Psychometric Quality of Measures of Learning Outcomes in Low- and Middle-Income Countries

Psychometric Quality of Measures of Learning Outcomes in Low- and Middle-Income Countries Psychometric Quality of Measures of Learning Outcomes in Low- and Middle-Income Countries Masha Bertling, Abhijeet Singh, and Karthik Muralidharan Abstract We investigate the properties of measures of learning outcomes, as these are the tools commonly used to monitor the progress toward identifying the most e. [...] All authors participated in conceptualization of the study; MB led the analysis of the microdata in the paper; MB and AS co-wrote the draft. [...] Third, our analysis of the psychometric properties of the test booklets for the studies for which we received item-level information does provide some reassurance about the actual state of practice in most of the literature. [...] For example, PARCC (2017) reports reliability of .91 for literacy and .92 for mathematics, ACT reports reliability of .95 on their literacy test based on a national sample of students who took the ACT in 2018 as part of the state and district testing, or GRE for their 2020 version of the test reports the reliability of .92 for literacy and .95 for mathematics. [...] In addition to the large variation in difficulty and content coverage, we wanted to bring our baseline and endline assessments on the same scale: so, at the time of designing the endline assessments, we kept a substantial share of common items from the baseline, retaining especially those items which were highly discriminatory at all levels of proficiency at the baseline.