Cardus Education Survey: Do the Motivations for Private Religious Catholic and Protestant Schooling in North America Align with Graduate Outcomes?

Two US probability panels from Knowledge Networks: the 2010 Science of Generosity Survey (~2,000 randomly sampled Americans, restricted to high school graduates aged 23-49) and a second KN panel of ~1,500 graduates aged 23-40 with a private-school oversample (~1,000 private, ~500 public). The homeschool cells are tiny: 61 religious and 21 non-religious homeschool graduates.

US adults aged 23-40 (and 23-49 in the generosity panel) who graduated from high school, classified by the sector they primarily attended. Homeschoolers are split by whether the respondent's mother attended religious services regularly.

The design's one real virtue is the sampling frame: this is a random internet panel, not a recruited advocacy sample, so it does not carry the selection defect that voids the Rudner/Ray convenience studies. Everything else is weak. Analysis is nested OLS/logit/multinomial regression with public-school graduates as the reference: model 1 has school-sector dummies only, model 2 adds gender, race, age, both parents' education, religious tradition, service attendance and volunteering, closeness to each parent, whether each parent pushed academically, and family structure. There is NO measure of prior achievement or childhood ability, so the adjusted sector coefficients are not treatment effects — everything that led a family to home educate a particular child is still in the residual, and reverse causation (a struggling child withdrawn from school) is untestable. Coefficients are reported ONLY as unlabelled bar graphs, with significance declared at p<.1 because of small cells; no numeric tables, standard errors or CIs are published, so effect magnitudes cannot be extracted from the report. The homeschool n of 61 religious graduates is the binding constraint on everything reported about them. Cardus is a Christian think tank whose case for religious schooling the survey was designed to test; the homeschool findings are incidental to that purpose and are reported for religious homeschoolers only.

Homeschool graduates were more likely than every other sector to attend open-admission (least selective) universities; markedly less likely to attend Carnegie Research I or II universities; showed essentially no advantage on the mean SAT of the institutions they attended; showed no advantage in years of education (Catholic and non-religious private graduates did show one); and were less likely to obtain advanced degrees. On the sponsor's own religious and moral-formation outcomes the homeschool result is positive — stronger evangelical identity, belief in biblical inerrancy, traditional views on sexuality and family — and on those the report reads homeschooling as succeeding. On the achievement and attainment side the direction is flatly opposite to the advocacy literature, from a random sample rather than a recruited one. Downstream reviewers add that homeschool graduates reported lower college preparedness, more feelings of helplessness and less clarity of life direction, and lower political engagement.

High. Parental education, religiosity, family structure, closeness and academic pressure are all adjusted, which is more than most of this literature manages, but no measure of the child's prior ability or achievement exists, and the decision to home educate is endogenous to exactly that. The tiny homeschool cell also means the adjusted coefficients are estimated with enormous imprecision, which the p<.1 threshold papers over rather than fixes.

The 2014 wave of the same survey deliberately withheld its homeschool findings. Sikkink & Skiles 2015 pooled the 2011 and 2014 waves (n=201 homeschoolers) and reproduced the attainment result with a larger cell — about one year less postsecondary education than public school graduates, net of controls. That is the same instrument, same team and overlapping data, so it is extension rather than independent replication. Chen, Hinton & VanderWeele 2021, on a completely different cohort (GUTS) with baseline adjustment, independently finds the same direction on college degree attainment (RR 0.77).