Homeschooling and Young Adult Outcomes: Evidence from the 2011 and 2014 Cardus Education Survey

pooled 2011 + 2014 Cardus Education Survey: 141 religious home-schoolers, 60 nonreligious home-schoolers, 1,771 public-school graduates, out of 3,000+ respondents

US adults aged 24-39 who graduated high school between roughly 1990 and 2008; probability panels fielded by Knowledge Networks/GfK

The methodological point that matters: these are randomly sampled respondents, not volunteers recruited through homeschool networks, which is what separates Cardus from almost every other adult-outcome dataset in this field (Ray's alumni surveys in particular). That makes it the natural check on the favourable consensus. Everything else is weak — cross-sectional retrospective self-report, schooling type recalled by the adult, a homeschool cell of ~200 split across two motivational subgroups, and no ability to model the fluidity between homeschooling and institutional schooling that characterises most homeschooling careers. The 2011 wave's homeschool cells were tiny (61 religious, 21 nonreligious); pooling with 2014 is what makes the estimates usable at all. Grade D. Reviewers also note a plausible selection story running the other way from the usual one: children are often withdrawn to homeschool BECAUSE they were not thriving, which would bias this sample toward worse outcomes. The archive should hold both selection stories simultaneously and conclude the design cannot adjudicate. One further limit on what is extractable: the report publishes NO coefficient tables, standard errors or confidence intervals, so magnitudes beyond "about one year" of postsecondary education are not recoverable. The control set is nonetheless substantial — both parents' education, parental academic involvement, religious attendance, civic involvement, intact family, race, gender, age, sibship, citizenship, and an ever-attended-public-school indicator. The authors' own caveat, verbatim: "Selection effects cannot be ruled out. Home education is often an alternative for families and children who are not thriving in institutional schooling... Panel survey data would help sort out selection from schooling effects; without that, we can only report an association."

Homeschool graduates in a random sample look worse on the social and civic outcomes, not better. They volunteer less often and give less to charity than public-school graduates; they report lower trust in neighbours, in strangers, and in institutions including government and media; they participate less in religious institutions than their upbringing would predict. Psychologically, they are more likely to report a weak sense of direction and purpose in life and to feel helpless in dealing with life's problems. On family formation the picture is mixed — they marry at similar ages with comparable divorce rates, but a larger share never marry, which the authors gloss as a distaste for institutions possibly extending to marriage. They also complete fewer years of postsecondary education and attend smaller, less selective institutions. This is the single most important counterweight in the socialisation literature: it is the only large random-sample adult dataset outside Cardus's own later wave, and it does not reproduce the favourable picture that parent-report and alumni-volunteer studies produce.

High and untreated. Selection into homeschooling is not modelled, and the authors concede selection bias cannot be ruled out in either direction. Reduced institutional trust is precisely the trait that plausibly caused the parents to homeschool in the first place, so the most striking civic finding may be the parents' disposition inherited and transmitted rather than an effect of the arrangement.

Directionally replicated on volunteering by Hill & den Dulk 2013 in a completely different national panel (NSYR), which is the strongest cross-dataset agreement in this literature. Directly CONTRADICTED on volunteering, giving and helplessness by Cheng & Watson's 2025 Cardus wave, which uses the same survey brand and a fresh 2023 Ipsos panel and reports long-term homeschoolers volunteering MORE and feeling helpless LESS. Same instrument family, opposite sign, a decade apart — the most plausible reconciliations are cohort change in who homeschools and the 2015 report's failure to separate short-term from long-term exposure, but neither has been tested. Flagged `mixed` for exactly this reason.