Predictors and Consequences of Adolescents' Norms Against Teenage Pregnancy

Abstract

African American and Latino teenagers and communities are frequently assumed to have weaker norms against teenage pregnancy than whites. Despite their importance, adolescents' norms about teenage pregnancy have not been measured or their correlates and consequences documented. This study examines individual-level and contextual variation in adolescents' embarrassment at the prospect of a teenage pregnancy and its relationship with subsequent teenage pregnancy. Descriptive analyses find that norms vary by gender and individual- and neighborhood-level race, ethnicity, and socioeconomic status (SES). In multivariate analyses, neighborhood-level racial/ethnic associations with embarrassment are explained by neighborhood-level SES. Embarrassment is associated with a lower likelihood of subsequent teenage pregnancy but does not mediate racial, ethnic, or socioeconomic influences, underscoring the importance of both norms and structural factors for understanding teenage fertility.

ACKNOWLEDGMENTS

The author thanks Cecilia Ridgeway, Susan Olzak, Ross Macmillan, Robert Crosnoe, Christine Horne, Jason Boardman, Fred Pampel, Justine Tinkler, and the members of the Stanford Social Psychology and Youth and Community Development workshops for their helpful comments.

NOTES

Notes

¹ This study uses neighborhoods, operationalized as census tracts, as a proxy for teenagers' local contexts. While other contexts such as peers and schools are also important for adolescent sexual behaviors and norms (Teitler and Weiss 2000), the neighborhood contains a variety of relationships and institutions in which adolescents are embedded and which frequently encompass peer relationships and schools.

² While norms may also differ among other racial and ethnic groups besides African Americans, Latinos, and whites, this study reflects the focus in the literature on racial/ethnic cultures, as well as addressing pragmatic concerns about subsample sizes for other races, by comparing these three groups.

³ Less than 8 percent of eligible cases were missing data on any variable in the analyses of variation in norms. Missing data on any variable was negatively associated with reporting embarrassment at the prospect of a teenage pregnancy (p < .01).

⁴ Compared with the earlier sample, Wave III included higher proportions of female and nonblack respondents who were in younger grades at Wave I. Wave III weights used in the analysis adjust for such patterns of nonresponse, and analyses have shown that bias from nonresponse is very low for a variety of outcomes (Chantala, Kalsbeek, and Andraca n.d.). My supplemental analyses found that compared with those who did not respond in Wave III, a higher proportion of respondents was embarrassed at the prospect of a teenage pregnancy, a lower proportion was black and a higher proportion was white, a lower proportion was poor and a higher proportion had a high household income, and respondents came from neighborhoods with lower proportions of black residents and higher neighborhood SES.

⁵ Because Add Health only includes a mean of seven and a median of two respondents per census tract (Cubbin et al. 2005), accounting for the slight clustering of respondents within tracts using this method is reasonably accurate (see Harris, Duncan, and Boisjoly 2002 for a similar approach).

⁶ Booth, Rustenbach, and McHale (2008) compared biological children listed in the roster of household members with respondents' reported births and found that births appeared to be somewhat undercounted in Add Health. Therefore, our count of teenage pregnancies is likely an underestimate.

⁷ Responses were coded into approximate years of education: no schooling = 0, eighth grade or less = 8, some high school = 10, trade/vocational/business school instead of high school = 11, high school graduate or GED = 12, trade/vocational/business school after high school = 13, some college = 14, college degree = 16, graduate/professional training = 18. The highest education levels of the mother and her spouse/partner were averaged using parent reports. If the spouse's education was missing, I substituted the teenage respondent's report of his education level, and lacking that, the mother's education stood in for both. If no parent completed the survey, the adolescent respondent's report of both parents' education levels was substituted. Cubbin et al. (2005) found 75 percent agreement between parents' and adolescents' reports of parental education levels when data for both were available.

⁸ Finding suitable items was difficult because among whites, African Americans, and Latinos, race/ethnicity and neighborhood SES are so intertwined that there is little overlap in the distributions of neighborhood SES indicators across these three groups. Without overlapping distributions, analysis is not meaningful (Brewster 1994b). Following Brewster, I included only items for which each group's (whites', African Americans', and Latinos') median fell within the 20th and 80th percentiles of each of the other two groups. This criterion led to the exclusion of four measures of neighborhood SES that did not have sufficiently overlapping distributions across racial/ethnic groups: the proportion of respondents receiving welfare income, the unemployment rate, the proportion of residents aged 25 or older without a high school degree, and the proportion of families living below the poverty line.

⁹ Standardized effect sizes from bivariate logistic regression analyses (not reported in tables) illustrate the relative size of the associations between individual-level racial and socioeconomic factors and embarrassment at the prospect of a teenage pregnancy. Among girls, socioeconomic factors were more important than race/ethnicity. An increase of one standard deviation in parents' education was associated with an approximately 50 percent greater likelihood of reporting embarrassment at the prospect of a teenage pregnancy. Among boys, this socioeconomic relationship was even greater at about 70 percent, but they were joined by a similarly strong negative association between African American racial identity and reported embarrassment.

¹⁰ Here and elsewhere, respondents' odds or likelihoods of reporting embarrassment at the prospect of a teenage pregnancy are based on odds ratios, which exponentiated the coefficients reported in the tables.

¹¹ Criteria for mediation followed Baron and Kenny (1986): The addition of the mediating variable must reduce or eliminate the relationship between the independent and dependent variables, the mediating variable must significantly predict the dependent variable, and the independent variable must significantly predict the mediating variable.

¹² Was race/ethnicity or SES a more important predictor of embarrassment at the prospect of a teenage pregnancy, and was the neighborhood or individual level more important? Supplemental analyses based on Model 2 that used standardized variables to compare effect sizes found that for both genders, two predictors had the largest, roughly equally sized effects: individuals' African American race and parental education. The Hispanic and neighborhood SES coefficients were smaller, and neighborhood racial/ethnic composition was not significant.