Predicting Youth Out-of-School Time Participation: Multiple Risks and Developmental Differences

By Wimer, Christopher Simpkins, Sandra D; Dearing, Eric; Bouffard, Suzanne M; Caronongan, Pia; Weiss, Heather B

Youth out-of-school time (OST) programs and activities can provide developmental benefits for participating youth. Yet little research has examined the contextual predictors of youth OST participation. To address this issue, we examined a collection of child-, family-, school-, and neighborhood-level characteristics as predictors of OST participation using data from the Panel Study of Income Dynamics-Child Development Supplement. In summary, child and family characteristics were most useful in predicting participation such that children least likely to participate were those characterized by high levels of developmental (e.g., low achievement, behavior problems, poor health) and family (e.g., parent psychological distress and low emotional support) problems. These relations, however, emerged only during middle school and high school. For certain types of activities, namely athletics and lessons, problems measured across various contexts were more strongly associated with OST participation for higher-income families than for lower-income families. These findings point to the importance of considering multiple developmental domains and developmental periods in understanding predictors of youth OST participation. There has been a surge of interest among policymakers, practitioners, and researchers in the impact of youth’s participation in out-of-school time (OST) programs and activities (Mahoney, Harris, & Eccles, 2006). OST programs include after-school programs with a defined structure, boundaries, and goals, whereas OST activities include athletics, lessons, and extracurricular activities that may or may not take place within the context of a defined program. High-quality, organized OST programs and activities have the potential to support and promote youth development because they (a) situate youths in safe environments; (b) prevent youths from engaging in delinquent activities; (c) teach youths general and specific skills, beliefs, and behaviors; and (d) provide opportunities for youths to develop relationships with peers and mentors (Eccles & Gootman, 2002). Little research, however, has investigated what factors beyond demographic characteristics are associated with youths’ OST participation. This study begins to fill this gap by examining whether greater numbers of risks at the child, family, school, and neighborhood levels predict youths’ OST participation in programs and activities.

A recent report by the National Academy of Sciences revealed that participation in OST activities is predictive of academic success as measured through test scores, absenteeism, school dropout rates, homework completion, school grades, and course enrollment (e.g., Barber, Eccles, & Stone, 2001; Bartko & Eccles, 2003; Cooper, Valentine, Nye, & Lindsay, 1999; Mahoney & Cairns, 1997; Posner & Vandell, 1994, 1999). Participation in OST programs and activities is also related to multiple indicators of positive social development, including more prosocial and less aggressive behavior with peers and close friends as well as lower levels of depressive symptoms and behavior problems (e.g., Eccles & Templeton, 2002; Petit, Laird, Bates, & Dodge, 1997; Simpkins, Fredricks, Davis- Kean, & Eccles, 2006; Vandell & Shumow, 1999). Importantly, the developmental benefits of OST participation appear greatest for youths at risk for poor developmental outcomes such that OST participation serves a protective function for these youths (Mahoney, 2000; Marsh & Kleitman, 2002; Roeser & Peck, 2003).

It is likely that both individual and contextual factors predict participation for youths. The notion that individuals affect and are affected by their physical and psychosocial surroundings was first offered by Bronfenbrenner’s (1979) ecological systems model and was later extended to emphasize the vital reciprocal influence between individuals and contexts (Sameroff & Chandler, 1975; Sameroff, 1983). OST programs and activities are one such developmental context shaping and being shaped by youths (Mahoney, Larson, & Eccles, 2005; Posner & Vandell, 1999). For instance, although poor children are less likely to participate in after-school programs for a variety of contextual or individual reasons, they may also benefit more from such participation due to the relative dearth of other enriching opportunities in their families, schools, and communities (Mahoney, Lord, & Carryl, 2005).

Youths’ participation in OST programs and activities is likely to result from a complex set of factors including availability, parents and programs functioning as gatekeepers, and youth interest. In other words, OST selection factors function at the individual and program levels as well as at the family, school, and neighborhood levels. That is, the propensity to enroll in OST programs or activities and the resultant outcomes may be shaped by individual (child-level) factors and features of the multiple overlapping contexts that youths experience in their families, schools, and neighborhoods. Operating within a developmental systems framework (Sameroff, Bartko, Baldwin, Baldwin, & Seifer, 1998), we examine these selection processes in the present study with a focus on developmental risks present at the individual child level and the family, school, and neighborhood levels.

Developmental Risk and OST Participation: Child, Family, School, and Neighborhood Factors

Previous research has defined risk factors as behaviors or actions that can compromise various aspects of youths’ successful functioning during a particular developmental period (Perkins & Borden, 2003). Researchers generally concur that risk is a multivariate construct and that developmental outcomes are usually predicted by the number of risk factors rather than particular discrete factors, with children at greater risk as the number of risk factors increases (e.g., Sameroff et al., 1998). In other words, children vary in terms of the specific factors and combination of factors that place them at risk for poor outcomes. Two children can have dramatically different risk factors but have the same poor outcomes. Unfavorable outcomes are thought to be exacerbated by exposure to multiple risk factors (Sameroff et al., 1998), although this does not rule out the possibility that there may be thresholds of numbers of risk factors that may be critical or particular constellations of risks that are particularly detrimental to youths. In line with this perspective, the central question addressed in the current analysis is: How do risk factors, at the child, family, school, and neighborhood levels, predict youths’ OST participation in a variety of OST activities and programs?

Child risk factors. Youths with lower academic, behavioral, and physical adjustment may be less likely to become involved in OST programs and activities. For instance, researchers have found that academic indicators are powerful, positive predictors of children’s participation (Jordan & Nettles, 2000; Marsh & Kleitman, 2002). Children’s poor adjustment may have direct ramifications on participation if children cannot satisfy the grade or other requirements set by the programs and activities. In addition, children with poor adjustment may not participate in OST programs or activities due to difficulties engaging in a group context, fear of peer rejection, or lack of encouragement or active discouragement by program staff. Despite their potential importance, these barriers have received little empirical attention. We expect that as the number of child psychosocial, academic, and physical difficulties increase, the likelihood of youths’ participation in organized OST settings will decrease. Developmental difficulties may be more important during adolescence than childhood for two reasons. First, program and activity barriers in addition to entry requirements may increase with age (e.g., competition to participate in sports activities typically increases with age). Such declines in access may translate into decreased participation of youths with greater numbers of problems because they may possess fewer of the requisite skills and may be less likely to be encouraged to participate. Second, because increases in autonomy and decision making suggest that adolescents are the primary decision makers about how they spend their nonschool hours (Gauvain & Perez, 2005), adolescents with more problems may be more likely than their peers to opt out of OST activities.

Family risk factors. Parents play a significant role in children’s development through many avenues (Furstenberg et al., 1999; Parke & Buriel, 1998). For example, parents provide developmental opportunities by choosing schools, providing books in the home, and being involved in children’s schools (Lareau, 1989). Research increasingly shows that parents also use community-based activities as part of a compilation of strategies to support their children’s success (Dearing, McCartney, Weiss, Kreider, & Simpkins, 2004; Weiss et al., 2003). Stable, engaged, and supportive fami lies headed by efficacious parents are more likely to be able to engage youths in constructive use of their nonschool time (Larson, Dworkin, & Gillman, 2001; Ardelt & Eccles, 2001). Research also suggests that these parenting behaviors as well as youth outcomes are influenced by parents’ psychological, financial, and social resources (McLoyd, 1998). As such, families with greater challenges or disadvantages, such as parents who are less engaged and have fewer resources, may be less likely to have youths involved in OST programs and activities. School risk factors. Given that many OST activities occur within the context of children’s schools, examining school- level determinants of youth participation in OST extracurricular activities is of prime importance. Schools with high concentrations of poor and minority students are in worse states of disrepair with more inhospitable and inadequate facilities (US General Accounting Office, 1996) and also offer more in-school academic and less enrichment-focused content, a pattern that may extend to schools’ OST offerings (Roth, Brooks-Gunn, Linver, & Hofferth, 2003). Similarly, researchers have found that larger schools and schools with high pupil-teacher ratios have lower rates of student participation in community and extracurricular activities (Jordan & Nettles, 2000; Lindsay, 1984; Marsh & Kleitman, 2002). This relationship is likely due to the fact that there are typically a maximum number of slots available in youth activities such that in smaller schools, there is a larger percentage of the overall student body who can participate (Barker & Gump, 1964; Quiroz, 2000). Research also shows that students in disadvantaged urban schools perceive more daily hassles at school and that these negative feelings decrease school engagement (Seidman, Aber, Allen, & French, 1996). It is likely that greater numbers of such school-level risks could result in lower activity participation.

Neighborhood risk factors. Neighborhood characteristics are also likely to be key determinants of activity participation. Neighborhood indicators, such as disadvantage, may impact participation because activities are unavailable or because parents limit youths’ participation (Leventhal & Brooks-Gunn, 2000; Jarrett & Jefferson, 2003). Neighborhood disadvantage (e.g., poverty, crime) is associated with more restrictive family management strategies to protect youths from dangerous neighborhood environments (Furstenberg et al., 1999). In support of this hypothesis, Furstenberg and colleagues (1999) found that indicators of neighborhood disadvantage were linked with lower participation in OST community activities. Another neighborhood factor relevant to youths’ OST participation is community-level collective efficacy, which refers to the “linkage of trust and cohesion with shared expectations for control” (Morenoff, Sampson, & Raudenbush, 2001, p. 520). Neighborhoods with higher collective efficacy are more likely to include networks of community members who will intercede when youths are engaging in delinquent behavior and monitor each other’s children (Rankin & Quane, 2002). The resultant safety in such neighborhoods may facilitate parents’ and youths’ use of OST settings. We therefore expect that the number of risks at the neighborhood level will be negatively associated with youths’ participation.

The Moderating Influences of School Age and Family Income

It is likely that the importance of some of these risk factors will increase as youths age into adolescence. As children age they are granted more autonomy, and their identity and goals become more stable, suggesting that youths are likely to play increasingly larger roles in making OST-related choices as they get older (Collins, Gleason, & Sesma, 1997; Gauvain, 1999; Gauvain & Perez, 2005). Research has also suggested that there are critical differences in OST programs and activities. Restricted access to high-quality programs and activities (e.g., grade requirements, necessary athletic skills) and various gatekeepers may keep high- risk youths out of programs as youths enter adolescence (McLaughlin, Irby, & Langman, 1994). Thus, youths’ academic, psychosocial, and physical problems may subsequently have a larger impact on participation during adolescence than during childhood. Correspondingly, youths’ families may have a larger influence on participation during childhood than during adolescence, when adolescent and gatekeeper decision making should be lower. We therefore expect that child-level indicators will become stronger predictors as children age, whereas family-level indicators will become weaker as children age. Similarly, as children age they should be less bound by defined spaces such as schools and neighborhoods, as they should be more able to traverse such boundaries and respond to contextual risks at, for instance, the school level by increasing participation in other settings, such as community programs. Thus, we expect school- and neighborhood-level indicators to be more strongly associated with participation among younger youths.

Another potential moderator is families’ income. While some research has conceptualized low income as directly impacting youth development (e.g., Gutman, Sameroff, & Eccles, 2002), we treat it here as a potential moderator given the public concern with engaging low-income youths in prosocial activities. In particular, we focus on the potential moderating effects of family income. In short, youths living in families with low levels of family income are more likely than their higher-income peers to experience poor outcomes across most developmental domains, including academic achievement, health, and social-emotional functioning (McLoyd, 1998). Given this, youths from poorer families may be more vulnerable to the negative influences of other risk factors in their lives. On the other hand, many OST programs entail significant monetary costs for participating families (Lind, Relave, Deich, Grossman, & Gersick, 2006) such that poorer families may be unable to participate, regardless of other risk factors present, in some types of activities. Thus, there is the potential for family socioeconomics to moderate the effects of other risk in one of two ways: (a) low family income may exacerbate the limiting effects of other risks on OST participation, or (b) the propensity for low-income children to participate may be so low that even coming from less risky contexts may not bolster participation. If (a) were true, we would expect greater numbers of risks to be more strongly related to participation among lower-income families, while if (b) were true we would expect greater numbers of risks to only be predictive for higher-income families.

Research Questions

In this study, we assess the relationship between cumulative risk at the child, family, school, and neighborhood levels and enrollment in multiple types of OST activities above and beyond demographic factors. In general, we expect that greater risks will be associated with lower odds of enrolling in OST activities. Second, we examine whether risk at each of the four levels is moderated by youths’ age. We expect that child-level indicators will be more strongly associated with OST participation for older youths than for younger youths, but family-, school-, and neighborhood-level indicators will be more strongly associated with OST participation for younger youths than for older youths. Third, we examine whether cumulative risk at each of the four levels is moderated by family socioeconomic status with the expectation that risk will more strongly predict OST participation among lower-socioeconomic (SES) youths than high-SES youths, as they may be more vulnerable; however, for activities that are more costly (e.g., athletics), risk may matter only for higher- SES youths, as lower-SES youths may be unable to enroll in these activities regardless of risk levels.

Method

Participants

This study used data from the Panel Study of Income Dynamics- Child Development Supplement (PSID-CDS) (Mainieri, 2004). This longitudinal study follows children of parents who participated in the larger PSID study, which began in 1968. During the first wave of CDS data collection in 1997, families with children between the ages of 1 day old and 12 years were eligible for participation in the PSID-CDS. Up to 2 randomly selected children in each family were recruited to participate, yielding 3,563 children from 2,394 households. PSID-CDS families were reinterviewed at Wave 2 during the school year of 2002-2003 (n = 2,908 children/adolescents from 2,017 families; 91% of Wave 1 families). The PSID-CDS is a nationally representative sample of children with an oversampling of poor, African American, and Latino families. Data for the current study come from questionnaires completed by youths and their primary caregivers (PCG), census tract data, and school-level data obtained from the National Center for Education Statistics.

Participants in this study included the 2,435 youths who were in either elementary school (first through fifth grades; 44.3%), middle school (sixth through eighth grades; 24.3%), or high school (ninth through twelfth grades; 31.4%) during Wave 2 and who had completed data for the risk indices (e.g., some youths were too young at Wave 1 to have completed the assessments used to construct the child risk index). They ranged between 6 and 19 years of age (M = 11.7, SD = 3.5 years). The sample was split evenly by gender. The majority of participants were either African American (42%) or European American (45%). The remaining participants were Latino (8%) or were from another racial or ethnic group (5%). Families’ mean income per year in 2000 was $67,733 (SD = 85,530). Parent education also ranged widely. The higher education level in the parent dyad was used to represent parent education (Shumow & Lomax, 2002). Parent education ranged from a high school degree or less (43%) to at least some graduate education (11%), with a mean of 13.3 total years (SD = 2.52). The majority of the PCGs who completed questionnaires (95%) were mothers. Measures

OST participation. We utilized seven measures of OST participation asked of PCGs and children in Wave 2 to determine whether the youths participated in the following activities during the previous 12 months:

(a) Community Programs (PCG report): Was child a member of any groups or programs in the community?

(b) Athletics (combined from PCG and child report1): Did child participate on an athletic team?

(c) Scouting (PCG report): Did child participate in scouting?

(d) Lessons (PCG report): Did child take lessons such as music, dance, or drama?

(e) School activities (child report): Did child participate in any extracurricular activities at school?

(f) Service clubs (child report): Did child participate in any volunteer or service clubs?

We also computed a seventh OST measure that measured whether the youths participated in any of the above six OST contexts.

Demographic indicators. We included a variety of sociodemographic background factors in all analyses. These included (a) income (the natural log of an income-to-needs ratio averaged over five years, with the income-to-needs ratio defined as total family income from all sources divided by the number of persons in the family unit), (b) education (parents’ highest level of education), (c) race or ethnicity (dummy variables for Black, Latino, and Other racial or ethnic groups, with the comparison group being White), (d) child’s gender (1 = male, 0 = female), and (e) school-age (dummy variables for middle school and high school youths, with the comparison group being elementary school youths).

Risk indices. We created four risk indices, one for each of the four levels (child, family, school, and neighborhood). These indices were based partially on previous research on risk (Sameroff, Seifer, Baldwin, & Baldwin, 1993; Rutter, 1979). In line with this work, individual risk factors were dichotomized. Although dichotomizing entails some loss of information for continuous measures, we nevertheless dichotomize in order to operationalize our risk indices as the number of risks present at each level. Consistent with the multiple risk framework, we therefore test whether youths facing a greater number of risks at any given level are less likely to participate in a variety of OST programs and activities. All risk indices were standardized to have a mean of 0 and a standard deviation of 1 within each level. For the family, school, and neighborhood levels we relied on 2002 data, as these indicators were unlikely to be influenced by 2002 OST participation. However, for the child-level risks, 2002 child factors were as likely to be outcomes of OST participation than they were to be determinants. We thus operationalized child risk using data from the 1997 Wave 1 of the CDS, five years prior to youths’ 2002 OST participation. Reliability was reported for all applicable multi-item scales.

Child risk index. The child risk index came from the 1997 (Wave 1) PCG interview and child assessments. The child risk index was composed of six indicators: (a) low reading achievement score (over 1 SD below the mean on the Woodcock-Johnson Letter-Word Recognition subtest) (Woodcock & Johnson, 1989); (b) low math achievement (over 1 SD below the mean on the Woodcock-Johnson Math Applied Problems subtest) (Woodcock & Johnson, 1989); (c) presence of a significant physical health problem, reported by the PCG; (d) high internalizing problems (over 1 SD above the mean on a composite from PCG report of the Behavior Problems Index, alpha = .84) (Peterson & Zill, 1986); (e) high externalizing problems (over 1 SD above the mean on a composite from the Behavior Problems Index, alpha = .88); and (f) low levels of positive behavior (over 1 SD below the mean on a composite from the PCG report of the Positive Behavior Scale, tapping autonomy, compliance, and social competence; alpha = .73) (Polit, 1998).

Family risk index. The family risk index was composed of eight indicators from the 2002 PCG interviews. The eight indicators were: (a) low emotional support (over 1 SD below the mean on a composite from the Home Observation and Measurement of the Environment (HOME) scale tapping the PCG’s provision of emotional support; alpha = .41)2 (Caldwell & Bradley, 1984); (b) low cognitive stimulation in the home (over 1 SD below the mean on a composite from the HOME scale tapping family provision of a cognitively stimulating home environment; alpha = .62) (Caldwell & Bradley, 1984); (c) low family involvement in the school (family members were not involved in any of eight activities at the child’s school, such as talking with the teachers, principal, etc.; alpha = .53); (d) low level of social support parents received from others (over 1 SD below the mean on a composite measuring the PCG’s practical and emotional support from other caregivers, family, and friends; alpha = .79); (e) low parental self-efficacy (over 1 SD below the mean on a composite measuring the PCG’s self-efficacy; alpha = .79) (Pearlin, Eieberman, Menaghan, & Mullan, 1981); (f) high family financial strain (over 1 SD above the mean on a composite measure of parents’ perceptions of their family’s level of financial strain; alpha = .69) (Conger & Elder, 1994); (g) high parental psychological distress (over 1 SD above the mean on a composite measure of the PCG’s self-reported levels of distress; alpha = .68) (Kessler et al., 2002); and (h) low parent community involvement (over 1 SD below the mean on a composite measure of the PCG’s involvement in the community, such as neighborhood meetings, parenting classes, or community centers; alpha = .70).

School risk index. Data for the school risk index were taken from data collected in the U.S. Department of Education’s Common Core of Data (CCD) on features of the schools PSID-CDS youths attended during Wave 2. The school risk index was composed of five indicators at Wave 2: (a) large school size (over 1 SD above the mean on the total number of pupils enrolled in the school), (b) high-poverty school (over 60% of the student body eligible for free or reduced price lunch), (c) high pupil to teacher ratio (over 1 SD above the mean), (d) Title I (a federal program that distributes funds to schools with a high percentage of pupils from low-income families) eligibility, and (e) high dropout rate in the school district (over 1 SD above the mean).

Neighborhood risk index. The neighborhood risk index was computed from measures taken from 2000 census data at the tract level as well as PCG reports of perceptions of the neighborhood environment in 2002. The neighborhood risk index was composed of six indicators: (a) high poverty neighborhood (over 1 SD above the mean on a variable measuring the disadvantage level of the PCG’s census tract), (b) high neighborhood disorder (over 1 SD above the mean on a composite of interviewer ratings of neighborhood conditions, such as the extent of litter, garbage, or broken glass on the street; alpha = .87), (c) low collective efficacy (under 1 SD below the mean on a composite of the PCG’s perception of collective efficacy in the neighborhood; alpha = .85), (d) dangerous neighborhood (the PCG’s report that the neighborhood is either somewhat or extremely dangerous), (e) neighborhood was a poor place to raise children (the PCG’s report that the neighborhood is either a fair or poor place to raise children, as opposed to a good, very good, or excellent place to raise children), and (f) difficulty distinguishing strangers (the PCG’s report that it is very difficult to tell a stranger in the neighborhood from a resident).

Results

Analysis Plan

After examining descriptive statistics, we estimated a series of multivariate logistic regression models predicting each of the seven OST participation indicators, all using the appropriate sampling weights included with the PSID-CDS. In every model, we controlled for school age, gender, race or ethnicity, family income, and parental education. First, we estimated the association between the four risk indices and OST participation in Model 1. This first model was used to determine whether, net of background factors and risks at the other three levels, the number of child-, family-, school-, or neighborhood-level risks was associated with reductions in the odds of participating in a variety of OST programs and activities. Second, in Model 2 we added interaction terms3 to Model 1 that were used to examine school age as a moderator.4 Significant interactions in this model indicate whether the associations between risk varied by children’s school age, as hypothesized. Third, in Model 3 we added interaction terms to Model 1 that were used to examine the moderating effects of family income. In all tables, logistic regression coefficients are presented as odds ratios. To parse the direction of relationships within the three age groups given the large number of significant interactions that we found, we refit Model 2 for each indicator separately within a subsample of youths in each school-age level. Our description of the associations within these different age groups is based on the direction of associations in these subsample analyses.5 Coefficients that are less than 1 indicate that youths were less likely to participate in the particular activity. Coefficients greater than 1 indicate that youths were more likely to participate in the particular activity.

Descriptive Results

Table 1 shows the percentage of youths participating in OST programs and activities by various demographic background characteristics. Elementary students were more likely than older youths to participate in scouting, while middle school students were more likely to participate in lessons and athletics than were other age groups, and high school students were more likely to participate in school activities and service clubs. White students consistently participated at higher levels than Blacks and Latinos, as did youths from more educated and higher-income families. Table 2 presents bivariate correlations between OST participation and the four risk indices. Given the large sample size, most correlations were statistically significant. The four measures of risk tended to be negatively correlated with most measures of OST participation, indicating that youths with higher levels of risk at the child, family, school, and neighborhood levels were less likely to participate in OST programs and activities; these correlations, however, tended to be relatively modest. Most OST participation indicators were also positively correlated with each other, although these correlations also tended to be rather modest, indicating that participation in one OST context did not necessarily mean participation in other OST contexts. Findings

Child risk. Table 3 presents the multivariate results broken down by risk index and type of OST program or activity. Of the seven OST program and activity outcomes we examined, child risk was unrelated to three types of activities: community programs, scouting, and service clubs. For the remaining four outcomes, however, greater numbers of child risks negatively predicted OST participation. Youths with more physical, psychosocial, and academic problems in 1997 were less likely than their peers to participate in athletics in 2002, even after controlling for sociodemographic differences and differences in family-, school-, and neighborhood-level risks. The significant interactions between child risk and middle school and high school status suggest that middle schoolers and high schoolers with such problems were especially less likely to participate in athletics in comparison to elementary school youths. Similarly, there was a stronger negative relation between child problems and likelihood of participating in lessons for high school students than for elementary school youths.6

For school activities, youths with more academic and social problems were less likely to participate in school-based activities at all school ages. The child risk index became nonsignificant once the interaction terms with school age were added to the model, and neither of the child interactions were significant. Further analyses within school-age subgroups revealed that the negative relation between child problems and school activity participation was evident in both the middle school and high school subsamples. Finally, for participation in any of the OST programs or activities, the interaction between high school status and the child risk index was significant and negative. These results confirm that older youths with greater social and academic problems are less likely to participate in any OST activities than youths with fewer such problems.

Family risk. Community programs, scouting, and school activities evidenced no significant relation to family risk. Family risk significantly predicted participation in the remaining four participation measures. First, the results suggested that older children from more disadvantaged family environments were less likely to participate in athletic programs. The significant negative middle school and family risk factor interaction indicated that coming from less involved and engaged families was associated with decreased odds of participation for middle school youths but was not associated with decreased odds for elementary school youths. Although the high school-family risk interaction was not statistically significant in the full model (as shown in Table 2), the subsample analysis revealed a significant negative relation between the family risk index and athletic participation among high school-aged youths as well.

Second, family risk was also negatively related to the odds of youths’ participation in lessons for children at all school ages. After including the interactions, however, neither the family risk index nor the interaction terms were still significant. This reduction in statistical significance was likely to emerge from inflated standard errors from the inclusion of the interaction terms. Indeed, separate analyses within the subsamples confirmed a significant and negative relation between the number of family risks and the odds of participating in lessons among middle school youths. This indicated that the overall negative relationship found between family risk and participation in lessons is particularly concentrated among the middle school sample. Third, high school youths from riskier family environments were less likely to participate in service clubs and any OST program or activity, while this was not the case for younger youths. Taken together with the findings in the previous section on child risk, these results reveal that older youths with greater academic and social problems and those from less supportive, involved, and engaged families are less likely to become involved in OST programs and activities than are other youths.

School risk. Many of the OST indicators were significantly predicted by the number of school risks, but patterns were inconsistent and varied by the type of activity. No significant relations were found for either athletic programs or lessons. For community programs and scouts, the interaction between high school status and risk was positive and significant, whereas the overall school risk index coefficient became negatively related to the odds of community program participation. Subsample analyses indicated that for elementary school youths, participation in community programs decreased as school risks increased, whereas school risks were not significantly associated with the odds of community program or scouting participation for high school youths.

For both school activities and service clubs, the school risk index was strongly and positively linked to school activity participation for elementary school youths, but these relations were attenuated among middle and high school youths. In fact, subsample analyses revealed no relation between school risk and participation in school activities and service clubs for middle school and high school youths, but the subsamples did reveal an increase in school activity and service club participation for elementary school youths as the number of school-level risks increased. Finally, for any OST participation, school risk was unrelated to participation overall. After including the school-age interactions, however, the interaction between high school status and the school risk indicator was significant but positive. This indicates that for high school students, greater numbers of school risks were associated with increased likelihood of participating in any OST activity. Taken together, youths in larger high-poverty schools were sometimes more likely to participate in certain activities and less likely to participate in other activities, but findings were inconsistent and dependent on the type of activity and age group of youths.

Neighborhood risk. Neighborhood risk was unrelated to participation in lessons, service clubs, and participation in any OST activity but was significantly related to the four remaining OST indicators. For community programs, the interaction between middle school status and neighborhood risk was positive and significant. This indicated that for middle school students, living in a higher- risk neighborhood was associated with an increased likelihood of participating in community programs; there was little or no association for elementary students or high school students. For athletics, the interactions as well as the subsample analyses indicated that for elementary school youths, athletic participation decreased as neighborhood risks increased, while this was not true for middle school and high school youths. Increases in neighborhood risks were also associated with lower odds of participating in scouts overall. The main effect of neighborhood risk was no longer statistically significant, however, once the interactions with school age were entered in Model 2, although this is likely due to a decrease in statistical power. Youths from riskier neighborhoods were also more likely to select into school-based activities overall, although like with scouting this relation disappeared once interactions with school age were included, again likely because of the decrease in statistical power. In sum, living in neighborhoods with greater risks was associated with decreased odds of scouting and athletics (especially for younger youths) but increased odds of participating in community programs and school activities (the former for middle school students only).

Socioeconomic Status As a Moderator

To examine whether risk operated differently for families of varying socioeconomic status, we refit Model 1 including interaction terms for each risk index by family income. For brevity, Table 4 presents only the interaction terms for each of the OST indicators. The three OST activities for which there were consistent and numerous significant interaction effects were athletic programs, lessons, and participation in any OST program or activity. Here, virtually all interactions between neighborhood, school, and family risk indices and the socioeconomic status indicators were significant and negative, indicating that as family income rises, the relation between risk and OST participation becomes more negative. That is, as income increased, the number of problems seemed to depress participation to a greater extent. Plots of these interactions7 revealed that risk at the family, school, and neighborhood levels was generally unrelated to participation in athletics and lessons for low-income students, but risk at these levels was negatively related to participation for higher-income students.

Discussion

The current study adds to existing literature by showing that youths with the most developmental problems and least supportive contexts-youths who some have argued are most in need of OST opportunities and can most benefit from them (e.g., Mahoney, 2000; Marsh & Kleitman, 2002)-are least likely to participate, especially in adolescence. Our study also confirms the importance of focusing on developmental and economic differences when examining OST participation. Specifically, we found that child and contextual factors are associated with the propensity to participate among youths in middle school and high school but not younger youths. Child and family factors were most consistently linked with OST participation, but only among adolescents. These relations were concentrated in athletics, lessons, school-based activities, and service clubs. The child risk index was composed of indicators such as low achievement, high behavior problems, and poor physical health. Previous research points to some potential explanations at the child and program levels. Given that adolescents are often granted more autonomy and decision making in their OST participation choices (Gauvain & Ferez, 2005), it makes sense that youths who are experiencing social, academic, or physical difficulties would be more likely to decide not to participate in prosocial contexts such as OST activities. Youths with social difficulties, such as being withdrawn, may choose solitary activities due to uneasiness with group situations or fear of rejection. Youths with other social difficulties, such as delinquent behavior or tendencies, may spend more time in unsupervised contexts (Osgood & Anderson, 2004). At the programmatic level, youths with academic or physical difficulties may be unable to meet the minimum requirements of activities (e.g., grade requirements), may not be encouraged to participate, or potentially may be actively discouraged from participating by relevant gatekeepers. The consistency in these findings was still somewhat surprising, however, given that child risk indices were measured five years earlier than participation.

The family risk index was composed of indicators that assessed the lack of parents’ investments in children (e.g., low investment in youths’ education) and parental resources (e.g., low support, greater mental health vulnerabilities). We had expected that family factors would be less critical for adolescents’ participation given increases in adolescent autonomy, but our analyses revealed that the opposite was the case. Here, we draw upon family research to speculate why these relations emerged. One possible explanation is that the various family risk measures (e.g., parent distress, conflict, economic strain) are all indicators that predict adolescent adjustment as evident in research by McLoyd (1998). Thus, our measure of family risk may have been a powerful predictor in adolescence because it captured additional developmental risks not captured in the child risk index. Another possibility is that family risk operates in a cumulative fashion, such that the influence of family-level factors in suppressing OST participation does not emerge until after childhood. While this is speculative, other research examining relations between the family context of OST participation (Larson, Dworkin, & Gillman, 2001; Mahoney & Stattin, 2000; Simpkins, 2005; Simpkins, Davis-Kean, & Eccles, 2005) confirms that family or parenting factors matter for OST participation but in complex ways. Taken together, these findings suggest that adolescents with higher clusters of disadvantage at the personal and family levels were the least likely to participate in a variety of structured OST contexts. In other words, those who could benefit most from the resources and structure of OST activities in adolescence are least likely to participate.

Neighborhood and school factors were less likely to be associated with participation than child and family characteristics. Neighborhood risk was, however, linked with lower athletic participation, particularly for younger youths, and higher participation in school activities for older youths. The results for neighborhoods and athletics may result from the fact that older youths are more mobile and able to take advantage of participation opportunities regardless of neighborhood boundaries and conditions. In contrast, elementary school youths may be more constrained by the availability of opportunities in their immediate neighborhood. School factors were inconsistently and idiosyncratically related to youths’ OST participation, although in some cases greater numbers of school-level risks were associated with increased propensity to enroll in certain OST activities.

Positive findings between greater numbers of risks at the school- and neighborhood-levels may suggest that controlling for socioeconomic and child- and family-level factors, families and youths may actively select into certain OST settings to compensate for the lack of opportunities in youths’ schools and neighborhoods. Alternatively, the weaker findings might emerge because our risk indicators may not have captured important school or neighborhood influences. Our school risk measure is the weakest measure of risk. Because the PSID-CDS has minimal school data, we had to rely on data from the National Center for Education Statistics (NCES), which does not contain as rich data on schools as the general PSID-CDS questionnaires do for the child, family, and neighborhood contexts. Unfortunately, distinguishing between these or other possible explanations for the school findings is beyond the scope of the current study and must be examined in more depth in further research.

Finally, we found that the number of contextual risks present at the family, school, and neighborhood levels seemed to matter less for lowincome youths’ participation in athletics and lessons, which are activities that likely entail significant monetary costs for families. This is not to say that families, schools, and neighborhoods do not matter for low-income youths’ OST participation but rather that for these specific activities, lowincome families may simply not be able to afford these costs regardless of the levels of risk in their families, schools, and neighborhoods. For lowincome youths and their families, access to high-quality affordable athletics and lessons may be problematic, even if they come from more engaged families, live in safer neighborhoods, and attend less disadvantaged schools. School- and community-based organizations offering these types of activities should seriously consider offering scholarships and sliding scale fees in order to overcome any financial barriers to youths’ participation. Policymakers could also consider requiring programs that receive public funding to guarantee a certain number of free slots to low- SES youths. Because of the theoretical and policy relevance of family income, we chose to focus only on family income as a potential moderator of personal and contextual risks, but future research should also consider the potential moderating effects of other demographic factors, such as race, ethnicity, or child gender.

In interpreting the results of this study, a few limitations should be acknowledged. First, the data are primarily cross- sectional. While it seems more likely that contextual risks would influence propensities to participate than vice versa, it is theoretically possible that the causal direction could flow the other way. For example, OST participation may lead to increases in parents’ involvement in youths’ education and development (e.g., U.S. Department of Education, 2003). Thus, longitudinal data will be necessary to further clarify the temporal direction of the relationships. However, we did model child risk five years prior to OST outcomes, leading us to conclude that the many child-risk related findings were even more impressive given this substantial time lag. second, future research should also build on our findings about the number of risks present at each level by examining which specific risks are important for influencing families’ and youths’ OST participation decisions. Person-oriented analyses, which create categories of people with different patterns of risk factors, may prove fruitful for understanding exactly how different aspects of a particular context code termine youths’ OS T participation.

Third, it is important to note that our OST measures tap a variety of program and activity participation types that are not mutually exclusive. For instance, school activities may encompass certain types of lessons, athletics, or service clubs. As such, our analysis is unable to make major contributions to understanding differences across types of programs and activities that might be possible with more localized and specific data. This is a tradeoff involved in analyzing national-level survey data on OST programs and activities. A related point is that our study only taps whether or not youths participate in a variety of OST contexts over an extended period of time (12 months) and does not consider the quality, intensity, or indicators of youths’ OST participation. These kinds of questions could be better addressed with smaller-scale studies of specific programs, activities, and their participants. Considering these dimensions of youths’ OST participation will be critical areas for future research.

This study highlights the need for more research on how and why risks at multiple levels (child, family, neighborhood, and school) influence youths’ participation. Our findings demonstrate that risk factors are a significant predictor of participation, particularly for adolescents. Given the opportunities and benefits afforded by OST activities, these findings raise concerns for all stakeholders working to improve the lives of youths. The findings have important implications for recruitment and retention efforts, policies and resource allocation, and future research. 1. About 70% of parent- child dyads agreed about the child’s participation in athletics the previous 12 months. Rather than trust one respondent over the other, we thought it more reasonable to assume that if one of the respondents reported participation, then participation likely occurred.

2. The emotional support reliability coefficient from the HOME scale is low. However, we do not view this as overly problematic. Bradley (2004), one of the authors of the HOME scale, noted that the standard reliability coefficient is often not applicable to the HOME subscales because each subscale taps a variety of indicators that, taken together, capture multiple aspects of emotional support that may or may not be interrelated within homes but nonetheless provide a comprehensive indicator when combined. In other words, the various elements of an emotionally supportive home environment need to be part of a unitary construct to provide an indication of the overall emotional climate in the home. The low reliability of the family involvement in school composite likely results from the same issue, as different parents engage with the school in different ways.

3. All interaction terms were created using centered variables (Aiken & West, 1991).

4. We also completed the same set of analyses with an indicator of overall risk, which was created by summing neighborhood, school, family, and child risk indices. In general, these findings replicated those found when using the four individual indicators of risk. These results are available from the first author upon request.

5. These results are available from the first author upon request.

6. For scouting, likely because only 11% of the full sample reported participating in this OST setting, the subsample analyses were less stable than for other OST contexts after splitting the sample by developmental stage. Thus, it is unclear whether within the high school subsample school risk is actually positively associated with the odds of participation. The coefficient is positive, but the standard error is so large (because scouting is less common among older youths) that the coefficient is statistically indistinguishable from 0.

7. Plots are available from the first author upon request.

References

Aiken, L. S., & West, S. G. (1991). Multiple Regression: Testing and interpreting interactions. Thousand Oaks, CA: Sage.

Ardelt, M., & Eccles, J. S. (2001). Effects of mothers’ parental efficacy beliefs and promotive parenting strategies on inner-city youth. Journal of Family Issues, 22, 944-972.

Barber, B. L., Eccles, J. S., & Stone, M. R. (2001). Whatever happened to the jock, the brain, and the princess? Young adult pathways linked to adolescent activity involvement and social identity. Journal of Adolescent Research, 16, 429-455.

Barker, R. G., &Gump, P. V. (1964). Big school, small school: High school size and student behavior. Stanford, CA: Stanford University Press.

Bartko, W. T., & Eccles, J. S. (2003). Adolescent participation in structured and unstructured activities: A person-oriented analysis. Journal of Youth and Adolescence, 32, 233-241.

Bradley, R. H. (2004). Chaos, culture, and covariance structures: A dynamic systems view of children’s experiences at home. Parenting: Science and Practice, 4, 243-257.

Bronfenbrenner, U. (1979). The ecology of human development: Experiments by nature and design. Cambridge: Harvard University Press.

Caldwell, B. M., & Bradley, R. H. (1984). The Home Observation for Measurement of the Environment. Little Rock: University of Arkansas at Little Rock.

Collins, W. A., Gleason, T., & Sesma, A., Jr. (1997). Internalization, autonomy, and relationships: Development during adolescence. In J. E. Grusec & L. Kuczynski (Eds.), Parenting and the internalization of values (pp. 78-99). New York: Wiley.

Conger, R., & Elder, G. H., Jr. (1994). Families in troubled times: Adapting to changes in rural America. New York: Aldine de Gruyter.

Cooper, H., Valentine, J. C., Nye, B., & Lindsay, J. (1999). The relationship between five after-school activities and academic achievement. Journal of Educational Psychology, 91, 369-378.

Bearing, E., McCartney, K., Weiss, H. B., Kreider, H., & Simpkins, S. (2004). The promotive effects of family educational involvement for low-income children’s literacy. Journal of School Psychology, 42, 445-460.

Eccles, J. S., & Gootman, J. (Eds.). (2002). Community programs to promote youth development. Washington, DC: National Academy Press.

Eccles, J. S., & Templeton, J. (2002). Extracurricular and other after-school activities for youth. Review of Research in Education, 26, 113-180.

Furstenberg, F. F., Cook, T. D., Eccles, J., Elder, G. H., & Sameroff, A. (1999). Managing to make it: Urban families and adolescent success. Chicago: University of Chicago Press.

Gauvain, M. (1999). Everyday opportunities for the development of planning skills: Sociocultural and family influences. In A. Goncu (Ed.), Children’s engagement in the world: Sociocultural perspectives (pp. 173-201). Cambridge: Cambridge University Press.

Gauvain, M., & Perez, S. M. (2005). Parent-child participation in planning children’s activities outside of school in European American and Latino families. Child Development, 76, 371-383.

Gutman, L. M., Sameroff, A. J., & Eccles, J. S. (2002). The academic achievement of African American students during early adolescence: An examination of multiple risk, promotive, and protective factors. American Journal of Community Psychology, 30, 367-399.

Jarrett, R. L., & Jefferson, S. (2003). “A good mother got to fight for her kids”: Maternal management strategies in an urban housing project. Journal of Children and Poverty, 9, 21-39.

Jordan, W. J., & Nettles, S. M. (2000). How students invest their time outside of school: Effects on school-related outcomes. Social Psychology of Education, 3, 217-243.

Kessler, R. C., Andrews, G., Colpe, L. J., Hiripi, E., Mroczek, D. K., Normand, S. L. T., Walters, E. E., & Zaslavsky, A. (2002). Short screening scales to monitor population prevalences and trends in nonspecific psychological distress. Psychological Medicine, 32(6), 959-976.

Lareau, A. (1989) Home advantage: Social class and parental intervention in elementary education. New York: Falmer.

Larson, R., Dworkin, J., & Gillman, S. (2001). Facilitating adolescents’ constructive use of time in one-parent families. Applied Developmental Science, 5, 143-157.

Leventhal, T., & Brooks-Gunn, J. (2000). The neighborhoods they live in: Effects of neighborhood residence on child and adolescent outcomes. Psychological Bulletin, 126, 309-337.

Lind, C., Relave, N., Deich, S., Grossman, J., & Gersick, A. (2006). The costs of out-of-school time programs: A review of the available evidence. Washington, DC: The Finance Project.

Lindsay, P. (1984). High school size, participation in activities, and young adult social participation: Some enduring effects of schooling. Educational Evaluation and Policy Analysis, 6, 73-83.

Mahoney, J. L. (2000). School extracurricular activity participation as a moderator in the development of antisocial patterns. Child Development, 71, 502-516.

Mahoney, J. L., & Cairns, R. B. (1997). Do extracurricular activities protect against early school dropout? Developmental Psychology, 32, 241-253.

Mahoney, J. L., Harris, A. L., & Eccles, J. S. (2006). Organized activity participation, positive youth development, and the over- scheduling hypothesis. Social Policy Report, 20(4), 1-32.

Mahoney, J. L., Larson, R. W., & Eccles, J. S. (2005). Organized activities as contexts of development: Extracurricular activities, after-school and community programs. Mahwah, NJ: Lawrence Erlbaum.

Mahoney, J. L., Lord, H., & Carryl, E. (2005). An ecological analysis of afterschool program participation and the development of academic performance and motivational attributes for disadvantaged children. Child Development, 76, 811-825.

Mahoney, J. L., & Stattin, H. (2000). Leisure activities and adolescent antisocial behavior: The role of structure and social context. Journal of Adolescence, 23, 113-127.

Mainieri, T. (2004). The Panel Study of Income Dynamics Child Development Supplement: User Guide for CDS-II. Ann Arbor, MI: Institute for Social Research.

Marsh, H. W., & Kleitman, S. (2002). Extracurricular school activities: The good, the bad, and the nonlinear. Educational Review, 72, 464-514.

McLaughlin, M. W., Irby, M. A., & Langman, J. (1994). Urban sanctuaries: Neighborhood organizations in the lives and futures of inner-city youth. San Francisco: Jossey-Bass.

McLoyd, V. C. (1998). Socioeconomic disadvantage and child development. American Psychologist, 53, 185-204.

Morenoff, J., Sampson, R. J., & Raudenbush, S. (2001). Neighborhood inequality, collective efficacy, and the spatial dynamics of urban violence. Criminology, 39, 517-560.

Osgood, D. W., & Anderson, A. L. (2004). Unstructured socializing and rates of delinquency. Criminology, 42, 519-549.

Parke, R. D., & Buriel, R. (1998). Socialization in the family: Ethnic and ecological perspectives. In W. Damon (Series Ed.) & N. Eisenberg (Vol. Ed.), Handbook of child psychology: Vol. 3. Social, emotional, and personality development (5th ed., pp. 601-689). New York: Wiley.

Pearlin, L. I., Lieberman, M. A., Menaghan, E. G., & Mullan, J. T. (1981). The stress process. Journal of Health and Social Behavior, 22, 337-356.

Perkins, D. F., & Borden, L. M. (2003). Positive behaviors, problem behaviors, and resiliency in adolescence. In R. M. Lerner and A. M. Easterbrooks (Eds.), Handbook of psychology: Developmental psychology, Vol. 6 (pp. 373-394). New York: Wiley.

Peterson, J. L., & Zill, N. (1986). Marital disruption, parent- child relationships, and behavioral problems in children. Journal of Marriage and the Family, 48, 295-307.

Petit, G. S., Laird, R. D., Bates, J. D., & Dodge, K. A. (1997). Patterns of after-school care in middle childhood: Risk factors and developmental outcomes. Merrill-Palmer Quarterly, 43, 515-538. Polit, D. (1998). The Positive Behavior Scale. Saratoga Springs, NY: Humanalysis.

Posner, J. K., & Vandell, D. L. (1994). Low-income children’s after-school care: Are there beneficial effects of after-school programs? Child Development, 65, 440-456.

Posner, J. K., & Vandell, D. L. (1999). After-school activities and the development of low-income urban children: A longitudinal study. Developmental Psychology, 35, 868-879.

Quiroz, P. (2000). A comparison of the organizational and cultural contexts of extracurricular participation and sponsorship in two high schools. Educational Studies, 31, 249-275.

Rankin, B. H., & Quane, J. M. (2002). Social contexts and urban adolescent outcomes: The interrelated effects of neighborhoods, families, and peers on African-American youth. Social Problems, 49, 79-100.

Roeser, R. W., & Peck, S. C. (2003). Patterns and pathways of educational achievement across adolescence: A holistic- developmental perspective. New Directions for Child and Adolescent Development, Number 101, Person-Centered Approaches to Studying Development in Context, 39-62.

Roth, J. L., Brooks-Gunn, J., Linver, M. R., & Hofferth, S. L. (2003). What happens during the school day? Time diaries from a national sample of elementary school teachers. Teachers College Record, 105, 317-343.

Rutter, M. (1979). Protective factors in children’s responses to stress and disadvantage. In M. W. Kent & J. E. Rolf (Eds.), Primary Prevention of Psychopathology, Vol. 3: Social Competence in Children (pp. 49-74). Hanover, NH: University Press of New England.

Sameroff, A. J. (1983). Developmental systems: Contexts and evolution. In P. H. Mussen (Series Ed.) & W. Kessen (Vol. Ed.), Handbook of child psychology: Vol. 1. History, theory, and methods (4th ed., pp. 237-294). New York: Wiley.

Sameroff, A. J., Bartko, T., Baldwin, A., Baldwin, C., & Seifer, R. (1998). Family and social influences on the development of child competence. In M. Lewis & C. Feiring (Eds.), Families, risk, and competence (pp. 161-185). Mahwah, NJ: Lawrence Erlbaum.

Sameroff, A. J., & Chandler, M. J. (1975). Reproductive risk and the continuum of caretaking casualty. In F. D. Horowitz, M. Hetherington, S. Scarr-Salapatek, & G. Siegel (Eds.), Review of child development research: Vol. 4 (pp. 187-244). Chicago: University of Chicago Press.

Sameroff, A. J., Seifer, R., Baldwin, A. & Baldwin, C. (1993). Stability of intelligence from preschool to adolescence: The influence of social and family risk factors. Child Development, 64, 80-97.

Seidman, E., Aber, J. L., Allen, L., French, S. E. (1996). The impact of the transition to high school on the self-system and perceived social context of poor urban youth. American Journal of Community Psychology, 24, 489-515.

Shumow, L., & Lomax, R. (2002). Parental efficacy: Predictor of parenting behavior and adolescent outcomes. Parenting: Science and Practice, 2, 127-150.

Simpkins, S. (2005, October). Supporting children’s development in and out of the classroom. In P. Davis-Kean (Chair), The changing role of parenting across the life span. Symposium conducted at the annual meeting of the Society for the Study of Human Development, Asilomar, CA.

Simpkins, S. D., Davis-Kean, P. E., & Eccles, J. S. (2005). Parents’ socializing behavior and children’s participation in math, science, and computer out-of-school activities. Applied Developmental Science, 9, 14-30.

Simpkins, S. D., Fredricks, J., Davis-Kean, P., & Eccles, J. S. (2006). Healthy minds, healthy habits: The influence of activity involvement in middle childhood. In A. Huston and M. Ripke (Eds.), Developmental contexts in middle childhood: Bridges to adolescence and adulthood (pp. 283-302). New York: Cambridge University Press.

U.S. Department of Education, Office of the Undersecretary. (2003). When schools stay open late: The national evaluation of the 21st-century Learning Centers program, first year findings. Washington, DC: U.S. Department of Education.

U.S. General Accounting Office. (1996). School facilities: America’s schools report differing conditions. GAO/HEHS-96-103. Washington, DC: U.S. General Accounting Office.

Vandell, D. L., & Shumow, L. (1999). After-school child care programs. Future of Children, 9, 64-80.

Weiss, H. B., Mayer, E., Kreider, H., Vaughan, M., Dearing, E., Hencke, R., & Pinto, K. (2003). Making it work: Low-income working mothers’ involvement in their children’s education. American Educational Research Journal, 40, 879-901.

Woodcock, R. W, & Johnson, M. B. (1989). Tests of Achievement, Standard Battery. Chicago: Riverside.

Christopher Wimer, Stanford University

Sandra D. Simpkins, Arizona State University

Eric Dearing, Boston College

Suzanne M. Bouffard, Pia Caronongan, and Heather B. Weiss, Harvard University

Christopher Wimer, Center for the Study of Poverty and Inequality, Stanford University; Sandra D. Simpkins, School of Social and Family Dynamics, Arizona State University; Eric Dearing, Department of Counseling, Developmental, and Educational Psychology, Lynch School of Education, Boston College; Suzanne M. Bouffard, Harvard Graduate School of Education, Harvard University; Pia Caronongan, Harvard Graduate School of Education, Harvard University; and Heather B. Weiss, Harvard Graduate School of Education, Harvard University.

This research was supported with a grant from the William T. Grant Foundation to Heather Weiss, Sandra Simpkins, and Eric Dearing. We would like to thank Robert Granger, Priscilla Little, and Holly Kreider for their comments on earlier drafts. An earlier version of this paper was presented at the Society for Research on Adolescence Biennial Meeting in San Francisco, California, March 2006.

Correspondence concerning this article should be addressed to Christopher Wimer, Stanford University, Sociology Department MC2047, Main Quad 450 Serra Mall, Building 120, Room 160, Stanford, CA 94305- 2047. E-mail: cwimer@stanford.edu.

Merrill-Palmer Quarterly, April 2008, Vol. 54, No. 2, pp. 179- 207. Copyright (c) 2008 by Wayne State University Pres