Elsevier

Journal of Public Economics

Volume 107, November 2013, Pages 63-78
Journal of Public Economics

Impact of bilingual education programs on limited English proficient students and their peers: Regression discontinuity evidence from Texas

https://doi.org/10.1016/j.jpubeco.2013.08.008Get rights and content

Highlights

  • Exploits a Texas policy rule to estimate the causal impact of bilingual education

  • Shows school programs for LEP students have spillover effects to non-LEP students

  • Providing BE (versus providing only ESL) increases non-LEP students' achievement

  • No significant impacts on Spanish home language (mostly LEP at one point) students

Abstract

Texas requires a school district to offer bilingual education when its enrollment of limited English proficient (LEP) students in a particular elementary grade and language is twenty or higher. Using school panel data, we find a significant increase in the probability that a district provides bilingual education above this 20-student cutoff. Using this discontinuity as an instrument for district bilingual education provision, we find that providing bilingual education programs (relative to providing only English as a Second Language programs) does not significantly impact the standardized test scores of students with Spanish as their home language (comprised primarily of ever-LEP students). However, we find significant positive impacts on non-LEP students' achievement, which indicates that education programs for LEP students have spillover effects to non-LEP students.

Introduction

One of the major challenges facing educators and policymakers today is the large and growing number of limited English proficient (LEP) children in U.S. public schools. About 1 in 9 students enrolled in pre-kindergarten to grade 12 were classified as LEP in 2008–09, a marked increase from the ratio of 1 in 13 recorded one decade earlier (National Clearinghouse for English Language Acquisition, 2011). These LEP students are present not only in big cities and other traditional immigrant-receiving areas, but also across the country; even by 2001–02, when U.S. immigrants were less geographically dispersed than they are today, about half of public schools in the U.S. had at least one LEP student (Zehler et al., 2003). Lack of proficiency in English presents a significant barrier to learning in U.S. schools, and given these recent trends in LEP student population and geographic dispersion, how to educate LEP students is likely to remain an important policy issue in the coming years.

School districts are required by federal law to provide special assistance to LEP students.1 They typically offer Bilingual Education (BE) or English as a Second Language (ESL) to help LEP students. While there is considerable variation in how these programs are implemented in the classroom, a defining feature of BE is the use of the student's native language for at least some of the academic instruction; other programs such as ESL teach only in English. Given this feature, LEP students participating in BE tend to be placed in a self-contained classroom with classmates who share the same home language and a dedicated bilingual education teacher who can teach in that language. In contrast, LEP students participating in ESL tend to be placed in mainstream classrooms with pullout time with an ESL teacher to improve their English skills.

In this paper, we identify the causal effect of BE on the academic achievement of LEP students and their non-LEP peers using quasi-experimental variation in BE exposure generated by a policy rule governing the provision of bilingual education programs in Texas. The policy rule requires a school district to offer BE when its enrollment of LEP students in a particular elementary grade level and language is twenty or higher. Below this 20-student cutoff, districts are free to offer BE or ESL, with most choosing to offer only ESL.2 This suggests a regression discontinuity (RD) design in which the effect of providing BE (relative to ESL) on student achievement can be obtained by comparing student outcomes in districts just above the 20-student cutoff (and therefore more likely to provide BE) and student outcomes in districts just below the cutoff. We elaborate on this RD strategy in Section 5.

This paper adds to a large literature evaluating educational programs for LEP students, which we briefly summarize in Section 2.2. It addresses two major gaps in this literature. First, this literature has focused exclusively on the impacts on the intended beneficiaries themselves (i.e., the LEP students) and ignored any effects that these programs might have on non-LEP students. Yet, because these programs change the student composition of mainstream classrooms and school budgets, among other things, there is potential for spillover effects to non-LEP students. To our knowledge, our study is the first to test for spillover effects of educational programs for LEP students, and to the extent that they exist, to quantify them. Quantifying these spillover effects is necessary for a complete cost–benefit analysis of the various LEP programs; all else equal, policy makers might prefer the program that benefits non-LEP students more (or, stated differently, harms non-LEP students less).

Second, most of the studies in this literature do not address the potential problem of endogeneity in student exposure to the educational programs for LEP students. In general, student exposure to a program is not random, and instead is the result of decisions made by students, parents, schools and districts. Thus, it is likely correlated with unmeasured and unobserved characteristics of the students, parents, schools and districts, some of which might in turn be correlated with student achievement. Estimates of program effects that do not take this into account tend to be biased. Our research adds to the handful of studies that provide estimates of the impacts of LEP programs with a causal interpretation.3

We implement our RD strategy using panel data on elementary schools in districts near the 20-Spanish-LEP-student cutoff defined by the policy rule. We describe these data in Section 6. We restrict our attention to the policy rule vis-à-vis Spanish LEP students for a practical reason: Spanish is the home language of 90% of Texas' total LEP enrollment, and is the only language for which there is enough variation across districts to implement our empirical strategy. Due to this restriction, our results pertain to the effect of district provision of Spanish bilingual education programs (relative to providing only ESL for Spanish LEP students). However, considering that Spanish is the language of over three-quarters of total LEP enrollment in the U.S. and accounts for an even higher share of bilingual education programs operating in the U.S. (Zehler et al., 2003), it is especially policy relevant to understand the effects of Spanish BE programs.

To preview the results of Section 5, we find a significant increase in the probability that a district provides BE above the 20-Spanish-LEP-student cutoff. We do not find any significant jumps at the cutoff in covariates unrelated to BE provision, nor do we observe “stacking” of districts below the cutoff, which validates the interpretation of differences in student outcomes just above and just below the cutoff as due to district BE provision.

We proceed by using the variation in district provision of BE induced by the policy rule as an instrumental variable to identify the causal impact of district provision of BE on student achievement. While the impact of school provision of BE would also be of interest, we focus on district provision of BE because this is directly linked to the policy rule. These instrumental variable estimates provide the local average treatment effect of district provision of Spanish BE among districts whose decision to offer Spanish BE is constrained by the policy rule. Our main findings are as follows. First, district provision of BE raises the standardized math and reading test scores of students who are non-LEP and whose home language is not Spanish. In our preferred specification controlling for a linear spline of the running variable, the district-wide Spanish LEP count in a student's first grade cohort, the positive impacts on non-LEP achievement are statistically significant. Students who are non-LEP and whose home language is not Spanish would never have been candidates to participate in Spanish BE programs, thus this finding is indicative of spillover effects.

Second, district provision of BE has generally positive but smaller and statistically insignificant effects on students whose home language is Spanish. A vast majority of Spanish home language students (89%) are classified as LEP in first grade, and so would have been eligible to participate in educational programs for Spanish LEP students. Hence, our results suggest that the intended beneficiaries of the LEP programs fare similarly in BE and ESL programs. Finally, we find that district BE provision increases test scores on all students taken together. The positive net impact indicates that on average, the test score gains due to district BE provision exceed test score losses.

Section snippets

Legislative background on educational programs for LEP students

The Bilingual Education Act passed in 1968 was the first federal law expressly addressing the educational needs of LEP students in American schools, and did so by providing a financial reward—federal grants awarded on a competitive basis—for providing help to LEP students.4 Later federal laws made it a legal

Conceptual framework

The direction of the impact of BE programs (compared to ESL programs) on academic achievement is theoretically ambiguous for both LEP and non-LEP students. LEP students are the students eligible to participate in BE and ESL. BE and ESL are the two most common programs offered by schools to address the learning needs of LEP students, and each has advantages and disadvantages. For example, native language instruction might delay English acquisition, but it might also enable LEP students to better

Data

To implement our RD strategy, we use publicly-available data on the standardized test scores and demographic characteristics of students enrolled in Texas public elementary schools.13 To maintain data confidentiality, the Texas Education Agency provided us with grouped student data rather than individual-level student data. In particular, we obtained mean data at the school-grade-year level for

OLS estimates of the effect of district provision of bilingual education

In Table 2, we present the results from estimating Eq. 1 via OLS using the same three samples that we use below to implement our RD strategy.28 The OLS coefficients for district BE provision are not significantly different from zero using the “non-LEP, non-Spanish home language” sample (columns 1 and 2), which if interpreted causally

Conclusion

In this paper, we examine the effects of bilingual education programs (versus ESL programs alone) on the achievement of intended beneficiaries (LEP students) and their classmates (non-LEP students). To address potential bias due to the endogeneity of student exposure to BE, we use a regression-discontinuity approach that exploits a policy rule governing the provision of BE in Texas. We find that district BE provision has significant positive effects on the standardized test scores of “non-LEP,

References (34)

  • Julian R. Betts

    Educational crowding out: do immigrants affect the educational attainment of american minorities?

  • Scott Carrell et al.

    Externalities in the classroom: how children exposed to domestic violence affect everyone's kids

    Am. Econ. J. Appl. Econ.

    (2010)
  • Aimee Chin et al.

    Impact of Bilingual Education Programs on Limited English Proficient Students and Their Peers: Regression Discontinuity Evidence from Texas.” NBER Working Paper No. 18197

    (2012)
  • Rosa Cho

    Are there Peer Effects Associated with Having English Language Learner (ELL) Classmates?: Evidence from the Early Childhood Longitudinal Study Kindergarten Cohort (ECLS-K)

    (2011)
  • James Crawford

    Bilingual Education: History, Politics, Theory and Practice

    (1989)
  • David Figlio

    “Boys Named Sue: Disruptive Children and Their Peers”. NBER Working Paper No. 11277

    (2005)
  • Charlotte Geay et al.

    Non-native Speakers of English in the Classroom: What are the Effects on Pupil Performance?” IZA Discussion Paper No. 6451

    (2012)
  • Cited by (44)

    • Mother tongue reading materials as a bridge to literacy

      2022, Economics of Education Review
      Citation Excerpt :

      Much of the literature on bilingual versus immersion education comes from the US, with a focus on integrating Spanish-speaking children into mainstream English education. After decades of debate (see Slavin and Cheung (2005) for an overview), recent evidence from methodologically rigorous studies finds little enduring difference between the two (Chin et al., 2013; Slavin et al., 2011). Based on this evidence, policy advice from the US emphasises the importance of education quality, rather than the immersion/bilingual dichotomy, as a primary driver of English language learner success (Chin, 2015).

    • Do refugee students affect the academic achievement of peers? Evidence from a large urban school district

      2022, Economics of Education Review
      Citation Excerpt :

      However, a related strand of literature suggests that effects differ by the type of English as a Second Language service provided to ELs. For example, Chin et al. (2013) find that providing bilingual education programs aimed at increasing achievement among ELs has positive spillover effects for non-EL students.35 Thus, it is plausible that an increase in EL students, due to a higher share of refugees, has implications for non-EL students that vary with the type of ESL instruction.

    • English learners reclassification and academic achievement: Evidence from Minnesota

      2020, Economics of Education Review
      Citation Excerpt :

      The only aberration is a decrease in the proportion of students attending charter schools and in prior ACCESS scores among 5th-grade EL students scoring just above the threshold in the analytic sample (Columns 3 and 11 of Table 2).29 To check whether this aberration poses a threat for our research design, we estimate a seemingly unrelated regression of the eleven equations for each grade and conduct a chi-squared test of the null hypothesis that the parameters across eleven equations are jointly equal to zero (Carlson & Knowles, 2016; Chin, Daysal, & Imberman, 2013). Panels B and C of Tables A6 to A9 present the p-values for this test.

    View all citing articles on Scopus

    We thank Erzo F.P. Luttmer, Chinhui Juhn, Danielle Li, and workshop participants at the 2010 and 2011 National Academy of Education Annual Meeting, 2011 ASSA Meetings, 2012 SOLE Meetings, and 2012 APPAM Fall Research Conference for helpful comments and discussion. Imberman gratefully acknowledges financial support from the National Academy of Education/Spencer Foundation Postdoctoral Fellowship. A University of Houston Small Grant was used to purchase the data, for which we are grateful. The authors bear sole responsibility for the content of this paper.

    View full text