Suessmilch Lecture

Uncovering College Effect Heterogeneity using Machine Learning

Jennie E. Brand
Online Presentation, September 29, 2020

As part of the Suessmilch Lecture series, Jennie E. Brand from University of California, Los Angeles (UCLA), gave a talk on college effects on wages. Doing so, she explores variation in effects to recursive partitioning based on causal trees.

Title

“Uncovering College Effect Heterogeneity using Machine Learning”

Abstract

Individuals do not respond uniformly to treatments, such as events or interventions. Sociologists routinely partition samples into subgroups to explore how the effects of treatments vary by selected covariates like race, gender, and income. In so doing, analysts determine the key subgroups based on theoretical priors. Data-driven discoveries are also routine, yet the analyses by which sociologists typically go about them are often unreported, problematic, and seldom move us beyond our expectations and biases to explore new meaningful subgroups. Emerging machine learning methods based on decision trees allow researchers to explore sources of variation that they may not have previously considered or envisaged.

Assessing a central topic in the social inequality literature, college effects on wages, I compare what I learn from covariate and propensity-score-based partitioning approaches for exploring variation in effects to recursive partitioning based on causal trees. I expand on the existing causal tree literature by applying leaf-specific effect estimation strategies to adjust for observed confounding, including inverse propensity weighting, nearest neighbor matching, and doubly robust causal forests. I also use sensitivity analyses to address the possibility of unobserved confounding.

About the speaker

Jennie E. Brand is Professor of Sociology and Statistics. She is also Director of the California Center for Population Research and Co-Director of the Center for Social Statistics (CSS). Prof. Brand studies social stratification and inequality, and its implications for various outcomes that indicate life chances. Her research agenda encompasses three main areas: (1) access to and the impact of higher education; (2) the consequences of disruptive events, such as job displacement; and (3) causal inference and quantitative and computational methods for observational data.