Parent Tutoring: A Meta-Analysis

37 studies — 32 comparisons from 20 group-design studies and 25 comparisons from 17 single-subject design studies

K-12 children whose own parent delivered a specific academic intervention at home, mostly in reading and mathematics; heavy representation of struggling learners and school-psychology referrals.

Education and Treatment of Children 29(1), 79-106. The only meta-analysis in the literature whose unit of analysis is exactly this archive's question — the PARENT as the instructor delivering a specified academic procedure to their own child, rather than "parental involvement" in the diffuse sense. Two separate syntheses were run because the underlying literature is split between group designs (standardized mean difference) and single-subject designs (percentage of non-overlapping data). Outcomes were analysed by grade level, skill area, training features, treatment fidelity, and assessment type — which is precisely the moderator set needed here, and precisely what could not be extracted without the full text.

Parents trained to deliver a specific academic intervention to their own child produce a mean weighted effect of +0.55 across 20 group-design studies, and a median PND of 94% across 17 single-subject studies — effects "generally positive across both types of studies". A secondary source (Fernandez-Alonso et al. 2022, citing Erion) reports the subject breakdown as mathematics d = 0.59 and reading comprehension d = 0.57. Read against the tutoring literature this archive already holds, +0.55 sits ABOVE Nickow, Oreopoulos & Quan's nonprofessional/volunteer tutor cell (g = 0.17) and above their pooled 0.288 — which is the reason to distrust it rather than to celebrate it. The corpus is small-n, single-subject-heavy, largely unrandomized, and typically uses curriculum-based or treatment-aligned measures rather than independent standardized tests; PND in particular has no defensible mapping onto a standardized-test effect size.

High. Parents who volunteer to be trained and then actually run a structured tutoring protocol several times a week for months are a strongly selected group, and the single-subject designs have no between-family comparison at all. Group designs with randomization are a minority of the corpus. The +0.55 should be read as an upper bound on the best case of parent delivery under supervision, not as an estimate of what parent delivery does on average.

Kupzyk, LaBrot & Collins 2023 updated the review for 2006-2020 and found the literature had barely grown or improved — 30 studies, still 20 of them single-subject. Prior reviews classified parent tutoring as a "promising practice" with methodological limitations, and the 2023 update did not upgrade that. The magnitude does NOT replicate against the better-designed tutoring literature: Nickow, Oreopoulos & Quan 2024 (89-96 RCTs) puts nonprofessional tutors at g = 0.17 and the whole field at 0.288, and Ritter et al. 2009 (RCT-only volunteer tutoring) at 0.30. Erion's +0.55 is a small-study, aligned-measure number and the archive's own effect-size discipline treats it as a red flag rather than a result.