Family, Academic, and Peer Group Predictors of Adolescent Pregnancy Expectations and Young Adult Childbearing

Abstract

Compared with previous generations, today’s young people increasingly delay parenthood. Having children in the late teens and early 20s is thus a rarer experience rooted in and potentially leading to the stratification of American families. Understanding why some adolescents expect to do so can illuminate how stratification unfolds. Informed by theories of the life course, social control, and reasoned action, this study used the National Longitudinal Survey of Youth 1997 cohort (n = 4,556) to explore outcomes and antecedents of adolescent pregnancy expectations with logistic regressions. Results indicated that those expectations—including neither low nor high (i.e., split) expectations—predicted subsequent childbearing. These apparently consequential expectations were, in turn, most closely associated with youth’s academics and peer groups. These findings illustrate how different domains can intersect in the early life course to shape future prospects, and they emphasize split pregnancy expectations reported in a nationally representative sample of young women and men.

Despite worries about a supposed decline of the American family, most young people still plan to become parents. They just do so at older ages and in different contexts than previous generations. Partly as a result of these changing norms and the broad structural factors that contributed to them, pregnancy in one’s late teens or early 20s sharply differentiates life course chances and needs to be better understood (Cherlin, 2009). One way to pursue this understanding is to examine how what happens during the transition into young adulthood is rooted in experiences during adolescence, when young people are actively and passively making plans for their futures in ways that reflect their present (Johnson, 2005; Schneider, 2003). In this way, adolescent expectations to become pregnant or to get someone pregnant in the near future—a marker of current circumstances and future prospects—likely forecasts the increasing fertility-related divergence of young people’s lives (McLanahan, 2004).

In this spirit, I use the National Longitudinal Survey of Youth 1997 cohort (NLSY97) to answer two research questions: (1) How do adolescent pregnancy expectations matter for future childbearing? (2) How do adolescents’ families, academics, and peer groups shape the full spectrum of these expectations? Expectations tend to predict later behavior across a variety of domains, including fertility (Barber, 2001), but these patterns likely go beyond the basic idea that high expectations lead to fertility and low expectations do not. I incorporate split expectations, which represent the middle ground of pregnancy expectations that could complicate these links (Jaccard, Dodge, & Dittus, 2003; Zabin, Astone, & Emerson, 1993). In considering whether youth whose pregnancy expectations fall in between the high and low ends are a distinct group or lean toward one pole or another, I recognize pregnancy expectations reflect where adolescents currently are and where they are going, and so they need to be examined in relation to important settings (i.e., family, school, peers) that position adolescents on life course trajectories.

Exploring a broader spectrum of pregnancy expectations is important for characterizing the adolescents holding those expectations and for identifying young parents in need of extra support. Illuminating the gray area between the two poles of pregnancy expectations (definitively yes vs. definitively no) adds a new layer to our understanding of how adolescents anticipate major life events such as pregnancy and parenthood. Additionally, research of this kind informs how researchers and practitioners should study and assist young people, suggesting new ways to talk to them about the future by using their own perspectives on decisions that may be unclear or undefined. This study’s findings could also identify two targets for policy interventions: adolescents who are uncertain about their future fertility as well as young parents with limited opportunities. If young people with fewer social supports are the same ones who have children early in young adulthood, then additional assistance and resources may be particularly beneficial for them and their children.

Pregnancy Expectations in Adolescence and Childbearing in Young Adulthood

Originally used to forecast population-level fertility, fertility expectations refer to the likelihood of a pregnancy or birth in the future, and they are driven by an individual’s fertility goals as well as the personal resources and opportunity structures at one’s disposal to achieve those goals (Hendershot & Placek, 1981). A pioneer in the field of fertility motivations, W. B. Miller (1992) argued expectations are a very similar yet more passive form of intentions, which he described as “what someone actually plans to do” (W. B. Miller, 1994, p. 228). Adolescent pregnancy expectations, therefore, represent how young people envision their future fertility based on what they want and what realities might hold them back from achieving it. Often operationalized as continuously increasing from low levels (negative expectations, i.e., strong feelings about not wanting a pregnancy) to high levels (positive expectations, i.e., strong feelings about wanting a pregnancy), expectations are important to study because they often translate into actual behavior, meaning adolescents’ expectations have implications for what they do in adulthood (Johnson, 2005; Schneider, 2003). Indeed, expectations of an adolescent or nonmarital birth are both positively associated with later childbearing among youth from a variety of socioeconomic backgrounds (Trent & Crowder, 1997). Pregnancy expectations are also negatively associated with educational and socioeconomic attainment regardless of actual fertility outcomes (Raley, Kim, & Daniels, 2012). This literature usually conceptualizes pregnancy expectations linearly based on the likelihood of a pregnancy (see Raley et al., 2012) or dichotomously based on the age at which respondents expect a pregnancy (see Trent & Crowder, 1997).

Fertility desires are the focus of an alternative literature that is distinct from yet very much related—and able to offer lessons—to the expectations literature. Contrary to expectations, desires pertain more to individuals’ feelings and wishes about the occurrence and timing of a pregnancy or birth, and they are driven by separate and distinct positive and negative childbearing motivations (W. B. Miller, 1992, 1994, 1995). A large body of research has explored the combination of positive and negative desires, revealing people can express ambivalence (i.e., high desires to have a child and high desires to not have a child) or indifference (i.e., low desires to both have a child and not have a child; W. B. Miller, Barber, & Gatny, 2013), as well as strong desires to avoid future pregnancy coupled with expression of happiness at the prospect (Aiken, Dillaway, & Mevs-Korff, 2015; Aiken & Potter, 2013). Ambivalence is particularly important because it is negatively associated with consistent contraceptive use (W. B. Miller, Trent, & Chung, 2014; Yoo, Guzzo, & Hayford, 2014) and positively associated with early childbearing (Jaccard et al., 2003; W. B. Miller et al., 2013; Zabin et al., 1993) such that the contraceptive practices and fertility of young people with ambivalent desires approach those of youth with positive desires.

This study combines those two literatures by taking the idea of a middle ground from research on fertility desires and applying it to split pregnancy expectations. I use a nationally representative sample of adolescents to explore expectations of early pregnancy and actually experiencing a pregnancy, during a life course stage when expectations take on added significance given societal norms and consequences associated with early childbearing, especially among teenagers (Furstenberg, 2003). Attitudes may be especially impactful during adolescence and young adulthood (Barber, 2001) when the opportunity structure changes in important ways and when choices and constraints set young people on pathways that can be difficult to reroute. Furthermore, young people who become parents despite not fully expecting to are likely a unique group faced with distinct challenges. Ambivalent adolescents who become parents experience more emotional distress (East & Barber, 2014; Francis, Malbon, Braun-Courville, Lourdes, & Santelli, 2015), and unintended pregnancies are associated with numerous risks for mothers’ and children’s physical and mental health (Brown & Eisenberg, 1995). Because pregnancy expectations in general can be thought of as “desires constrained by reality” (W. B. Miller, 1994, p. 228) and based on personal resources and opportunity structures (Hendershot & Placek, 1981), split expectations specifically may represent the tipping of that desires-versus-reality balance. On one hand, an adolescent with split expectations may desire a pregnancy in the near future yet in reality not have the resources to support a child; on the other hand, another adolescent with split expectations may desire to avoid a pregnancy yet not have adequate access to effective contraception, making a future pregnancy a very real possibility.

Following the basic imagery of life course theory and using several of its principles to orient this investigation, pregnancy expectations can be subsumed within developmental, social, and institutional trajectories that translate early experiences into future outcomes (Crosnoe & Johnson, 2011; Elder, 1998). The principle of timing in lives—that the impact of events depends on when they happen—indicates when adolescents expect a pregnancy to occur relates to their expected sequencing of life events and when such a pregnancy comes to fruition has implications for other aspects of young people’s lives, such as schooling and employment. When pregnancy stalls or cuts off adolescents’ educational goals, for example, young mothers report more negative feelings about their pregnancy and parenthood (East & Barber, 2014). The principle of human agency—that individuals actively construct their own lives yet are subject to broader opportunities and constraints—suggests pregnancy expectations and childbearing are matters of personal choice as well as structural constraints, which can interfere with the ability to act on motivations while individual capacities are also capable of overcoming environmental forces on behavior. A life course perspective applied to adolescents’ norms against teenage pregnancy, for example, reveals that norms are shaped by individual- and neighborhood-level socioeconomic status, each of which independently predict the risk of teenage pregnancy (Mollborn, 2010).

This study’s first research question, therefore, is, how do adolescent pregnancy expectations—especially those that fall in between high and low—forecast early childbearing in young adulthood? This study sheds new light on an old question by incorporating adolescents’ split pregnancy expectations—as opposed to desires—to predict later fertility behavior.

Family, Academic, and Peer Group Predictors

As age at first birth in the U.S. population has increased, young people who expect to get pregnant in the near future are increasingly different from their peers in past, current, and future ways that map onto socioeconomic status, race/ethnicity, and other systems of stratification (Amato et al., 2008; McLanahan, 2004). Returning to life course theory, social pathways such as the educational institution and social convoys such as interpersonal relationships with parents and friends come together to shape adolescence (Crosnoe & Johnson, 2011). In this study, I argue stratification is likely to play out in those central domains of adolescent life—families, schools, and peer groups—to influence whether adolescents expect to follow the norm of delaying childbearing until after young adulthood. To be sure, early childbearing is neither antisocial nor inherently problematic (e.g., Geronimus, 1991), but drawing on perspectives from research on problem behavior can supplement the orienting life course perspective to explain why individuals follow or stray from societal conventions surrounding life transitions.

Social control theory (Hirschi, 1969) proposes that attachment to prosocial others and institutions, commitment to valued social relationships, and involvement in prosocial activities promote normative behavior, especially during critical periods such as adolescence. Taking delayed childbearing as a normative behavior, adolescents who are more attached, committed, and involved in their families, schools, and peer groups—central institutions and relationships in their lives—would be less likely to expect an early pregnancy. Similarly, the theory of reasoned action outlines a cost–benefit analysis of opportunities and sacrifices (Ajzen & Fishbein, 1980). If young people’s aspirations are reasonable and achievable, they have more to lose by engaging in behaviors that compromise those futures, whereas adolescents with no such aspirations have “nothing to lose” (Driscoll, Sugland, Manlove, & Papillo, 2005). Adolescents with higher opportunity costs would thus expect to delay childbearing whereas adolescents with a lack of costs would develop pregnancy expectations. Adolescents with neither high nor low (i.e., split) pregnancy expectations may represent youth for whom social control is present yet does not exert the level of influence required to lead them to absolutely want to delay a pregnancy for the time being. In other words, they may have “something to lose” but the cost–benefit analysis is not enough to fully push their expectations one way or another.

Previous research has highlighted the importance of looking across multiple domains of young people’s lives to identify risk and protective factors for teenage pregnancy (East, Khoo, & Reyes, 2006). Here, I apply that framework to pregnancy expectations. Within families, social control theory suggests that interpersonal processes increase the supervisory control of parents over children and make children more likely to accept that control. Indeed, the parent–child relationship and parents’ management of children’s time are associated with the risk of adolescent pregnancy (B. C. Miller, Benson, & Galbraith, 2001). Operationalization of this concept could include household routines and monitoring to measure how parents regulate children’s sexual activity, as well as parent–child closeness and parenting style that would influence whether children plan to follow their parents’ wishes.

In the educational domain, factors that increase adolescents’ stakes in the academic hierarchy of school give them more to lose in tangible rewards. Students with higher opportunity costs such as good grades and expectations to graduate college also expect to have children later in life (Plotnick, 2007) and are less likely to become teen parents (Driscoll et al., 2005). Conversely, students who are less attached to school—and less likely to buy into the legitimacy of school’s control over their lives—would have less to lose as a result of a pregnancy in young adulthood, a cost–benefit analysis that could be measured with suspensions or school mobility.

Adolescents’ peer groups, however, could offer an alternative rewards and status structure in which adolescents gain if they do things that would make them lose in other domains. In such structures, perceived popularity is a reward associated with alcohol use and sexual activity (Mayeux, Sandstrom, & Cillessen, 2008) and increased substance use and delinquency (Allen, Porter, McFarland, Marsh, & McElhaney, 2005), despite those behaviors being costs or risks at home or school. Given that risky behaviors such as delinquency, substance use, and early sexual debut independently predict adolescent pregnancy (Hockaday, Crase, Shelley, & Stockdale, 2000) and that they tend to co-occur, peer groups—especially romantic relationships and sexual activity—have the potential to influence whether adolescents expect to experience a pregnancy in the near future.

This study’s second research question, therefore, is, how do adolescents’ families, academics, and peer groups shape their outlooks on early childbearing across the spectrum of pregnancy expectations? I hypothesize adolescent pregnancy expectations will be higher and early childbearing more common among youth who have fewer routines, less monitoring, and lax or controlling parenting styles within their families; have lower grades and educational expectations as well as more suspensions and mobility at school; and engage in more romantic relationships, earlier sexual activity, delinquent behavior, and substance use with peers. Expectations and subsequent childbearing will be lower and less common among youth who experience more supervisory control at home, have higher stakes and investments in academics, and are not involved in alternative peer status structures. Based on past research on ambivalence, I hypothesize that adolescents who report split pregnancy expectations will follow early childbearing patterns similar to youth with high expectations, yet family, academic, and peer group predictors of those expectations will be weaker because they do not fully push adolescent expectations to one pole or another (i.e., neither high nor low). In sum, whether adolescents expect to experience a pregnancy at a young age is likely influenced by current experiences in their families, schools, and peer groups, which set them on life course trajectories that either present a consequential cost—benefit analysis of supports and opportunities or promote a “nothing to lose” perspective.

Method

Data and Sample

To address the two research questions, this study used the Bureau of Labor Statistics’ NLSY97, a nationally representative longitudinal data set that followed almost 9,000 respondents (N = 8,984 in 1997) from midadolescence into young adulthood, tracking their experiences and behaviors in the labor market, as well as in their families, schools, and peer groups. Each year beginning in 1997, targeted respondents, who were aged 12 to 16 years as of December 31, 1996, answered an electronic questionnaire using a computer-assisted personal interview system as part of a core sample, with an oversample of Black and Latino/Latina respondents.

The analytical sample was limited to respondents aged 16 to 18 years for practical and conceptual reasons. Pregnancy expectations were only asked of the full sample in the year 2000 survey when respondents were between the ages of 16 and 20 years. I excluded 19- and 20-year-olds because they represented a different life stage than adolescence. This exclusion also allowed me to use information from previous waves to create standard measures of predictors at age 15 before pregnancy expectations were captured in the year 2000, accounting for potential issues endogeneity. Among 16-year-olds, for example, data from the 1999 wave captured predictors at age 15, whereas survey data from 1997 was used for 18-year-olds. Focusing on 16- to 18-year-olds’ expectations for a pregnancy within the next 5 years also meant that such a pregnancy would occur well before the national average age at first birth of 25 in 2000 (Martin, Hamilton, Ventura, Menacker, & Park, 2002): the youngest respondents were 21 by the end of that window and the oldest ones were 23. Last, respondents who had children before the year 2000 survey (n = 251) were excluded from the analytical sample because their expectations, childbearing, and predictors were likely qualitatively different from adolescents who were not parents and related to other factors not included in the analytical models or measured in the NLSY97. Ancillary analyses were conducted without this exclusion and yielded substantively similar results.

These selection criteria resulted in an analytical sample of 4,556 adolescents with available data on their pregnancy expectations. Panel A in Table 1 describes the sample. Respondents were almost evenly distributed across the three age groups from 16 to 18 years old. About half of the sample was female, and more than two thirds of respondents were White, 15% were Latino/Latina, and 12% were Black. About 30% of the sample respondents had at least one parent who graduated college, 55% of respondents lived with both biological parents at age 15 years, and 22% of respondents were born to teen mothers.

Table 1.

Descriptive Statistics for All Study Variables (n = 4,556).

	Frequency, %	Mean (SD)
Panel A: Covariates
Age in years at 2000
16	33.45
17	33.43
18	33.12
Gender is female	48.08
Race/ethnicity
White	67.72
Latino/Latina	14.93
Black	12.41
Other race/ethnicity	4.94
College-educated parent(s)	30.40
Intact family structure at age 15 years	55.16
Maternal teen birth	22.43
Religion is Evangelical	28.14
Panel B: Family, school, and peer group predictors
Family predictors at age 15 years
Household routines		10.20 (5.35)
Parental monitoring		9.91 (3.22)
Parenting style
Uninvolved	14.14
Permissive	31.39
Authoritarian	15.48
Authoritative	38.99
School predictors
Grades in eighth grade		5.79 (1.75)
College expectations in 2000		52.60 (38.64)
Ever suspended by age 15 years	26.15
Number of schools by age 15 years		1.93 (0.74)
Peer predictors at age 15 years
Frequency of dating
Never this year	38.60
A few times	20.41
Less than once per month	12.60
Once or twice per month	14.01
Once per week or more	14.38
Delinquency scale		1.57 (2.02)
Substance use scale		1.16 (1.13)
Early sexual debut	17.95
Panel C: Pregnancy and childbearing outcomes
Pregnancy expectations for next 5 years
Nonexistent	34.99
Low	37.41
Split	17.88
Elevated	9.72
Any children within 5 years (i.e., early childbearing)	30.50

Note. Weighted descriptive statistics shown.

Measures

Pregnancy Expectations.

Respondents reported the percent chance of an event from 0 to 100 for a series of questions about their expectations for the future. To preface that series, the interviewer explained that: an impossible event would have a 0% chance; a possible but unlikely event would have a 3% or a 15% chance, for example; a pretty even chance would be 46% or 52%; a likely but not certain event would have a 78% or a 94% chance, for example; and a certain event would have a 100% chance (Bureau of Labor Statistics, 2002). Young men reported their chance of getting someone pregnant and young women reported their chance of becoming pregnant within the next 5 years, which I combined into one continuous measure.

As Figure 1 shows, however, the distribution of that measure did not follow the normal distribution of a continuous variable. Instead, pregnancy expectations were clustered around specific values similar to the groupings outlined in the interviewer prompt before the series of expectations questions (impossible, possible but unlikely, pretty even, likely but not certain, and certain). Informed by the shape and patterns in the distribution of adolescents’ expectations of an early pregnancy—which capture adolescents’ own descriptions of their chances—I created a categorical measure to meaningfully capture pregnancy expectations as nonexistent (0% chance), low (1%-49% chance), split (50% chance), and elevated (51%-100% chance).

Figure 1.

Distribution of adolescents’ expectations for early pregnancy.

Early Childbearing.

A binary variable indicated whether the respondent had a child between the 2000 and 2005 waves. Based on survey records of live births, respondents were coded as 1 if they became a first-time parent between 2000 and 2005 and coded as 0 if they were not parents by 2005.

Family Predictors.

Adolescent reports contributed to an index of household routines ranging from 0 to 28 that measured the number of days per week respondents did things together with their families (e.g., eating meals, homework, something fun, something religious). Additionally, Child Trends created several composite measures of respondents’ relationships with their parents, including monitoring (a continuous variable ranging from 0 to 16) and parenting style (a categorical variable for uninvolved, permissive, authoritarian, or authoritative). This study relied on data about respondents’ residential mothers primarily, using residential fathers if data on mothers were missing.

Academic Predictors.

Adolescents were asked about their overall grades received in 8th grade, with response options ranging from mostly below Ds (a value of 1) to mostly As (a value of 8). Similar to pregnancy expectations, the year 2000 survey gathered data on college expectations as respondents’ reported percent chance that they would be in school 5 years from the 2000 interview. Respondents also reported at the initial 1997 survey wave if they had ever been suspended and provided updates each subsequent wave, which were coded into a binary indicator of ever being suspended by age 15. With a similar baseline number and updates, respondents reported the number of schools they had attended by age 15.

Peer Predictors.

Adolescents reported their frequency of dating in the past year, with responses ranging from never (value of 1 representing both respondents who had never been on a date before as well as respondents who had not dated in the past year) to once or more per week (value of 5). To measure delinquency, I followed Child Trends’ index summing whether (1 = yes, 0 = no) youth had engaged in 10 criminal (e.g., selling drugs, stealing) and delinquent (e.g., running away from home, fighting) activities. As with suspension and number of schools, coding for delinquency relied on baseline reports and updates to measure number of activities by age 15 (range: 0-10). I used a similar coding scheme to create an index of substance use following Child Trends’ index of whether respondents had smoked a cigarette, drank alcohol, or used marijuana by age 15. Respondents who had ever engaged in sexual intercourse reported (in the 2000 survey wave) the age at which they first had sex. A binary variable measured early sexual activity, with respondents who had sex before age 15 years coded as 1 and respondents who had never had sex or who had sex for the first time at age 15 years or older coded as 0.

Covariates.

The following set of covariates accounts for characteristics potentially influencing youth’s pregnancy expectations, early childbearing, and predictors. Respondents reported their age at the date of interview in 2000 and their gender (1 = female) and their race and ethnicity (dummy variables for White, Black, Latino/Latina, and some other race/ethnicity) in the initial 1997 survey. Respondents also reported in the initial survey the religion in which they were raised, which I used to create a binary indicator of being raised evangelical Protestant. Reports of parents’ highest grade completed were used to measure at least one parent with a college degree (1 = yes, 0 = no). Family structure was measured as living with both biological parents at age 15 (a value of 1) versus in some other family structure (a value of 0). Adolescents reported maternal age at birth, from which I created a binary indicator for teenage birth.

Analytical Strategy

The first set of multivariate analyses used logistic regression to predict the odds of respondents having a child during the 5 years after reporting their pregnancy expectations. In addition to the full set of covariates, Model 1 included dummy variables for pregnancy expectations, with nonexistent as the reference group. Model 2 added the full set of earlier family, academic, and peer group predictors. A small number of cases (n = 75, 2% of the analytical sample) was lost between Models 1 and 2 because of missing values on the predictors added in Model 2. The second set of multivariate analyses then switched pregnancy expectations from predictor to outcome. Because of my decision to analyze adolescents’ expectations for early pregnancy as clustered categories, multinomial logistic regression was the most appropriate statistical technique. These models regressed the categories of expectations on each set of predictor variables and the full set of covariates. In all multivariate models, the Stata command mi estimate used multiply imputed data to create adjusted estimates of coefficients and standard errors for variability between imputations (StataCorp, 2015, p. 41).

Results

Sample Characteristics and Pregnancy Expectations

As a starting point, Panel B in Table 1 describes the analytical sample in terms of the three focal contexts. Within families, household routines were somewhat limited (M = 10.20 on an index with a maximum of 28), and parental monitoring was midrange with the index score slightly above the midpoint (M = 9.91 on an index with a maximum of 16). The most common parenting style was authoritative (39% of parents), followed closely by permissive (31%), and authoritarian and uninvolved the least common (16% and 14%, respectively). In terms of academics, adolescents reported, on average, receiving mostly Bs in eighth grade, and slightly more than half of respondents expected to be in college in 5 years. By age 15 years, 26% of adolescents had been suspended from school, and the average number of schools attended was close to two. In their peer groups, 39% of adolescents had never been on a date in the past year and 20% had only been on dates a few times. Delinquent and criminal activity was low (M = 1.57 out of a maximum 10 activities ever) and, on average, adolescents had ever used one out of three substances (cigarettes, alcohol, and marijuana). Last, almost one in five adolescents in the sample had sexual intercourse before age 15.

Turning to pregnancy expectations, Panel C in Table 1 shows that most adolescents had nonexistent or low expectations for pregnancy in the next 5 years, 35% and 37% respectively. Nevertheless, about 18% of adolescents reported a 50-50 chance of pregnancy (i.e., split expectations), and less than 10% of the sample reported elevated expectations. Looking forward, about 30% of young people did go on to have a child in that timeframe.

Pregnancy Expectations in Adolescence and Later Childbearing

To address the first research question of this study about how adolescent pregnancy expectations were associated with early childbearing in young adulthood, logistic regressions estimated childbearing within the 5 years after pregnancy expectations were assessed (see Table 2). In Model 1, net of all covariates, the odds of having a child within 5 years were 28% and 62% higher, respectively, among respondents who reported split or elevated pregnancy expectations relative to respondents with nonexistent expectations in adolescence. Ancillary analyses set each category of pregnancy expectations as the reference group, revealing that the association between pregnancy expectations and early childbearing was statistically significant across all category comparisons, except for low expectations and nonexistent expectations.

Table 2.

Odds Ratios From Logistic Regression Models Estimating Early Childbearing.

	(1)	(2)
Pregnancy expectations (ref: nonexistent)
Low	0.903 (0.074)	0.888 (0.075)
Split	1.283** (0.121)	1.145 (0.113)
Elevated	1.621*** (0.178)	1.281* (0.148)
Family predictors
Household routines		1.009 (0.008)
Parental monitoring		1.028* (0.013)
Parenting style (ref: authoritative)
Uninvolved		1.228† (0.141)
Permissive		1.073 (0.093)
Authoritarian		1.166 (0.130)
School predictors
Grades in eighth grade		0.864*** (0.020)
College expectations		0.996*** (0.001)
Ever suspended		1.342*** (0.115)
Number of schools		1.008 (0.050)
Peer predictors
Frequency of dating		1.070* (0.029)
Delinquency		0.987 (0.023)
Substance use		1.038 (0.041)
Early sexual debut		1.317** (0.123)
Constant	0.055*** (0.039)	0.101* (0.094)
Pseudo R2	.062	.095
N	4,448	4,373

Note. Odds ratios are presented. Standard errors are in parentheses. Data are unweighted. All models include controls for age, gender, race/ethnicity, parent education, family structure, maternal teen birth, and religion.

^†p < .1.

^*p < .05.

^**p < .01.

^***p < .001.

In Model 2, the addition of the full set of family, academic, and peer group predictors explained the baseline association between split pregnancy expectations and early childbearing. Although still a statistically significant predictor of early childbearing, the higher odds among respondents who reported elevated pregnancy expectations declined from 62% to 28% relative to nonexistent expectations, or a 55% decrease in magnitude of the effect size. Across predictors, only one family measure was statistically significantly associated with early childbearing (a positive association with parental monitoring), whereas multiple academic and peer group factors mattered. Academically, the odds of early childbearing decreased as both grades in eighth grade and college expectations increased. The odds of early childbearing were 34% higher, however, among respondents who had ever been suspended. In terms of peer groups, frequency of dating and early sexual debut were significantly (positively) associated with early childbearing yet delinquency and substance use were not.

To attend to the first research question, therefore, pregnancy expectations operated in a categorical rather than a linear way to forecast early childbearing: Relative to nonexistent expectations, only elevated expectations were consistently associated with a significantly greater risk of early childbearing, whereas family, academic, and peer group predictors explained the association between split expectations and childbearing. Additionally, most academic factors were associated with childbearing, yet only parental monitoring within families and only dating and sexual debut within peer groups mattered for early childbearing.

Family, Academic, and Peer Group Predictors of Pregnancy Expectations

The above results—that pregnancy expectations predicted early childbearing yet not uniformly across all categories of expectations—offered support for testing the second research question of this study: How do adolescents’ families, schools, and peer groups shape their pregnancy expectations across various levels and types of expectations? To answer that question, multinomial logistic regressions estimated expectations for experiencing a pregnancy within the next 5 years (nonexistent expectations set as the reference group) with each set of family, academic, and peer group predictors (see Table 3).

Table 3.

Relative Risk Ratios From Multinomial Logistic Regression Model Estimating Adolescent Pregnancy Expectations.

	Reference: Nonexistent expectations
	Low	Split	Elevated
Family predictors
Household routines	0.984† (0.008)	0.975* (0.010)	0.995 (0.012)
Parental monitoring	0.990 (0.013)	0.990 (0.016)	0.993 (0.019)
Parenting style (ref: authoritative)
Uninvolved	0.987 (0.125)	0.895 (0.140)	1.030 (0.186)
Permissive	1.143 (0.103)	1.101 (0.125)	1.178 (0.162)
Authoritarian	0.996 (0.119)	1.046 (0.151)	1.224 (0.209)
School predictors
Grades in eighth grade	1.067* (0.028)	1.006 (0.031)	0.962 (0.035)
College expectations	0.996*** (0.001)	0.992*** (0.001)	0.988*** (0.002)
Ever suspended	0.822* (0.082)	1.160 (0.131)	1.029 (0.136)
Number of schools	1.133* (0.065)	1.224** (0.081)	1.151† (0.090)
Peer predictors
Frequency of dating	1.049 (0.031)	1.074* (0.039)	1.089† (0.048)
Delinquency	0.967 (0.026)	1.038 (0.031)	1.075* (0.037)
Substance use	1.192*** (0.051)	1.122* (0.058)	1.134* (0.070)
Early sexual debut	0.899 (0.101)	0.955 (0.121)	1.502** (0.209)
Constant	0.014*** (0.014)	0.000*** (0.000)	0.000*** (0.000)

Note. n = 4,373. Pseudo R² = .056. Relative risk ratios are presented. Standard errors are in parentheses. Data are unweighted. All models include controls for age, gender, race/ethnicity, parent education, family structure, maternal teen birth, and religion.

^†p < .1.

^*p < .05.

^**p < .01.

^***p < .001.

For family predictors, household routines were the only component associated with pregnancy expectations, such that routines were significantly negatively associated only with the risk of reporting split expectations relative to nonexistent expectations. A one standard deviation increase in the household routines index was associated with 13% lower risk of split expectations compared with nonexistent expectations. That association was not statistically significant for low or elevated expectations relative to nonexistent expectations.

Next, each academic predictor was significantly associated with at least one category of pregnancy expectations. The risk of reporting low pregnancy expectations relative to nonexistent expectations increased as grades in eighth grade improved, yet that risk was 18% lower among adolescents who had been suspended. Relative to nonexistent pregnancy expectations, as college expectations increased, the risk of reporting any other pregnancy expectations category decreased. A 10-point increase in the percent chance of college attendance, for example, was associated with a 4% decrease in the risk of reporting low pregnancy expectations relative to nonexistent expectations, yet that decrease was three times bigger (12%) for the comparison between elevated and nonexistent pregnancy expectations. Last, number of schools attended was significantly positively associated with reporting low or split pregnancy expectations, an association which was greatest for split expectations. Attending one additional school was associated with 22% higher risk of reporting split pregnancy expectations relative to nonexistent expectations, compared with 13% higher risk of low expectations.

All peer group predictors were also significantly associated with pregnancy expectations, but with clear patterns differentiating the expectations categories. Frequency of dating was only statistically significantly associated with the (higher) risk of reporting split pregnancy expectations relative to nonexistent expectations. Delinquency and early sexual debut were also significantly associated with only one pregnancy expectations category; however, those peer group aspects were positively associated with the risk of elevated pregnancy expectations relative to nonexistent expectations. A one standard deviation increase in the delinquency index was associated with 15% greater risk of elevated expectations relative to nonexistent ones. The risk of elevated pregnancy expectations relative to nonexistent expectations was also about 50% higher among adolescents who had sex before age 15 years. In addition to the distinct trends for split expectations and elevated expectations, a general trend emerged for substance use, which was positively associated with all categories of expectations.

Contrary to my general hypotheses of higher pregnancy expectations among adolescents with less family supervisory control, lower academic stakes and investments, and involvement in alternative reward peer status structures, results suggest academics and peer groups mattered most, with interesting findings for the split pregnancy expectations category. The one significant family predictor (household routines) was associated only with reporting split pregnancy expectations relative to nonexistent expectations. For most academic measures, split expectations followed patterns similar to elevated expectations. One exception was school mobility mattered for both low and split pregnancy expectations relative to nonexistent expectations, yet not for elevated expectations. In peer groups, on the other hand, the split expectations category looked more like the low expectations category. The link between more frequent dating and the higher risk of split pregnancy expectations, however, was one exception to that general pattern. My hypothesis that family, school, and peer predictors would be weaker for split expectations was thus not supported across most of these results.

Discussion

As Americans increasingly push back having children until their mid to late 20s and early 30s, young people who do not follow that trend have become a distinct (and disadvantaged) group (McLanahan, 2004). This study looked further back in the life course to explore whether that differentiation in childbearing originates in adolescence, a crucial period when youth begin to make concrete plans for their futures (Johnson, 2005; Schneider, 2003). Informed by life course theory and theories of social control and reasoned action, I argued that adolescent pregnancy expectations are complex, shaped by earlier family, academic, and peer group predictors, and set youth on trajectories toward or away from pregnancy in young adulthood. The relevance of this study is that it illuminates how multiple configurations of expectations are rooted in one’s own disadvantage early in life and can translate into stratification in the next generation when the early childbearing linked to expectations occurs in the absence of strong supports and resources.

With special attention to potential nonlinearity, this study posed two research questions: (1) How do adolescent pregnancy expectations forecast childbearing in young adulthood? (2) How do adolescents’ families, academics, and peer groups shape the full spectrum of expectations? The overwhelming majority of this nationally representative sample of adolescents reported nonexistent or low pregnancy expectations, and far fewer reported elevated expectations. Yet about 18% of respondents reported a 50-50, or split, chance of becoming or getting someone pregnant within the next 5 years. Those expectations were important because almost one in three respondents became young parents, with the odds of early childbearing significantly higher among respondents with split or elevated expectations compared with nonexistent expectations. Overall, pregnancy expectations were more sensitive to academic opportunity costs and alternative peer group rewards. In investigating those research questions, two themes emerged.

First, split expectations (i.e., neither elevated nor nonexistent) represented a group of adolescents who were distinct in their risk of early childbearing as well as their family, academic, and peer group experiences. Adolescents with elevated or split pregnancy expectations were at greater risk of early childbearing, although that association between split expectations and fertility was explained by academics and peer groups. As far as predicting pregnancy expectations, the risk of reporting split relative to nonexistent expectations was distinct in several aspects of the family and peer domains, and most academic predictors were associated with a greater risk of low compared with nonexistent expectations. Informed by research on a middle ground of young people’s desires (e.g., ambivalence) about pregnancy and childbearing (Jaccard et al., 2003; W. B. Miller, 1994; W. B. Miller et al., 2013; Zabin et al., 1993), this study explored the full continuum of pregnancy expectations, a combination of plans and constraints. By adding another layer to how young people envision their future fertility—and how those expectations unfold in young adulthood—this study took steps toward understanding expectations that “may conform to a logic different from that familiar to the statistician … [but] must be based on a better understanding of the original logic that produced them” (Hendershot & Placek, 1981, p. 307).

Difficulty in determining the exact meaning of expectations for experiencing a pregnancy within the next 5 years is one limitation of this study. Qualitative research could offer a more nuanced picture of pregnancy expectations by using adolescents’ own descriptions of their experiences and outlooks on their futures. Additionally, future studies aimed at understanding pregnancy expectations should investigate the mechanisms (e.g., contraceptive use, abortion, sexual activity) that fully explain how expectations translate into childbearing. Adolescents engaging in unprotected sex, for example, may report elevated expectations because they do not want a pregnancy yet their behavior makes it a likely possibility. Nonetheless, this study makes a strong argument for a nonlinear conceptualization of pregnancy expectations.

Second, adolescents in the most precarious situations were the most likely to report higher pregnancy expectations and to experience early childbearing, with key differences. In general, adolescent pregnancy expectations predicted young adult parenthood. Family, academic, and peer group predictors, however, fully explained the association between early childbearing and reporting split compared with nonexistent pregnancy expectations, and they accounted for more than half of the association between early childbearing and elevated expectations. In other words, academics and romantic and sexual relationships influenced the translation of expectations into behavior for young people who may be uncertain about their futures yet not for those adolescents with definitively high expectations. Another measurement-related limitation of this study, however, is the category of elevated expectations was defined broadly as a 51% to 100% chance of experiencing a pregnancy within the next 5 years, and it was reported by less than 10% of the analytical sample. Future research should unpack the characteristics and fertility behavior of this particularly at-risk group and establish a threshold at which pregnancy expectations become most salient for early childbearing. Despite that limitation, this study’s findings suggest that policy efforts to reduce early childbearing should target adolescents with “nothing to lose” attitudes and greater expectations of becoming young parents, outlooks which tend to co-occur.

Furthermore, adolescents held such expectations because of academic factors that raised the opportunity costs an early pregnancy would present to their academic success and peer group factors that offered an alternative status structure rewarding behaviors associated with adolescent or early pregnancy. Interpersonal processes in the family that could increase supervision and shape the parent–child relationship in ways that might affect expectations about sexual activity and actual sexual behavior were not consistently associated with early childbearing or adolescent pregnancy expectations. Although parents’ influence on adolescent behavior and well-being should not be underestimated (see, Helsen, Vollebergh, & Meeus, 2000; Kandel, 1996), the results of this study suggest that youth’s fertility expectations and later behavior are strongly shaped by their schools and friends. Future research should thus acknowledge that expectations and behaviors surrounding early childbearing are more complicated than a simple calculation of risks and benefits and should seek to understand when and why certain domains and factors influence how young people anticipate their futures.

Adolescents’ expectations as well as their families, academics, and peer groups carry weight for how their futures will unfold, as this study has shown through the link between adolescent pregnancy expectations and early childbearing. However, pregnancy expectations—and the factors that influence those expectations—are complex, with split expectations representing a group of young people that is distinct in terms of who they are and what behavior they exhibit. This study’s findings that pregnancy expectations tend to translate into childbearing and that expectations are higher among more vulnerable adolescents highlight the need to consider the families, schools, and peer groups in which young people live their day-to-day lives. Investigation of multiple domains and various facets of those domains could identify risks as well as protective factors associated with early pregnancy (East et al., 2006). Importantly, this study suggests the need for pregnancy prevention programs targeting the most disconnected and uncertain adolescents, as well as additional resources for young parents who may be struggling.