The Evidence on Teaching

Effects of Interim Assessments Across the Achievement Distribution

Konstantopoulos S, Li W, Miller SR, van der Ploeg A · 2015

grade Brctindependentunreplicated
Sample
Same Indiana 2009-2010 experiment (~57 schools, K-8)
Population
Indiana K-8 students; mathematics and reading achievement distributions.
Design
Quantile-regression re-analysis of the Indiana interim-assessment experiment, asking whether effects differ across the achievement distribution rather than only at the mean. Secondary analysis of the same randomized data, so it inherits the trial's internal validity.
Key findings
Quantile effects of interim assessments are essentially flat (TOT 0.16-0.22 in math). The widely repeated 'low achievers gain more from benchmark testing' reading is not supported by the size of the differences.
Genetic confound
Minimal (randomized). Lens-relevant: a genuinely compensatory intervention should show a downward-sloping quantile profile, and this one does not — which is evidence against the 'data systems close gaps' story rather than for it.
Replication notes
Secondary analysis of the same trial reported in Konstantopoulos et al. 2016 — treat the two records as one experiment.

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
Treatment-on-treated mathematics effects across quantilesTOT0.16-0.22, roughly FLAT across the ability distributionstandardizedend of yearbusiness-as-usualend-of-treatmentdomain-skill
The paper's own framingclaimthe authors report that lower achievers in grades 3, 5 and 6 appear to benefit more; on the flat quantile band above, this archive reads the differences as too small to support that claimstandardizedend of yearbusiness-as-usualend-of-treatmentdomain-skill

Cited by