COMPUTATIONAL SOCIAL SCIENCE – A Simple Direct Estimate of Rule-of-Thumb Consumption using the Method of Simulated Quantiles and Cross Validation – Palmer
COMPUTATIONAL SOCIAL SCIENCE FRIDAY SEMINAR
Nathan M. Palmer, Ph.D. Candidate
Computational Social Science Program
Department of Computational and Data Sciences
George Mason University
A Simple Direct Estimate of Rule-of-Thumb Consumption using the Method of Simulated Quantiles and Cross Validation
Friday, November 3, 3:00 p.m.
Center for Social Complexity Suite
3rd Floor, Research Hall
Campbell and Mankiw (1989, 1990) famously demonstrated that aggregate data supported a model of household consumption in which roughly 50% of agents followed an optimizing strategy while the other 50% followed a “rule of thumb” strategy, consuming their current income. This paper revisits that hypothesis using structural, micro-level, semi-parametric estimation and formally selecting between different models of agent behavior. I find strong evidence supporting a generalization of Campbell and Mankiw (1989, 1990)’s original conclusion: roughly 50% of the population behaves in a way similar to “rule of thumb” consumers, even when the data is allowed to dictate how severe that rule of thumb behavior is. In addition, this paper demonstrates the usefulness and flexibility of both the Method of Simulated Quantiles and K-fold cross validation for selecting between of agent behavior. This type of model selection is crucial for creating agents to populate robust, richly-featured agent-based models of macroprudential and macro-financial systems.