The Unintended Consequences of Common Core

• In Event: The Impacts of Standards and Curriculum on Education Policy Outcomes

Abstract

Much debate on education policy has focused on potential consequences of the Common Core State Standards (CCSS) for various stakeholders. However, there is limited empirical evidence about its effects on student achievement. In this paper, we estimate the causal impact of the introduction of the CCSS on student achievement. We combine information on the introduction and implementation of the CCSS with individual and school level student achievement data from the National Assessment of Educational Progress (NAEP) and the Stanford Education Data Archive (SEDA).

So far, the literature has addressed this research question using state-level data on student achievement (Song et al., 2019). Our approach with student and school level data has multiple methodological advantages. First, individual-level data does not suffer from aggregation bias (Goodfriend, 1992). Second, we can determine effect heterogeneities at the levels of individuals and schools. Third, individual-level data allows us to employ more refined identification strategies to estimate a causal effect. For example, the availability of individual-level information on the school type allows to estimate quadruple-difference regressions, as explained below.

Given that the adoption of the CCSS did not happen at random, a correlational analysis of CCSS and student performance would likely suffer from omitted variable bias. Hence, we employ a quasi-experimental estimation strategy, performing quadruple-difference estimations across states, years, school types, and subjects. The first difference exploited in this approach refers to the difference between states. In a given year the CCSS are in place only in some states. As of 2019, more than 40 states have adopted (at least some elements of) the CCSS. The second difference makes use of the different timing of the introduction and implementation of the standards. Some states introduced the standards earlier than others, and even conditional on the adoption year the implementation phase differs between states. The third difference is based on the fact that the CCSS are introduced in public schools, while private schools are not compelled to implement them. The fourth difference refers to the subjects affected by the reform. The CCSS focus on mathematics and English language art/literacy, while other subjects such as science are not addressed by the CCSS. In sum, we can control for constant differences between states, years, school types and subjects.

The paper uses two series of nationally representative datasets for estimating the causal impact of CCSS on student performance. The first dataset is restricted individual-level dataset of the NAEP. We have access to the entire series of NAEP data for grades 4, 8 and 12 for all subjects (including Math, Reading and Science) from 1990-2016. Second, we use all available school level data from the SEDA from 2009-2016. We merge these two datasets with the state-level data on the introduction of the CCSS.

The NAEP test is comparable across states and over time. It includes individual level control variables such as student race, gender and socio-economic status. In addition, it has a rich set of school level variables, such as an indicator about the school type which allows to estimate differential effects for students in public schools vs. private schools. The 2009-2015 NAEP data also allow us to test for the fidelity of implementation of the common core, i.e. whether the schools’ standards were structured according to eight different entities (including state standards).

The SEDA comprises data on educational outcomes such as academic achievement levels and achievement gaps at the school, school district, and state level. Academic achievement is collected for mathematics and English/Language Art (ELA) for grades 3 through 8. It is derived from aggregated test score data from standardized testing programs of US states. In addition, SEDA includes comprehensive background information such as measures of racial and socioeconomic composition and segregation patterns.

While the NAEP test is considered a “low stakes” test, the student achievement data from SEDA are based on high stakes exams. Also, the state tests underlying SEDA are aligned to the CCSS, while the NAEP is not. Hence, estimating the effect of the introduction of Common Core using NAEP and SEDA permits us to assess the robustness of effects with respect to testing formats. To conclude, the NAEP and SEDA enable us to estimate the effect of the Common Core Standards on student performance in a quadruple-difference estimation design.



References:

Goodfriend, M. (1992)"Information-aggregation bias." The American Economic Review (1992): 508-519.

Song, M., R. Yang, and M. Garet (2019). “Effects of States’ Implementation of College- and Career Ready Standards on Student Achievement.” American Institutes for Research. Unpublished draft: https://www.c-sail.org/sites/default/files/Effects%20of%20CCR%20standards%20on%20stu%20achievement_4-2019_AERA%20DRFAT.pdf.

Authors

• Benjamin Arold, University of Munich (LMU) and ifo Institute

• M. Danish Shakeel, University of Buckingham