Estimating the Effect of Florida’s Low-100 Reading Program: Summarizing Regression Discontinuity Models with Bayesian Model

Abstract

In 2013, the state of Florida mandated an additional hour of intensive reading instruction for the lowest-performing 100 elementary schools across Florida. This requirement was implemented during the 2013–2014 academic year based on the schools’ ranking in 2012–2013. This study assesses the effectiveness of the extra-hour intervention by using a regression discontinuity design (RDD). Often RDD analyses fit multiple models and then select a single best model using stepwise regression leading to overestimation of the effect size and underestimation of the standard error. This study used the Bayesian model averaging approach, which incorporates uncertainty about the best model. The estimated treatment effect, averaged over the six models and weighted by the models’ posterior probabilities, is 6.1 points (d = .25) with a 95% confidence interval of 5.8 to 6.4 points.