Multiple-level and multiple-group gap decomposition using mediation analysis

Speaker(s):

Beatriz Gallo Cordoba, Monash University

Abstract:

Gap decomposition has been used as a strategy to better understand such gaps and propose policy initiatives to narrow such gaps. Gallo Cordoba, Leckie and Browne (2022) propose using mediation analysis as a tool to decompose achievement gaps in multiple levels (e.g. students, schools, local authorities) and multiple groups (e.g. White, Black, Asian). This interactive session will explain the intuition behind this decomposition technique and how to implement it in R and Stata in four steps: first, showing the equivalence between within- and between-level 2 (e.g. schools) gap decomposition and mediation analysis; second, applying mediation analysis to decompose the gap across multiple levels; third, applying mediation analysis to decompose the gap across multiple groups, and finally combining the last two steps to decompose the gap across multiple levels and groups. Participants are encouraged to bring their own data but mock data will be provided.