Primary Submission Category: Applicants in Social Sciences
The Causal Effect of Volatility: Estimation by Marginal Structural Models
Authors: Ian Lundberg, Nanum Jeon, Hao Liang,
Presenting Author: Ian Lundberg*
We apply marginal structural models to a social science question about the causal effect of economic volatility. Our question is motivated by social science literature showing a major shift in the U.S. labor market. While it was historically common for a U.S. worker to experience a long-term, stable careers with a single employer (e.g. in the 1950s), the labor market experiences of present-day workers are more often characterized by frequent job changes. Thus the life course trajectory of economic well-being often involves more rises and falls today than in the past. We would like to understand how economic volatility at ages 25–34 affects the probability of being married at age 35. Because volatility is a treatment that unfolds across time periods, we use marginal structural models. But standard MSMs that focus on the cumulative treatment cannot answer our applied question: a trajectory that rises and falls repeatedly is more volatile than one that is consistently average, even though the cumulative treatment is the same. We therefore generalize the usual functional form of MSMs to directly incorporate the volatility of the treatment, measured by a summary of year-to-year changes. We produce estimates using panel data on a probability sample of U.S. adults. Broadly, our study illustrates one example where generalizations of cumulative treatment MSMs can be useful in a new class of social science applications.