Shadow Education, American Style: Test Preparation, the SAT and College Enrollment

NELS:88, US high-school class of 1992; OLS with controls including prior achievement

US high-school students, self-reported test-preparation use

Observational analysis of the SAME NELS:88 data Briggs used, with a sociological framing (test preparation as "shadow education" and a stratification mechanism) rather than a causal-inference framing. Its treatment of self-selection is a rich covariate set, which Briggs had already shown is insufficient - each covariate block shrinks the estimate without stabilising it.

Reported effects of about +30 SAT composite points for a private test-preparation course and about +37 for a private tutor, with books, video, computer software and high-school courses at roughly nil after controls. Preparation use is strongly stratified by income and parental education, and predicts selective-college enrolment. The null on books and software is the substantively interesting part and is consistent with the Hao 2025 finding that drilling sample items alone does little. Recorded as excluded because it estimates the same parameter on the same dataset as Briggs with a weaker treatment of selection, and its larger point estimates are what a weaker treatment of selection predicts.

High. Purchase of test preparation is one of the most income- and parental-education-stratified behaviours in the dataset, and both correlate with heritable ability; controls cannot remove this.

Same dataset as Briggs (2001), so agreement between them is not independent replication. No independent dataset has reproduced the +30/+37 magnitudes.