Talk of Family: How Institutional Overlap Shapes Family-Related Discourse Across Social Class

Area of Science:

• Sociology
• Computational Social Science
• Family Studies

Background:

• The centrality of family in adults' lives is a key aspect of social stratification.
• Existing research on family centrality often relies on traditional qualitative methods, limiting large-scale analysis.
• The deinstitutionalization of family is a significant debate in contemporary American society.

Purpose of the Study:

• To explore social class variation in the centrality of family in adults' lives.
• To introduce and demonstrate a novel machine learning application for analyzing qualitative interview data at scale.
• To investigate educational disparities in family centrality using discourse atom topic modeling.

Main Methods:

• Application of a novel machine learning approach, discourse atom topic modeling.
• Analysis of interview transcripts from the American Voices Project (N = 1,396).
• A two-phase analytical approach: person-level and line-level transcript analysis.

Main Results:

• Family, as represented by discourse, is more central in the lives of adults without a college degree than among college-educated adults.
• The degree of institutional overlap between family and other key institutions (health, work, religion, criminal justice) does not vary by education level.
• The study highlights the value of machine learning for large-scale qualitative data analysis.

Conclusions:

• Educational attainment is associated with variations in the centrality of family in adults' lives.
• Findings contribute to the debate on the deinstitutionalization of family in the United States.
• The developed machine learning method offers a powerful tool for future research on educational disparities and qualitative data.