A longitudinal study of achievement outcomes in a privatized public school: A growth curve analysis

Date of Award




Degree Name

Doctor of Philosophy (Ph.D.)

First Committee Member

James D. McKinney, Committee Chair


Current initiatives in school reform include contracting educational services to private, for-profit firms. The purpose of this dissertation was to ascertain whether the students who attended one elementary school in Miami-Dade County, run by The Edison Project, made greater academic progress than comparable students who attended other district schools. It provides a longitudinal examination of the students' academic achievement during the school's first three years of operation, 1996--97 through 1998--99. In addition, it extends the range of statistical techniques typically used in program evaluations. Hierarchical linear modeling (HML) was employed in an individual growth curve analysis. The results from this procedure were compared to those from a traditional repeated measures analysis of variance (ANOVA).A quasi-experimental, 2 x 3 factorial design (i.e., treatment by time) was used to compare the project and control groups' reading and mathematics scores on the Stanford Achievement Test. Two panels of students in the project school were followed across the three school years. Panel A was comprised of 114 students initially enrolled in grade 2, and Panel B of 159 students initially enrolled in grade 3. Using stratified random sampling, equal numbers of students who attended other district schools were selected as control groups. The selected control groups remained comparable to the project groups in terms of demographic characteristics and performance on a pretest, despite attrition.Four sets of analyses were completed: for Panel A in reading and mathematics, and Panel B in reading and mathematics. The results of both the repeated measures ANOVA and the HLM analyses indicated that significant levels of growth were achieved across the three-year period in both subject areas for all groups of students. In reading there were no statistically significant differences between the project and control groups that could be attributed to group membership. The results were less consistent in mathematics. The repeated measures ANOVAs identified significant treatment by time interactions for both panels in mathematics. The HLM analyses pinpointed the source of the variation as within-group variation in Panel A, but as between-group variation in Panel B. Thus, the only statistically significant finding identified in the HLM analyses was that the rate of growth over time was greater for the Panel B project students than for the control students. Nonetheless, after the project students had received three years of instruction by means of the Edison model, their final levels of performance in reading and mathematics did not differ significantly from the students' in other district schools.


Education, Administration; Education, Elementary

Link to Full Text