Diverse Outcomes for a Diverse Population: Findings About Homeschooled Adults from the Cardus Education Survey

2350 adults aged 24-39, of whom 181 were home-schooled for at least one year (roughly 60 short-term, 60 medium-term, 61 long-term)

Nationally representative sample of US adults aged 24-39, fielded 10-22 October 2023 by Ipsos Public Affairs on the KnowledgePanel probability panel

A probability-panel sample rather than a homeschool-network volunteer sample, which again puts it in the small set of adult-outcome studies whose sampling frame is not selected on enthusiasm for home education. Its methodological contribution is dose: instead of a single homeschool dummy it splits exposure into short-term (1-2 years), medium-term (3-7) and long-term (8-13), which turns out to matter enormously and which almost no other study in the field does. Otherwise weak — cross-sectional, retrospective self-recalled schooling history, all outcomes self-reported at one sitting, subgroup cells of about 60, no adjustment reported for family background in the headline comparisons, and a grey-literature report from an organisation (Cardus) that publishes on the value of religious and alternative education. Grade D. The non-monotonic pattern is the tell: medium-term homeschoolers show the WORST mental-health indicators while long-term show the best, which is what you expect if duration is proxying for why the family homeschooled — crisis-driven withdrawal in the middle band, conviction-driven home education at the long end — rather than measuring a dose of treatment.

Long-term homeschoolers (8-13 years) look better than never-homeschooled adults on nearly every social, civic and psychological outcome measured: 47% volunteered in the past year vs 27%; 50-58% of short- and long-term homeschoolers gave to charity vs 42%; 54% of medium-term homeschoolers had held a volunteer leadership position; 20% of long-term homeschoolers say they often feel helpless dealing with life's problems vs about 33% of everyone else; they report the lowest depression and anxiety, the highest optimism, gratitude (97%) and life satisfaction (4.38 on a 6-point scale); 65% are currently married vs 40%, with a 5% divorce rate vs 9% and more children on average (1.23 vs 0.84). They are also far more religious (72% believe in God vs 41%; 54% attend weekly). The critical qualifier is that these advantages attach to LONG-DURATION homeschooling only and do not increase monotonically with dose — medium-term homeschoolers post the lowest mental-health scores in the study. Read as evidence about socialisation, this contradicts the isolation concern in its strong form for the committed-homeschooling subgroup, and says nothing reassuring about the substantial fraction of children who are homeschooled for a few years.

High and essentially untreated. Long-duration home education is chosen by intact, religious, high-conviction families, and the outcome list — marriage, low divorce, more children, religious attendance, life satisfaction — is close to a direct readout of that family type. No sibling, within-family or instrumented comparison; the headline contrasts are group means, not covariate-adjusted estimates.

Directly contradicts Sikkink & Skiles' 2015 analysis of the earlier Cardus waves on volunteering, giving and helplessness — same survey programme, opposite sign, and the contradiction is unexplained. Also contradicts Hill & den Dulk 2013 (NSYR panel, homeschool volunteering odds ~0.42 of public). Consistent with Chen et al.'s outcome-wide Nurses' Health Study II analysis on adult volunteering. Three probability samples now disagree on the single most-measured adult social outcome in this literature; nothing has adjudicated between them, and the duration split proposed here is the most promising but wholly untested reconciliation.