Casual Inference and Decompositions using AI Models: Statistical Theory and Applications to Worker Transitions

This talk will review recent work adapting tools from causal inference, including tools for estimating decompositions, average treatment effects, and heterogeneous treatment effects, to problems involving sequence data, such as sequences of words in text, sequences of jobs in worker careers, and sequences of measured behaviors and actions in customer journeys. We provide new theory tailored to these problems and apply the methods to the problem of estimating the gender wage gap in worker careers as well as decomposing the sources of changes in gender wage gaps over time. We illustrate approaches to derive insight about causal effects, including approaches to answer causal questions about how individual trajectories evolve over time.

Speaker: Susan Athey, Stanford University

Attend in person or watch online (see weblink)