SGER: Developing Statistical Models for the Diffusion of Educational Policies and Interventions

University Of Delaware, Newark DE

Investigators

Abstract

The work described in this proposal is to catalog existing mathematical models of diffusion that are likely to be of relevance to educational evaluation and relate them to existing statistical methodologies that account for the non-linear form of the diffusion process as well as the hierarchical organizational structure of educational systems. In addition, this research will examine how units (schools, individuals) transition across various states of innovation adoption. Finally, the implications of this work for the quantitative evaluation of educational interventions, and in particular the problem of scale-up, will be explored. Three exploratory studies will be conducted. The first study will begin by reviewing mathematical models of diffusion that are of relevance to the study of educational innovations. These models derive from the seminal work of Mahajan & Peterson (1985). Following the cataloging of relevant diffusion models, we will examine the utility of a series of statistical models that capture various aspects of the diffusion process. Specifically, we will study extensions of discrete time event-history modeling as a method to account for among group variation in the rate of adoption of educational innovations. The second study is based on the premise that understanding how policies and interventions diffuse within and across levels of the system requires an understanding of the characteristics of individuals occupying adoption classes at all levels of the system and over time. This study will examine latent transition analysis as a methodology to more fully understand classes of adoption behavior and how to model how adoption class membership can change over time. The third study examines the implications of diffusion modeling for the evaluation of educational innovations. That is, the evaluation of an educational policy or intervention at any given point in time is dependent on the numbers and types of adopters at that time point. The problem of evaluating policies and interventions within the framework of diffusion is akin to evaluating systems in disequilibria. It is argued that the evaluation of a policy or intervention will be sensitive to the fact that there may not exist a stable equilibrium of adoption during the time of measurement. This sensitivity may have profound effects on quantitative evaluations of policies and interventions. It is argued that the proposed research will provide rigorous links between the established literature on the diffusion of innovations and the equally established literature on educational evaluation and policy analysis. Furthermore, the work proposed here is timely, given the national focus on determining what works in the context of educational interventions and the problem of scale-up. This proposal seeks to develop methodologies that will provide insights into the dynamics of intervention diffusion that can aid in the evaluation of interventions.