COMPUTATIONAL SOCIAL SCIENCE

Doctoral Dissertation Proposal Defense: Palmer

Monday, August 5, 10:00 a.m.
162 Research Hall

Dissertation Proposal: Learning to Consume: Individual versus Social Learning

Nathan Palmer, Ph.D.
Computational Social Science
George Mason University

Abstract: Sims (1980) famously stated that non-rational behavior is a "wilderness:" There is only one way for economic agents to be "rational" but infinitely many ways to be non-rational. While this may be strictly true, some paths through the wilderness are more familiar and well-trod than others. I highlight a growing economic literature which employs tools from reinforcement learning and approximate dynamic programming to impose bounded rationality in intertemporal consumption-savings problems. These tools generalize dynamic programming by easing the information requirement on decision makers, who instead learn behavior from trial-and-error experience. This behavior can converge to optimal behavior if the environment is appropriately stationary and experience is long enough; in the meanwhile, learning may introduce persistent aggregate dynamics. My dissertation contributes to this emerging literature by extending one of its earliest models, Allen and Carroll (2001). My first chapter extends Allen and Carroll's original agent-based consumption-savings model in two ways: first, by incorporate social learning, and second, by introducing an new intuitively motived estimator of the value of a linear consumption function. Both additions improve the time required for agent learning by many orders of magnitude. My second and third chapters will seek evidence for this social learning mechanism in microeconomic, macroeconomic, and experimental data. An alternate fourth chapter may explore general equilibrium effects of social learning, as well as properties of a new reinforcement learning algorithm inspired by Allen and Carroll (2001) which incorporates agent expectations into the learning process. After describing these chapters in more detail, I outline my work plan and describe the intellectual merit and broader impact of my work.