Youth living in one-parent households are more likely to engage in sexual activity and related risk behaviors than are those in two-parent households. For example, compared with youth living in two-parent households, those living in one-parent households are significantly more likely to report sexual experience; young age at first sex; truancy; tobacco, alcohol or drug use; a history of having been arrested or detained by law enforcement; and fighting.¹

These findings are particularly important because the proportion of youth living in one-parent households in the United States has generally increased over the past three decades. For example, the proportion of youth younger than 18 living in one-parent households rose from 12% in 1970 to 28% in 1996 and stabilized at 27-28% from 1997 to 2002. From 1970 to 2002, the proportion of youth living with their mothers only increased from 11% to 23%, and the proportion of youth living with their fathers only increased from 1% to 5%.²

Studies have examined how family structure (i.e., one- or two-parent household) may influence youth risk behaviors. One study found that parental emotional support did not significantly interact with family structure to predict sexual risk behavior for males or females.³ Other research has suggested that the quality of the mother-child relationship (defined as the amount of communication, warmth, love and closeness between mother and child) may be more important than family structure in determining young women's age at first sexual intercourse.⁴ Still another study found that fathers' absence from the household was significantly linked to early sexual activity and teenage pregnancy.⁵ These results suggest that further research is necessary to investigate significant factors that may influence the sexual risk behaviors of teenagers residing in one-parent households.

One strategy for addressing adolescent risk behavior that is currently receiving substantial attention is youth development, a concept defined by Werner and Smith and later elaborated by Catalano and Hawkins.⁶ As a programmatic approach, youth development is based on adolescent behavior research that incorporates social constructs associated both with risk behaviors (e.g., drug and alcohol abuse, delinquency) and with positive outcomes (e.g., academic achievement, self-control) and resiliency.⁷

A fundamental premise of youth development is that youth should be viewed as resources instead of as potential problems.⁸ This approach places the public health emphasis on providing youth with an environment that encourages growth and development, despite potential adverse situations such as poverty or parents' divorce. The focus on youth development has promoted a research and program approach that links risk reduction with efforts to increase protective factors—or youth assets—such as valuing achievement, positive peer role models and close ties to caring adults.⁹

Assets can help youth avoid potentially harmful behaviors.¹⁰ Catalano and colleagues reported that participants in 24 of 25 programs that had been appropriately evaluated showed significant improvements in such problem behaviors as drug and alcohol use, truancy, violence, sexual risk behavior and smoking.¹¹ Less is known about the role of assets among youth who are at increased risk of engaging in potentially harmful behavior, especially those living in one-parent households. Typically, researchers control for family structure's potential influence on the relationship between youth assets and risk behavior in much the same way as they control for parents' socioeconomic status and youths' age, race and ethnicity. Although the relationship between youth assets and youth risk behavior has been studied,¹² no published research investigates how this relationship may vary by family structure. Because family structure is clearly associated with youth risk behaviors, including sexual risk behaviors, it seems plausible that youth assets may also have a role in the relationship between family structure and youth risk behavior. For example, previous investigations using the data analyzed here suggested that some assets were associated with a reduced likelihood of fighting or alcohol use only for youth living in one-parent households.¹³ In this study, we systematically investigate how the relationships between nine youth assets and four sexual risk behaviors may vary according to family structure, with an emphasis on assets that were significant in one-parent or in two-parent households, not in both.

METHODS

Data Collection and Study Design

Data were collected in 1999-2000 from 1,350 parent-teenager pairs residing in households in inner-city areas of two Midwestern cities, both of which had populations of approximately 500,000. We randomly selected households using a CD-ROM city directory produced by Metromail Corp.¹⁴ The CD-ROM included residences without telephone service or with unlisted numbers, as well as households listed in the phone book. Residences were randomly sampled from households on the CD-ROM that had zip codes representing inner-city areas. One parent and one teenager from each household were randomly selected for interviewing if the household contained more than one teenager or parent.

Teenagers and parents were interviewed simultaneously in different rooms of the residence with a computer-assisted data entry system. Basic demographic information was collected from both teenagers and parents, but data on youth assets and risk behaviors were collected only from teenagers. Teenagers self-administered the risk behavior questionnaire by listening to tape-recorded items with headphones and entering responses into the computer. The survey response rate was 51%, which accounts for all refusals and a proportion of households in which we were unable to determine if an eligible teenager lived. An extensive description of the study methods has been published elsewhere.¹⁵

Measures

•Demographic variables. Depending upon the analysis, youth age was included either as a continuous variable or as a variable stratified into three categories: 13-14, 15-17 and 18-19. The majority of parents (87%) identified themselves as biological parents. Other parental categories were adoptive parent, legal guardian, foster parent, grandparent and adult responsible for the teenager. Family structure was based on parents' reports of whether one or two parents were living in the household. Yearly household income was stratified into three categories: less than $20,000, $20,000-35,000 and greater than $35,000. Parental education was also stratified into three categories: both parents had less than a high school education; at least one parent had completed high school, a GED or some college; and at least one parent had a bachelor's degree or higher. The educational level of the parent not interviewed was obtained from the participating parent, regardless of whether the household contained one or two parents.* When appropriate, for items that concerned the other parent, the definition of parent (residing or not residing in the household) was left to the interviewed parent.

•Youth sexual risk behaviors. We assessed four sexual risk behaviors, using established items recommended in teenage pregnancy prevention research.¹⁶ Whether youth had had sexual intercourse was assessed by the question "Have you ever had sexual intercourse ('done it,' 'had sex,' 'made love,' 'gone all the way')?" Youth responded either yes or no. The other three sexual behaviors applied only to youth who had had intercourse. Current sexual activity was assessed by the item "During the last four weeks, about how many times have you had sexual intercourse?" Youth were categorized as currently sexually active if they reported having had intercourse at least once during that time. We measured age at first intercourse with the question "How old were you when you had sexual intercourse for the first time?" Youth responded by indicating the applicable age. To create an indicator of having delayed first sex until age 17, we dichotomized these responses (16 or younger vs. 17 or older) and analyzed data only for youth aged 17 or older. Use of birth control was determined by the question "The last time you had sexual intercourse, did you or the other person use birth control?" Youth responded either yes or no. Using these data, we created four outcome measures (coded one if present and zero otherwise): never had intercourse, not currently sexually active, delayed intercourse until age 17 and used birth control at last intercourse.

•Youth assets. The research team utilized focus group and needs assessment data to determine key assets to examine in the study. A literature search was conducted to identify appropriate items for asset measurement. Items with established reliability and validity from previously published research were used when possible. Items were created and pretested if appropriate items were not available in the literature.

Factor analyses and reliability tests were used in the construction of scales. Scales were constructed with items that loaded .40 or higher on one factor and with a Cronbach's alpha of at least .60. The nine assets are listed in Table 1 along with the number of items representing the asset, Cronbach's alpha and an example item. The development and construction of the assets are detailed elsewhere.¹⁷

An asset was coded as one (present) or zero (absent) according to youths' mean scores on responses to the items constituting the asset. Items were generally scored from one to four, four being the most positive response, and an individual was classified as having an asset if his or her mean score was three or higher. This indicated that the positive behavior had been reported "usually or almost always," "very important or extremely important" or "agree or strongly agree."

Analysis

Statistical analyses of assets associated with never having had intercourse were based on the overall sample of 1,253 youth, of whom 651 lived in one-parent households and 602 lived in two-parent households. Forty-nine youth were missing demographic data, 20 gave a race or ethnicity other than those listed in Table 2 and 28 did not answer the question that determined sexual experience; given some overlap between these groups, a total of 97 youth were excluded from the sample. Only youth who had had sexual intercourse were included in the other analyses. The maximum sample size for these analyses was 469 youth, of whom 291 lived in one-parent households and 178 lived in two-parent households. The sample of youth aged 17 and older, on which we based our analysis of delaying first intercourse, comprised 126 respondents in one-parent households and 79 in two-parent households.

All analyses were performed with SPSS 10.0.¹⁸ An alpha of .05 or lower was used unless otherwise stated. Analyses were performed separately for one- and two-parent households. Chi-square tests were used to examine univariate associations between dichotomous risk factors and demographic variables. Logistic regression analyses were conducted to calculate the unadjusted odds ratio indicating the relationship between each asset and the absence of each risk factor; multiple logistic regressions were used to assess these relationships while controlling for possible confounders. Multivariate models controlled for youth age (continuous), gender, race and ethnicity, and household income and parent education only if univariate analysis indicated a p-value of .20 or lower. Interactions between assets and demographic variables were assessed in each logistic regression with the alpha level set at .01 or lower. Logistic regression was conducted for significant interactions, stratifying by the demographic variable that had shown the significant correlation.

To assess the potential effects of youths' cumulative number of assets (0-9), we included a variable measuring total assets in the regression analyses. The analyses included only those youth for whom all assets had been measured. Adjusted and unadjusted odds ratios were calculated with the same methodology as that used for individual assets.

RESULTS

The study sample of 1,253 youth had a mean age of 15.4 (standard deviation, 1.7), and 52% were female. Forty-nine percent were non-Hispanic white, 23% non-Hispanic black, 19% Hispanic and 10% non-Hispanic Native American. Fifty-two percent of the youth lived in one-parent households, and 79% of single-parent households were mother-only. Sixty-six percent of the youth lived in households with reported income levels of $35,000 or less, and 12% had parents who had both not graduated from high school.

Never Had Intercourse

Fifty-five percent of youth in one-parent households reported never having had sexual intercourse, a significantly lower proportion than the 70% in two-parent households (Table 2). The proportion who had not had sexual intercourse decreased as youth age increased, and this trend was significant in both one- and two-parent households. In two-parent households, the proportion of youth who reported never having had intercourse was significantly higher among those who had at least one parent with a bachelor's degree than among those with less educated parents (79% vs. 67-77%).

Table 3 shows only those assets that were significantly associated with reports of risk behavior for youth in one-parent or two-parent households, not both. Two youth assets met this criterion for the outcome reflecting sexual inexperience. Youth living in one-parent households who had the aspirations asset were significantly more likely to have never had intercourse than were youth in one-parent households who did not have the asset (odds ratio, 1.8). Among youth living in two-parent households, those with the community involvement asset were significantly more likely than those without the asset to report never having had sex (1.9).

The nonparental adult role models, peer role models and use of time (religion) assets were significant in the unstratified sample and in both household strata. The family communication, good health practices (exercise/nutrition) and responsible choices assets were significant in the whole sample but not in either stratum. Analyses of the data from the unstratified sample have been published elsewhere.¹⁹

The adjusted odds ratios for the total number of assets were significant for one-parent as well as for two-parent households (1.3 and 1.4, respectively—Table 3). This indicates that, depending on family structure, youths' odds of never having had sexual intercourse increased about 30-40% with the addition of any one asset.

Not Currently Sexually Active

The proportion of youth in one-parent households who were not currently sexually active did not differ significantly from that in two-parent households (46% and 52%, respectively—Table 2). For youth residing in two-parent households, the proportion not currently sexually active decreased significantly with age. Among youth residing in one-parent households, the proportion who were not currently sexually active was significantly lower among whites than among youth of other races and ethnicities (38% vs. 44-58%).

The interaction between parent education and the peer role models asset was significantly associated with not being currently sexually active among youth in one-parent households (Table 3). In this stratum, youth with peer role models whose parents both had less than a high school education were significantly more likely to report not being currently sexually active than were those without the asset whose parents had a comparable level of schooling (odds ratio, 21.2). None of the other youth assets were significant in the unstratified sample or in the one- or two-parent household stratum.

Logistic regression models that included the total assets variable did not yield significant results. This suggests that youth assets had no significant cumulative effect on youths' odds of current sexual activity, regardless of family structure.

Delayed First Intercourse

The proportion of youth who delayed first sexual intercourse until age 17 or older was significantly lower among those in one-parent households than among those in two-parent households (15% vs. 30%—Table 2). In multivariate analyses, youth residing in one-parent households who reported good health practices were significantly more likely to have delayed sexual intercourse than were youth residing in one-parent households who did not have this asset (odds ratio, 7.3—Table 3). None of the other youth assets were significantly associated with this outcome in either household stratum.

For youth living in one-parent households, the odds of having delayed first sexual intercourse increased by nearly 80% with each additional asset. The adjusted odds ratio for the total number of assets for youth living in two-parent households was nonsignificant.

Used Birth Control at Last Intercourse

The proportion of youth who had used birth control at last intercourse did not differ significantly by household stratum (63-65%—Table 2). For youth residing in two-parent households, it increased significantly with age. Among youth residing in two-parent households, a significantly lower proportion of Native Americans than of youth in other racial or ethnic groups had used birth control at last intercourse (40% vs. 64-73%). In two-parent households, a significantly higher proportion of males than of females reported this behavior (71% vs. 60%).

The peer role models and family communication assets were significantly associated with birth control use among youth living in one-parent households (Table 3). Youth in one-parent households with either of these assets were significantly more likely to have used birth control at last sexual intercourse than were youth in one-parent households without these assets (odds ratios, 1.9-2.0). None of the other youth assets were significant in the unstratified sample or in either household stratum. For youth living in one-parent households, the odds of having used birth control at last intercourse increased by about 20% with the addition of any one asset.

DISCUSSION

Our findings suggest that associations between risk behaviors and demographic factors may vary according to household structure. Youth in one-parent households tended to report risk behaviors in greater proportions than did youth in two-parent households, although this difference was significant for only two risk behaviors. In addition, most of the significant relationships between demographic variables and risk behaviors were found only in the two-parent household stratum. For example, demographic factors (age and race) were significantly linked to birth control use among youth in two-parent households but not among those in one-parent households. One interpretation of these findings is that youth in one-parent households may be more likely than those in two-parent households to participate in risk behaviors, regardless of their demographic profile; in contrast, some demographic variables may be associated with participation in sexual risk behaviors among youth in two-parent households. Identifying youth who are at increased risk is important, although none of the demographic factors examined in this study are likely to be the focus of intervention activities. Appropriate interventions should be developed and implemented for youth who are at elevated risk; programs focusing on youth development, and more specifically on youth assets, are promising approaches.

The results also suggest that some youth assets may be more important for youth residing in one-parent households than for youth in two-parent households. This is supported by the fact that significant relationships between youth assets and all four risk behaviors emerged in analyses of the one-parent stratum, whereas just one youth asset was significant in the two-parent stratum.

As previously mentioned, earlier studies that control for the potential influence of family structure have found significant associations between youth assets and risk behaviors. Our findings suggest that those results may have been at least partially driven by the influence of one-parent household data. All of the significant and nonsignificant relationships between assets and risk behaviors found in this study for one-parent households (except for the link between assets and never having had sexual intercourse) were also found in the unstratified analyses.

Differences in sample size and the resulting differences in statistical power to detect significant associations could explain why several relationships between youth assets and risk behaviors were significant in one-parent households but not in two-parent households. Compared with youth living in two-parent households, those in one-parent households were more numerous in each analysis. As a result, there was more statistical power to detect significant differences in the one-parent household stratum than in the two-parent stratum. However, the potential bias is mitigated by two facts. First, the significant odd ratios for the one-parent household youth were fairly substantial. Second, the point estimates of all significant odds ratios for youth living in one-parent households were higher than those for youth in two-parent households, although this difference was not statistically tested. These results may suggest that although all youth can benefit from assets, youth living in one-parent households may stand the most to gain from assets' potential protective influence.

Analyses of the association between youths' total number of assets and the likelihood that they had participated in sexual risk behaviors also seem to support the notion that assets may be more important for youth residing in one-parent households than for youth in two-parent households. The total number of assets was associated with three of the four sexual risk behaviors among youth in one-parent households, but with only one risk behavior among youth in two-parent households.

One relevant question is whether results that suggest the importance of family structure for associations between assets and sexual risk behaviors hold true for other youth risk behaviors. Unpublished data from our work suggests that the pattern may persist for youth aggression and delinquent behavior, but not for alcohol, drug and tobacco use. Future studies should more thoroughly investigate how relationships between youth assets and sexual risk behaviors vary according to family structure, and should also examine associations between assets, family structure and nonsexual risk behaviors.

Findings for the family communication asset provide some support for the idea that an increase in the quality of parent-child relationships may be associated with a reduction in sexual risk behavior. In this study, the asset was a multi-item construct that assessed aspects of the parent-child relationship such as parents' attempts to understand youths' point of view and youths' discussion of their problems with parents. This construct included items similar to those used in other studies that have assessed the quality of mother-child relationships and parental emotional support. Among youth in one-parent households, having the family communication asset was significantly associated with increased odds of birth control use at last sexual intercourse but not with other outcomes, including several that were outcomes of interest in other studies.

One particularly interesting finding was that parental education and the peer role models asset interacted to predict a lack of current sexual activity, although the small sample size of 32 youth indicates that the results should be interpreted with caution. The asset was significant only for youth residing in one-parent households whose parents had less than a high school education. Having the peer role models asset was also significantly associated with birth control use at last sexual intercourse among youth in one-parent households. Other researchers have suggested that susceptibility to peer influence may be greater for youth who live in one-parent or stepparent households.²⁰ Our results support that inference for one-parent households and, more notably, suggest that peer influence on sexual risk behavior may be positive, rather than negative, for youth in this environment.

The link between the good health practices asset and delay of first intercourse among youth in one-parent households corresponds with reported associations between participation in organized sports and a reduction in sexual risk behavior.²¹ In our study the asset was narrowly defined as health behaviors related to exercise and nutrition. Further research is needed to determine why and how this and other assets are associated with a reduction in risk behaviors. Regardless of how one interprets the role of family structure in the relationships between assets and risk behaviors, the overall findings of this study suggest that youth in one-parent households may benefit more from these assets than youth in two-parent households.

Finally, it is important to note that most of the parents (79%) in one-parent households were mothers. Parent gender was thus examined for significant associations with the outcomes investigated in this study; no such associations were found. The question of whether risk behaviors vary with gender differences between parents and youth (e.g., risk behaviors of boys in mother-only households versus those of boys in father-only households) is important but was not pursued because of the limited number of male parents within the one-parent households in our study.

One possible implication of our findings for practitioners is that specific youth assets may be particularly effective in reducing the sexual risk behaviors of youth residing in one-parent households. Youth development programs that focus on building positive peer role models, aspirations for the future, family communication and good health practices may be most beneficial for youth residing in one-parent households. However, our data are cross-sectional and do not indicate the causal directions of observed relationships; therefore, using these results to design interventions may be premature.

Findings from this study should be viewed in light of other limitations. Youth may have given socially acceptable answers to the sexual risk behavior questions. However, the fact that youth entered their answers into a computer in private may have helped to reduce the potential reporting bias. The question "Have you ever had sexual intercourse?" did not discriminate between consensual and nonconsensual intercourse. Thus our sample of youth who had had sexual intercourse may not have been limited to those who had chosen to do so. Finally, the moderately low response rate raises questions about the generalizability of the results. However, we compared the distribution of races, ethnicities and parental income levels in the study sample with census data from the same zip codes and found no significant differences;²² therefore, the sample was likely representative of the neighborhoods.

This is the first published study to find that associations between youth assets and a reduction in several sexual risk behaviors were significant for youth residing in one-parent households but not for youth residing in two-parent households. These results are particularly relevant because they focus on youth living in one-parent households, who have elevated odds of participating in sexual risk behaviors. The results also suggest some of the mechanisms (i.e., youth assets) that might positively influence the behaviors of youth living in one-parent households. Further research is needed to assess potential causal relationships between youth assets and sexual risk behaviors, and to more thoroughly explore why some youth assets may be more important for youth living in one-parent households than for those in two-parent households.