Background: Over the last twenty years, retention has increased in popularity with policymakers who are concerned with student achievement and aim to end social promotion (i.e., the practice of advancing a student to the next grade level although they have not yet met the academic expectations of their current grade level). Despite calls to expand retention policies, prior research shows mixed results regarding the benefits of retention (Jimerson 2001; Lorence 2014; Schwerdt, West, and Winters 2017). Furthermore, research focused on retention often suggests that retention disproportionately impacts Black and Latino students, as well as students from low socioeconomic backgrounds (Xia Kirby 2009; Locke Sparks 2019). Although previous research has examined the efficacy of grade retention, there is much to learn about the way that state policies around retention affect who is retained and its efficiency in improving student performance on standardized tests. Context: Indiana is one of 19 states in the United States that has a law specifically focused on the literacy instruction of third graders that impacts retention decisions (Cummings Strunk and De Voto 2021). Indiana's test-based policy went into effect in 2012, mandating all third-grade students take an additional standardized test to assess their foundational reading skills called Indiana Reading Evaluation and Determination or IREAD-3. If students did not pass the test after two attempts, they were to be retained. Prior to 2012, each school corporation was responsible for determining its policy regarding retention, with many leaving it to individual schools to determine retention criteria. In 2017, Indiana shifted this policy once again, asking schools to consider students performance on the standardized test (IREAD-3) alongside students overall academic performance to determine whether retention is the best course of action for the student. Figure 1 demonstrates how retention rates shift across the policy periods outlined below. All organizations, especially schools, grapple with finding the balance between rules, regulations, and discretion. This project will contribute to the literature by using administrative data to examine the impact of changes in Indiana's retention policies on which students are most likely to be retained and subsequently, its impacts on student achievement. Research Questions: First, does test-based retention policy reduce inequality in the probability of retention between racial and ethnic minority students and white students or the inequality between high and low-SES students? And second, does test-based retention policy reduce inequality between retained and non-retained students English language arts and mathematics test scores? Data Collection & Analysis: The data for this study come from the Indiana Department of Education and includes information on all public-school students attending schools in Indiana. The first question regarding shifts in the probability of retention utilizes data from 11 cohorts of students. Because there is no consistent measure of academic achievement across the three policy periods, they are separated into two different analyses. The first analysis (n= 588,930) compares the probability of retention during the pre-policy period (2009-2011) to the test-based retention period (2012-2016). The second analysis (n= 590,701) compares the test-based retention period (2012-2016) to the discretionary test-based retention period (2017-2019). Both analyses use logistic regression to predict the outcome of interest, retention, in third grade. Each of these models includes an interaction between the policy period and racial/ethnicity or socioeconomic status as well as a variety of controls. I follow Mize (2019) procedure for analyzing the significance of the interactions. The second question regarding academic achievement outcomes over time utilizes data from 6 cohorts of students (n=406,637), beginning with the 2008-09 cohort of third graders to the 2013-14 cohort of third graders. The last three cohorts of students (2012-2014) matriculated through third grade under the test-based retention policy, while the students in the first three cohorts (2009-2011) matriculated through third grade prior to it. Each cohort is followed for 4 years after their third-grade year. The second analysis uses ordinary least-squares (OLS) regressions and school fixed effects models to examine differences in academic achievement across retained and non-retained students across policy periods as measured by the annual ISTEP+ (Indiana Statewide Testing for Educational Progress) assessment in Language Arts and Math. Preliminary Findings: Inequity in Probability of Retention: Preliminary results suggest that Black students are more likely to be retained across all policy periods when controlling for academic achievement in third grade. Upon the introduction of student and school-level controls, Black students are still more likely to be retained than their white peers under test-based retention. Under periods of increased discretion (pre-policy period & the discretionary test-based period) white students are the most likely to be retained. With regards to socioeconomic status, preliminary results suggest that students who qualify for free or reduced lunch are more likely to be retained under all three policy periods, but especially during the test-based retention period. Inequity in Academic Achievement: With regards to the second question comparing the achievement of retained and non-retained students across the pre-policy and test-based policy, I am particularly interested in the interaction term between retention and policy period. These results suggest that students retained under the test-based policy period consistently score lower than their peers retained under the pre-policy period for four outcomes analyzed: 4th grade English, 4th grade math, 5th grade English and 5th grade math. Conclusions: Given the increased popularity of test-based retention policies, this project examines how shifts in discretion shape which students are more likely to be retained. Preliminary results suggest that Black students are more likely to be retained under the test-based retention period, a pattern that does not hold under either discretionary period upon the introduction of controls. Results from linear regression models suggest that students retained under the test-based period score lower than their peers retained under the pre-policy period. Moving forward, this project aims to further examine the probability of retention across different groups including by race/ethnicity-gender and socioeconomic status-race. Furthermore, this project will also examine how retained students' disciplinary outcomes vary across policy periods.