Assessing the Impact of an ESL/Bilingual Program by Means of Instrumental Variable Estimation

The purpose of this study was to assess the impact of the Portland (Oregon) public school district\u27s ESL/bilingual program on the academic performance of limited English proficiency (IEP) students. The study attempted to correct a statistical bias that might lead to underestimating the effectiveness of ESL/bilingual programs. This statistical bias is caused by a negative correlation between student achievement and the characteristics which result in a student being placed in the ESL/bilingual program. Several variables and proxies representing characteristics of the school, the neighborhood, and the student\u27s personal background were examined for their contribution to explaining the academic progress of LEP students in reading, mathematics, and English language usage. This dissertation attempts to answer the following major questions: Is the Portland school district\u27s ESL/bilingual education approach effective in increasing LEP students\u27 academic progress in reading, mathematics, and English language usage? Does the amount of FSL/bilingual instruction influence the academic achievement of LEP students in reading, mathematics, and English language usage? Do the personal background characteristics of LEP students influence their academic gains in reading, mathematics, and English language usage? Do neighborhood factors influence LEP students\u27 gains in reading, mathematics, and English language usage? Achievement gains of LEP students in Grades 3-11 from the Portland (Oregon) Public School district were examined. Data on pertinent characteristics relating to school, neighborhood, and personal background information were collected. The data were analyzed using multiple regression analysis and instrumental variable estimation. Instrumental variable (IV) estimation was found to be appropriate to deal with the serious problem of selection bias in evaluating achievement gains of LEP students in ESL/bilingual programs. The problem of selection bias occurs when learners are selected for a program or for evaluation study because of characteristics which will also influence their scores on a test. Subsequent effects of this type of selection, and possible solutions to this type of problem, are discussed. The findings suggest that the ESL/bilingual education approach had a strong and statistically significant impact in improving mathematics achievement. The program\u27s impact on language usage achievement was weak, and it showed no consistent results relating to reading achievement. The findings indicate that the greatest impacts are in academic areas rather than in language areas. The results were not strong, but the ESL/bilingual program appeared to have some positive benefits in terms of achievement gain in mathematics and language usage which simpler statistical techniques tend not to show. However, because of the statistical problems and the methods used to address them, confidence in estimates of the specific parameters is not great. Studies covering different geographic areas and longer periods of time are recommended