Improving Math and Reading Outcomes for English Learners

• In Event: Promising Scholarship in Education: Dissertation Fellows and Their Research

Abstract

Hispanic students tend not to perform as well as their White peers on academic tests, on average, and the gap is even wider for Hispanic English Learners (ELs), students who do not pass district criteria for English proficiency (Hemphill & Vanneman, 2011; NCES, 2015; Reardon & Galindo, 2009). Many different approaches exist with the intent to help ELs attain English proficiency while also accessing the core curriculum content (e.g., English as a Second Language [ESL] classes, bilingual programs, dual-language settings). Importantly, policy differences regarding EL classification and available services lead to very similar students receiving very different services across states and districts. Criteria such as the method of assessment of students’ English proficiency and the level of proficiency used to classify students as ELs are not universal; instead, they are established by local policies. Consequently, a child who is classified as an EL in one location, and thus provided supplemental language instruction, may not be classified as such in another, and would therefore receive no services. Further, the type of services offered differs across settings. There is some evidence that students in bilingual and dual-language settings may exhibit lower achievement than their peers in other settings in some early grades, particularly in reading, then catch up at later grades, though this work has not generally looked at national trends (August & Shanahan, 2008; Rolstad, Mahoney, & Glass, 2005; Umansky & Reardon, 2014; Valentino & Reardon, 2015). In this study, I examine the language services provided to Hispanic Spanish-speaking ELs to investigate which approaches are most effective in raising student achievement in math and reading in early elementary school using two nationally representative data sets, the Early Childhood Longitudinal Studies, kindergarten classes of 1998 and 2011 (ECLS-K and ECLS-K:2011, respectively), to examine trends across the US.

Study design and methods
To compare the achievement of ELs in various instructional settings, ordinary least squares regression is used to examine achievement at each wave from kindergarten to third grade, and hierarchical linear modeling is used to examine longitudinal trajectories. In the ECLS-K, students were identified as participating in one of three settings in the fall of kindergarten: no services, in-class ESL services, or pullout ESL services. Because both in-class and pullout settings are classified as ESL, the two were collapsed into two categories for comparison, participation in either ESL or no services. In ECLS-K:2011, services included ESL, bilingual, and dual-language programs, leading to four categories for comparison (the three types of services as well as no services). I then also use propensity score matching to identify virtually identical students receiving different services to allow for comparison of the achievement of very similar students in these different instructional settings. Students are matched on observable characteristics such as achievement data and demographic information existing prior to implementation of the treatment and weighted by their propensity to be in each setting. OLS regression and hierarchical linear models using the resulting inverse probability weights are then examined to compare these matched samples.

Author

• Jennifer Timmer, University of Illinois at Urbana-Champaign