Abstract

The present study employs discrete-time hazard regression models to investigate the relationship between student loan debt and the probability of transitioning to either marital or nonmarital first childbirth using the 1997 National Longitudinal Survey of Youth (NLSY97). Accounting for nonrandom selection into student loans using propensity scores, our study reveals that the effect of student loan debt on the transition to motherhood differs among white, black, and Hispanic women. Hispanic women holding student loans experience significant declines in the probability of transitioning to both marital and nonmarital motherhood, whereas black women with student loans are significantly more likely to transition to any first childbirth. Indebted white women experience only a decrease in the probability of a marital first birth. The results from this study suggest that student loans will likely play a key role in shaping future demographic patterns and behaviors.

Introduction

A growing body of literature has consistently demonstrated that greater levels of education are strongly associated with childbearing behaviors, such as delayed ages at first childbirth, reduced unintended births, and increased likelihood of birth within the context of marriage in the United States (Hayford and Guzzo 2016; Lamidi 2016; Manning 2013; Smock and Greenland 2010). Concurrently, although racial/ethnic disparities in childbearing have converged since the 1990s, research has shown that racial/ethnic differences in fertility outcomes persist within educational groups even after accounting for sociodemographic characteristics (for a review, see Smock and Greenland 2010). For example, although college education is associated with reduced nonmarital childbearing, a greater proportion of non-Hispanic black (hereafter “black”) and Hispanic college graduates have unwed births than their non-Hispanic white (hereafter “white”) peers (Keels 2014). Indeed, Keels (2014) showed that the greatest racial differences in overall childbearing are in fact among women with some college education or more, suggesting that further investigation of racial/ethnic differences in fertility among highly educated women may provide particularly important insights into the variability in fertility behaviors within social class groups.

Women’s educational attainment increased substantially over the past 50 years. Young women from all racial/ethnic backgrounds outpace their male peers in terms of college enrollment and graduation (Aud et al. 2010). Coinciding with women’s gains in educational attainment, however, is the emergence of student loan debt. A report conducted by the American Association of University Women (AAUW) found that female graduates have larger outstanding student loan balances than male graduates and face greater difficulty in paying down student debt (Corbett and Hill 2012). Difficulty servicing student loan obligations is particularly pronounced among students of color, who are highly overrepresented among borrowers, especially black students. In 2012, 86 % of black students and 73 % of Latino students earning a bachelor’s degree graduated with student loan debt compared with 68 % of whites (Baum et al. 2015). Therefore, the growing student loan burden among women attending college, particularly minority females, is an increasingly salient topic in terms of family formation. The current study examines whether racial/ethnic disparities in student loans contribute to the racial/ethnic gap in childbearing.

Student Loan Debt and Family Formation

Ratio of Student Debt to Income

The empirical literature examining the relationship between student loans and family formation show that they share a variable relationship with individual outcomes and behaviors, perhaps due to the financial credit constraints individuals experience early in their careers. Our study thus includes a measure of student loan debt as a proportion to annual incomes. Respondents report any income received in the past year, including full-time (30 hours per week or more), part-time, temporary, or seasonal work. The student-debt-to-income (SDTI) variable is constructed as the ratio of annual student debt owed to total annual income. High SDTI values indicate greater levels of student debt relative to total income, and thus greater relative loan burden. SDTI ratios vary based on the amount of debt repaid each year and annual changes in income. Among respondents, years containing missing annual income are imputed with the respondent’s average income from age 18 until year of exit in the models. We impute respondents missing all data on annual income (3 %) using mi impute pmm in Stata to retain sample size.

Race and Ethnicity Interactions

We include binary variables representing black and Hispanic race/ethnicity as main effects and separately interact them with the binary indicator of ever receiving student loans to test whether the association between student loans and the probability of transitioning to marital and nonmarital birth varies by race/ethnicity.

Control Variables

Additional factors affecting first birth include time-invariant indicators for immigrant status, postsecondary degree (i.e., bachelor’s and postgraduate), respondent’s region (at baseline in 1997), the log value of parental household net worth (in 1997), logged parental income (in 1997), intact family (biological or adoptive parents married in 1997), biological mother’s and father’s highest education (in 1997), and mother’s age at first childbirth. Time-varying indicators include educational enrollment (1 = yes; 0 = no); respondent’s employment status, coded 1 if she reports having worked 30 or more weeks and at least 30 hours per week in the previous year, and 0 otherwise; and respondent’s logged annual income. We impute missing information using multiple imputation in Stata for paternal education (7.8 %), parental net worth and income (15.6 %), and living with both parents in 1997 (8.4 %).

Plan of Analysis

Because of the separate processes underlying marital and nonmarital first birth, we separately evaluate the dynamic relationship between student loan debt and transitioning to first birth using a discrete-time hazard framework, with discrete intervals measured in person-years. More specifically, we use a piecewise semiparametric approach that runs a series of logistic regressions to separately estimate baseline hazards of each individual covariate, time-varying and time-invariant, on the probability of marital and nonmarital first birth separately for each duration interval (Allison 2010). Baseline hazard function estimates are generated using maximum likelihood. We use the Huber-White (Huber 1967; White 1980) method for clustering standard errors within individuals.

The growing body of literature on student debt has frequently acknowledged the limitation of empirical inferences due to nonrandom assignment of student loans (e.g., Bozick and Estacion 2014; Dowd 2008; Gicheva 2016; Grinstein-Weiss et al. 2016). Much of the economic literature has used an instrumental variable approach to address these limitations, whereas others (e.g., Dowd 2008; Titus 2007) have strongly urged researchers to use a propensity score framework to account for the endogeneity of student debt and individual outcomes. Thus, in absence of a strong instrumental variable and an experimental design to account for selection bias into the treatment, and in concordance with previous research, we further isolate the effect of student loans utilizing a propensity score approach. This approach makes two assumptions: (1) no additional confounders exist between treated and untreated subjects, and (2) every subject has a nonzero probability of either treatment (for a review, see Austin 2011). Defined by Rosenbaum and Rubin (1983), propensity scores (PS) represent the probability of the treatment assignment conditional on observed baseline covariates. PS represent a balancing score, where the distribution of measured baseline covariates is similar between treated and untreated subjects. In the case of our study, PS are the probability of ever taking student loans (treatment), conditional on a set of 18 baseline covariates measured prior to treatment (see Table S1, Online Resource 1). Adolescent characteristics include working as a teenager (0 = no; 1 = yes), immigrant status (0 = no; 1 = yes), rural area 1997 (0 = no; 1 = yes), living with both parents in 1997 (0 = no; 1 = yes), mother or father with less than a high school education (0 = no; 1 = yes), and indicators for whether parental net worth and parental income were less than the median (0 = no; 1 = yes). Measures related to college attendance capture type of college institution (i.e., private nonprofit, for profit) (0 = no; 1 = yes), age at first college enrollment, receipt of financial aid (i.e., government financial aid, work study, grant) (0 = no; 1 = yes), financial aid provided by family members (0 = no; 1 = yes), college preparatory courses (0 = no; 1 = yes), logged average college tuition, indicator of ever enrolling part-time (0 = no; 1 = yes), and an indicator for ever attending remedial math or English courses (0 = no; 1 = yes). Students taking remedial classes take longer, on average, to graduate from college than their peers (Barry and Dannenberg 2016; Complete College America 2015), extending the number of years students are at risk of accumulating student debt and simultaneously delaying transitioning to traditional adulthood roles (e.g., adult career, marriage, parenthood). Last, PS models control for race/ethnicity.

We estimate PS and subsequent weights using inverse probability weighted regression adjustment (ipwra) models with teffects in Stata 14. We use this approach because of its doubly robust property: unbiased estimates of treatment effects can be obtained if either the outcome model or the propensity model (but not both) is misspecified (Wooldridge 2007). We compared estimated PS of treated and untreated women to ensure balanced covariates, finding a PS range of 0.2 to 0.7 (M = 0.48) (Online Resource 1, Fig. S1). Results from our treatment model suggest that race/ethnicity and parental wealth are strongly associated with selection into student loans, closely followed by postsecondary experiences, such as attending a for-profit institution. These results suggest that these particular sociodemographic characteristics are the strongest predictors of taking educational loans, consistent with previous literature (e.g., Addo et al. 2016).

In our study, average treatment effects (ATE) represent the probability of first birth among women who are untreated if these women had received the treatment. We estimate ATEs using Cox hazard regressions adjusting for their propensity of taking student loans. ATEs are treatment effects at the individual level; because we wish to measure the effect of student loans on the probability of first birth at the population level, we use marginal treatment effects (MTE)—the difference in outcomes between treated and untreated populations—to weight final analyses (for more details on ATE vs. MTE, see Austin 2013). Upcoming Figs. 1 and 2 plot the estimated Nelson-Aalen cumulative hazard of marital and nonmarital first birth, respectively, by race/ethnicity and student loan status weighted by MTE. Results from the selection equation are provided in Table S1, Online Resource 1. The results from this table show that the probability of marital and nonmarital first birth among untreated populations is estimated to decrease by 0.03 and 0.07, respectively, had these women taken student loans.

The final data are structured in person-year format, with each woman contributing one observation for each survey year without birth after leaving their higher education institution. After a respondent reports a birth, that data point contributes to the model, and she is censored from the model thereafter. Because of differences in exposure to risk of motherhood—risk of marital birth is conditional on marriage—analyses examining the hazard of first birth are analyzed separately for marital and nonmarital births. Models are adjusted for marginal treatment effects. Unadjusted results are provided in Table S2, Online Resource 1. We estimate alternate models by comparing marital with nonmarital births directly in a competing risk framework using combined data (models not shown) with similar findings. We estimate additional sensitivity analyses examining potential biases stemming from the conditional outcome of marriage on estimates of marital births using a sequential hazard approach as described by Wu and Martin (2009); these analyses are provided in Tables S3 and S4 in Online Resource 1. Overall, the results suggest that our presented findings on marital births may be sensitive to the indirect effect of differential timing of first marriage in certain race/ethnic groups. The results from these sensitivity analyses should, however, be interpreted with caution because of the relatively short period between marriage formation and marital birth observed among our sample. The mean number of person-years contributed to the model is 5.2 years.

Results

Table 1 presents the descriptive characteristics of women by birth outcome. Results show that a smaller proportion of women with nonmarital births took student loans (46 %), compared with those with marital first births. Relative to married mothers, unmarried mothers borrowed slightly greater amounts—$6,170 compared with$6,142. Average SDTI ratios are similar between women who transition to first birth: 16.1 % among married mothers and 16.6 % among unmarried mothers. Women transitioning to marital first births reported higher average annual incomes than women with nonmarital births: $25,461 compared with$19,675.

Black women account for less than 8 % of marital births and almost 41 % of the nonmarital births, representing the largest share of unmarried mothers. In contrast, Hispanic women account for 18 % of marital births and 22 % of nonmarital births. Women with some college but no degree make up the vast majority of women who reported nonmarital first births (75 %). Last, women reporting nonmarital births are more likely to have been raised in the South, and are less likely to have parents with some college education or more, and are from families with less income and wealth. In contrast, women reporting marital births are more likely to have been raised in homes with both biological parents and are from relatively advantaged families, as indicated by parental education, income, and net worth.

Figures 1 and 2 chart the Nelson-Aalen cumulative hazard of marital and nonmarital first birth by race/ethnicity and student loan receipt, respectively, adjusting for the aggregate propensity to take student loans (the MTE). Figure 1 suggests that women without student loans are more likely to transition to marital first birth, with the exception of black women, who show a slightly greater risk of marital first birth with loans. Conversely, Fig. 2 indicates that women with student debt have a greater risk of nonmarital birth relative to their debt-free counterparts. Black and Hispanic women exhibit the greatest risk of nonmarital first birth compared with their white counterparts, net of student debt.

Table 2 displays the results from the hazard analyses adjusted for the propensity to take student loans. Model A displays the hazard ratios for the relationship among taking student loans, student loan balance, SDTI, and race/ethnicity, controlling for immigrant status, college degree, and current enrollment status on the probability of transitioning to marital and nonmarital first births. Model B includes interactions between race/ethnicity and ever taking student loans. Model C displays the results from the full model, which controls for sociodemographic characteristics.

The results in Model A show that taking student loans is negatively associated with the transition to marital first birth but positively associated with nonmarital first birth, presenting early support for Hypothesis 3. The risk of transitioning to marital first birth is 2 % [(1 – 0.98) × 100] lower among young women who took student loans compared with their peers without student debt, whereas the risk of nonmarital first birth is 1 % greater for women with student loans relative to debt-free women. Remaining measures of educational debt show that outstanding student debt balances are significantly negatively associated with the transition to first birth (partial support for Hypothesis 1). Counter to Hypothesis 1, however, the effect of SDTI is not statistically significant. These results suggest that net of adjustments for the propensity to take educational loans, holding student loans likely continues to capture the effect of family socioeconomic resources, whereas outstanding debt obligations may increase the financial costs associated with childbearing net of income.

Corroborating previous findings, the results in Model A indicate that black and Hispanic women are significantly less likely to experience marital first births and more likely to experience nonmarital births relative to their white counterparts. The inclusion of interaction terms in Model B, however, suggests that the relationship between student debt and first birth differs by race/ethnicity. Consistent with Hypothesis 2a, the association between student loan debt and first marital and nonmarital birth is disproportionately greater among Hispanic women. Hispanic women with student loans experience a disproportionately reduced risk of both marital and nonmarital first births (hazard ratio (HR) = 0.70 (0.67 × 1.05)). In comparison, white women with loans experience only a 2 % reduction in the hazard for marital births (HR = 0.98) and experience no significant difference in nonmarital births (HR = 1.05) in Model B. This suggests that white women with debt may have higher nonmarital fertility schedules than Hispanic women with debt. Overall, our results indicate that the overall risk associated with nonmarital first births among Hispanic women observed in Model A would be larger had these women not taken student loans.

The results in Model B show that in contrast to both white and Hispanic women, black women with educational debt experience significant increases in the risk of both marital and nonmarital first births. Specifically, among black women, student debt roughly doubles the risk of marital birth (HR = 2.28 (2.33 × .98)) and increases the risk of nonmarital birth by 22 % (HR = 1.22 (1.16 × 1.05)). These results are largely consistent with Hypothesis 2b, which states that the relationship between loan receipt and first birth—particularly, nonmarital first birth—will be weaker for minority women than for white women. In fact, education debt clearly increased (rather than decreased) the risk of both marital and nonmarital births in this group. Sensitivity analyses interacting debt balances rather than the global indicator of taking student loans showed a similar positive association between student loans and transitioning to a nonmarital first birth among black women (not shown). Overall, these findings suggest that rather than inhibiting the transition to first birth, student loans are positively associated with a pathway to motherhood among black women.

A clearer picture of the interactions between race/ethnicity and student debt on the risk of marital and nonmarital birth observed in Model B are provided in Fig. 3, panels a and b, respectively. Panel a shows that white and Hispanic women without student debt have a greater hazard of marital first births, relative to all other groups, peaking five to six years after exiting postsecondary programs. Hispanic women with student loans exhibit a delayed peak in marital birth hazard rates, approximately seven years after leaving postsecondary programs. Black women without student debt are the least likely to transition to marital first birth, whereas black women with student debt exhibit a greater hazard of marital first births compared with their debt-free counterparts every year postcollege. In contrast, panel b shows that the hazard rates for nonmarital first birth are greater among black women with and without student loans, whereas the hazard rates for nonmarital births among white women with and without student loans fall between the hazards rates of indebted and debt-free Hispanic women.

The results in the final model (Model C) show that net of controls for sociodemographic characteristics, student loans continue to exhibit an independent effect on the risk of transitioning to motherhood, with differential effects by race/ethnicity. Regarding marital first births, taking student loans is significantly associated with a reduced risk of motherhood among white and Hispanic women, relative to whites without loans. Consistent with Model B, the effect is more pronounced among Hispanic women than among their white counterparts. Overall, the magnitude of the negative association between student loans and the transition to motherhood among Hispanics appears stronger for marital first births (HR = 0.40 (0.41 × 0.98)) than nonmarital first births (HR = 0.65 (0.66 × 0.99)) (Hypothesis 3). In contrast, the positive interaction between black women and student loans remains significant for either type of birth. Indeed, the impact of student loans on marital first births becomes stronger in magnitude among black women after we control for sociodemographic characteristics. The remaining measures of student loans indicate that for every one-unit increase in logged outstanding student debt, the risk of transitioning to marital first birth decreases by 1 %.

Regarding nonmarital first births, the effect of student loans remains significantly negative for Hispanic women, and this group continues to experience the greatest reduction in risk of birth from holding educational debt in this model (Hypothesis 2a). Conversely, the risk of a nonmarital first birth remains increased for black women compared with their white peers, and student debt increases this risk further (Hypothesis 2b). Outstanding student debt balances are a significant deterrence to nonmarital first births, such that the risk of transitioning to nonmarital first birth decreases by 5 % for each one-unit increase in logged amount of outstanding student debt. Together, the findings in Model C suggest that the associations between student debt and first birth apparent in Model 2 remain robust after other background and socioeconomic factors are held constant.

Summary and Discussion

Rapid increases in college enrollment among young women from diverse racial/ethnic backgrounds have created the need to further explore possible associations between aspects of college attendance and family formation. One emerging feature of college attendance is the increasing probability of graduating with student loan debt. The consequences of carrying educational debt for family formation, however, are largely unknown. Emerging literature has revealed that student loan debt negatively affects union formation and childbearing and that women are more affected than men (Addo 2014; Bozick and Estacion 2014; Nau et al. 2015). However, little is known about the relationship between student debt and family formation among racial/ethnic groups. The literature has consistently documented racial/ethnic variation in family formation, patterns that diverge among advantaged women (e.g., Keels 2014; Sweeney and Raley 2014). The current study builds on the established literature by providing evidence that student loans play an important and variable role in shaping fertility behavior among racial/ethnic minority women who attend higher education institutions.

Overall, we found a pronounced effect of student loans on timing of first birth among minority women, albeit in opposite directions. The risk of transitioning to motherhood, particularly through a marital birth, was lower for those taking student loans among white and especially Hispanic women. These findings suggest that student loans slightly delay white and Hispanic married women’s abilities to transition to parenthood compared with their counterparts without loans. Indeed, competing risk analyses showed that the negative association between taking student loans and transitioning to first marital and nonmarital births was especially pronounced among Hispanic women, who were more likely to remain child-free (models not shown). This finding is notable in light of studies finding that Hispanic women continue to exhibit higher fertility rates compared with other racial/ethnic groups when education and background characteristics, such as education and nativity, are controlled for (Bogue 2010; Lichter et al. 2012). Further, Hispanic women continue to bear children earlier than their white peers even after the inclusion of extensive controls for sociodemographic characteristics; psychosocial resources; and sexual behavior, knowledge, and attitudes (Guzzo et al. 2014). This finding suggests that Hispanic women who finance their educations potentially pay the greatest price in terms of overall childbearing. If student loans act as a barrier to marriage in a market where women face a shortage of desirable mates, women with debt may be significantly inhibited in their ability to form families until they are more financially established (Cherlin 2004; Furstenberg 2014; Oppenheimer 1988). Alternatively, the increased mobility associated with higher levels of educational attainment, potentially afforded by student loans (e.g., as found among black students by Jackson and Reynolds 2013), may enable this group to postpone or forgo family formation in the absence of desirable mates or desire for marriage. These results suggest that researchers should focus greater attention on the outcomes of Hispanic students seeking higher education.

In contrast to our findings among Hispanic women, our results showed that the risk of transitioning to a first birth increased among black women who financed their education. Indeed, our results indicated that black women with student loans were more likely to transition to marital and nonmarital first births than black women without student debt (e.g., panels a and b of Fig. 3). Further, this positive association between education loans and a marital first birth remained significantly positive in sensitivity analyses that accounted for differential risk due to the timing and duration of first marriage (Table S4, Online Resource 1). These results suggest that student loans potentially help black women achieve the upward mobility necessary for marital births, indicating that selection into marriage may play an important role. In support, research by Min and Taylor (2016) showed that black women with student loans are also more likely to marry than their counterparts without student debt. These findings may be partly due to the disproportionately positive association between student loans and college persistence and completion for black students (Jackson and Reynolds 2013). Thus, the positive association between student loans and motherhood among young black women could reflect their increased odds of college completion, thereby conferring indirect benefits in the realm of family formation, particularly marital births. However, it is important to note the limits of positive associations between student loans and adult achievements. High levels of debt are negatively associated with college completion, and black students are at the greatest risk of defaulting on student loans (Jackson and Reynolds 2013). Moreover, the black-white disparity in student loan amount doubles within four years following college graduation, such that black graduates owe \$25,000 more on average than their white counterparts (Scott-Clayton and Li 2016). Therefore, future studies may potentially reveal negative associations between particularly large amounts of educational debt and demographic behaviors.

Last, it is important to note the differential effect between the three measures of student loan debt used in our study. Our results showed that the global indicator of taking student loans and outstanding student loan debt obligations exhibited an independent effect on marital first births beyond sociodemographic controls and adjustments for women’s propensity to incur educational debt, whereas SDTI was nonsignificant. On the other hand, debt balances and their corresponding ratio to income remained negatively associated with nonmarital first births, net of controls. Although explaining the causal mechanisms underlying the relationship between holding student loans and family formation behavior is beyond the scope of our study, we believe that it is possible that these different measures of debt captured different aspects of the student loan experience. For example, the global indicator of taking student loans may capture a latent aspect of socioeconomic status or an aspect that is intrinsic to the borrowing experience itself that propensity score adjustments cannot fully account for. Black women who married and subsequently become mothers, for instance, may in fact be the most relatively advantaged of their peers in their ability to successfully navigate higher education institutions and find marriageable partners despite the structural disadvantages that many black Americans commonly face. Additionally, our finding that outstanding loan balances negatively influenced the transition to marital motherhood while the effect of SDTI was null suggests that women delaying births may wish to fully pay off their debt obligations, regardless of the proportion of income their debt consumes, before transitioning to a new phase of adulthood. Prioritizing repayment may also be in response to changing interest rates. Prior to July 2006, interest rates on Stafford student loans were variable, ranging from just over 1 % to 8.25 %, with wide variability across each year (FinAid n.d.). In 2006, however, Stafford loans were replaced with loans of a single fixed rate of 6.8 %, and that rate has steadily decreased in more recent years to 3.76 % for undergraduate loans and 5.31 % for graduate loans in 2016 (Federal Student Aid n.d.). Alternatively, SDTI among married women may be less influential on family formation decisions if their spouse’s earnings offset the wages needed to service loan obligations.

Regarding nonmarital births, the negative association between SDTI and debt balances may indicate a delayed ability to acquire important assets (such as purchasing a home) because of credit constraints, which may subsequently delay childbearing (for a discussion on the relationship between homeownership and timing of fertility, see Mulder 2006a, 2006b). Overall, our study suggests that ever taking student loans, outstanding debt balances, and the proportion of income needed to service debt obligations were not equivalent in their consequences for family formation and that these measures of debt operate differently between racial/ethnic groups. Future research should conduct more in-depth analyses of the various aspects associated with holding student loans, including fixed versus variable interest rates, and their implications for the timing of future life course transitions. A particularly fruitful avenue of research may be to examine the association between student loans and multiple life course events within racial/ethnic groups, which could potentially parse out the observed heterogeneity in response to taking student loans and servicing debt obligations in young adulthood.

Although our study attempted to be comprehensive, it is important to note limitations. We were unable to fully eliminate the potential of unmeasured variables yielding differences in the risk of a first birth. For example, receiving parental assistance or outside guidance in selecting college courses or completing financial aid applications plays an important role in selecting individuals into student debt, which may ultimately influence their subsequent trajectory into parenthood. These processes are an important consideration, particularly with the recent expansion of higher education, given that the pathway to college completion has become increasingly diverse, particularly among the socioeconomically disadvantaged (Goldrick-Rab 2006). Our study did not account for potential differences in postsecondary educational experiences. Class-based differences in college experiences lead to cumulative disadvantages among students from lower socioeconomic positions that shape and alter future opportunities (Aronson 2008). Thus, future studies should examine how socioeconomic disparities shape postsecondary experiences, including selection into student loan debt, and how these dynamic processes influence future adulthood transitions.

In addition to potential limitations of our models, our study was limited by data regarding deferment, forbearance, and default on student loans. Respondents in deferment, forbearance, or default are likely to experience competing factors that may produce different results. However, default rates have been decreasing since the Great Recession (Wright and Serrato 2015), although black and Hispanic graduates remain more likely to default than their white peers (Scott-Clayton and Li 2016). In addition, the effect of student loans on deferment or forbearance was likely to be partly captured by the student debt-to-income ratio, which was generally nonsignificant. A limitation also worth noting is that information on student loan debt was self-reported. If these measures are subject to misreporting, our findings would likely differ.

One final note: research has indicated that the fertility behaviors of those with some college education more closely resemble the fertility of those with a high school education than those who receive a bachelor’s degree or higher (Gibson-Davis and Rackin 2014; Lundberg et al. 2016). Indeed, analyses limited to women earning a bachelor’s degree or higher showed that white graduates with student loans experience a significantly increased risk of nonmarital first births (not shown); however, these findings are limited because of the small sample sizes of highly educated women who had a nonmarital first birth. Future analyses should examine whether the association between student loans and union context of first birth persists across educational groups among a larger sample of women, and potentially with women’s completed fertility schedules (i.e., women aged 50 and older). Although the vast majority of U.S. women become mothers by the maximum ages of the sample included in our study, it is likely that analyses encompassing completed fertility schedules would yield slightly different findings.

In closing, our study illuminates the impact of student loan debt on women’s fertility decisions, revealing that this association depends on race/ethnicity. Our study found that student loans were associated with delayed motherhood concentrated among Hispanic borrowers, whereas black borrowers were relatively undeterred by incurring debt; indeed, loans may have increased the risk of motherhood for black borrowers. Additional research is needed to determine whether student loans present a pathway for independence or further challenges for the upward mobility of women and offspring. Among single mothers especially, managing the costs of childcare in addition to student loan repayments may contribute to decreased intergenerational transfers, aiding in the reproduction of economic and social inequalities. Although women with greater levels of education experience increased employment opportunities and wages, few studies have investigated whether women can leverage advanced degrees to fully offset the costs of raising children in the presence of student loans. The results of this study point to the policy importance of identifying the consequences of student loan debt in order to mitigate the impact of potential financial hardship faced by young adults who are forming, or wish to form, families.

Acknowledgments

A previous version of this paper was presented at the 2016 annual meeting of the Population Association of America. This study was supported in part by the National Science Foundation Research Fellowship Program under Grant No. 2016-1449440. Any opinions, findings, and conclusions or recommendations expressed in this study are those of the authors and do not necessarily reflect the view of the National Science Foundation.

References

2014
).
Debt, cohabitation, and marriage in young adulthood
.
Demography
,
51
,
1677
1701
. 10.1007/s13524-014-0333-6.
2016
).
Financial integration and relationship transitions of young adult cohabiters
.
Journal of Family and Economic Issues
,
38
,
84
99
. 10.1007/s10834-016-9490-7.
Addo, F. R., Houle, J., & Dwyer, R. (
2016
).
Young, black, and (still) in the red: Parental wealth, race, and student loan debt
.
Race and Social Problems
,
8
,
64
76
. 10.1007/s12552-016-9162-0.
Allison, P. D. (
2010
).
Survival analysis using SAS: A practical guide
(2nd ed.).
Cary, NC
:
SAS Institute
.
American Association of University Women (AAUW)
. (
2016
).
The simple truth about the gender pay gap
.
Washington, DC
:
AAUW
Aronson, P. (
2008
).
Breaking barriers or locked out? Class-based perceptions and experiences of postsecondary education
. In J. T. Mortimer (Ed.),
Social class and transition to adulthood: New directions for child and adolescent development
(Jossey-Bass Education Series No. 119, pp.
41
54
).
San Francisco, CA
:
Wiley Subscription Services
.
Aud, S., Fox, M., & KewalRamani, A. (
2010
).
Status and trends in the education of racial and ethnic groups
(NECES Report No. 2010-015).
Washington, DC
:
U.S. Department of Education, National Center for Education Statistics
Austin, P. C. (
2011
).
An introduction to propensity score methods for reducing the effects of confounding in observational studies
.
Multivariate Behavioral Research
,
46
,
399
424
. 10.1080/00273171.2011.568786.
Austin, P. C. (
2013
).
The use of propensity score methods with survival or time-to-event outcomes: Reporting measures of effect similar to those used in randomized experiments
.
Statistics in Medicine
,
33
,
1242
1258
. 10.1002/sim.5984.
Avery, C., & Turner, S. (
2012
).
Student loans: Do college students borrow too much—Or not enough
.
Journal of Economic Perspectives
,
26
(
1
),
165
192
. 10.1257/jep.26.1.165.
Barry, M. N., & Dannenberg, M. (
2016
).
Out of pocket: The high cost of inadequate high schools and high school student achievement on college affordability
(Policy brief).
Washington, DC
:
Education Reform Now
Baum, S., & Johnson, M. (
2015
).
Financing public higher education: The evolution of funding
(Report).
Washington, DC
:
Urban Institute
.
Baum, S., Ma, J., & Payea, K. (
2013
).
Education pays 2013: The benefits of higher education on society
(Trends in Higher Education Series report).
New York, NY
:
College Board
.
Baum, S., Ma, J., Pender, M., & Bell, D. (
2015
).
Trends in student aid 2015
(Trends in Higher Education Series report).
New York, NY
:
College Board
.
Becker, G. S. (
1981
).
A treatise on the family
.
Cambridge, MA
:
Harvard University Press
.
Bogue, D. J. (
2010
).
Contraception, attitude-practice, and fertility differentials among US Hispanic, African-American, and white women
.
Journal of Population Research
,
27
,
275
292
. 10.1007/s12546-011-9052-7.
Bozick, R., & Estacion, A. (
2014
).
Do student loans delay marriage? Debt repayment and family formation in young adulthood
.
Demographic Research
,
30
(
article 69
),
1865
1891
. https://doi.org/10.4054/DemRes.2014.30.69
Cherlin, A. J. (
2004
).
The deinstitutionalization of American marriage
.
Journal of Marriage and Family
,
66
,
848
861
. 10.1111/j.0022-2445.2004.00058.x.
Child Trends
. (
2015
).
Births to unmarried women: Indicators on children and youth
(Data Bank report).
Bowling Green, OH
:
Child Trends
Complete College America
. (
2015
).
Four-year myth: Make college more affordable. Restore the promise of graduating on time
.
Indianapolis, IN
:
Complete College America
Corbett, C., & Hill, C. (
2012
).
Graduating to a pay gap: The earnings of women and men one year after college graduation
.
Washington, DC
:
American Association of University Women
Dowd, A. C. (
2008
).
Dynamic interactions and intersubjectivity: Challenges to causal modeling in studies of student loan debt
.
Review of Educational Research
,
78
,
232
259
. 10.3102/0034654308317252.
Dynarski, S. M. (
2016
).
The dividing line between haves and have-nots in home ownership: Education, not student debt
(Evidence Speaks Reports Vol.
1
, No.
17
).
Washington, DC
:
Brookings Institute
Emmons, W. R., & Noeth, J. B. (
2015
).
Why didn’t higher education protect Hispanic and black wealth?
(In the Balance No. 12).
St. Louis, MO
:
Federal Reserve Bank of St. Louis
Federal Student Aid
. (n.d.).
Interest rates and fees
FinAid
. (n.d.).
Historical interest rates
Fry, R. (
2014
).
The changing profile of student borrowers: Biggest increase in borrowing has been among more affluent students
(Social & Demographic Trends report).
Washington DC
:
Pew Research Center
Furstenberg, F. F. (
2014
).
Fifty years of family change: From consensus to complexity
.
Annals of the American Academy of Political and Social Science
,
654
,
12
30
. 10.1177/0002716214524521.
Gibson-Davis, C., & Rackin, H. (
2014
).
Marriage or carriage? Trends in union context and birth type by education
.
Journal of Marriage and Family
,
76
,
506
519
. 10.1111/jomf.12109.
Gicheva, D. (
2016
).
Student loans or marriage? A look at the highly educated
.
Economics of Education Review
,
53
,
207
216
. 10.1016/j.econedurev.2016.04.006.
Gladeaux, L., & Perna, L. (
2003
).
Borrowers who drop out: A neglected aspect of the student loan trend
(National Center Report No. 05-2).
San Jose, CA
:
National Center for Public Policy and Higher Education
.
Goldrick-Rab, S. (
2006
).
Following their every move: An investigation of social-class differences in college pathways
.
Sociology of Education
,
79
,
67
79
. 10.1177/003804070607900104.
Goldscheider, F. K., & Waite, L. J. (
1986
).
Sex differences in the entry into marriage
.
Journal of Marriage and the Family
,
92
,
91
109
.
Grinstein-Weiss, M., Parentie, D. C., Taylor, S. H., Guo, S., & Raghavan, R. (
2016
).
Racial disparities in education debt burden among low- and moderate-income households
.
Children and Youth Services Review
,
65
,
166
174
. 10.1016/j.childyouth.2016.04.010.
Guzzo, K. B., Nash, S. P., Manning, W. D., Longmore, M. A., & Giordano, P. C. (
2014
).
Unpacking the “black box” of race-ethnic variation in fertility
.
Race and Social Problems
,
7
,
135
149
. 10.1007/s12552-014-9141-2.
Hayford, S., & Guzzo, K. B. (
2016
).
Fifty years of unintended births: Education gradients in unintended fertility in the US, 1960–2013
.
Population and Development Review
,
42
,
313
341
. 10.1111/j.1728-4457.2016.00126.x.
Hayford, S., Guzzo, K. B., & Smock, P. J. (
2014
).
The decoupling of marriage and parenthood? Trends in the timing of marital first births, 1945–2002
.
Journal of Marriage and Family
,
76
,
520
538
. 10.1111/jomf.12114.
Hayford, S. R., & Morgan, S. P. (
2008
).
The quality of retrospective data on cohabitation
.
Demography
,
45
,
129
141
. 10.1353/dem.2008.0005.
Hofferth, S. L., & Goldscheider, F. (
2010
).
Family structure and the transition to early parenthood
.
Demography
,
47
,
415
437
. 10.1353/dem.0.0102.
Huber, P. J. (
1967
).
The behavior of maximum likelihood estimation under nonstandard conditions
. In L. M. Le Cam & J. Neyman (Eds.),
Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability. Vol. 1: Statistics
(pp.
221
233
).
Berkeley
:
University of California Press
.
Huelsman, A. (
2015
).
The debt divide: The racial and class bias behind the “new normal” of student borrowing
.
New York, NY
:
Demos
.
Huinink, J., & Kohli, M. (
2014
).
A life-course approach to fertility
.
Demographic Research
,
30
(
article 45
),
1293
1326
. https://doi.org/10.4054/DemRes.2014.30.45
Jackson, B., & Reynolds, J. (
2013
).
The price of opportunity: Race, student debt, and college achievement
.
Sociological Inquiry
,
83
,
335
368
. 10.1111/soin.12012.
Jones, J., & Schmitt, J. (
2014
).
A college degree is no guarantee
(CEPR Report).
Washington, DC
:
Center for Economic and Policy Research
Kamenetz, A. (
2006
).
Generation debt: Why now is a terrible time to be young
.
New York, NY
:
.
Keels, M. (
2014
).
Choosing single motherhood
.
Contexts
,
13
(
2
),
70
72
. 10.1177/1536504214533504.
Kroeger, T., Cooke, T., & Gould, E. (
2016
).
The class of 2016: The labor market is still far from ideal for young graduates
(Report).
Washington, DC
:
Economic Policy Institute
Lamidi, E. (
2016
).
A quarter century change in nonmarital births: Differnces by educational attainment
(Family Profiles Report No. FP-16-05).
Bowling Green, OH
:
National Center for Family & Marriage Research
.
Lichter, D. T., Johnson, K. M., Turner, R. N., & Churilla, A. (
2012
).
Hispanic assimilation and fertility in new destinations
.
International Migration Review
,
46
,
767
791
. 10.1111/imre.12000.
Lundberg, S., Pollak, R., & Stearns, J. E. (
2016
).
Family inequality: Diverging patterns in marriage, cohabitation, and childbearing
.
Journal of Economic Perspectives
,
30
(
2
),
79
102
. 10.1257/jep.30.2.79.
Manning, W. D. (
2013
).
Trends in cohabitation: Over twenty years of change, 1987–2010
(Family Profiles Report No. FP-13-12).
Bowling Green, OH
:
National Center for Family & Marriage Research
.
Mathews, T. J., & Hamilton, B. E. (
2016
).
Mean age of mothers is on the rise: United States, 2000–2014
(NCHS Data Brief No. 232).
Hyattsville, MD
:
U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Health Statistics
.
Mezza, A., Daniel, R., Shane, S., & Sommer, K. (
2016
).
(Finance and Economics Discussion Series 2016-010).
Washington, DC
:
Board of Governors of the Federal Reserve System
.
Min, S., & Taylor, M. (
2016
,
August
).
Estimating the effect of student loan debt on timing of marriage among race/ethnic groups
.
Presented at the annual meeting of the American Sociological Association
,
Seattle, WA
.
Minicozzi, A. (
2005
).
The short term effect of educational debt on job decisions
.
Economics of Education Review
,
24
,
417
430
. 10.1016/j.econedurev.2004.05.008.
Mulder, C. H. (
2006a
).
Population and housing: A two-sided relationship
.
Demographic Research
,
15
(
article 13
),
401
412
. https://doi.org/10.4054/DemRes.2006.15.13.
Mulder, C. H. (
2006b
).
Home-ownership and family formation
.
Journal of Housing and the Built Environment
,
21
,
281
289
. 10.1007/s10901-006-9050-9.
Nau, M., Dwyer, R. E., & Hodson, R. (
2015
).
Can’t afford a baby? Debt and young Americans
.
Research in Social Stratification and Mobility
,
42
,
114
122
Oppenheimer, V. K. (
1988
).
A theory of marriage timing
.
American Journal of Sociology
,
94
,
563
591
. 10.1086/229030.
Patten, E. (
2016
).
Racial, gender wage gaps persist in U.S. despite some progress
(Fact Tank report).
Washington, DC
:
Pew Research Center
Payne, K. K. (
2012
).
Timing of first marital birth
(Family Profiles Report No. FP-12-11).
Bowling Green, OH
:
National Center for Family & Marriage Research
Rosenbaum, P. R., & Rubin, D. B. (
1983
).
The central role of the propensity score in observational studies for causal effects
.
Biometrika
,
70
,
41
55
. 10.1093/biomet/70.1.41.
Rothstein, J., & Rouse, C. E. (
2011
).
Constrained after college: Student loans and early-career occupational choices
.
Journal of Public Economics
,
95
,
149
163
. 10.1016/j.jpubeco.2010.09.015.
Schwartz, C., & Mare, R. (
2005
).
Trends in educational assortative mating from 1940 to 2003
.
Demography
,
42
,
621
646
. 10.1353/dem.2005.0036.
Scott-Clayton, J., & Li, J. (
2016
).
Black-white disparity in student loan debt more than triples after graduation
(Evidence Speaks Reports Vol.
2
, No.
3
).
Washington, DC
:
Brookings Institute
Smock, P. J., & Greenland, F. R. (
2010
).
Diversity in pathways to parenthood: Patterns, implications, and emerging research directions
.
Journal of Marriage and Family
,
72
,
576
593
. 10.1111/j.1741-3737.2010.00719.x.
South, S. J., & Crowder, K. (
2010
).
Neighborhood poverty and nonmarital fertility: Spatial and temporal dimensions
.
Journal of Marriage and Family
,
72
,
89
104
. 10.1111/j.1741-3737.2009.00685.x.
Sweeney, M., & Raley, R. (
2014
).
Race, ethnicity, and the changing context of childbearing in the United States
.
Annual Review of Sociology
,
40
,
539
558
. 10.1146/annurev-soc-071913-043342.
Titus, M. (
2007
).
Detecting selection bias, using propensity score matching, and estimating treatment effects: An application to private returns using a masters degree
.
Research in Higher Education
,
48
,
487
521
. 10.1007/s11162-006-9034-3.
Wei, C. C., & Horn, L. (
2013
).
Federal student loan debt burden of noncompleters
(Statistics in Brief No. 2013-155).
Washington, DC
:
Institute of Education Sciences, National Center for Education Statistics, U.S. Department of Education
.
White, H. A. (
1980
).
Heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity
.
Econometrica
,
48
,
817
838
. 10.2307/1912934.
Wooldridge, J. M. (
2007
).
Inverse probability weighted estimation for general missing data problems
.
Journal of Econometrics
,
141
,
1281
1301
. 10.1016/j.jeconom.2007.02.002.
Wright, G., & Serrato, S. (
2015
).
Default rate declines, yet 611,000 defaulted on federal student loans. Despite decreases, for-profit colleges account for the greatest share of defaults
[Press release].
Oakland, CA
:
The Institute for College Access and Success
Wu, L. L., & Martin, S. P. (
2009
).
Effects of exposure on prevalence and cumulative relative risk: Direct and indirect effects in a recursive hazard model
.
Sociological Methodology
,
39
,
185
232
. 10.1111/j.1467-9531.2009.01212.x.
Yang, Y., & Morgan, S. P. (
2003
).
How big are educational and racial fertility differentials in the U.S.?
Social Biology
,
50
,
167
187
.