Preparing students for college admissions tests

about 1,900 clients of the first author's own commercial tutoring company

Fee-paying SAT tutoring clients in New York, Washington DC and Atlanta

Archival analysis of a commercial tutoring company's own client records, conducted by the company's founder. There is NO uncoached control group of any kind, which means the headline number cannot be separated from ordinary retest gain and maturation - and Powers & Rock measured that counterfactual at 12-16 points per section for uncoached students, which is most of what is reported here as a coaching effect. Recorded as excluded, not because the analysis is careless but because the design cannot answer the question it is used to answer.

Reported gains of 22.6 points per official SAT administration against about 5 points per unofficial practice test, with dose-response for tutoring time-on-task and homework completion, an hour of individual tutoring about 57% more productive than an hour of group instruction, and larger gains from starting earlier in junior year and spacing sessions. Two of those - the individual-versus- group gap and the spacing benefit - are consistent with this archive's tutoring and spaced-practice topics, which is why the paper is worth recording rather than ignoring. But with no control arm the 22.6 points is an upper bound containing the entire retest effect, and the first author owns the company being evaluated.

High. Clients who buy tutoring differ systematically from those who do not on every variable that predicts test performance, and there is no comparison group at all.

No independent replication; no post-2005 randomised trial of any name-brand commercial test-preparation course was located, which is itself a finding about this literature.