Too Much Information: When Does Additional Testing Benefit Schools?
62 Pages Posted: 6 Aug 2021 Last revised: 4 Nov 2022
Date Written: November 3, 2022
Abstract
U.S. K-12 school districts that traditionally utilized ongoing "formative" assessments of student progress increasingly rely on additional "interim" assessments to predict student performance on standardized tests. Moreover, some districts are experimenting with merit-based teacher incentives tied to standardized test scores.
We examine the relationship between predictive midyear assessments and teacher incentives using a two-period principal-agent model. The school district (principal) decides whether to implement interim assessments and how much merit pay to offer, while teachers (agents) choose how much effort to exert each period. We use two-state ("proficient" vs. "not proficient") Markovian dynamics to describe the evolution of student test readiness. The transition probability has two components: one independent of and another dependent on teachers' effort levels, the "baseline proficiency factor" and "response-to-effort parameter," respectively. For schools starting the period in the not-proficient state, the "improvement potential factor" captures the increased difficulty of success.
Our results indicate that interim assessments are not always beneficial, driven by the difficulty of incentivizing teachers for too high or low an improvement potential factor, and the demotivating effect of accurate unfavorable results. For schools that start the year behind, an interim assessment is valuable if the baseline proficiency factor is high and the budget can only support incentivizing second-period effort after proficient results, either interim only or under any assessment decision. For schools that start the year in the proficient state and have a baseline proficiency factor of zero, an interim assessment is only optimal for a low improvement factor and moderate budget.
Keywords: education operations, service operations, dynamic principal-agent, standardized testing
Suggested Citation: Suggested Citation