How Multilevel Modeling Techniques Underestimate the Effects of Racial Oppression

• In Event: Critical Methodological Innovations in Quantitative Research

Abstract

Through a series of simulation studies that model operations of racial oppression at the individual, meso, and systemic levels on students and their learning experiences, this paper explores whether and to what extent Hierarchical Linear Modeling (HLM) techniques under-estimate the effects of racial oppression. The analyses provide evidence that HLM techniques accurate estimate the simulated relationship between pre-test and post-test scores and the simulated intervention effect on post-test scores. However, analyses find that the HLM techniques systematically under-estimate the effects of various forms of racial oppression. This under-estimation becomes worse as sample sizes and variation in simulated effects of oppression increases.

Authors

• Mike K. Russell, Boston College

• Daniel Raphael, Boston College