In this article, we use newly available data from the Relationship Dynamics and Social Life (RDSL) study to compare a wide range of attitudes related to pregnancy for 961 black and white young women. We also investigate the extent to which race differences are mediated by, or net of, family background, childhood socioeconomic status (SES), adolescent experiences related to pregnancy, and current SES. Compared with white women, black women generally have less positive attitudes toward young nonmarital sex, contraception, and childbearing, and have less desire for sex in the upcoming year. This is largely because black women are more religious than white women and partly because they are more socioeconomically disadvantaged in young adulthood. However, in spite of these less positive attitudes, black women are more likely to expect sex without contraception in the next year and to expect more positive consequences if they were to become pregnant, relative to white women. This is largely because, relative to white women, black women had higher rates of sex without contraception in adolescence and partly because they are more likely to have grown up with a single parent. It is unclear whether attitudes toward contraception and pregnancy preceded or are a consequence of adolescent sex without contraception. Some race differences remain unexplained; net of all potential mediators in our models, black women have less desire for sex in the upcoming year, but they are less willing to refuse to have sex with a partner if they think it would make him angry and they expect more positive personal consequences of a pregnancy, relative to white women. In spite of these differences, black women’s desires to achieve and to prevent pregnancy are very similar to white women’s desires.
The unequal distribution of teen and unintended pregnancies by race is an important health-related disparity, alongside others, including infant mortality, life expectancy, and chronic disease (e.g., Williams and Jackson 2005). Black women have earlier first births (age 21 versus 24) and have larger families (2.4 versus 2.2) than white women (Martinez et al. 2012: tables 6 and 7). The teen pregnancy rate is more than twice as high for black women (43.9 per 1,000) than for white women (20.5 per 1,000) (Martin et al. 2013:Table A). The black-white disparities in unintended pregnancy are large. Black women evaluate 69 % of their pregnancies as unintended, while the corresponding number for white women is 42 % (Finer and Zolna 2014); and the unintended pregnancy rate for black women is more than double (92 per 1,000) the rate for white women (38 per 1,000) (Finer and Zolna 2014). The abortion rate is more than three times higher for black teens than for white teens (44.3 and 13.8 per 1,000, respectively) (Kost et al. 2010: table 1.0).
The U.S. Department of Health and Human Services, in its Healthy People 2020 goals, has identified the reduction of unintended pregnancy as a key nationwide health improvement priority. Understanding the substantial racial variation in unintended pregnancy is fundamentally important for its overall reduction, given the disproportionate share that occurs to black women. But the black-white disparity in unintended childbearing is puzzling: why do black women have so many more pregnancies that they evaluate as unintended, relative to white women? This article contributes to our understanding of this question by assessing black-white differences in pregnancy desires within the context of other attitudes related to pregnancy.
We focus on a particularly important period in the life course: the transition to adulthood. Although the proportion of pregnancies that are unintended is highest in the early teen years (e.g., 98 % for those under age 15 and 89 % for 15- to 17-year-olds), the highest rates of unintended pregnancy are concentrated in the late teens and early 20s. The rate per thousand is 42 for ages 15–17, 105 for ages 18–19, 101 for ages 20–24, and 69 for ages 25–29 (Finer and Zolna 2011). The data we use here—from the Relationship Dynamics and Social Life (RDSL) study—are based on a sample of 18- and 19-year-olds.
We focus on an unusually large set of attitude measures, drawn from 34 distinct questions asked of 961 young women in the RDSL. Other sources of survey data on the transition to adulthood, such as the National Longitudinal Study of Adolescent Health (Add Health), include many fewer measures of attitudes. Race differences in pregnancy-related attitudes—for example, toward sex, marriage, contraception, premarital childbearing—have been discussed in the qualitative literature (e.g., Anderson 1990; Burton 1990; Edin and Kefalas 2005; Levine 2013; Stack 1974), but little empirical research has systematically examined race differences in attitudes using population-based samples.1 Previous research about race differences in pregnancy desire itself has been somewhat inconsistent, sometimes finding that black women have more desire for pregnancy, and sometimes finding that black women have more ambivalence than their white peers (Abma et al. 2010; Jaccard et al. 2003; Schwarz et al. 2007).
We focus on (1) general attitudes, (2) individual desires, (3) expectations, and (4) willingness to engage in unplanned or undesired behaviors. As shorthand, we refer to all four concepts as “attitudes.” These concepts are drawn from the prototype/willingness model, a dual-processing model developed by Gibbons and Gerrard (1997) in part to explain adolescents’ risky behavior, such as sex without contraception. In this model, decisions are made in two ways. One decision pathway, similar to the theory of reasoned action (Ajzen and Fishbein 1980), is analytic and responsive to general attitudes and desires, which in turn influence individual expectations, intentions, and behavior. This is “deliberate” reasoning. The second decision pathway in the dual model is reactive, or “automatic.” The prototype/willingness model incorporates the concept of behavioral willingness, which is an indicator of openness to risky behaviors, to reflect this second pathway in order to predict whether even those individuals who do not have positive attitudes toward the risky behavior, do not want the risky behavior, and do not expect the risky behavior will nonetheless engage in the risky behavior in some situations.
Proximate Determinants of Pregnancy: Sex and Contraception
We investigate race differences in these concepts from the prototype/willingness model in the domains of sex, contraception, and pregnancy. Attitudes, desires, expectations, and willingness in these domains may affect pregnancy via their connection to pregnancy desires, or via sexual and contraceptive behavior. Desire to avoid pregnancy reduces risk of pregnancy (Miller 2011; Miller et al. 2010, 2013)—via both sex and contraception—by delaying or reducing the frequency of sexual intercourse among some women (i.e., the decision to delay serious relationships or to delay sexual intercourse within a relationship), and by increasing contraceptive use and consistency (Moreau et al. 2013). Further, independent of pregnancy desires, attitudes toward sex and contraception as well as other attitudes related to pregnancy may influence the risk of pregnancy more directly through women’s sexual and contraceptive behavior.
Determinants of Attitudes Toward Sex, Contraception, and Pregnancy
In the following subsections, we describe three sets of ideas about why we might expect black-white differences in attitudes related to sex, contraceptive use, and pregnancy: (1) family background and adolescent experiences, (2) economic opportunity and attainment, and (3) the legacy of medical experimentation and forced sterilization on low-SES, African American, and other minority populations in the United States.
Family Background and Adolescent Experiences
Attitudes toward sex, contraception, and pregnancy are formed at least partly during childhood, and childhood familial experiences vary significantly by race. In the United States, black children grow up with higher rates of teen and single parenthood (Martin et al. 2013; Smock and Greenland 2010), and black children experience more family instability (Abma et al. 2010; Fomby et al. 2010; Kennedy and Bumpass 2008; Raley and Wildsmith 2004; Smock and Greenland 2010). Furthermore, relative to white women, black women have earlier sexual debut (Martinez et al. 2011) and use contraception less effectively (Kusunoki et al. 2014). Collective socialization theories—focused on family and neighborhood role models—stress that exposure to these behaviors encourages young people to behave similarly and perhaps even to prefer these behaviors (e.g., Brewster 1994; Brooks Gunn et al. 1993; East et al. 2007, 2009; Trent 1994; Trent and Harlan 1994; Wilson 1987). In addition, young women who have sex, do not use contraception, and/or experience a pregnancy are likely to develop even more positive attitudes toward those behaviors because of cognitive dissonance, regardless of their attitudes before those experiences (Festinger 1957).
In the United States, black children grow up in substantially more religious families than white children (Steensland et al. 2000). The vast majority belong to historically black churches (i.e., evangelical denominations of Baptist, Methodist, or Pentecostal), which are particularly conservative on family issues, such as sex and contraception (Chatters et al. 2009; Lincoln and Mamiya 1990). Religious opposition toward sex and contraception may explain why black women have less knowledge about contraception and are less likely to use contraception (Frost and Darroch 2008; Frost et al. 2007; Guzzo and Hayford 2012; Mosher and Jones 2010; Rocca and Harper 2012; Shih et al. 2011), and may also explain differences in attitudes.
In addition, black-white differences in attitudes may be related to differences in union formation, particularly lower marriage rates among black women. High levels of unemployment and incarceration among black men, coupled with especially strong racial homogamy preferences among white men (Lin and Lundquist 2013), leave black women with fewer partnership options than white women (Bulcroft and Bulcroft 1993; Harknett and McLanahan 2004; Lichter et al. 1992; Pettit and Western 2004; Wilson 1987).2 Sex ratio theory suggests that those of the more abundant gender (women in this case) will lower their standards and accept mates and family formation strategies that they might otherwise consider unacceptable (Guttentag and Secord 1983; Tucker and Mitchell-Kernan 1995). A study of women at historically black colleges and universities found that women tolerated refusal to use condoms and nonmonogamous sexual behavior in their male partners specifically because of the paucity of potential male partners (Ferguson et al. 2006). In addition, skewed sex ratios in favor of women have been empirically linked to earlier first births, especially nonmarital births (South and Trent 1988). Coupled with black men’s particularly low desire for marriage (Anderson 1990; South 1993), this may translate into more positive attitudes toward nonmarital sex and parenthood among black women than among white women.
Economic Opportunity and Attainment
Black women experience more poverty, less education, more unemployment, lower-quality early education, more discrimination, more disadvantaged neighborhoods, and fewer opportunities for both education and employment, relative to white women (Avery and Rendall 2002; Conley 1999; Isaacs 2007; Oliver and Shapiro 2006; Orr 2003; Proctor and Dalaker 2002). Poverty itself may affect attitudes toward contraception: poor women are less likely to have insurance to cover the financial costs of contraception (Ebrahim et al. 2009) and thus may perceive barriers to its use. In addition to differences in the affordability of contraception, the concentrated poverty and disadvantage experienced by more blacks than whites might affect attitudes related to pregnancy for at least three reasons: uncertainty, opportunity costs, and structural disadvantage.
Uncertainty and instability of all types, which are endemic to concentrated poverty, may contribute to differences in attitudes between blacks and whites. Burton and Tucker (2009) elegantly described the instability and insecurity that are so pervasive in the lives of poor African American women: employment opportunities that are limited to intermittent and low-wage jobs, few alternatives (i.e., wealthy husbands) to reduce their breadwinner burden, transient living conditions, anxiety about serious relationships, and fear of death. Uncertainty about family formation, including about the long-term chances of a marriage surviving, is high for American women in general, and it is particularly high for poor women. Because children are viewed as an attainable path to stability for the women themselves and, they hope, for the fathers, children are valued and sought (Burton 1990; Edin and Kefalas 2005). Statistical analyses have also documented that there is, indeed, little perceived stigma for nonmarital childbearing among poor women (although they do not necessarily prefer to have babies before getting married) (Cherlin et al. 2008).
Lower opportunity costs for young black women than for white women may also lead to more positive attitudes toward early sex and pregnancy and less positive attitudes toward contraception (East 1998). Armstrong and Hamilton’s (2013) in-depth qualitative study found that even among those enrolled in a four-year college, young women with the strongest educational and career aspirations (and thus with the highest opportunity costs) purposely delayed serious relationships (and thus pregnancy) in explicit recognition that their opportunity costs of family formation would be high.
Finally, researchers have argued that structural disadvantages—such as lower-quality schools, fewer neighborhood associations, and less cohesive neighborhood networks in general—may lead to a set of “ghetto-related” attitudes and behaviors (see, e.g., Anderson 1990; Wilson 1996). Statistical analyses suggest that, indeed, the neighborhood economic conditions of blacks explain a substantial amount of variance in nonmarital pregnancy (South and Baumer 2000) and some attitudes related to early sex (Browning and Burrington 2006).
Legacy of Medical Experimentation and Forced Sterilization
Several potential reasons for race-based differences in attitudes toward contraception and (to some extent) pregnancy focus not on family background or socioeconomic explanations but instead on race itself. The Tuskegee Syphilis Study is one of the most well-known examples of unethical medical experimentation in the United States (Reverby 2009). Operated by the U.S. Public Health Service from 1932 through the 1970s, poor black men with late-stage syphilis were observed closely but were never treated for syphilis. Societal discourse about the study remains; in 1997, President Clinton offered a public apology to the men, their families, and the African American community.
There is also a strong history of reproductive abuse in the United States, and it is particularly relevant to black women. The federally funded, involuntary sterilization of poor women in the United States (and Puerto Rico) was documented as recently as the 1970s (Boonstra et al. 2000; Gibson-Rosado 1993; Malat 2000; Presser 1969; Roberts 1997; Stern 2005). Extremely common in the South, forced sterilization earned the nickname “Mississippi Appendectomy” (Roberts 2000; Washington 2006). Even more recently, between 2005 and 2013, many female prisoners in California received tubal ligations without proper informed consent (California State Auditor 2014). Further, the issue of whether poor women should bear children is still at the heart of ongoing welfare debates (e.g., see Harris and Wolfe 2014; Jencks 2001) and is disproportionately felt by minorities. These debates have been linked to the eugenics movement, whose goal was to “improve the inborn qualities of a race” (Galton 1904) through selective breeding and sterilization. Eugenics was quite popular in the United States in the late nineteenth and early twentieth century, prior to World War II (Osborn 1937). Within the black community, some argue that contraception is a “genocidal tool,” whereas others see it as a way to improve health and well-being in the black population (Roberts 2000).
Distrust of contraception and of medical professionals in general among the African American population was documented by many studies in the 1970s (e.g., Darity and Turner 1972; Farrell and Dawkins 1979; Schnittker 2004; Turner and Darity 1973) as well as in more recent studies (Rocca and Harper 2012; Thorburn and Bogart 2005). We expect this distrust to translate to more negative general attitudes toward contraception.
The Relationship Dynamics and Social Life (RDSL) study is based on a random sample of 1,003 young women (ages 18–19) residing in a Michigan county, drawn from driver’s license and state ID card records. A 60-minute face-to-face baseline survey interview was conducted between March 2008 and July 2009 by the professionally trained survey interviewers of the Survey Research Center at the University of Michigan’s Institute for Social Research. The interview assessed family background, demographics, attitudes, romantic relationships, education, and career trajectories. Respondents were paid $35 to participate in the interview. The incentive scheme, coupled with the cooperative nature of this age group and their interest in the subject matter, resulted in an 84 % response rate (93 % among located women). The main strength of this data set for the analyses presented here is its large and broad set of attitude measures. The data also include a relatively large population of young black women and a range of sociodemographic characteristics among both black and white women in the sample.
Column 1 in Table 1 provides the proportions (the mean for religiosity) for all independent variables included in these analyses. We analyze the 961 respondents with complete data on the independent variables described in the following subsections. Although these measures are correlated, the correlations are not high. The strongest correlation is between childhood public assistance and parental homeownership, which is only –.35.
Race was measured with the following question: “Which of the following groups describes your racial background? Please select one or more groups: American Indian or Alaska Native, Asian, Native Hawaiian or other Pacific Islander, Black or African American, or white.” Those who selected more than one group were asked a follow-up question: “Which of these groups, that is [groups chosen], would you say best describes your racial background?” For this follow-up question, 3 % of respondents did not choose a single race category. In all, 35 % of the sample reported their race as African American. A preceding question about Hispanic ethnicity yielded 69 Latinas, who were coded according to their answer to the race question; 28 selected African American, and 41 selected white. We recognize the limitations of this binary race measure: race is socially constructed and varies with time and location, and thus our dichotomy is an imperfect indicator. This substantially limits our ability to uncover nuance in these relationships—for example, variation within the self-identified black population, or the way dynamic and/or contextualized racial identities more subtly shape these attitudes. We hope that these statistical analyses motivate future analyses with more nuanced measures of race.
We use three indicators of childhood family background. The first is from the question, “How important if at all is your religious faith to you?” The mean score for this item was 2.7 on a scale from 1 (not important) to 4 (more important than anything else). Although this question refers to the present, not childhood, religiosity in young adulthood is highly related to the religiosity of the childhood home (Regnerus et al. 2004). On the second family background question, “How old was your biological mother when she had her first child?,” 37 % answered that she was younger than 20. The third question asked about whom the respondent primarily lived with while growing up. About one-half of the respondents reported growing up with two parents (either two biological parents, or one biological parent and one stepparent), 40 % lived with one biological parent only (no stepparent), and 8 % lived in some other arrangement (e.g., with grandparents or an aunt)
Childhood Socioeconomic Status
Respondents were asked a series of questions to assess childhood socioeconomic status (SES). In response to, “While you were growing up, did your family ever receive public assistance?” 36 % of respondents answered yes. When asked to report their mother’s highest level of education, 9 % reported less than a high school education, 34 % indicated high school graduate, and 57 % reported at least some college. In response to the same question regarding father’s highest level of education, 11 % reported less than a high school education, 41 % indicated a high school diploma, and 48 % reported at least some college. Because our analyses are stratified by race and the inclusion of both parental education variables results in some very small cross-tabulated cells, we use a dichotomous measure that combines maternal and paternal education. In the full sample, 66 % had at least one parent with at least some college education. In response to, “While you were growing up, did your parents or guardians own their own home?,” 71 % said yes.
Adolescent Experiences Related to Pregnancy
We use five indicators of past experiences related to pregnancy. Fifty-one percent of respondents reported that they were 16 or younger when asked, “How old were you the first time you had sexual intercourse?” When asked, “With how many total partners have you had sexual intercourse?,” 60 % reported two or more sexual partners. When asked, “Have you ever had sexual intercourse without using some method of birth control, such as condoms, pills, or another method?,” 48 % answered affirmatively. In response to, “How many times have you been pregnant in your life?,” 22 % reported a prior pregnancy, which was most often only one prior pregnancy. Finally, respondents were asked a series of questions about their current relationship status and living arrangements, which we used to create a measure with four mutually exclusive categories: currently married or engaged (9 %), cohabiting (10 %), dating (55 %), and no relationship (26 %).
We measure current SES with six questions. First, we asked for the respondent’s total income in the past 12 months with 16 categories ranging from <$1,000 to $25,000 or more. Because many of these 18- and 19-year-old women still lived with their parents and/or were still enrolled in school, their incomes are low. We divide this variable into four quartiles: <$1,000 (35 %), between $1,000 and $2,999 (21 %), between $3,000 and $5,999 (19 %), and ≥$6,000 (24 %). Respondents were asked, “At the end of the month, do you usually have some money left over, just enough money to make ends meet, or not enough money to make ends meet?”; 48 % said “some money left over,” 34 % said “just enough,” and 19 % said “not enough.” When asked if they owned a car, 49 % said yes. They were also asked, “Are you currently receiving public assistance from any of the following sources? WIC (Women, Infants, and Children Program), FIP (Family Independence Program), cash welfare, or Food Stamps.” In all, 27 % of respondents indicated “yes” for at least one category of public assistance. Finally, because respondents were sampled at age 18 or 19, many were still enrolled in school and few had completed any postsecondary education. When asked, “Are you going to school at all now?” and “Did you graduate from high school, get a GED, or neither?,” 58 % reported being currently enrolled full-time, and 77 % reported completing high school.
Columns 2 and 3 of Table 1 present these proportions separately for black and white respondents. Bold numbers indicate the independent samples t test (religiosity) or chi-square tests (all other variables) that revealed statistically significant differences. The vast majority of variables differ significantly by race, with the black sample having more disadvantaged family background and childhood SES, riskier adolescent experiences related to pregnancy, and more disadvantaged current SES.
The dependent variables are constructed from 34 measures of attitudes toward sex, contraception, and pregnancy. The specific question wording, response options, valid N, range, mean, and standard deviation for each measure are presented in Table 2.3 Regarding sex and contraception, we include four types of measures: general attitudes, desires (personal preferences), expectations, and willingness. The RDSL data set does not include a measure of willingness to become pregnant; thus, for pregnancy, we include measures of attitudes, desires, and expectations. Because we allow item-specific missing data for each measure, the valid N ranges from 919 to 961. Of the 961 respondents, 822 (86 %) had answers to all of the attitude questions, 7 % were missing one answer, 4 % were missing two answers, and only 3 % were missing three or more answers.
In addition, although Table 2 presents each individual measure, we use theoretical reasoning and exploratory factor analysis to combine some of the measures into scales. For sex, the five measures of general attitudes form a strong scale, with an eigenvector of 2.10 and all factor loadings greater than .30. For contraception, the 10 measures of general attitudes form a strong scale, with an eigenvector of 3.85 and all factor loadings greater than .37. For pregnancy, there are nine measures of expectations for the consequences of pregnancy and one measure of overall expectation of pregnancy. We keep the overall expectation measure separate to facilitate comparisons with expectations in the other two domains. The nine items about personal consequences reveal a strong scale, but one measure does not fit well with the scale: “If you got pregnant now, your family would help you raise the child.” Thus, we leave that measure separate but include it in our analysis because expectation of family support is sometimes proffered as an explanation for higher rates of teen pregnancy among black women (e.g., see Burton 1990; McDonald and Armstrong 2001). Eigenvalues and factor loadings are presented for each domain in Table 4 in the appendix.
We perform two types of analyses with these data. First, we conduct two-tailed independent samples t tests on each attitude measure to assess whether the mean response differed for black and white respondents. Those results are also presented in Table 2.
Second, we estimate a series of Oaxaca-Blinder decomposition models, which are widely used to study mean differences between groups.4 For example, economists developed these models to decompose gender differences in wages (Blinder 1973; Oaxaca 1973), and sociologists previously used similar models to decompose race differences in wages (Duncan 1969). We use these models to decompose the racial differences in attitudes into two parts: one part that is explained by differences in individual-level characteristics (e.g., family background and childhood SES), and another part that is unexplained (and is due to the effects of those and other variables on attitudes).5
We use a model-building approach, adding independent variables in blocks. We include them in approximately temporal order: childhood family background and childhood SES, followed by adolescent pregnancy-related experiences, and then current SES. However, the temporal ordering is not precise. The measures of family background and childhood SES may be thought of as interrelated rather than temporally ordered. In addition, adolescent pregnancy-related experiences and current SES may not be exogenous of attitudes at ages 18 or 19, if earlier attitudes influence those experiences and are also related to later attitudes. We conducted three sensitivity tests. First, we added the childhood SES variables to the models before the family background variables. The results did not differ from those shown in Table 3. Second, we estimated the models without current SES. Those results did not differ from those in Table 3. Finally, we estimated the models without adolescent pregnancy-related experiences. We discuss those models in the text that follows. The extent to which the overall race difference in the mean is explained by the independent variables can be thought of as explanations for the race difference in attitudes, rather than signaling that there are no race differences.
Race Differences in Attitudes
Table 2 presents means of attitude measures separately by race, with bold numbers indicating differences that are statistically significant at the .05 level.
The black young women in our sample hold significantly less positive attitudes about and have less desire for young nonmarital sex than white women. Black women tend to view premarital sex as wrong, to more strongly agree that they are not ready for a sexual relationship, and to express less desire for sex in the upcoming year.6 However, in spite of these relatively negative feelings, they report less willingness than white women to refuse sex if doing so would make their male partner angry.
Race differences in women’s attitudes toward contraception are more complicated. Although nearly all the attitude measures differ significantly by race, compared with white women, black women have more positive attitudes toward some aspects of contraception and less positive attitudes toward others. On average, black women perceive greater access to contraception than do white women: fewer consider birth control expensive, fewer believe that they cannot afford birth control, and more consider birth control easy to get. However, black women generally hold less favorable attitudes toward using contraception, more strongly agreeing that it takes too much planning to have birth control available, is too much of a hassle to use, and makes women feel sick. Finally, more black than white women perceive moral dilemmas with contraception—that requesting condom use raises issues of trust in the relationship, that birth control is morally wrong, and that girls who use contraception are “looking for sex.”7 Black and white women have a similarly strong desire to use birth control if they do have sex, but black women perceive a 7 % higher probability than white women of having sex without contraception in the coming year.
Black women tend to have more negative attitudes about nonmarital childbearing in general than white women, but desire for pregnancy and desire to avoid pregnancy are very similar across race. However, black women expect more positive personal consequences of a pregnancy. More black than white respondents believe that pregnancy would reduce their loneliness, that they could handle the responsibility, that it would make their partner happy, and that it wouldn’t be all that bad to get pregnant. Fewer black than white women think that they would have to grow up too fast, that they would have to quit school, or that they could not afford to raise the child. Black and white women are similar in terms of expectations for family support.8 Despite these overall more positive expectations among black women and national trends suggesting that black teen’s pregnancy rates are about twice as high as white teen’s (Zolna and Lindberg 2012), black respondents do not differ from white respondents in their expectations of a pregnancy during the upcoming year.
Multivariate Oaxaca-Blinder Decomposition Models
Table 3 presents the Oaxaca-Blinder decomposition models for the race differences in attitudes. Column 1 presents differences in means between blacks and whites (same as in Table 2), along with standard errors and statistical significance of the differences. Columns 2 through 5 present the decompositions, by blocks of the aforementioned independent variables: family background, childhood SES, adolescent experiences related to pregnancy, and current SES. These numbers represent the amount of the mean difference that is explained by race differences in individual characteristics, in terms of the specific blocks of independent variables. Columns 6 and 7 present the total amounts and percentages, respectively, of the mean race difference that is explained by race differences in individual characteristics in the model. The numbers in column 8, representing the unexplained portion of the mean difference, are equivalent to the coefficients for race in an OLS regression model including all of the independent variables; these numbers indicate the extent to which the race differences are net of the individual characteristics represented by the independent variables in the model, and their statistical significance indicates whether there is a significant race difference net of the race differences in the explanatory variables in the model.
Table 3 demonstrates that family background—mainly religiosity (not shown in tables; see Table 5 in the appendix)—explains much of black women’s less positive general attitudes toward young nonmarital sex and lower desire to have sex in the next year, relative to white women. Current SES (mainly lower income and educational attainment; not shown in tables) further explains black women’s lower desire for sex in the upcoming year.
Adolescent pregnancy-related experiences, although they are statistically significant variables in the models, do not explain differences in attitudes or desires for sex because those experiences are associated with more positive attitudes and more desire, but black women are more likely both to have those experiences and to be less positive toward and less desirous of nonmarital sex. Thus, the positive number indicates that black women have even less positive attitudes toward young nonmarital sex and are less desirous of sex than we would expect, given their adolescent pregnancy-related experiences.
The independent variables explain 77 % of the mean difference in general attitudes toward young nonmarital sex, leaving no statistically significant race difference net of those variables. Race differences in the desire for sex in the upcoming year, however, are statistically significant net of these potential explanatory factors, which explain only 40 % of the race difference. None of the independent variables—either clustered or individually—explain why black women are less willing than white women to refuse sex.9
Family background and current SES explain much of black women’s overall more negative general attitudes toward contraception compared with white women. Religiosity and income are key (not shown in tables): black women are more religious and poorer (see Table 1), and more religious and poorer women have more negative attitudes toward contraception than the less religious and wealthier. The model explains 91 % of the mean race difference in general attitudes toward contraception.10
In contrast, black women’s higher expectation for sex without contraception in the upcoming year, relative to white women’s, is largely explained by adolescent pregnancy-related experiences. Specifically, black women were more likely than white women to have had sex without contraception in adolescence, which translates into higher expectations for sex without contraception in the future. The model explains 67 % of the mean race difference. If we omit adolescent pregnancy-related experiences from the decomposition model, the race difference is not as well explained, with a borderline significant race difference remaining net of the explanatory factors (total explained = 2.27, p > .10; total unexplained = 4.61, p < .10). However, the race difference in expectations might actually be net of adolescent experiences if different prior expectations were causally related to those experiences.
African American women’s less positive general attitudes toward nonmarital childbearing, relative to white women’s, are almost fully explained by family background—again, specifically by religiosity. Adolescent pregnancy-related experiences, as was the case for attitudes toward sex, do not explain the race difference in general attitudes toward nonmarital childbearing. Black women are more likely than white women to have experienced a pregnancy in adolescence, but that experience is associated with more positive general attitudes toward nonmarital childbearing.
Finally, if they were to get pregnant, young black women expect more positive (or fewer negative) personal consequences than white women, which is substantially explained by family background and adolescent experiences—namely, growing up with a single parent and having sex without birth control.11 If we exclude adolescent pregnancy-related experiences from the decomposition model, the explanatory role of family background increases slightly (.14, p < .01), the explanatory role of current SES increases dramatically (.10, p < .01), and the total portion of the race difference that is explained increases slightly (.24, p < .001). (If both sets of variables are omitted, the total explained portion of the race difference remains similar, with more explanatory power shifted to family background.) This suggests that compared with white women, black women expect more positive consequences of a pregnancy because of experiences in childhood, which influence subsequent experiences in adolescence and early adulthood.
Young women—both black and white—espouse moderate attitudes toward nonmarital sex in adolescence, desires for a sexual relationship, and expectations of having sex. They report a relatively strong willingness to refuse sex with a partner, even if it would make him angry. Their overall attitudes toward contraception are positive. They have negative attitudes about unmarried childbearing in the abstract, have very low desire for pregnancy, have very high desire to avoid pregnancy, and perceive a very low chance that they will get pregnant in the upcoming year. However, they have strong expectations that their family would help them raise the child if they had one, and overall they perceive only moderate negative personal consequences of pregnancy. These attitudes are largely nonconducive to pregnancy.
We found consistent black-white differences in attitudes toward sex, contraception, and pregnancy, which may partly explain why young black women’s pregnancy rates, and particularly their unintended pregnancy rates, are higher than white women’s. Compared with white women, black women have less positive attitudes about sex and are less desirous of sex in the coming year, but they are also less willing to refuse to have sex with a partner if it would make him angry. Compared with white women, black women have more negative attitudes toward contraception and are more likely to expect to have sex without contraception in the upcoming year. Young black women have more negative general attitudes toward pregnancy but anticipate more positive consequences of becoming pregnant at this time in their lives. Together, this constellation of race differences in attitudes is consistent with a path toward earlier pregnancy for black women.
If they are less willing to refuse sex, black women may be more likely than white women to have sex even when they don’t desire pregnancy and don’t have contraception available. In addition, young black women’s perceptions of contraceptive use as difficult, fraught with side effects, and morally questionable may contribute to their lower use of contraception, use of effective methods, and consistent use of contraceptives compared with white women (Jacobs and Stanfors 2013). Finally, although both black and white young women state a strong desire to avoid pregnancy and very little desire to become pregnant, young black women’s greater tolerance for the personal consequences of pregnancy may reduce their commitment to implementing those desires. Through these three processes related to sex, contraceptive use, and pregnancy desire, even these small but consistent differences in attitudes may accumulate into substantial behavioral differences that could explain some of the disparity in unintended pregnancy rates.
Future research should address the extent to which attitudes explain racial disparities in behavior. Although recent research by Rocca and Harper (2012) demonstrated that perceptions about contraception (safety, side effects, and overall knowledge) do not explain race differences in contraceptive use, the attitudes toward contraception examined in the current study may have more explanatory power. For example, because condoms are the most commonly used method among black women, their perception that asking a partner to use a condom is a signal of distrust may explain why they are, on average, less consistent contraceptors (Kusunoki et al. 2014). In addition, young black women’s negativity toward nonmarital sex may explain why they tend to use a method that requires less advance planning (condoms) rather than more effective methods, such as birth control pills. Our findings that relative to white women, black women perceive contraception to be less expensive and easier to access but to require more planning and to interfere more with pleasure suggest that black women may be answering these questions in reference to condoms, while white women may be more likely to be thinking of oral contraceptive pills (which are more expensive, harder to access, require less planning, and do not interfere directly with pleasure).
We found substantial support for our ideas about religiosity—that black women may have more negative attitudes toward sex, contraception, and nonmarital pregnancy because they are more religious than white women. Religiosity, however, does not seem to play a major role in black women’s expectations of having sex in the next year, their willingness to refuse sex, or their perceptions of the personal consequences of pregnancy. We also found support for the role of socialization. Growing up with a single mother partially explains black women’s more positive expectations of the personal consequences of pregnancy. Although we did not directly test ideas about collective socialization, we suspect that black women’s tolerance of the personal consequences of a pregnancy at this young age may be due in part to mothers’ experiences or the experiences of extended-family role models or neighbors. The role, albeit small, of income and educational attainment in explaining black women’s more negative attitudes toward contraception than white women’s is consistent with ideas about the role of race and poverty. However, we cannot determine whether it is the instability, lower opportunity costs, or structural disadvantages associated with poverty, or poverty itself, that is related to attitudes toward contraception. Finally, we found that young women who have had sex without contraception in the past expect to have sex without contraception in the future and expect more positive personal consequences of a pregnancy. Of course, it is unclear whether these expectations preceded sex without contraception or instead sex without contraception produced these expectations. We suspect that both processes are at work.
In addition, some key race differences in attitudes toward sex, contraception, and pregnancy remained statistically significant even after we accounted for family background, socioeconomic characteristics, and adolescent pregnancy-related experiences. Black women have less desire for sex in the next year, are less willing to refuse sex if it would make their partner angry, and have more positive attitudes about the potential personal consequences of a pregnancy—all regardless of their family background, childhood or current SES, or adolescent experiences. We did not provide evidence for or against the role of sex ratios in producing these unexplained attitudinal differences. However, we speculate that they may be important, given empirical research demonstrating that black women both notice and respond to low sex ratios of men to women (Ferguson et al. 2006). Census data show that in U.S. counties with at least 50,000 African Americans, the sex ratio for 18- to 19-year-old women is 107 men per 100 women for whites versus 100 men per 100 women for blacks (authors’ calculations). Sex ratios become smaller and the race disparities become wider with age: 99 versus 87 for ages 20–24, and 100 versus 81 for ages 25–29 (authors’ calculations). Given that women tend to date older men and that white men have strong preferences for racial homogamy in dating (Lin and Lundquist 2013), it is plausible that these differences may produce different attitudes for black and white women. Future research should explore whether differing sex ratio contexts are associated with attitudes or behaviors related to sex, contraception, and pregnancy, and whether those differences further explain the race differences described here.
We also did not provide evidence for or against the role of past medical experimentation and forced sterilization in producing these unexplained differences in attitudes toward contraception. Although previous research confirms that black women are more likely than white women to hold suspicious beliefs about contraception (Thorburn and Bogart 2005), we do not know why. We speculate that this history may play a role. This is an important topic for ongoing and future research.
In spite of attitudes that are relatively nonconducive to pregnancy, low levels of desire for pregnancy, and low expectations of pregnancy, about one-fifth of the 961 women analyzed here became pregnant during the subsequent 2.5 years: 25 % of the black women and 17 % of the white women (authors’ calculations). As other researchers have concluded, there is a gap between young women’s stated desires and subsequent behavior, and the gap is larger for black women than for white women (Finer and Zolna 2011). Others have speculated that black women may want their pregnancies but may feel societal pressure to label them as unintended because social norms place higher value on white fertility than on black fertility (Harris 2010; Harris and Wolfe forthcoming). We found no evidence here that the questions used to assess attitudes—including pregnancy desires—are interpreted differently by black and white women. Factor analyses revealed that the different attitude measures are correlated similarly for whites and blacks (not shown in tables), standard deviations of the measures are similar for black and white women (see Table 2), and independent variables are related to the attitude measures in similar ways for black and white women (not shown in tables). However, race differences in the other attitudes explored here might explain why black women have more unintended pregnancies, even if intentions for black and white women hold similar meaning, by understanding them within a broader personal context: if black women are less willing to refuse sex and are more negative about contraception, they may be more likely to become pregnant even without desire for a pregnancy. If they perceive more positive (or fewer negative) consequences of pregnancy, they may be less committed to implementing their pregnancy desires. We do not interpret this to mean that their pregnancy desires are qualitatively different. Future research must continue to explore reasons for this race gap in unintended pregnancies and to explore whether black and white women’s pregnancy intentions are meaningfully different.
Finally, we found a pattern of race differences in attitudes specifically related to partners. Black women are less willing to refuse sex if it would make their partner angry, more strongly believe that asking a partner to use a condom signifies distrust, and have greater expectations that their partner would be happy about a pregnancy. Considering male partners’ desires when assessing a pregnancy’s intention status may explain why some unintended pregnancies occur. That the attitudes related to partners represent the largest black-white differences we found in our analyses suggests the importance of this consideration for understanding race differences in unintended pregnancy. It also suggests that, at least in this age group, there may be important differences between the partners of white and black women. Future research should investigate these differences—for example, whether black women are more likely than white women to have romantic partners who desire pregnancy, or whether black men are more negative than white men toward women requesting condom use.
The present study has important limitations. The narrow geographic focus (a single county in Michigan) of the RDSL study is notable. However, although the sample is not nationally representative, Michigan falls around the national median in measures of cohabitation, marriage, age at first birth, completed family size, nonmarital childbearing, and teenage childbearing (see Lesthaeghe and Neidert 2006). More important, the county has a large black population (about 35 %), and the proportion of black residents in the major city within the county is even higher.12 The United States has 65 cities that are at least 25 % black, comprising at least 10 million of the 39 million black residents in the United States. Thus, the women in the RDSL sample live within a similar racial composition as many African Americans in the United States. On the other hand, the study includes only a small number of Latinas, who were classified as either white or black in our analyses—a limitation that we hope motivates future researchers to implement similar studies on larger and more diverse populations.
More importantly, this statistical portrait of black-white differences in attitudes ignores much of the nuance inherent in each of these concepts. Race is not a simple dichotomous variable. Attitudes are not easily reduced to Likert-scale questions. We have not identified variations in these overall patterns and have not uncovered anything about the young women whose attitudes do not fit these patterns. Thus, this analysis does not, for example, provide much insight into the processes that produce variation in attitudes within race (Jarrett 1997). The RDSL includes follow-up data on the young women summarized here, along with more than 70 in-depth, semi-structured interviews with black and white women. We consider the current analysis a necessary first step in the process of understanding these race differences in attitudes and related behaviors.
This research was supported by two grants from the National Institute of Child Health and Human Development (R01 HD050329, R01 HD050329-S1, PI Barber), a grant from the National Institute on Drug Abuse (R21 DA024186, PI Axinn), and a population center grant from the National Institute of Child Health and Human Development to the University of Michigan’s Population Studies Center (R24 HD041028). The authors gratefully acknowledge the Survey Research Operations (SRO) unit at the Survey Research Center of the Institute for Social Research for their help with the data collection, particularly Vivienne Outlaw, Sharon Parker, and Meg Stephenson. The authors also gratefully acknowledge the intellectual contributions of the other members of the original RDSL project team: William Axinn, Mick Couper, Steven Heeringa, and Yasamin Kusunoki, and the Advisory Committee for the project: Larry Bumpass, Elizabeth Cooksey, Kathie Harris, and Linda Waite. Finally, we thank Lisa Neidert for her expertise in computing the sex ratios with census data, and the anonymous reviewers for important feedback that greatly improved this article. We are particularly grateful to the Editor, who expertly and generously shepherded this article through an extensive review process.
Because women tend to date older men, even young black women whose cohort-mates do not yet face high rates of imprisonment face a relative shortage of partners.
Some of these measures were replicated from the National Survey of Family Growth Cycle 6 (contraception attitude 1.7 and pregnancy attitude 1), from the National Survey of Adolescent Health Waves I and II (sex attitudes 1.5 and 2; contraception attitudes 1.1, 1.3, 1.4, 1.5, 1.9, and 1.10; and pregnancy attitudes 3.1.1, 3.1.2, 3.1.5, 3.1.6, and 3.1.7), and the Intergenerational Panel Study of Parents and Children (sex attitude 1.1). Other items were created specifically for the RDSL study, some with explicit reference to the prototype/willingness model (Gibbons and Gerrard 1997).
We also estimated OLS models for each attitude measure and added sets of independent variables in the same way they are entered into the Oaxaca-Blinder models (see Table 5 in the appendix). We tested whether decreases in the coefficients across nested models were significant using the method described in Clogg et al. (1995). The results of those analyses were overall quite similar, and the coefficient in the full OLS model (including all controls) was identical to the unexplained part of the race difference after all characteristics were included in the Oaxaca-Blinder decomposition models.
The mean race difference can also be decomposed into three parts with the Oaxaca-Blinder method: race differences in endowments (values for the independent variables), race differences in the coefficients, and the interaction between race differences in endowments and coefficients. We present the two-part decompositions for four reasons. First, our hypotheses concern endowments; we have no hypotheses about how family background, adolescent experiences related to pregnancy, or SES differences would have different attitudinal consequences for blacks and whites. Second, there are very few significant race differences in the coefficients for the independent variables, with three exceptions: (1) although religiosity leads to more negative attitudes about sex, this is less true for blacks than for whites; (2) being in a relationship increases expectations of sex for whites, but this is less true for blacks; and (2) whites with two prior pregnancies have less negative attitudes toward contraception and more positive attitudes toward pregnancy than those without prior pregnancies, while blacks with two prior pregnancies have more negative attitudes toward contraception and less positive attitudes toward pregnancy than those without prior pregnancies. Third, for attitudes toward contraception and pregnancy, race differences in individual-level characteristics contribute dramatically more to the mean race differences in attitudes than race differences in coefficients or the interaction between the two types of race differences. In the case of attitudes toward sex, however, race differences in coefficients are more important. We describe those exceptions in the Results section.
We also examined differences between the 571 white and 205 black nonpregnant 18- and 19-year-old women in the 2006–2010 National Survey of Family Growth (NSFG). There were no race differences in responses to the items about sex: “Any sexual act between two consenting adults is all right”; “It is all right for unmarried 18-year-olds to have sexual intercourse if they have strong affection for each other”; and “It is all right for unmarried 16-year-olds to have sexual intercourse if they have strong affection for each other.”
In the NSFG, young black women expected less embarrassment than young white women in response to the question, “What is the chance that it would be embarrassing for you and a new partner to discuss using a condom?” There were no race differences in responses to two questions: (1) “What is the chance that if your partner used a condom during sex, you would feel less physical pleasure?” (a question somewhat similar to the RDSL contraception question 7, except that the RDSL question refers to “birth control” rather than condoms); and (2) “What is the chance that if a new partner used a condom, you would appreciate it?” Thus, NSFG analyses may indicate slightly more positive attitudes toward condoms among young black women relative to young white women; the only RDSL measure that focuses specifically on condom use is item 1.8.
In the NSFG, young black women expressed more positive attitudes than young white women in response to, “If you got pregnant now, how would you feel?” but there was no race difference in response to, “It is okay for an unmarried female to have a child” (similar to RDSL pregnancy question 1). Thus, compared with young white women, young black women have more positive attitudes about the personal consequences of a pregnancy in both data sets but have similar (NSFG) or more negative attitudes (RDSL) about single parenthood in general.
We also estimated three-part Oaxaca-Blinder decomposition models in which the explained difference includes differences attributable to race differences in individual characteristics, race differences in coefficients for the independent variables (i.e., race interactions in the pooled model), and the interaction between race differences in characteristics and race differences in coefficients. These models showed that the race differences in coefficients in models of desire to have sex and willingness to refuse sex account for more of the mean race difference in attitudes than did the race differences in the individual characteristics. However, in a pooled model, none of the coefficients are statistically different for blacks and whites. Thus, the explanatory power of the race differences in coefficients is due to the accumulation of many small race differences that are individually indistinguishable from zero.
However, if the nine measures of general attitudes toward contraception are modeled separately, a more varied picture of race differences emerges. Net of all mediating factors in our models, black women find contraception cheaper and easier to access than white women; in addition, however, they find contraception to be more hassle to use and more likely to make women sick, and they believe that asking for condom use signals distrust. Thus, although the factor analysis strongly suggests that these items measure a single underlying construct for both black and white women, there are clearly race differences within that single construct.
The difference that is net of the mediators in the model is largely driven by net differences in responses to three items: (1) “If you got pregnant now, you would have to quit school,” (2) “If you got pregnant now, your partner would be happy,” and (3) “If you got pregnant now, you could not afford to raise the child.”
We do not give the percentage in order to protect the anonymity of the study location.