The national evaluation of Title V, Section 510 abstinence education programs collected survey data on study youth over a four to six year period, depending on the year that they began to participate.В At the time of enrollment in the study, youth were of middle school age or youngerВ in most cases, too young to be sexually active.В Over the course of the evaluation, they aged into mid-to-later adolescence, when many youth are making decisions about their own sexual activity.В To gain insight into the unfolding of these decisions over time and the effect of program participation on these decisions, the evaluation has included analyses of both short-term and longer-term program impacts.
A previous DHHS study report found that the four focal programs had an impact on several of their intended short-term outcomes, which were hypothesized to lower rates of teen sexual activity (Maynard et al. 2005).В В Most notably, relative to their peers in the control group, youth in the program group reported views more supportive of abstinence and less supportive of teen sex, and they demonstrated a heightened awareness of the possible negative consequences of teen sex.В Program group youth were also significantly more likely than youth in the control group to make formal pledges to abstain from sex until marriage.
This chapter explores two potential explanations for the apparent inconsistency between short-term impacts on outcomes believed to be predictive of abstinence and the lack of longer-term impacts on abstinence:В (1) these outcomes failed to affect, or mediate, sexual abstinence as hypothesized, and (2) the short-term impacts on these outcomes were simply too small or did not persist for long enough to have an impact on eventual sexual activity.
Notably, while the chapter provides insight into the links between potential mediators and sexual abstinence, it cannot establish causality.В An observed relationship between a mediator and sexual abstinence might reflect the effect of unobserved factors correlated with that mediator rather than the causal impact of the mediator itself.В For example, peer pressure could have a causal effect on sexual abstinence, in which case peer pressure and sexual abstinence would be correlated.В But an observed correlation between peer pressure and sexual abstinence could also arise from youth with an unobserved propensity to engage in sexual activity selecting into peer groups in which peer pressure is high.В The analytic approach, presented below, cannot disentangle these two explanations for any correlations between potential mediators and sexual abstinence.
The evaluation draws on a rich longitudinal data set that includes multiple measures of attitudes and other possible mediators of youth behavior as well as of their behavioral outcomes. These data not only allow the analysis of program impacts over time, but also enable us to examine the pathways through which programs might have affected behaviors.
The logic model for the evaluation, presented in Chapter I and reproduced below (Figure VI.1), presents the pathways through which program effects were hypothesized to occur. Programs aimed to alter the level and nature of services youth received in ways that would influence potential mediators of teen sexual activity. Examples of these potential mediators include youth views toward abstinence, their relations with peers, and their perceived consequences of teen sex. The first year impact report examined program impacts on receipt of services (Box C) as well as on several potential mediators (Box D). The earlier chapters of this report estimate program impacts on long-term behavioral outcomes (Box E). This chapter focuses on the potential links between selected mediators (Box D) and teen sexual behavior (Box E).
Logic Model for Evaluating the Impact of Title V, Section 510 Programs
The relationships between mediators and sexual abstinence are estimated using a multiple regression model. The outcome of interest whether youth have remained abstinent (measured at the time of the final follow-up survey) is regressed against a set of covariates that measure several potential mediators of abstinence. These mediators are based on data from an initial follow-up survey, conducted six to nine months after youth enrolled in the study. Findings from the regression thus provide an estimate of whether a potential mediator of behavior, such as relations with peers, does in fact predict whether youth have abstained from sex three to five years later (the period between the initial and final follow-up surveys).
The analysis focuses on five groups of potential mediators, all measured from the initial follow-up survey data. They include (1) youth views toward abstinence, sex, and marriage; (2) peer influences and relations; (3) self-concept, refusal skills, and communication with parents; (4) perceived consequences of teen sex; and (5) pledging to abstain from sex. Ideally, the analysis would be expanded to include the other potential mediators of teen sex shown in Figure VI.1, such as knowledge of STD and pregnancy risks. However, measures of these potential mediators were not collected from the initial follow-up survey.
Two potential mediators from the initial follow-up survey views supportive of abstinence and friends support for abstinence were significantly predictive of reported sexual abstinence on the final follow-up survey (Table VI.1). Specifically, youth reporting views more supportive of abstinence were more likely to report abstaining from sexual intercourse on the later survey. The magnitude is large; a one-unit increase in the measure is associated with an eight percentage point increase in the likelihood of remaining abstinent (p-value = 0.01). Likewise, having a network of close friends who are supportive of abstinence was strongly associated with increased sexual abstinence. A one-unit increase in support for abstinence among friends is associated with a five percentage point increase in the likelihood of remaining abstinent.
Links Between Potential Mediators and Later Sexual Abstinence
||Change in Rate of Abstinence for
One Unit Change in Potential Mediator
|Views on Abstinence, Teen Sex, and Marriage
|Support for abstinence
||0-3 [least to most supportive]
|Support for teen sex (reversed)
||0-3 [most to least supportive]
|Support for marriage
||0-3 [least to most supportive]
|Peer Influences and Relations
|Friends support for abstinence
||0-5 [least to most supportive]
|Peer pressure to have sexa
||0-3 [least to most pressure]
|Self-Concept, Refusal Skills and Communication with Parents
|Self-esteem and control
||0-3 [lowest to highest level]
||0-2 [lowest to highest skills]
|Communication with parents
||0-3 [least to most communication]
|Perceived Consequences of Teen and non-Marital Sex
|Perceived personal consequences
||0-3 [least to most consequences]
|Perceived general consequences
||0-2 [least to most consequences]
|Pledged to abstain
||0 or 1 [yes or no]
|Sources: The measures of potential mediators and the measure of sexual abstinence are based, respectively, on the Wave 2 and Wave 4 Surveys of Teen Activities and Attitudes (Mathematica Policy Research, Inc., 2000, 2005) administered to youth 6 to 12 months and 42 to 78 months, after enrollment in the Title V, Section 510 Abstinence Education Program study sample.
Note: See Appendix E for complete information on these measures. All estimates are adjusted based on weighted regression models. The estimated change represents an association between the two measures and should not be interpreted as causal, since it might be explained by other, unmeasured factors.
aTeens in Control and FUPTP samples were not asked the questions used to construct these measures because of their young ages at the time of the Wave 2 survey. As a result, these estimates are based on a model pooling data across only the two older sites. All other estimates are based on a model pooling data across all four sites.
***p-value (of change shown) < 0.01; **p-value < 0.05; *p-value < 0.10, two-tailed test.
Of the remaining potential mediators, none is associated with sexual abstinence in the direction hypothesized in the logic model (Table VI.1). One measure, support for marriage, has a negative association with sexual abstinence, which is inconsistent with the logic model. The remaining measures including self-concept, refusal skills, and communication with parents; perceived consequences of teen sex; and pledging all bear no statistically significant association with later sexual abstinence. Perhaps the most surprising of these findings concerns the pledge, which two previous studies (Bearman and Bruckner 2001; Rector et al. 2004) found to be associated with delayed sexual initiation but this study finds to have no statistically significant association with later sexual abstinence.
Given that support for abstinence by youth and peer support for abstinence are the only significant long-term predictors of sexual abstinence found in this study, the remainder of this chapter focuses on these two measures (defined in Table VI.2) and how they changed over time. For findings on the other measures shown in Table VI.1, see Appendix E.
Several potential mediators of teen sexual abstinence commonly addressed by Title V, Section 510 program curricula are found to have no association with sexual abstinence three to five years later. Notable among these are self-concept, refusal skills, and communication with parents; perceptions of negative consequences from teen sex; and pledging to abstain from sex. Two other potential mediators are found to be significantly associated with future sexual abstinence: youth support for abstinence and their friends support for abstinence. Although the analysis cannot determine whether either of these associations is causal, findings suggest that promoting support for abstinence, both among youth and their friends, should be an important feature of future abstinence programs.
The programs evaluated in this report had at most a small impact on support for abstinence in the short term, and they had no impact on support for abstinence in the longer term. However, levels of support among both program and control group youth did change significantly over time. For example, at the end of their first school year in the study sample (the time of the first follow-up survey), most program group youth reported having a majority of friends supportive of abstinence. But by the time of the final follow-up survey when most program youth had entered middle to late adolescence and all youth had completed the programs only a small fraction had maintained this high level of peer support.
 See Appendix Table E.1 for definitions of these measures. Pledging to abstain is shown in Box C of the logic model (services received) because it is often a component of programs curricula; however, the act of pledging may act as a mediator of future behavior, making it relevant for this analysis.