Reestablishing Empirical Class Analysis Using the Dictionary of Occupational Titles

• In Event: Economic Sociology Section Roundtables
  In Refereed Roundtable (60 minute): Roundtable 11: Socioeconomic Inequality

Abstract

Rising inequality in the neoliberal era has a Marxian flavor. The labor share of income is declining, the middle class is shrinking, and labor markets are polarizing. Yet empirical class analysis has been largely abandoned (with a few notable exceptions) from the sociology of stratification. I argue that empirical, Marxian class analysis has been advanced and pursued relatively little since the publication of Erik Wright's Class Counts (1997) because of the high cost of fielding Wright's required survey. In this presentation I demonstrate that, contrary to what Wright and his collaborators believed, the new-Marxian class categories which Wright theorized can be distinguished within public survey data such the Current Population Survey. To accomplish this task, I leverage the Dictionary of Occupational Titles (DOT), which contains rich information on workplace social relations. I merge a digitized and cleaned version of the 4th edition DOT (1977) with Wright's original (1980) class survey. Using machine learning prediction, I show that DOT variables can predict the Marxian class categories in Wright's own data but without need for Wright's bespoke questions on social relations of power, authors, and property. I do this by training a LASSO model to predict class, save the selected DOT variables and the optimal coefficients, and use these variables and coefficients to impute class categories in public data. Doing so only requires the DOT variables and merging these to detailed occupational categories. I hope that reestablishing a reproducible, empirical basis for class analysis will renew quantitative class analysis and open many opportunities to explain the roots of rising income and wealth inequality.

Author

• Edo Navot, University of Florida