Correcting for Bias in School Composition Effect Estimates: An Application of the Regression Discontinuity Approach

• In Event: Park Central Roundtable Session 11
  In Roundtable Session: School Effectiveness and School Improvement SIG Roundtables

Abstract

Two facets of under-specification have been identified in studies investigating the school composition effect: the existence of measurement error in the level 1 predictors (measurement error bias) and the lack of level 1 covariates (omitted variable bias). Both of these factors have shown to lead to misleading estimates with conventional educational effectiveness models. I incorporate English years four and five data from TIMSS-95 to demonstrate how the regression discontinuity approach can be used when seeking to separate the effects of school composition from school effectiveness and measures and obtain estimates net of measurement error and omitted variable bias. This is the first study that integrates regression discontinuity models and multilevel structural equation models to assess the school composition effect.

Authors

• Ioulia Televantou, University of Nicosia

• Herbert W. Marsh, Australian Catholic University