Marginalized Learners in International and Regional Test Data: The Extent of Floor Effects

International assessments inform education policy debates, yet little is known about their floor effects: To what extent do they fail to differentiate between the lowest performers, and what are the implications of this? TIMSS, SACMEQ, and LLECE data are analyzed to answer this question. In TIMSS, floor effects have been reduced through the introduction of a greater number of easy items. SACMEQ and LLECE, despite being specifically designed for developing countries, display substantial floor effects, a problem obscured by their strong reliance on multiple-choice questions. Consequently, some learners are officially classified as proficient who effectively score zero points beyond random guessing. Though floor effects do not substantially alter country rankings, they are large enough to distort trends over time. Designers of assessment programs need to limit floor effects through a greater number of easier multiple-choice items and more constructed-response items. The former solution is easiest to implement.