Identifying Students at Risk Using a New Math Skills Assessment

Abstract

Math skills are critical for success in economics courses. However, instructors often lack information about the level and extent of their students' underpreparedness in mathematics. In this paper, we describe the development of two new assessments of math skills relevant for introductory and intermediate economics courses. The assessments may be used early in the semester to evaluate students and identify those with weak math skills. With data from introductory and intermediate microeconomics courses, we employ two different methodologies, a simple threshold analysis and a more sophisticated LASSO logit approach, to illustrate how our assessments identify students at risk of underperforming.

Econ-assessments.org: Automated Assessment of Economics Skills

“…The test helps instructors teach students in their economics courses based on the skills they have instead of the skills we hope they have. In addition, the test can identify early in the term those students who will need extra support during the term, and it has been shown to be a strong predictor of performance in introductory courses (Orlov et al 2021b). MESA-Foundations was co-developed by Irene Foster (George Washington University), George Orlov (Cornell), and Doug McKee (Cornell), and has been piloted at Cornell University and George Washington University.…”

Using Machine Learning to Create an Early Warning System for Welfare Recipients*

“…Ex‐ante , it is unclear whether supervised ML algorithms applied to the available administrative data will outperform simpler models or currently used early warning systems in the Australian social welfare system. Certainly, the literature documents many cases where ML has failed to outperform simpler models (Beattie, Laliberte, and Oreopoulos, 2018; McKenzie and Sansone, 2019; Salganik et al ., 2020; Orlov et al ., 2021). Limited predictive performance can arise because researchers have insufficient number of observations or miss out the relevant inputs to properly calibrate a ML algorithm.…”