Demographics are Predictive of Child Abuse and Neglect but Homeschool Versus Conventional School is a Non-issue: Evidence from a Nationally Representative Survey

1253 adults (527 public, 293 private, 150 homeschool with more than six years in the sector, plus mixed-sector respondents)

US adults reporting retrospectively on abuse and neglect experienced during their school-age years; online survey administered by Barna Group, September-October 2021

The only peer-reviewed study that puts a homeschool comparison group next to a conventional-school comparison group on maltreatment, which is why it gets recorded rather than dismissed — and it is also the study with the largest conflict of interest in this file. The first author founded and directs the National Home Education Research Institute, the advocacy research body whose output the reviews in this archive repeatedly flag as activist scholarship; `developer-led` is the correct flag even though the enum was designed for programme developers. Design problems independent of who wrote it: outcomes are adult retrospective self-report of childhood maltreatment, a measure with well-known recall and disclosure bias that runs in the direction of under-reporting by exactly the population least likely to be surveyed at all; recruitment was through an online panel with snowball elements that the published critique says overrepresents certain demographics; the abuse instrument was short and not independently validated; the design is cross-sectional; and the most severely maltreated homeschooled children — the ones who died, or who never reached adulthood in a condition to answer a survey — are structurally absent from an adult-respondent frame. Grade D.

The authors report that school sector is a non-issue for child abuse and neglect once demographics are accounted for: family structure, years in foster care, large family size and household poverty carry the association, and homeschool versus conventional school does not. They further report that the (non-significant) abuse and neglect of homeschooled respondents was perpetrated outside the home rather than by family members. Taken at face value this is the only quantitative evidence with a comparison group on either side of the maltreatment argument, and it points to no elevated risk. Taken at design value it is a self-report survey commissioned by the homeschooling movement's leading advocate, with a frame that cannot observe the cases the concern is actually about. Both readings should be held: the study is the best available on this question AND it is not good enough to settle it. Luck et al. 2025, reviewing the whole field independently, reach that same two-sided conclusion.

Not the operative confound. The operative ones are selection into the survey frame, reverse causation (withdrawal to homeschool often follows a maltreatment investigation, which this design cannot sequence), and survivorship — the most severely harmed children cannot be respondents.

Not replicated. A formal published response (Stewart & McCracken 2023, same journal) attacks the sampling and the instrument; Ray and Shakeel published a further 2023 analysis which Luck et al. 2025 say did not address the methodological objections. There is no independent group that has run this comparison, and the two organisations that have run anything resembling it — NHERI and CRHE — are the advocacy poles of the debate.