parents as normative could lead to better reporting of paternity and also partially explain our findings.

We also find that the proportions of both teenage fathers and mothers who were unmarried at the time of their children’s birth increased significantly between the two cohorts, reflecting the broader increasing pattern of non-marital births. Particularly, our examination of the marital and non-marital cohabitation of teenage mothers and fathers with their partners suggests that there was a significant shift from marital cohabitation to non-marital cohabitation, although the overall partner cohabitation rates were steady across the cohorts (Chandra et al. 2005; Martinez et al. 2012). Such trends imply that the increased number of children who are born to teenage fathers may be less likely to have sufficient financial support from their fathers, because teenage fathers are less likely to work or pay child support (Cancian et al. 2011, 2013; Nepomnyaschy and Garfinkel 2010; Pirog-Good and Good 1995). Also, an increase in non-marital cohabitation suggests increasingly unstable family environments for children of teen parents, because non-marital cohabitation is less likely to provide the full legal protection to children that marital cohabitation provides (Bartfeld 2000; Cancian et al. 2011, 2013).

Our additional findings on changes in educational attainment and labor earnings of teenage parenthood provide more insights into the changing socio-economic circumstances of teenage parents. The absolute rate of college degree attainment for teen fathers and mothers in the NLSY97 cohort has increased compared to the rate in the NLSY79 cohort. However, the real change is close to nil compared with men who were not teen fathers. The average inflation-adjusted income of teen fathers and mothers at age 29 was higher (both statistically significant) in the NLSY97 cohort compared with the NLSY79 cohort. However, accounting for the progress made by men and women who delayed parenthood between cohorts, the labor earnings of those who parented in their teens regressed. In many critical respects, this analysis paints a picture of increasingly disadvantaged socio-economic circumstances for teenage parents, which is consistent with previous findings (Barber et al. 1999; Becker 1981; Cancian et al. 2011, 2013; Fletcher and Wolfe 2012; Hotz et al. 2005; Manning et al 2014).

In sum, it seems that more coercive and coordinated child support enforcement did not discourage teenage fatherhood (see, for example, Cancian et al. 2011, 2013; Nepomnyaschy and Garfinkel 2010), though it may have encouraged more mothers to inform the fathers of their paternity. Increasing paternity establishment rates undoubtedly account for part of the increase in reports of paternity by young men between the two NLSY panels. The work-oriented welfare reform passed in the latter half of the 1990s seemed to improve labor market and education outcomes of teenage fathers and mothers in absolute terms, but these outcomes regressed compared with those of peers who were not teen parents. The same pattern was observed with the incomes of young adults who did and did not become parents in their teen years. Coupled with the insecurity of increased non-marital cohabitation, these patterns point to an increase in relative disadvantages for children born to teenagers.

As a direct and immediate aid to children of teenage fathers and mothers, government funded Head-Start programs and free pre-K programs may support children’s early development and provide more time and resources for teenage mothers to resume their study or work. Additionally, the increased absolute education levels of teenage fathers and mothers suggest that those at risk of becoming teen parents are more likely to remain in high school than in previous years, which indicates that schools have a greater opportunity to design and implement programs targeted toward those at greatest risk of unintended pregnancy and unprepared parenting. For those who are already teen parents, there are opportunities to create better programs to facilitate continued educational attainment. To date, many of the programs targeted at teen parents have been at the high school level, but the increasing rates of college degree obtainment suggest that institutions of higher education should also consider programmatic improvements to facilitate enrollment by students who became parents in their teens.