Born Without a Silver Spoon: Race, Wealth, and Unintended Childbearing

Abstract

The United States has a surprisingly high rate of unintended fertility, particularly among women of color. Although studies have examined socioeconomic correlates of unintended fertility, the role of economic resources remains unclear. Wealth may provide an important context for whether a birth was intended or unintended. Moreover, staggering racial wealth disparities may contribute to racial/ethnic patterns of unintended childbearing. This study examines the linkages between wealth and unintended first births, drawing on data from the NLSY79 (N = 1508). Results suggest that net wealth is negatively related to the probability of having an unintended first birth, controlling for a host of sociodemographic characteristics. We also use decomposition analysis to quantify wealth’s contribution to racial/ethnic disparities in unintended childbearing. Second only to marital status, differences in net wealth account for 9–17% of racial/ethnic disparities in unintended childbearing. Our results suggest that wealth is a significant and heretofore overlooked correlate of unintended childbearing.

Introduction

The United States has a surprisingly high rate of unintended fertility—about one-third of all births were considered mistimed or unwanted at the time of conception (Mosher et al. 2012). Unintended childbearing is socially patterned, and is especially prevalent among women of color and populations with low education (Finer and Zolna 2016; Guzman et al. 2010; Musick et al. 2009). Unintended fertility is potentially a vehicle for social inequality, given that the ability to control the timing and spacing of children is also related to several aspects of social mobility and economic resources for women (Sonfield et al. 2013). For example, the ability to delay childbearing has been linked with higher wages and lifetime career earnings (Miller 2011; Taniguchi 1999). Moreover, unintended childbearing is associated with poor health and development for both parents and children (Axinn et al. 1998; Barber and East 2009; Herd et al. 2016, Su 2012; for exceptions see; Joyce et al. 2000; Kost and Lindberg 2015; Marsiglio and Mott 1988). A deeper understanding of the correlates of unintended childbearing may shed light on this process of social stratification.

Research often focuses on education or race and ethnicity to delineate social class disparities in unintended fertility (Finer and Zolna 2011, 2016; Guzman et al. 2010; Musick 2002; Musick et al. 2009), but the role of economic resources remains unclear. Although low-income women are much more likely to experience unintended fertility compared to women with higher income (Finer and Zolna 2016), this pattern is largely attributed to other characteristics, such as marriage, race/ethnicity, age, and education (Abma and Mott 1994; Kost and Forrest 1995; Musick et al. 2009). It is possible that income, which measures the flow of money into a household, is an incomplete measure of economic resources. Wealth—a measure of economic resources that accounts for both assets and debts—is a significant barometer of economic, social, and cultural resources, and may represent a more complete measure (Oliver and Shapiro 2006). Wealth also contributes to racial and ethnic inequality. In 2010, the racial wealth gap was three times larger than the racial income gap (McKernan et al. 2013). Whites had six times as much wealth as Blacks or Latinas (McKernan et al. 2013). Women of color may be doubly disadvantaged given their socioeconomic status and lower position in the wealth distribution (Addo and Lichter 2013). In light of this staggering racial and ethnic inequality, wealth may be a salient dimension of socioeconomic status that contributes to patterns of unintended fertility.

Indeed, wealth is an important consideration in Americans’ decisions about family formation. An influential body of qualitative research has suggested that lacking financial security and assets such as a home or car leads low-income women to delay marriage, but not necessarily childbearing (Augustine et al. 2009; Edin and Kefalas 2005; Gibson-Davis et al. 2005; Smock et al. 2005). Empirical research has supported this theory, providing evidence that wealth, income, and debt are important predictors of family structure (Addo 2014; Edin and Kefalas 2005; Oppenheimer 1988; Sassler 2004; Schneider 2011; Smock et al. 2005; Sweeney 2002), but not fertility (Gibson-Davis 2009). Although wealth may be unrelated to fertility overall, it is unclear whether it is associated with unintended childbearing specifically. It is possible that less wealthy women are more likely to classify their births as unintended, but empirical research has not yet investigated this issue.

Our study builds on prior research by conceptualizing wealth as a central socioeconomic correlate of fertility intentions. Drawing on a sample of mothers from the NLSY79 who conceived their first child from 1985 to 2006, we first examine whether wealth is an independent predictor of unintended birth, after accounting for other known associated factors such as race, marital status, education, and income. Second, given large racial and ethnic disparities in both pregnancy intentions and wealth, we employ decomposition techniques to quantify the contribution of wealth and other population characteristics to racial/ethnic disparities in unintended childbearing. Our paper sheds light on the intersection of race and class early in the life course by focusing on the role of wealth in unintended childbearing.

Background

Conceptual Framework: Linking Wealth and Fertility Intentions

Wealth may be related to fertility intentions for several reasons. First, wealth may be a symbolic marker of financial stability and a cultural prerequisite for intended childbearing. Women may hesitate to characterize a birth as intended if there are questions about whether they can afford to raise a child or provide a stable home environment. A body of research on nonmarital births among low-income families provides some relevant insight on the cultural symbolism of wealth for family formation (Augustine et al. 2009; Edin and Kefalas 2005; Gibson-Davis 2009; Gibson-Davis et al. 2005). This research suggested that low-income parents delayed marriage, but not necessarily childbearing, because marriage was associated with a degree of emotional and economic stability that seemed unachievable due to poor education and job prospects. This reflected a cultural belief that married couples must attain a middle-class lifestyle that included stable employment, home ownership, and enough money to pay for a nice wedding (Edin and Kefalas 2005; Sassler and Cunningham 2008; Smock et al. 2005). At the same time, low-income parents reported that children instilled a sense of hope, purpose, and fulfillment that was missing from their lives (Augustine et al. 2009; Edin and Kefalas 2005). The high social value of children coupled with the low opportunity costs of nonmarital birth and nearly unattainable criteria for marriage lead low-income women to delay marriage but not childbearing (Edin et al. 2007; Edin and Kefalas 2005; Gibson-Davis 2009). Although this research suggested that wealth was unrelated to whether one has a child, it may be related to whether a child was considered unintended. Lacking sufficient wealth, parents may be more likely to classify their pregnancies as unintended.

Second, wealth may be related to unintended childbearing due to its association with self-efficacy—the belief that one has control over one’s life. Qualitative research on White working class and poor married couples in the 1950s suggested that people in the middle class were socialized to be future-oriented and have a strong sense of selfefficacy, which was, in turn, related to purposive fertility planning and use of contraception (Rainwater 1960). In contrast, hardships faced by people in lower social classes engendered uncertainty about the future and a sense that one did not have control over life events, which impeded efforts to control fertility. Indeed, studies have linked low self-efficacy with inconsistent and ineffective contraception among young unmarried women (England et al. 2011, 2016; Sassler and Miller 2014). Qualitative research also found that low-income women with nonmarital births were often ambivalent about their pregnancies, and hesitated to characterize them as completely planned or unplanned. Some women considered their pregnancies to be a matter of fate or God’s will, which reflects low self-efficacy (Edin et al. 2007; Edin and Kefalas 2005). These studies highlight the salience of our central research question, which examines whether wealth is associated with conventional quantitative measures of pregnancy intentions among a large nationally representative birth cohort sample of women.

Finally, from an economic perspective, women with little to no wealth have a weaker economic safety net to buffer periods of unemployment or economic hardship, which may compromise their ability to purchase and consistently use effective contraception (Dehlendorf et al. 2010; Sassler and Miller 2014). Lacking sufficient wealth, such as a savings account, may also be a barrier to obtaining an abortion. Abortions are expensive and the vast majority are paid for out-of-pocket (Boonstra et al. 2006). If the cost of an abortion is prohibitive, less affluent women with unintended pregnancies may be more likely to carry them to term.

Prior Research on Economic Resources and Unintended Childbearing

Empirical studies that have examined economic disparities in unintended fertility have relied on measures of family income expressed as a percentage of the federal poverty threshold (e.g., Abma and Mott 1994; Chandra et al. 2005; Finer and Henshaw 2006; Finer and Zolna 2014, 2016; Williams 1991). There are clear economic disparities in unintended fertility according to this measure. In 2011, poor women were twice as likely to have an unintended pregnancy compared to women above the poverty threshold (Finer and Zolna 2016). Poor women were not only more likely to have unintended pregnancies, they were also less likely to terminate them, resulting in even wider economic disparities in unintended births (Finer and Zolna 2011, 2016).

Yet most empirical studies documenting economic disparities in unintended fertility are based on descriptive statistics that do not account for potentially confounding factors, such as education or race/ethnicity (Chandra et al. 2005; Finer and Henshaw 2006; Finer and Zolna 2011, 2014, 2016; Mosher et al. 2012). Descriptive statistics may therefore overstate the role of income in unintended fertility. Indeed, multivariable regression estimates that controlled for characteristics such as race, age, education, religious affiliation, family background, and cognitive test scores, found that income and poverty status were statistically nonsignificant or very weak predictors of unintended fertility (Abma and Mott 1994; Kost and Forrest 1995; Musick et al. 2009).

Why is income a poor predictor of unintended fertility? It is possible that other characteristics, such as education or marital status, are more influential. Another potential explanation is that income is an incomplete measure of the economic and cultural resources that are related to pregnancy intentions. Income measures the flow of money into a household, but cannot provide a comprehensive picture of overall social status and financial wellbeing. In contrast, wealth represents a bundle of both capital and non-material resources. In addition to economic capital, wealth confers important social resources, such as access to social status, political power, selective educational institutions, and high-quality health care (Oliver and Shapiro 2006). Wealth provides an important safety net that can help families cope in times of crisis, such as job layoffs and medical emergencies (McKernan et al. 2009). It can be invested in education, further strengthening long-term career and earning potential (Keister and Moller 2000).

Wealth may also capture intergenerational legacies of affluence that are not fully measured by income and education (Charles and Hurst 2003). This is reflected in large and persistent racial wealth disparities in the US. Black families own five cents for every (median) dollar of wealth that a White family owns (Tippet et al. 2014). Across the income distribution, Black and Latino households hold less wealth than White households. For Black women, educational attainment and marital status have contributed very little to reducing the wealth gap with White women (Addo and Lichter 2013; Zaw et al. 2017). Wealth may therefore provide a more complete measure of current and future household economic and social resources than income.

Although income and wealth are related, they are not tightly correlated (Keister and Moller 2000). Wealth is a measure of real assets, such as homes and businesses, financial assets, such as savings accounts, bonds, stocks, mutual funds, and retirement accounts, and liabilities, such as mortgages and consumer debts. Wealth can be negative, indicating that debts are larger than assets, or it can be positive, indicating that assets are larger than debts. The relationship between income and wealth also changes across the wealth distribution (Barsky et al. 2002). In other words, low-income households are more likely to have zero or negative wealth, indicating a strong association between income and wealth at the lower end of the wealth distribution. At the higher end of the wealth distribution, in contrast, income and savings constitute only a small percentage of a household’s wealth portfolio. Wealth therefore provides a unique perspective on the household economic context.

Prior Research on Racial and Ethnic Disparities in Unintended Childbearing

There are large racial and ethnic disparities in unintended childbearing in the US. In 2001, over two-thirds of births to Black women were unintended (66%), compared to 46% among Latina women and 36% among non-Latina White women (Guzman et al. 2010). Many of these racial and ethnic differences in unintended childbearing were attributed to marital status, age at birth, parity, and education, however. After accounting for these factors, Black women still had more unintended births than White mothers, but Latina and White mothers had statistically equivalent rates of unintended childbearing (Guzman et al. 2010).

What accounts for racial disparities in unintended childbearing? There is some evidence that the Black-White gap in early childbearing—which is predominantly unintended—is partially attributed to early sexual initiation (sex at age 15 or younger) (Guzzo et al. 2015). The Black-White gap in unintended childbearing might also reflect racial differences in women’s willingness to terminate unwanted pregnancies (Guzman et al. 2010). Indeed, in 2011 unintended pregnancies among Black women were more likely to result in live births compared to Whites and Latinas (Finer and Zolna 2016). In contrast, nativity and marital status played an important role in shaping patterns of unintended childbearing among Latina women (Guzman et al. 2010). Social desirability bias in reporting unintended pregnancy may be particularly strong among foreign-born Latinas, many of whom come from Latin American countries whose cultures and religious traditions honor motherhood regardless of social class (Guzman et al. 2010; Landale and Oropesa 2007).

These observable characteristics have helped to explain the persistent racial disparities in unintended childbearing between Whites and Latinas and, to a certain extent, the Black-White gap. We argue, however, that the exclusion of wealth has yielded an incomplete explanation for observed patterns of fertility among young adult American women and, in particular, differences in unintended fertility by race/ethnicity.

Methods

Data

This study draws on data from the National Longitudinal Survey of Youth, 1979 (NLSY79), a nationally representative birth cohort study following individuals who were born between 1957 and 1964. Respondents were interviewed annually from 1979 to 1994, and biennially thereafter. The cohort sample and longitudinal data design allowed us to observe respondents through their entire childbearing period.

Our analytic sample is n = 1508 women who conceived their first child in 1985 or later¹ and had valid data on pregnancy intentions. Of the 12,686 respondents interviewed in the NLSY79, our analytic sample excluded n = 6403 male respondents (50%), n = 1351 female respondents who did not have any children (11%), n = 3257 female respondents who conceived their first child prior to 1985 (26%), and n = 167 respondents who were missing data on pregnancy intentions (1%). Missing data for other variables were imputed with chained equations (Royston 2004).

We limited respondents to those who conceived their first birth in 1985 or later for two reasons. First, asset information was collected among the full sample beginning in 1985, when all respondents were adults over the age of 18 and therefore likely to have assets and debts independent of their parents. Second, this sample allows us to focus on adult childbearing. This is advantageous because three quarters of teen pregnancies are unintended (Finer and Zolna 2016), making it difficult to fully disentangle the relationship between age and fertility intentions among teenagers. This limits our sample to women who had their first births at age 20 and older, and should be interpreted within this context. The exclusion of teen mothers removed a disadvantaged portion of the sample. Our results might therefore represent a conservative estimate of the relationship between net wealth and unintended childbearing.

Measures

Unintended First Birth

The NLSY79 asked female respondents to retrospectively report pregnancy intentions. Respondents who were using contraception or did not want to become pregnant despite stopping contraception were asked, “Just before you became pregnant the (first, second, third, etc.) time, did you want to become pregnant when you did?” If a respondent answered no, she was asked, “Did you want a(nother) baby but not at that time, or did you want (none/no more) at all?” Pregnancies were classified as intended if the respondent stopped using contraception because she wanted to get pregnant, reported that she wanted to become pregnant when she did, or reported that it “didn’t matter” whether she got pregnant. Pregnancies were classified as mistimed if the respondent wanted another baby but not at that time, and unwanted if she did not want a baby at all. We combine mistimed and unwanted births into a single category for unintended births (1 = unintended, 0 = intended).

We then identified the intention status of the pregnancy that resulted in the respondent’s first live birth². The empirical advantage of focusing on first births is that wealth may be endogenous to fertility at higher parities. For example, having an unintended first birth may disrupt wealth accumulation and limit economic resources, which in turn could be related to subsequent childbearing patterns. Focusing on first births allows us to more effectively isolate the predictive power of wealth on pregnancy intentions. In addition, women with an unintended first birth are more likely to have subsequent unintended births (Guzzo and Hayford 2011), which suggests that the first birth is an important indicator of a woman’s fertility trajectory.

Although self-reported pregnancy intention measures are commonly used in research, their validity is often debated. One concern is that these measures are susceptible to social desirability bias and retrospective reporting bias (Bachrach and Newcomer 1999; Rosenzweig and Wolpin 1993; Sable 1999). Unintended pregnancies and births are therefore likely to be under-reported. Nevertheless, there is evidence that retrospective accounts of pregnancy intentions in the NLSY79, the same data used in the current study, do not bias statistical estimates of the effects of unintended fertility (Joyce et al. 2002). Two studies that examined the retrospective reporting of pregnancy intentions found that approximately 80% of mothers consistently reported their pregnancy intentions in longitudinal surveys (Guzzo and Hayford 2014; Joyce et al. 2002).

Finally, the survey question likely over-simplifies the concept of pregnancy intentions, and does not capture any ambivalence women may feel about their births (Edin et al. 2007; McQuillan et al. 2011). The pregnancy intention measure may also conflate fertility plans with emotional reactions, such as desire, happiness, or disappointment about pregnancy (Bachrach and Newcomer 1999; Santelli et al. 2009). In other words, parents may feel joyful about a pregnancy even if it was unintended. Despite these concerns, the conventional measure of pregnancy intentions provides results that are generally consistent with more nuanced measures of pregnancy desirability (Kost and Lindberg 2015). This measure is also used as an indicator of fertility and population health in national population statistics.

Net Wealth

Our key independent variable is net wealth³, calculated as assets minus liabilities. The NLSY questionnaire includes approximately twenty questions about the value of financial assets and liabilities. Net wealth, a household-level measure, was based on respondent estimates of the worth of their home, savings, asset portfolios, businesses, or vehicles, minus any mortgage, property, or outstanding debts greater than $500. The measure was lagged so it captured net wealth in the year of conception. For respondents who conceived between survey waves, we used data collected in the year prior to conception. The values were adjusted for inflation (measured in 2010 dollars).

The distribution of net wealth is highly skewed, so we evaluated different functional forms of this measure to find the best fit (Killewald et al. 2017). In the main analyses, net wealth is reported in deciles (range 1–10), the best-fitting specification. In supplemental analyses we evaluated a log transformation, the inverse hyperbolic sine transformation, and quartiles of net wealth (available by request). These specifications provided a poorer model fit, but generally yielded results that were consistent with the main analyses.

Control Variables

We adjusted the analyses to account for characteristics that may be related to both unintended childbearing and net wealth. As such, all variables were measured prior to the mother’s first birth. Several variables captured time invariant demographic, cognitive, and psychological characteristics. Race/ethnicity was measured with a categorical variable that indicates whether she is White (e.g., non-Latina, non-Black), non-Latina Black, or Latina. Mother’s cognitive ability was measured in 1980 with the Armed Forces Qualification Test (AFQT); a dichotomous variable indicates whether she has a low AFQT score (in the 25th percentile or lower). The mother’s family background at age 14 was based on retrospective questions that indicate whether the respondent’s mother had low education (less than a high school), whether the respondent lived with both of her biological parents, was born outside the United States, a foreign language was spoken in the home, literacy materials were in the household, and her region of residence (South/non-South and urban/rural). Religion at birth indicates whether the mother was born into Catholic, liberal Protestant, conservative Protestant, or none/other religious denominations.

Another set of variables measured the mother’s sociodemographic characteristics at the time of conception. The mother’s age at conception was calculated using the mother’s and child’s birth dates, and is measured with categories for 20–24, 35–29, 30–34, and 35–45 to capture the nonlinear relationship between age and unintended fertility. The calendar year of conception is measured in years. We used the dates of marriage, divorce, and the child’s birth date to calculate the mother’s marital status at the time of conception. A categorical variable indicates whether she was (a) married at conception, (b) divorced, separated or widowed at conception, or (c) never married at conception.

Other time-varying characteristics were measured in the year of conception for respondents who conceived during an interview year, or the year prior to conception for respondents who conceived between interview years. The highest grade the mother completed in school at the time of conception was measured with categories for less than high school, high school, some college, or college or more education. A continuous variable measured the total household income from salary and wages in the calendar year prior to conception, reported in terms of 2010 thousands of dollars. Family size at conception was measured with a continuous variable.

Data Analysis

The first phase of analysis relies on logistic regressions to estimate the association between net wealth and the probability of having an unintended first birth. All analyses are weighted with sampling weights to account for the complex sampling design of the NLSY79. We first estimated a model that predicts the odds of unintended first birth as a function of race, marital status, and education at the time of birth, the key demographic characteristics that are often used to illustrate social stratification in patterns of unintended childbearing. This model allowed us to evaluate whether results from our sample are in line with other empirical estimates that did not account for net wealth. Next, we added the measure of net wealth to examine whether it is independently associated with unintended birth, and whether it accounts for some of the initial association between race, marital status, education, and the odds of having an unintended first birth. Finally, we included additional covariates that might be associated with both unintended birth and net wealth, such as the mother’s household income, cognitive ability, and characteristics of her childhood household at age 14.

In the second phase of analysis, we quantified the contribution of population characteristics to racial and ethnic disparities in unintended childbearing. Specifically, we decomposed the gap in unintended childbearing between Whites and Blacks, and between Blacks and Latinas, using the regression estimates from the logistic models from phase one. The difference in unintended childbearing between Whites and Latinas was not statistically significant, so we did not decompose that gap. We employed Fairlie’s decomposition technique, which is an extension of the Blinder-Oaxaca decomposition technique for nonlinear models (1999, 2005). This model is expressed in Eq. (1):

$${\overline{Y}}^W-{\overline{Y}}^B=\left[{\displaystyle \sum _{i=1}^{N^W}\frac{F\left(X_i^W{\widehat{\beta }}^W\right)}{N^W}}-{\displaystyle \sum _{i=1}^{N^B}\frac{F\left(X_i^B{\widehat{\beta }}^W\right)}{N^B}}\right]+\left[{\displaystyle \sum _{i=1}^{N^B}\frac{F\left(X_i^B{\widehat{\beta }}^W\right)}{N^B}}-{\displaystyle \sum _{i=1}^{N^B}\frac{F\left(X_i^B{\widehat{\beta }}^B\right)}{N^B}}\right],$$ (1)

where ${\overline{Y}}^j$ is the average probability of unintended birth, ${\overline{X}}^j$ is a vector of average values for the independent variables, ${\widehat{\beta }}^j$ is a vector of coefficient estimates, and N^j is the sample size for race N^j (where W represents Whites, B represents Blacks). F is the cumulative distribution function from the logistic distribution of Y. The first term in brackets represents the part of the racial gap in unintended childbearing that is due to group differences in distributions of × (observable characteristics), and the second term represents the part due to differences in the coefficients in the processes determining unintended birth (the “returns” to the observable characteristics). We focused on the portion of the gap that is “explained” by observable characteristics (the first term).

Given racial and ethnic differences in the processes that predict unintended childbearing, it was unclear which group should be adopted as the standard for weighting the first term of the decomposition. Therefore, we first estimated the decomposition using the White coefficient estimates as standard, and then re-estimated the decomposition using the Black coefficients as standard; we report both sets of results. We followed the same approach to decompose the Black-Latina gap in unintended childbearing.

We also estimated the extent to which net wealth and other specific observable characteristics explain racial and ethnic disparities in unintended birth. The contribution of net wealth to the racial gap is estimated by calculating the change in the average predicted probability of unintended birth when replacing the Black distribution of net wealth with the White distribution of net wealth, while holding other variables constant. The independent contribution of net wealth (X_l·) to the racial disparity in unintended childbearing is expressed in Eq. (2). We repeated this process to estimate the contribution of each observable characteristic in the model.

$$\frac{1}{N^B}{\displaystyle \sum _{i=1}^{N^B}F\left({\widehat{\alpha }}^W+X_{1i}^W{\widehat{\beta }}_1^W+X_{2i}^W{\widehat{\beta }}_2^W\right)-F\left({\widehat{\alpha }}^W+X_{1i}^B{\widehat{\beta }}_1^W+X_{2i}^W{\widehat{\beta }}_2^W\right).}$$ (2)

Results

Table 1 presents weighted descriptive statistics for the total sample, and by the pregnancy intention of the mother’s first birth. Seventeen percent of our sample had an unintended first birth. Among unintended births, 85% were mistimed and 15% were unwanted (not shown in table). This is unsurprising given our focus on first births. Recall that a birth is considered mistimed if the respondent reported that she wanted a child at some point in the future but her pregnancy happened too soon. Women at the end of their childbearing period are more likely to report unwanted births, likely because they are more certain that they do not want any more children in the future (D’Angelo et al. 2004). The majority of the sample is White (87%), 8% is Black, and 5% is Latina. There are several statistically significant differences between respondents who had unintended vs. intended first births. Women who had an unintended first birth had lower net wealth, and were more likely to be Black, younger, and unmarried at conception. Women with unintended first births also had lower AFQT scores, education, and income compared to women who had an intended first birth.

Table 1.

Descriptive statistics, NLSY79 sample of women who conceived their first birth in 1985 or later

Variable	Total		Intended first birth		Unintended first birth
	Mean	Std. dev	Mean	Std. dev	Mean	Std. dev
Unintended first birth	0.17
Net wealth decile	5.99	2.85	6.28	2.78	4.56	2 77***
Mother’s race
White	0.87		0.89		0.76	***
Black	0.08		0.06		0.18	***
Latina	0.05		0.05		0.06
Mother’s age at conception
20–24	0.20		0.17		0.35	***
25–29	0.46		0.48		0.39	**
30–34	0.26		0.28		0.17	***
35–45	0.08		0.07		0.09
Mother’s relationship at conception
Married	0.72		0.79		0.38	***
Divorced, separated, or widowed	0.13		0.12		0.20	***
Never married	0.15		0.09		0.42	***
Low AFQT score	0.16		0.15		0.21	*
Mother’s education at conception
Less than high school	0.04		0.04		0.07	*
High school	0.32		0.31		0.37
Some college	0.26		0.25		0.30
College or more education	0.37		0.39		0.26	***
Income at conception ($10,000)	9.81	20.43	10.45	20.95	6.62	17.26*
Mother’s family background
Mother had low education	0.23		0.22		0.27
Lived with bio parents	0.80		0.82		0.70	***
Born outside of US	0.04		0.05		0.04
Lived in urban area (age 14)	0.80		0.79		0.85	*
Lived in the South (age 14)	0.29		0.29		0.31
No literacy materials in HH	0.03		0.03		0.03
Catholic	0.41		0.43		0.32	**
Liberal protestant	0.16		0.16		0.15
Conservative protestant	0.30		0.28		0.38	**
None or other religion	0.13		0.13		0.14
Year of conception (1985–2005)	1990.05	4.03	1990.18	4.01	1989.44	4.06**
Family size at conception	2.08	0.97	2.08	0.87	2.04	1.38
Observations	1508		1197		311

Asterisks indicate statistically significant difference between intended and unintended;

^***p < 0.001,

^**p < 0.01,

^*p < 0.05

Table 2 presents results from logistic regressions predicting unintended birth, expressed in terms of log odds. Model 1 adjusts for race/ethnicity, marital status, and education. As expected, Black women have higher odds of having an unintended first birth compared to White women, adjusting for marital status and education. Women who were unmarried at conception also have significantly higher odds of having an unintended first birth compared to married women. These results are in line with existing research on the social patterns of unintended pregnancy (Finer and Zolna 2016; Guzman et al. 2010).

Table 2.

Logistic regressions predicting unintended first birth

Variables	Model 1		Model 2		Model 3
	B	(SE)	B	(SE)	B	(SE)
Net wealth decile			−0.13***	(0.03)	−0.11**	(0.04)
Black	0.51*	(0.21)	0.37	(0.21)	0.37	(0.27)
Latina	0.12	(0.22)	0.03	(0.23)	0.37	(0.27)
20–24					0.83	(0.60)
25–29					−0.02	(0.53)
30–34					−0.38	(0.43)
Divorced/separated/widow	1.23***	(0.23)	1.20***	(0.24)	1.35***	(0.24)
Never married	2.09***	(0.21)	1.89***	(0.22)	2.01***	(0.23)
Low AFQT score					−0.39	(0.26)
Less than high school	0.08	(0.41)	−0.22	(0.40)	−0.29	(0.46)
High school	0.22	(0.22)	0.03	(0.23)	−0.09	(0.26)
Some college	0.25	(0.24)	0.13	(0.24)	−0.01	(0.25)
Income at conception					−0.00	(0.01)
Mother had low education					−0.12	(0.22)
Lived with bio parents					−0.06	(0.22)
Born outside of US					−0.31	(0.37)
Lived in urban area					0.51*	(0.25)
Lived in the South					0.01	(0.21)
No literacy materials in HH					−0.55	(0.35)
Catholic					−0.24	(0.31)
Liberal protestant					0.08	(0.34)
Conservative protestant					0.12	(0.29)
Year of conception					0.02	(0.04)
Family size at conception					0.02	(0.07)
Constant	−2.50***	(0.18)	−1.61***	(0.28)	−45.90	(79.94)
Observations	1508		1508		1508

Robust standard errors in parentheses

^***p < 0.001,

^**p < 0.01,

^*p < 0.05

Model 2 includes the net wealth variable, measured in deciles. A decile increase in the net wealth distribution is associated with decreased odds of having an unintended first birth, controlling for race, marital status, and education (B = − 0.13, p < 0.001). This model provides evidence that net wealth has an independent and statistically significant association with unintended first births. This model also suggests that wealth accounts for some of the initial association between race and unintended first births. When we adjust for wealth in Model 2, the coefficient for the Black variable is attenuated and no longer statistically significant. Wealth does not seem to account for the initial relationships between marital status and unintended birth, however; the coefficients on these variables in Model 2 are similar in magnitude to Model 1 and remain statistically significant.

Model 3 includes a host of measures that potentially confound the relationship between net wealth and unintended birth. The coefficient on net wealth is similar to Model 2 and remains statistically significant (B = − 0.11 p < 0.01), even after adjusting for household income and a rich set of variables that capture the mother’s demographic and socioeconomic characteristics. This is further evidence that net wealth has an independent association with the probability of unintended first birth, and that it captures a unique dimension of economic resources that is not measured by income.

The results from Model 3 are further illustrated in Fig. 1, which presents the predicted probabilities of having an unintended first birth for each net wealth decile and holding all other control variables at their means. About 20% of women in the lowest net wealth decile are predicted to have an unintended first birth, compared to 14% in the 5th decile and 8% in the highest decile.

Fig. 1.

Adjusted predicted probabilities of unintended first birth by net wealth decile. Predicted probabilities calculated from coefficients in fully conditional model (Table 2, Model 3)

Table 3 presents descriptive statistics by race/ethnicity, and illustrates significant disparities in unintended childbearing. Fifteen percent of White mothers had an unintended first birth, compared to 37% of Black mothers and 21% of Latina mothers. The difference in unintended first births between Whites and Blacks is statistically significant (p < 0.001), as is the disparity between Blacks and Latinas (p < 0.001). The difference between Whites and Latinas is not statistically significant, which is consistent with prior research (Guzman et al. 2010). This table also reveals several compositional differences by race/ethnicity. Whites are more likely to have characteristics that are considered to be socioeconomically advantageous; on average they had higher net wealth, higher likelihood of being married at conception, higher cognitive test scores, and more education compared to Blacks and Latinas. They also grew up in households that were comparatively advantaged in terms of living with both biological parents and having literacy materials in the household at age 14. Blacks are the least advantaged relative to Whites and Latinas, on average. They had the lowest net wealth, were least likely to be married at conception, and least likely to have lived with both parents as a child. They were also more likely to be Conservative Protestant and to have lived in the South at age 14.

Table 3.

Descriptive statistics by race/ethnicity, NLSY79 sample of women who conceived their first birth in 1985 or later

Variable	White		Black		Latina
	Mean	Std. dev.	Mean	Std. dev.	Mean	Std. dev.
Unintended first birth	0.15		0.37	***	0.21	a
Net wealth decile at conception	6.23	2.80	4.05	2.46***	4.96	2.87***a
Mother’s age at conception
20–24	0.19		0.30	**	0.29	*
25–29	0.48		0.36	*	0.39
30–34	0.26		0.26		0.21
35–45	0.08		0.08		0.11
Mother’s relationship at conception
Married	0.75		0.44	***	0.61	**a
Divorced, separated, widowed	0.14		0.06	*	0.14	a
Never married	0.11		0.50	***	0.24	***a
Low AFQT score	0.12		0.51	***	0.43	***
Mother’s education at conception
Less than high school	0.04		0.06		0.14	***a
High school	0.33		0.26		0.31
Some college	0.24		0.41	***	0.35
College or more education	0.39		0.26	**	0.21	**
Income at conception ($10,000)	10.32	21.08	5.72	15.40*	7.53	14.48
Mother’s family background
Mother had low education	0.19		0.38	***	0.68	***a
Lived with bio parents	0.83		0.54	***	0.72	*a
Born outside of US	0.03		0.06		0.23	***a
Lived in urban area (age 14)	0.79		0.87		0.87
Lived in the South (age 14)	0.27		0.55	***	0.29	a
No literacy materials in HH	0.02		0.10	***	0.18	***a
Catholic	0.41		0.13	***	0.87	***a
Liberal protestant	0.18		0.09	*	0.01	***a
Conservative protestant	0.28		0.65	***	0.06	***a
None or other religion	0.13		0.13		0.06	a
Year of conception	1990.10	3.97	1989.63	4.14	1989.99	4.81
Family size at conception	2.01	0.77	2.41	1.65***	2.71	1.86***
Observations	990		269		249

Asterisks indicate statistically significant difference relative to White;

^***p < 0.001,

^**p < 0.01,

^*p < 0.05

^aStatistically significant difference between Blacks and Latinas, p < 0.05

In terms of socioeconomic advantage, Latinas are generally less advantaged than Whites, but more advantaged than Blacks. Notable exceptions include that Latinas had significantly lower education, were more likely to have mothers with low education, and were more likely to have had no literacy materials in their childhood household compared to both Whites and Blacks. Latinas were also more likely to have been born outside the US, and identify as Catholic.

To what extent do these compositional differences, such as net wealth, marital status, age, and education, contribute to racial and ethnic disparities in unintended first births? To answer this question we turn to decomposition results presented in Table 4. These decompositions include the same set of control variables as specified in the regression models. The effect of categorical variables, such as age, relationship status, and education, are represented as the total effect across all categories. Recall that we estimated each decomposition two times, changing the racial/ethnic group that serves as the “standard;” these estimates functionally serve as upper-and lower-bound estimates. Results indicate that racial differences in net wealth account for 9–17% of these gaps in unintended first births.

Table 4.

Decomposition of racial/ethnic gaps in unintended childbearing

	Black-White gap Group 1 = Black Group 2 = White				Black-Latina gap Group 1 = Black Group 2 = Latina
	Black as standard		White as standard		Black as standard		Latina as standard
	Coeff.	% explained	Coeff.	% explained	Coeff.	% explained	Coeff.	% explained
Net wealth decile	0.03	13	0.04	16	0.01	9	0.03	17
Mother’s age at conception	0.02	8	0.02	10	0.00	3	0.01	4
Mother’s relationship at conception	0.13	56	0.11	49	0.07	46	0.04	27
Low AFQT score	−0.04	−17	0.02	7	0.00	2	0.00	0
Mother’s education at conception	0.00	−2	0.02	8	0.00	0	0.02	10
Income at conception ($10,000)	0.00	1	−0.02	−7	−0.01	−3	0.01	5
Mother had low education	−0.01	−2	0.00	2	−0.01	−6	−0.02	−14
Lived with bio parents	0.00	0	−0.01	−2	0.00	−2	0.03	19
Born outside of US	0.00	0	0.00	0	−0.01	−7	0.00	0
Lived in urban area (age 14)	0.01	2	0.01	3	0.00	0	0.00	0
Lived in the South (age 14)	0.00	−1	0.00	1	0.00	2	0.00	3
No literacy materials in HH	0.00	−2	−0.01	−4	0.01	7	0.01	5
Religious denomination	0.02	7	−0.01	−6	−0.03	−19	0.11	69
Year of conception	0.00	0	−0.01	−4	−0.01	−4	−0.01	−7
Family size at conception	0.00	0	0.01	7	−0.01	−7	0.01	7
Observations	1259		1259		518		518
Group 1 Observations	269		269		269		269
Group 2 Observations	990		990		249		249
Share Unintended Group 1	0.37		0.37		0.37		0.37
Share Unintended Group 2	0.15		0.15		0.21		0.21
Gap in Unintended	0.22		0.22		0.16		0.16
Total explained	0.14	63	0.18	81	0.03	22	0.23	146

The Black-White decomposition estimates that differences in observable group characteristics explain 63–81% of the gap in unintended first births. Of particular relevance to our study, results indicate that if Black women had similar net wealth profiles of the White women in the sample, the gap in unintended first births would decrease by approximately 14% on average. Most of the Black-White gap in unintended first births is due to differences in marital status at conception. If Blacks had the same rate of marriage as Whites, the gap in unintended first births would decrease 49–56%. Combined, net wealth and marital status at birth explain about 65% of the gap in unintended first births.

Results from the Black-Latina decomposition indicate that when Blacks are used as the standard, 22% of the gap is explained by group differences in characteristics, whereas when Latinas are used as the standard, the gap is completely explained by compositional differences (> 100%). Differences in Black-Latina net wealth profiles explain between 9 and 17% of the intendedness differential. The decomposition results also indicate that the pattern of unintended first births for Latina women would be slightly more responsive to net wealth changes compared to Black women if differences between the two groups were eliminated. Latina women were also much more likely to be married at conception; therefore it is not surprising that differences in mother’s marital status contribute significantly to the gap.

Robustness Checks

Wealth and marriage are tightly intertwined, particularly for couples who can pool assets. Wealth is also positively linked with age throughout most of adulthood, given that people tend to acquire assets that generate wealth and accumulate over time. Because both marriage and age are also correlates of unintended childbearing, we examined whether the relationship between wealth and the probability of an unintended first birth is moderated by these factors. We found no evidence of this moderation (Appendix Table 5). As additional robustness checks, we estimated separate regressions on subsamples of married, unmarried, and younger respondents, respectively. These models yield very similar results to the full sample (Appendix Tables 6 and 7).

Table 5.

Moderation analysis

Variables	Model 1		Model 2
	Interaction with respondent’s age at birth		Interaction with respondent’s marital status at birth
	B	(SE)	B	(SE)
Net wealth decile	−0.03	(0.12)	−0.16***	(0.05)
20–24	1.77	(0.98)	0.87	(0.60)
25–29	0.60	(0.94)	0.05	(0.52)
30–34	0.01	(0.89)	−0.32	(0.43)
20–24 × Net wealth	−0.17	(0.13)
25–29 × Net wealth	−0.09	(0.13)
30–34 × Net wealth	−0.05	(0.12)
Divorced/separated/widow	1.33***	(0.24)	0.77	(0.55)
Never married	1.98***	(0.23)	1.58***	(0.39)
Divorced/separated/widow × Net wealth			0.11	(0.09)
Never married × Net wealth			0.09	(0.08)
Black	0.39	(0.27)	0.38	(0.26)
Latina	0.33	(0.27)	0.35	(0.27)
Low AFQT score	−0.37	(0.26)	−0.39	(0.25)
Less than high school	−0.31	(0.47)	−0.27	(0.46)
High school	−0.08	(0.26)	−0.09	(0.26)
Some college	0.00	(0.25)	−0.02	(0.25)
Income at conception ($10,000)	−0.00	(0.01)	−0.00	(0.01)
Mother had low education	−0.12	(0.23)	−0.12	(0.22)
Lived with bio parents	−0.06	(0.22)	−0.08	(0.21)
Born outside of US	−0.32	(0.37)	−0.34	(0.37)
Lived in urban area (age 14)	0.51*	(0.25)	0.51*	(0.25)
Lived in the South (age 14)	0.03	(0.21)	0.03	(0.21)
No literacy materials in HH	−0.57	(0.35)	−0.53	(0.35)
Catholic	−0.23	(0.32)	−0.23	(0.31)
Liberal protestant	0.08	(0.35)	0.07	(0.34)
Conservative protestant	0.11	(0.30)	0.10	(0.29)
Year of conception	0.02	(0.04)	0.02	(0.04)
Family size at conception	0.01	(0.07)	0.02	(0.07)
Constant	−44.85	(79.25)	−40.59	(80.09)
Observations	1508		1508

Robust standard errors in parentheses;

^***p < 0.001,

^**p < 0.01,

^*p < 0.05

Table 6.

Logistic regressions predicting unintended first birth, limiting sample by age

Variables	Model 1 Age<30		Model 2 Age<35
	B	(SE)	B	(SE)
Net wealth decile	−0.18***	(0.04)	−0.14***	(0.04)
Black	0.35	(0.31)	0.39	(0.28)
Latina	0.24	(0.32)	0.17	(0.29)
20–24	0.90***	(0.25)	1.44***	(0.37)
25–29			0.50	(0.29)
Divorced/separated/widow	1.04**	(0.33)	1.10***	(0.28)
Never married	1.87***	(0.27)	1.96***	(0.24)
Low AFQT score	−0.46	(0.30)	−0.39	(0.26)
Less than high school	−0.22	(0.52)	−0.35	(0.47)
High school	−0.09	(0.30)	−0.23	(0.27)
Some college	0.01	(0.30)	−0.08	(0.26)
Income at conception ($10,000)	0.01	(0.01)	0.00	(0.01)
Mother had low education	−0.36	(0.26)	−0.05	(0.24)
Lived with bio parents	0.04	(0.25)	−0.13	(0.23)
Born outside of US	−0.17	(0.41)	−0.26	(0.38)
Lived in urban area (age 14)	0.62	(0.32)	0.58*	(0.27)
Lived in the South (age 14)	0.05	(0.25)	0.15	(0.22)
No literacy materials in HH	−0.21	(0.39)	−0.53	(0.37)
Catholic	0.12	(0.37)	−0.04	(0.33)
Liberal protestant	0.46	(0.41)	0.19	(0.37)
Conservative protestant	0.46	(0.36)	0.13	(0.31)
Year of conception	0.05	(0.05)	0.06	(0.04)
Family size at conception	0.02	(0.08)	0.03	(0.07)
Constant	−108.74	(107.67)	−127.65	(85.36)
Observations	1026		1392

Robust standard errors in parentheses

^***p < 0.001,

^**p < 0.01,

^*p < 0.05

Table 7.

Logistic regressions predicting unintended first birth, limiting sample by marital status at conception

Variables	Model 1		Model 2
	Married		Unmarried
	B	(SE)	B	(SE)
Net wealth decile	−0.17**	(0.05)	−0.10†	(0.06)
Black	0.62	(0.38)	0.55	(0.34)
Latina	0.02	(0.42)	0.57	(0.37)
20–24	3.97**	(1.26)	0.46	(0.80)
25–29	3.02*	(1.19)	−0.41	(0.68)
30–34	2.59*	(1.13)	−0.69	(0.54)
Low AFQT score	−0.28	(0.37)	−0.46	(0.32)
Less than high school	0.28	(0.74)	−0.40	(0.58)
High school	−0.22	(0.36)	−0.07	(0.40)
Some college	−0.17	(0.34)	0.15	(0.39)
Income at conception ($10,000)	0.01	(0.01)	−0.03	(0.02)
Mother had low education	−0.29	(0.36)	0.04	(0.30)
Lived with bio parents	−0.53†	(0.30)	0.25	(0.28)
Born outside of US	−0.03	(0.50)	−0.53	(0.57)
Lived in urban area (age 14)	0.79*	(0.38)	0.30	(0.36)
Lived in the South (age 14)	0.12	(0.32)	−0.21	(0.29)
No literacy materials in HH	0.21	(0.51)	−0.82*	(0.41)
Catholic	−0.37	(0.45)	−0.03	(0.42)
Liberal protestant	−0.20	(0.49)	0.39	(0.49)
Conservative protestant	−0.15	(0.42)	0.38	(0.39)
Year of conception	0.02	(0.06)	0.01	(0.05)
Family size at conception	−0.13	(0.18)	0.10	(0.08)
Constant	−45.29	(124.98)	−27.28	(107.61)
Observations	1002		502

Robust standard errors in parentheses

^***p<0.001,

^**p<0.01,

^*p<0.05,

^†p<0.10

Supplementary Analyses

We conducted supplemental exploratory analyses to evaluate some of the mechanisms that might account for the link between wealth and unintended childbearing (results available by request). We focused on pregnancies rather than births, and examined respondents with first pregnancies that were conceived from 1985 to 1990 (n = 872). The observation window is limited to 5 years because the NLSY79 collected pregnancy intentions only among pregnancies that resulted in live births beginning in 1992. We estimated a series of logistic regressions to evaluate whether wealth is related to whether the pregnancy was: (a) unintended, (b) resulted in live birth, or (c) resulted in abortion. We also evaluated whether the respondent: (d) had any sex education, and (e) used any type of contraception before pregnancy. Similar to the results among births in our main analyses, we found a statistically significant negative relationship between wealth and the odds of unintended pregnancy (p < 0.01). Wealth is not statistically related to any of the other outcomes we examined, which provides some suggestive evidence that differential access to abortion, contraceptive use, or sex education are not driving the main results. We exercise caution in drawing conclusions from this supplemental analysis, however, due to the limited sample, truncated observation window, and significant under-reporting of self-reported abortion data.

Discussion

An influential body of qualitative and empirical research suggests that wealth contributes to unequal patterns of family formation. There is evidence that wealthier individuals are more likely to marry (Edin and Kefalas 2005; Gibson-Davis et al. 2005; Schneider 2011), and marital histories are associated with wealth accumulation in later life (Addo and Lichter 2013), reproducing patterns of inequality at the household level. Although an individual’s wealth is an important precursor to marriage, there is some evidence that it was not related to fertility in general (Edin and Kefalas 2005; Gibson-Davis 2009). Our study investigated whether net wealth is linked to a specific dimension of fertility—unintended first births—drawing on data from the NLSY79. We also investigated the role of wealth in racial and ethnic disparities in unintended childbearing.

Our study makes three key contributions to the literature. First, whereas most prior research on disparities in unintended childbearing focuses on income as the sole measure of economic resources, we investigated the unique and independent role of net wealth. Our results suggest that having less net wealth is associated with a higher likelihood of having an unintended first birth. Importantly, we adjust statistical models for household income to disentangle the effect of net wealth from other economic resources. Household income is not a significant predictor of unintended birth, which is consistent with prior research (Abma and Mott 1994; Kost and Forrest 1995; Musick et al. 2009). We also adjusted for several known correlates of unintended childbearing, including marital status, race, age, education, and a host of characteristics capturing the respondent’s own childhood environment. The statistical significance of net wealth is therefore noteworthy and suggests that it is a unique predictor of unintended birth.

Second, we build on prior research that suggests that wealth is unrelated to fertility by calling attention to heterogeneity among births, distinguishing between those that are intended and unintended. Our results add nuance to these prior findings, and suggest that wealth is indeed related to a distinct dimension of fertility—unintended childbearing. Although prior qualitative research suggests that many low-income women with nonmarital births were ambivalent about their pregnancy planning (Edin et al. 2007), our study suggests that wealth is linked to conventional measures of pregnancy intentions that are commonly used in demographic research.

Finally, net wealth partially accounts for racial/ethnic disparities in unintended first births. In regression models, wealth explains part of the initial association between race and unintended birth. In decomposition models, we found that racial/ethnic differences in net wealth explain approximately 14% of both the Black-White and Black-Latina gap in unintended childbearing. Although the share is relatively small, it is larger in magnitude than age at conception and education, two major correlates of unintended childbearing that receive much attention in scholarly literature.

It is striking that well-documented disparities in unintended childbearing by marital status persist even after controlling for net wealth. In fact, in our fully conditional models, net wealth, marital status, and living in an urban area at age 14 are the only statistically significant predictors of unintended first births. Furthermore, differences in marital status explain a significant share of racial/ethnic disparities in unintended childbearing. Although scholars argue that the link between marriage and childbearing has weakened given rising rates of nonmarital birth (Cherlin 2004; Smock and Greenland 2010), these results provide evidence that marital status remains an important correlate of unintended childbearing. This speaks to the salience of marriage in contextualizing how women think about their pregnancies and births.

Our study is not without limitations. This study focuses on a sample of adult women and their first births, which may limit the generalizability of the findings. For example, the vast majority of unintended births in our study were mistimed, and the results should be interpreted within this context. In addition, although our analysis accounts for a particularly rich set of background characteristics that may confound the relationship between wealth and unintended childbearing, our results remain susceptible to omitted variable bias. Wealth is not randomly distributed in the population, and if there are unobservable characteristics that are linked to both net wealth and selection into unintended childbearing the estimates will be biased. While the decomposition analysis easily lends itself to quantifying the contribution of explained characteristics to group differences, these results are susceptible to the same limitations of regression analysis.

This study is an important first step in establishing an empirical link between wealth and unintended childbearing, and it highlights several avenues for future research. For example, our study suggests that wealth has a significant relationship with the timing of a first birth, and future research should examine the link between wealth and unintended childbearing at higher parities. It is possible that wealth may have a different relationship with first births vs. subsequent births, although this presents an empirical challenge due to concerns about endogeneity. In addition, our study does not identify the causal mechanisms underlying disparities in unintended childbearing. For example, a financial safety net may play a critical role in shaping patterns of unintended fertility due to differential access to effective contraception and/or abortion. Supplementary exploratory analysis on a limited sample finds no support that wealth was associated with differential access to abortion, sex education, and contraceptives. Our results highlight the need for future studies to formally test the underlying mechanisms with richer data. Uncovering this process would provide important insights for social policy.

Family formation patterns have reified social class differences among American families. Fertility patterns that are associated with better economic, social, and developmental well-being for parents and children, such as marital childbearing and planned pregnancies, are more common among women with higher socioeconomic status (McLa-nahan 2004; Musick et al. 2009). Although scholars typically evaluate marriage and childbearing patterns using educational attainment or income as a proxy for socioeconomic status, our study suggests that wealth is a significant and heretofore overlooked correlate of unintended childbearing. Having a birth too soon potentially disrupts educational and career trajectories and has long-term impacts on economic and social mobility (Sonfield et al. 2013). The current study highlights another way in which wealth disparities can further stratify society through patterns of unintended childbearing.

Biography

Jessica Houston Su is an Assistant Professor of Sociology at the University at Buffalo-SUNY. Her research examines how patterns of fertility and family structure in the United States are related to the health and well-being of parents and children. She received her Ph.D. in Sociology from Cornell University. Her research has been published in Demography, the Journal of Marriage and Family, and the Journal of Health and Social Behavior.

Fenaba R. Addo is an Assistant Professor of Consumer Science in the School of Human Ecology at the University of Wisconsin-Madison. Her research primarily focuses on the role of assets and debt in relationships and family formation, wealth inequality, and health disparities. She received her Ph.D. in Policy Analysis and Management from Cornell University. Her research has appeared in American Sociological Review, Demography, Family Relations, Journal of Marriage and Family, and Social Science Research.

Appendix

See Tables 5, 6 and 7.