Association for Public Policy Analysis & Management

Federal and state laws require public school districts to provide support services to English Language Learners which help them “achieve English language proficiency and perform academically at the same high levels as their non-EL peers” (US Department of Education, 2016, 3). English Language Learners certainly can benefit from receiving instruction specific to language acquisition; studies have also found positive student outcomes associated with reclassification (Flores et al., 2009; Callahan, Wilkinson, & Muller, 2010; Kim & Hermann, 2010; Carlson & Knowles, 2016). Conversely, research suggests that extended classification as an English learner does not benefit, and indeed may harm, students (Jacobs, 2016; Menken, Kleyn, & Chae, 2012). Given the benefits of timely reclassification, schools have a process of review and reclassification each English learner goes through annually – the Language Proficiency Assessment Committee (LPAC). Generally, the factors going into LPAC decisions include English proficiency test scores, student grades, standardized test data, and teacher recommendations (Kerr, 2015). While test scores provide clear guidance on whether a school reclassifies a student, some subjectivity may enter using grades and teacher recommendations. As well, districts and schools may not follow procedures regarding test scores and reclassification. With this subjectivity and potential deviation from implementation standards, it is important to examine whether student characteristics, other than prior test scores, are predictive of reclassification.

In this paper, I evaluate which student characteristics, apart from academic achievement, affect probability of reclassification. I use statewide achievement and demographic data for academic years 2009 through 2015. The dataset contains basic student-level demographic information such as gender, race/ethnicity, Limited English Proficiency (LEP) status, Special Education status, and eligibility for free/reduced price lunch. In addition to demographic information, the dataset contains student-level state assessment data. This paper describes research conducted on a subsample of students in the complete dataset. The pooled years sample is restricted to the cohort of students who entered kindergarten in academic year 2009. It follows the students in that cohort through academic year 2015. Thus, the data will follow students in the cohort from kindergarten through their sixth grade year. To investigate whether specific student characteristics – other than prior achievement – predict reclassification, I use probit models with a binary variable – reclassified –as the dependent variable. The explanatory variables of interest are female, eligibility for free/reduced price lunch, and moving districts (within a school year). As well, I evaluate the relationship between academic achievement and socioeconomic status. Preliminary results suggest that, apart from lagged test scores, the most consistent determinant of reclassification is a student moving districts.