Youth with socioemotional problems initiate sexual intercourse at younger ages than other youth.1–10 This association holds for problems as diverse as antisocial behavior,1,2,6 depression,2–5,7 attention deficit hyperactivity disorder6 and low self-esteem,8–10 and has been observed in countries other than the United States.2,6,7 Because adolescence is a time of peak onset for mental health problems,11,12 their possible influence on sexual behavior deserves sustained attention from researchers and policymakers.
Two key limitations of prior research, however, diminish confidence in the general conclusion that socioemotional problems are associated with early sexual initiation. First, only a few studies1,2,6 have evaluated the associations of different types of problems with sexual initiation simultaneously, even though socioemotional problems often co-occur in adolescence. For example, youth who have high levels of internalizing problems (forms of emotional distress that are turned inward, such as depression, anxiety and excessive dependency) also tend to have high levels of externalizing problems (forms that manifest in acting-out behavior, such as disobedience, aggression and impulsivity).13,14 Unless studies consider both types of problems simultaneously, the independent contribution of each to early sexual initiation cannot be determined.
Second, prior research has not assessed gender, racial or ethnic differences in the associations. The social context of and normative expectations for sexual activity differ dramatically for young men and women, and for youth of various racial and ethnic backgrounds.15–18 These differences modify the associations of traditional predictors, such as parental support, with sexual initiation,5,19–21 and it is reasonable to hypothesize that they also modify the associations of socioemotional problems.
Our study examines two research questions that address these limitations. First, do internalizing and externalizing problems have independent associations with early sexual initiation? Second, do associations of internalizing and externalizing problems with early sexual initiation differ by gender or by race or ethnicity?
Socioemotional Problems and Sexual Initiation
A number of studies support the conclusion that a wide range of socioemotional problems may predispose youth to engage in sexual intercourse at early ages. Several studies have found associations between internalizing problems and early sexual initiation. For example, a cross-sectional analysis of the same data set used in this study found that older adolescents with high levels of -depressive symptoms are more likely than youth with low levels to have ever had sex,22 a result that has been replicated with other data sets.23,24 Depression also predicts sexual initiation prospectively: Youth who report high levels of depressive symptoms at one point in time are more likely than those who report low levels to subsequently engage in intercourse,3 as are youth with a diagnosed depressive disorder.2 Findings for externalizing problems are similar. Conduct disorder, attention deficit hyperactivity disorder and other externalizing problems have been associated with early sexual initiation both cross-sectionally25 and prospectively.1,6
However, in the absence of evidence from studies that consider multiple types of socioemotional problems simultaneously, these findings remain inconclusive. Estimating the associations of different kinds of problems is important for two reasons. First, identifying the types of problems that are most strongly associated with early sexual initiation provides an indirect test of competing explanations for the association of socioemotional problems with sexual behavior, as these specific types serve as proxies for the mechanisms implied by these explanations.
For example, one dominant explanation, based on problem behavior theory,26 asserts that delinquency, early sexual initiation and substance use form a syndrome of problem behaviors that has its origins in unconventionality.25 According to this theory, youth with “unconventional” personalities—those who have a high tolerance for deviance, who value independence over achievement and who do not expect to achieve academic success—are drawn to risky behaviors,27 particularly when they live in “unconventional” environments, characterized by lack of parental support, low religiosity and friends who have few parental controls.25 If the problem behavior explanation is accurate, antisocial behaviors (such as breaking things deliberately) and headstrong behaviors (such as arguing and being disobedient) would be expected to have stronger associations with early sexual initiation than such behaviors as hyperactivity and impulsivity. In contrast, if hyperactivity and impulsivity have stronger associations with early sexual initiation than antisocial and headstrong behaviors, problem behavior theory would not appear to explain the association of socioemotional problems with sexual initiation.
Another common explanation attributes associations of socioemotional problems with early sexual initiation to psychological deficits related to these problems, such as feelings of hopelessness2 and psychological neediness or dependency.7 According to this explanation, depressed youth are motivated to engage in sexual intercourse to gain esteem or to satisfy a need for belonging. Poor judgment, associated with depression and impulsivity, has also been proposed as an explanation for early sexual initiation.2,3 If psychological deficits explanations are accurate, depression, dependency and impulsivity would be expected to have stronger associations with early initiation than antisocial and headstrong behaviors.
A second reason to estimate the independent associations of different kinds of problems is that the types of problems implicated in early sexual initiation inform the development of preventive interventions by identifying appropriate target populations for and points of intervention. If, for example, antisocial behavior has stronger associations with early initiation than does depression or dependency, interventions would be most effective if oriented toward teaching youth positive and productive behaviors. In contrast, if depression has the stronger association with early initiation, programs that address the unique concerns and needs of depressed youth would be called for.
A key consideration in evaluating the associations of socioemotional problems with early sexual initiation is the relative importance of these problems in comparison with other predictors, such as socioeconomic status, gender, and race or ethnicity. Limited resources dictate that programs be directed toward problems that yield the greatest potential payoff in the reduction of STDs and teenage pregnancies. In this study, we gauge the importance of the distinct types of socioemotional problems not only by the significance of their associations, but also by their influence on the predicted probability of early sexual initiation.
Differences by Gender and Race
Sexual intercourse has different precursors and meanings for men and women, as well as for members of different racial and ethnic groups. For example, black adolescents hold more favorable attitudes toward premarital intercourse than do white adolescents.28–30 In addition, gender norms encourage sex for young men and discourage it for young women,17 and young men perceive themselves to have more decision-making power than their partners once they have become sexually active.31
These differences manifest themselves in differences in the associations of family and peer experiences with sexual decision making. In general, the less normative sexual intercourse is within a group, the stronger these associations. For example, school attachments are associated with a reduced risk of sexual intercourse for young women but not for young men,5,19 and the same is true for parental support.19 Furthermore, friends’ problem behaviors and lower levels of parental involvement are more strongly associated with precocious sexuality among white youth than among black or Hispanic youth.20,21
The limited existing evidence suggests that socioemotional problems may follow a similar pattern, but the evidence is far from conclusive. Depression has been more strongly associated with early sexual activity among young women than among young men5,22,24 (although not always23). However, the only study of conduct disorder that we could locate found a stronger association with early activity among young men than among young women.6 Because depression is more common among young women and conduct disorder is more common among young men, these findings suggest that sex--stereotypical problems predict sexual activity more strongly than do sex-atypical problems. Fewer studies of racial differences have been conducted, and the only one we found showed that depression had a stronger association with sexual debut among young white women than among young black women.3
We used data from the prospective Children of the National Longitudinal Surveys of Youth (NLSY); these data sets include measures of socioemotional problems at ages 10–11, and hence allow for the analysis of whether such problems are associated with sexual initiation before age 15. The comprehensive nature of these surveys makes it possible to control for correlates of socioemotional problems and sexual behavior that might otherwise render associations spurious.
The NLSY was designed to trace the labor market experiences of several nationally representative cohorts of men and women longitudinally. One cohort of women entered the study in 1979, when they were aged 14–21. In 1986, when these women were aged 21–28, the first of a series of developmental assessments was conducted with each of their children. Subsequent assessments have been conducted biennially; children were added to the sample as they were born. Assessments before age 10 relied on -interviewer-administered instruments and maternal reports; at ages 10–14, youth completed self-administered questionnaires on specific topics, and beginning at age 15, they were interviewed directly.
Because of the structure of the NLSY data collection, children in the earliest waves were born to disproportionately young mothers. To increase the representativeness and size of the sample, the present analysis combined samples of youth who were 10–11 years old at the surveys in 1992, 1994 and 1996; the combined sample consisted of 2,587 children. We focused on this age-group to ensure that the measures of socioemotional problems were derived from a point in time close, but prior, to early sexual initiation. Supplemental analyses confirmed the absence of cohort differences in the results. Forty-two youth who reported having initiated sex at age 10 or 11 were excluded from the analysis.
•Socioemotional problems. The measures of socioemotional problems were based on the Behavior Problems Index, a modified version of Achenbach’s Child Behavior Checklist.31 Mothers responded to 28 items by indicating whether they were “often true,”“sometimes true” or “not true” of their 10–11-year-old (scored as three points, two points and one point, respectively).* The items represented six types of problems: depression or anxiety (five items—e.g., “he or she feels worthless or inferior”), dependency (four items—e.g., “he or she is too dependent on others”), peer problems (three items—e.g., “he or she has trouble getting along with other children”), headstrong behavior (five items—e.g., “he or she argues too much”), hyperactivity (five items—e.g., “he or she has difficulty concentrating, cannot pay attention for long”) and antisocial behavior (six items—e.g., “he or she bullies or is cruel or mean to others”).
We used eight measures of socioemotional problems. The first six correspond to the subscales listed above (Cronbach’s alphas, 0.59–0.75), and two were summary scale scores calculated for internalizing and externalizing problems; items for these scales were chosen on the basis of previous factor analyses of these data.32 The internalizing scale consisted of seven items from the depression or anxiety and dependency subscales (alpha, 0.73). The externalizing scale comprised 15 items from the headstrong behavior, antisocial behavior, hyperactivity and peer problems subscales (alpha, 0.87).† Each scale score was calculated by standardizing the sum of the individual items to a mean of 0 and standard deviation of 1 in the complete sample of eligible youth. All items were coded so that higher scores indicated higher levels of problems.
Using these eight measures offers two advantages. First, although many of the subscales are highly correlated, the two summary scales reduce multicollinearity in the model. Second, the subscales conform closely to some of the specific explanations that have been given for the associations between socioemotional problems and sexual initiation. For example, the antisocial behavior subscale conforms to problem behavior interpretations, the dependency subscale to psychological neediness and the hyperactivity subscale to impulsivity.
•Early sexual initiation. We defined early sexual initiation as sexual intercourse before age 15. Although many studies have defined early initiation as first sex before age 16, the NLSY sample dictated a more restrictive definition, as 44% of these youth had had sex prior to turning 16. Because the cutoff point for early sexual initiation varied across previous studies, we also estimated models that used cutoffs of 14 and 16 years; results were similar to those reported here.
Age at sexual initiation was taken directly from youth reports. At each interview beginning at age 13, youth were asked whether they had already had intercourse and, if so, at what age they had first had sex. Because some gave inconsistent reports of their age at first sex at different surveys, the measure of early initiation was based on youths’ first reports, under the assumption that these were subject to fewer recall errors.‡
•Control variables. We controlled for several variables that have been associated with socioemotional problems and with the likelihood of early sexual initiation: the child’s gender and race or ethnicity (black, Hispanic or other); the mother’s age at the child’s birth, education level (in years) and marital status (married, never-married or formerly married, which included separated); and the proportion of years in which the child had lived in poverty up to age 10. All control variables were based on mothers’ reports at the time the child was 10 or 11 (the same year as the measures of socioemotional problems).
We used t tests to evaluate the bivariate associations between early sexual initiation and socioemotional problems and control variables. Multivariate logistic regression models estimated the associations of all of these variables with early initiation simultaneously. The first set of models included the two types of socioemotional problem (i.e., internalizing and externalizing problems) and the controls; subsequent models included the specific socioemotional problems and controls. All analyses were performed using Stata version 9.2. Only the 1,836 cases with valid values on all variables in the models were included. However, we repeated all analyses with the full sample of age-eligible youth, using mean-imputed missing values, and the substantive findings did not change.§
We extended the logistic regression models in two ways: by calculating predicted probabilities and by estimating models for group differences. In logistic regression models, coefficients and odds ratios convey the direction and significance of associations, but offer little insight into their relative contributions to the probability of experiencing the outcome. To gauge these contributions, we used estimated coefficients from the regression models to calculate the predicted probability that youth would initiate sex early. For characteristics represented by dummy variables (e.g., gender, race or ethnicity, marital status), we -calculated probabilities for youth with each value of the variable. For continuous variables, we calculated probabilities for youth with low and high values, defined as one standard deviation below and above the mean, respectively. In these calculations, all other variables in the model were held at their means.
To assess gender and racial or ethnic differences in associations, we used a modification of traditional logistic regression techniques, called heterogeneous choice regression. The most common approach to estimating group differences in a regression modeling framework is to calculate multiplicative interactions between dummy variables for group membership (e.g., female) and the other predictors in the model (e.g., externalizing problems), and to include those interactions as predictors. In logistic regression models, however, multiplicative interaction terms cannot be used to estimate the size and significance of group differences in effects, because these differences are confounded with group differences in residual variation (i.e., variation not accounted for by the predictors).33 We followed a procedure developed by Williams,34 which yields accurate estimates of group differences by estimating and then adjusting for unequal residual variation across groups; it uses the ordinal generalized linear model subprogram in Stata with a logit link function.35 The models identified predictors of residual variation using a stepwise procedure, and then adjusted for those predictors to estimate group differences in associations between socioemotional problems and early sexual initiation.
The sample was almost evenly divided by gender (Table 1). Thirty-six percent of the youth were black, 23% were Hispanic and 42% were of another race or ethnicity. On average, youth had spent 39% of their lives in poverty before age 10, which was a higher proportion than for all U.S. youth at the time these data were collected.36 Women in the sample were more likely to be formerly married than were U.S. women in the same age-group in 1995 (28% vs. 18%),37 were younger than average at the time their child was born (22 vs. about 26 years for women in 1984)38 and had had fewer years of schooling (12 vs. the median of 13 years in 1993).39 Overall, 26% of youth in the sample had had sexual intercourse before age 15 (not shown); this rate was higher than the 13% observed for the U.S. teenage population in 200240—a reasonable comparison year. Finally, the mean scores on the measures of socioemotional problems were all close to zero because of the standardization of the variables.
Youth who initiated sex before age 15 had a significantly higher average level of each socioemotional problem at age 10–11 than youth who initiated sex at older ages (0.09 to 0.27 vs. –0.02 to –0.10). The two groups were also different regarding social and demographic characteristics: Young women were underrepresented among youth who had initiated sex early (44% vs. 55%), and blacks were overrepresented (43% vs. 33%). Those who had early sex had spent a greater proportion of their lives in poverty than had other youth (50% vs. 35%), and were more likely to have an unmarried mother (57% vs. 38%); in addition, their mothers had less schooling than mothers of those who delayed first sex (mean, 11.6 vs. 12.2 years).
When the scales for internalizing and externalizing problems were considered separately, both were associated with early sexual initiation, after social and demographic characteristics were controlled for. A one-unit (i.e., one-standard-deviation) increase in internalizing problems was associated with a 14% increase in the odds of early sexual initiation (Table 2, page 96); a one-unit increase in externalizing problems was associated with a 22% rise in the odds of early sex. When both types of problems were included as predictors of early initiation, only the increased likelihood for externalizing problems remained significant (odds ratio, 1.2). Several of the social and demographic characteristics were significant in all three models: Being female and mother’s years of schooling were associated with a reduced likelihood of early initiation, while being black and having a mother who had never married or who was formerly married were associated with an elevated likelihood of early sex.
The same general pattern held with respect to the subscales for specific types of socioemotional problems (Table 3, page 97). When considered separately, all except -dependency and peer problems were associated with early sexual initiation, and the relative odds were similar to those observed for the internalizing and externalizing scales. The increase in the odds of early initiation associated with a one-unit increase in socioemotional problems ranged from 14% (for headstrong behavior) to 22% (for hyperactivity and antisocial behavior). When all of the types of socioemotional problems were included in the model, the only ones associated with early sexual initiation were two externalizing problems: hyperactivity and antisocial behavior (odds ratio, 1.2 for each).**
The difference in the predicted probabilities of having had sex before age 15 between youth with high and low levels of externalizing problems (0.28 vs. 0.21—Table 4, page 97) was -comparable with the differences between young men and young women (0.29 vs. 0.21), blacks and those of other race or ethnicity (0.28 vs. 0.22), and youth whose mothers had high and low levels of education (0.29 vs. 0.20). This -difference was only slightly smaller than the difference between youth with never-married or formerly married mothers and those with married mothers (0.30–0.32 vs. 0.20). These findings support the importance of externalizing problems as a correlate of the likelihood of initiating early sex.
Gender and Racial or Ethnic Differences
The associations of socioemotional problems with early sexual initiation were not significantly different for young women and men. The coefficients for the interactions between gender and the indicators of internalizing and externalizing problems were both positive, but neither difference was significant (Table 5). The same pattern held for the specific types of socioemotional problems (not shown).
Results for racial or ethnic differences in the associations of internalizing and externalizing problems with early sexual initiation were similar. For both scales, no significant differences were found in the coefficients for blacks and youth of other races or ethnicities or for Hispanics and other youth. Moreover, although a few racial or ethnic differences in associations between specific -socioemotional problems and early initiation were significant, they were not consistent: The association with dependency was weaker for blacks than for whites, and the association with headstrong behavior was weaker for Hispanics than for whites (not shown). An additional set of models compared results for blacks and Hispanics, and found no significant differences (not shown).
Our results confirm the importance of considering multiple types of socioemotional problems simultaneously in research on early sexual initiation. While most types of socioemotional problems were associated with early initiation when considered separately, the associations involving types of internalizing problems became nonsignificant in multivariate models. In short, observed associations of internalizing problems, such as depression, with sexual initiation appear to derive from their occurrence along with externalizing problems, such as antisocial behavior and hyperactivity. By implication, studies that fail to consider both types of problems may misestimate their relative importance in sexual development.
Accurate identification of the types of socioemotional problems that are associated with sexual behavior is important for evaluating theories of sexuality and problem behaviors. Many studies that have found associations between internalizing problems and early sexual initiation have attributed the associations to adolescents’ feelings of hopelessness or psychological dependency.2,7 However, among the types of problems we considered, dependency and peer problems were not associated with early sexual initiation, even when they were considered in separate analytic models. Explanations that rely on an image of sexually precocious youth as “needy” or “desperate” cannot account for this pattern of results. Indeed, according to our results, the key psychological correlates of early sexual initiation are hyperactivity and antisocial behavior. The items in these subscales denote problems in concentration and self-control, rather than lack of self-confidence or self-esteem.
The implications of our results for problem behavior theory are complex. Researchers who have applied this theory to the study of early sexual initiation posit that the consistently observed associations between externalizing behaviors, substance use and precocious sexuality can be attributed to psychological unconventionality, which predisposes youth to multiple, related forms of problem behavior.
In prior research on sexual initiation, unconventionality has been indexed by items such as valuing independence over achievement, low expectations for academic achievement and high tolerance of deviance.27 Whereas the items indicating antisocial behavior are consistent with this definition of unconventionality, the items indicating hyperactivity are not. Moreover, headstrong behaviors resonate with the concept of unconventionality but were not associated with early sexual initiation in our analysis. The discrepancy between the operational definition of unconventionality and the types of externalizing problems that predicted sexual initiation suggests that externalizing behaviors are associated with early sexual initiation not only because they share a psychological precursor, but also because socioemotional problems may be associated with poor or impulsive decision making. Thus, although problem behaviors may cluster because of an underlying psychological predisposition, the associations between externalizing problems and sexual initiation are attributable to factors both more -general and more specific than that.
The associations of socioemotional problems with early sexual initiation did not differ by gender or by race or ethnicity. This finding is surprising, given the dramatic group differences in social contexts and the prior evidence for gender and racial or ethnic differences in effects of family and peer predictors.5,19–21 We were especially surprised to find no gender specificity in the types of socioemotional problems that were associated with early sexual initiation. Prior studies that have found stronger associations of depression for young women than for young men have been based on older samples (typically 13–18-year-olds) and have used youth self-reports to measure depression.5,22,24 Our use of mothers’ reports may account for the differences in the results, particularly if mothers provide less accurate reports of depression than do youth themselves.
Strengths and Limitations
An important strength of our analysis is the use of longitudinal data to predict sexual initiation on the basis of prior socioemotional problems. The relative timing of the measures supports the interpretation that socioemotional problems may increase the likelihood of intercourse, rather than the reverse.41 However, we cannot establish causal priority even with longitudinal data. Socioemotional problems and romantic relationships develop throughout the stage of life considered in this study. Youth who engaged in sexual intercourse at a relatively early age may have been involved in earlier romantic or sexual activities that affected both their socioemotional development and their likelihood of subsequent intercourse.42
Limitations in the measurement of sexual initiation and socioemotional problems also may have influenced the pattern of results. For example, youth may have misreported the age at which they initiated intercourse. As noted earlier, some youth reported different ages at sexual initiation during different interviews; our decision to use their earliest reports increases the likelihood of accuracy, but does not ensure it. Random errors in the age at sexual initiation bias estimates of statistical significance downward and, thus, could account for some of the observed nonsignificant findings.
Another limitation is that our measure of sexual initiation made it impossible to distinguish between consensual and nonconsensual sexual behavior. We have interpreted the results as evidence that socioemotional problems may influence youth to actively seek or engage in intercourse, but they may reflect that such problems expose youth to environments that increase their risk of being forced into sexual activity. The initiation measure also did not distinguish between casual sex and sex in a dating relationship, which have different antecedents and consequences.43 This distinction may be especially important for characterizing the sexual choices of youth who engage in other problem behaviors as compared with the choices of those who do not.
Our necessary reliance on mothers’ reports of youths’ socioemotional problems may have contributed to the stronger associations observed for externalizing than for internalizing problems. Because externalizing problems are more visible than internalizing ones, mothers’ reports of the former may be more accurate. If so, estimates of associations with internalizing problems would be biased downward. While we cannot eliminate this possibility, measurement error is unlikely to fully account for the stronger associations observed for externalizing problems. Prior research has found substantial agreement in parents’ and children’s reports of children’s internalizing and externalizing problems,44,45 which suggests that parents report internalizing problems accurately. Moreover, depression was associated with early sexual initiation in our analysis before we controlled for the other types of problems. This suggests that the lack of associations between internalizing problems and early sexual initiation in multivariate models was due to the co-occurrence of internalizing and externalizing problems, rather than to weaker measurement.
Mothers’ reports of internalizing and externalizing problems are also limited in that they may be influenced by the mothers’ characteristics. For example, one study has found that depressed mothers overreport behavior problems in their children46 (although an earlier review of the literature found no evidence of distortion47). If the former finding applies to this sample, some of the association of externalizing problems with early sexual initiation could reflect the influence of mothers’ depression on reports of problematic behaviors and on their children’s sexual behaviors.
Early sexual initiation is one of several adverse outcomes associated with youth socioemotional problems. Studies of these other outcomes—including poor school performance and high school dropout—have also reported stronger associations for externalizing than for internalizing problems.47,48 This suggests that the processes through which socioemotional problems lead to poor school performance, high school dropout and early sexual initiation are similar. Studies of educational attainment have indicated that externalizing problems are associated with negative social responses that diminish youth’s connections to mainstream institutions and networks and that, thereby, increase youth’s likelihood of engaging in risky or harmful behavior.49 Future research on the sexual development of youth with socioemotional problems should consider the contributions of peer and school networks to the association.
Our results suggest that program planners should target interventions to youth with externalizing problems, particularly those who engage in antisocial or hyperactive behavior. Programs for youth who are depressed and anxious would, no doubt, improve their experiences in and out of school; however, from the perspective of programs aimed at sexual development, programs targeted to youth with externalizing problems are likely to have a greater impact. Such programs are best conceptualized as complements to, rather than substitutes for, efforts to address the social disadvantages that underlie early sexual initiation—an approach supported by our analysis of the independent relevance of externalizing behaviors and social and demographic characteristics.
This study represents a first step in analyzing the role of socioemotional problems in the sexual development of youth. Future studies that disaggregate the possible influence of distinct types of socioemotional problems on related outcomes, such as condom use and infection with STDs, hold promise for elucidating the mechanisms through which socioemotional problems may increase sexual risk.