Alternative Schooling Strategies and the Religious Lives of American Adolescents

3217 adolescents, of whom 76 were home-schooled

US adolescents aged 13-17, Wave 1 of the National Study of Youth and Religion (2002-2003), a nationally representative telephone survey

Nationally representative cross-sectional data with a rich control stack (parental religiosity and affiliation, marital status, education, homeownership, assets, income, adolescent demographics, autonomy, family and school transitions, educational aspirations). Its relevance here is indirect: the paper's outcomes are religious (service attendance, religious-education class, youth group, salience of faith, private devotional activity), not social functioning. What makes it worth recording is that its three MEDIATORS are socialisation variables measured in a probability sample — proportion of friends who are religious (friendship networks), intergenerational network closure, and the number of adults the adolescent can turn to for support. If home education systematically thinned adolescents' adult and peer networks, this is one of the few datasets in which it would show. The homeschool cell is 76, which is small for the interactions the paper tests, and the design is cross-sectional so no causal reading is available.

Home-schooled adolescents did not differ significantly from public-school adolescents on any religious outcome examined, and the paper reports that friendship networks, network closure and adult mentors played only a very limited role in mediating any schooling-type effect. The one homeschool-specific pattern reported is an interaction with parental religiosity rather than a main effect: home-schooled adolescents of nonreligious parents reported lower private religious activity, reversing as parental religiosity rose. For this archive's purposes the informative null is the negative space — a nationally representative dataset with explicit measures of network closure and available adult mentors produced no headline finding that home-schooled adolescents were network-deprived, but the paper never reports those descriptives by schooling type, so it cannot be used as positive evidence either way. It is recorded so the gap is visible rather than silently absent.

High for the religious outcomes (parental religiosity drives both the schooling choice and the child's faith) and untested for the network measures, which are not reported by schooling type at all.

Hill & den Dulk 2013 use the same NSYR panel and find a large negative homeschool coefficient on sustained volunteering, so the two papers are not independent samples and their apparent disagreement (null vs negative) is a difference of outcome, not of data.