Association for Public Policy Analysis & Management

A growing number of students choose to “opt out” from taking regularly administered standardized tests in California and across the country. This phenomenon has seen a recent surge in popularity for a variety of reasons, including concerns surrounding common core standards and the perceived narrow focus of testing. Growth in the opt out movement may have a direct impact on our ability to use student data in multiple important ways. In California, the ability to use Smarter Balanced summative assessments as a school accountability metric and to use interim assessments to support teaching and learning throughout the year are also susceptible to opt out growth. Even new, broader measures of student performance often include multiple measures of academic progress susceptible to bias in the presence of student opt outs, such as yearly student growth and school proficiency rates. This is particularly important because, as of the 2013 legislative session, these types of evaluations can be used to set statewide performance goals, and elements of the new accountability system can be used to justify intervention into chronically underperforming schools and districts.

In this paper I evaluate the impact of opting out in California’s CORE schools districts on a range of school and district accountability metrics. I utilize a panel of student level administrative data that includes teacher and school characteristics to simulate the impact of various opt scenarios, based on existing opt out patterns seen in the CORE districts and other large states (e.g. New York and Pennsylvania). I use linear regression and student fixed effects models (similar to those used in existing accountability measures) to identify how different proportions of students opting out would impact these measures. Further analyses identify which schools are most likely to fail to meet the minimum testing thresholds for accountability, as well as to recognize which schools are most likely to fall below existing accountability cut-offs as a result of growth in opting out in their schools. Additional exploratory analyses investigate the extent to which opt out growth is clustered at the classroom, school, and district level, and a survey of school climate is used to identify additional correlates of student opt out behavior. Finally, these data are used to determine if our ability to estimate achievement gaps is harmed by the non-random self-removal of certain student types from the testing data.

Preliminary evidence suggests that, while the existing level of opting out in the CORE districts is low, modest growth similar to that experienced in other states (e.g. New York and Colorado) would have substantial impacts on school and district accountability systems. Opt out rates as low as 5% can lower school’s performance in their district’s accountability systems. Perhaps even more troubling, modest rates of opting out (given existing demographics) can lead districts to underestimate the magnitude of achievement gaps with respect to both race and income, undermining a central measure of educational equity and progress.