Discovery of Social Beliefs About Ethnic Structure from Survey Data

Abstract

This paper introduces a new approach to discovering and exploring society-wide social beliefs about ethnic structure. Rooted in computational text analysis, it combines the strengths of both qualitative and survey-based approaches to the study of ethnicity. I use a structural topic modeling approach (Roberts, Stewart, and Airoldi 2016) to uncover general patterns in found in open-ended questions about ethnic groups. I then predict which patterns of responses tend to be associated with which ethnic groups, and use information about survey respondents to explore heterogeneity in social beliefs. To illustrate the method at work, I use original survey data from Malaysia, collected in 2017, to understand social beliefs about Malay, Chinese, Indian, and Arab ethnic groups in Malaysia.