Religion, Volunteering, and Educational Setting: The Effect of Youth Schooling Type on Civic Engagement

1717 (analysis sample = NSYR panel respondents who reported any volunteering at Wave 1, ages 13-17, followed to Wave 3, ages 18-23; homeschool subgroup is a small fraction of this and its size is never stated in the paper)

US adolescents in the National Study of Youth and Religion, a nationally representative telephone panel (Wave 1 n = 3,290 teens 2002-2003; Wave 3 2007-2008; W1-W3 retention 77.1%). Secondary schooling type reported by a parent at Wave 1: public, Catholic, Protestant, private nonreligious, homeschool, other.

The best-designed source in this archive on homeschooling and adult civic behaviour, and it is still observational. Strengths: a genuine national probability panel, a prospectively measured predictor, an adult outcome, and an unusually aggressive control stack — parental SES and ideology, race, age, gender, religious tradition, parent and youth worship attendance, private devotionalism, youth-group attendance, parental encouragement and modelling of volunteering, whether volunteering was required, parent-rated peer quality, Wave 3 religious activity, postsecondary enrolment, living arrangement and employment. Six nested logistic models test spuriousness, opportunity structures and motivation in turn. Weaknesses that cap it at C: no random or quasi-random assignment, so selection into homeschooling by unmeasured family traits remains the live alternative explanation; the sample is restricted to adolescents who already volunteered, so the estimand is *sustaining* volunteering, not generating it; the homeschool cell is small enough that the paper reports a 95% CI running roughly 0.08-0.40 on the predicted probability; and volunteering is self-reported at both waves.

Homeschooled adolescents were the least likely group to volunteer at all in adolescence (37% reported none at Wave 1 vs 34% public, 13% Catholic) and the gap widened sharply by young adulthood: at Wave 3, 77% of formerly homeschooled respondents reported no volunteering in the past year, against 53% of public-school, 34% Catholic and 23% Protestant. Among those who HAD volunteered as adolescents, the odds of still volunteering at 18-23 were about 0.42-0.46 that of public-school peers, falling to 0.33 in the fullest model — the effect survives, and slightly strengthens under, controls for family religiosity, parental modelling, organized-activity counts and friends' volunteering. Model 6 predicted probabilities: public 48%, Catholic 55%, Protestant 83%, private nonreligious 10%, homeschool 23%. The authors could not explain the gap through opportunity structures or motivation and explicitly say the decline among home-schooled and private-nonreligious emerging adults is not captured by any mechanism they measured. Note the direction: this is one of the few well-controlled datasets in the homeschool literature and it points AGAINST the favourable consensus built from advocacy surveys.

High. Schooling type is chosen by families, and the outcome is a family-transmitted behavioural disposition. The authors control for parental volunteering, parental encouragement and parental religiosity — better than most — but no genetically informative or within-family comparison exists, and the paper's own framing (H1, spuriousness) concedes the point is only partly testable with these data.

No independent replication of the homeschool coefficient exists. The direction is corroborated by a different analysis of a different random-sample dataset — Sikkink & Skiles' Cardus work also finds homeschool graduates volunteering and giving less than public-school graduates — but Cheng & Watson's 2025 Cardus wave reports the OPPOSITE sign for long-term homeschoolers, and Chen et al.'s Nurses' Health Study II analysis reports higher adult volunteering. Three national-ish datasets, three answers: the honest reading is that the adult civic outcome is unsettled and sensitive to cohort, sampling frame and how homeschool exposure is defined.