Grounding Agent Based Modeling in Population Reality: Empirical Calibration and Validation

• In Event: 203 - Regular Session. Population Process - Migration

Abstract

Agent-based modeling (ABM) is a set of methodological tools with increasing popularity in social science research, particularly on population phenomena such as fertility, migration, marriage, health, social networks, and residential segregation. Because ABM simulates individual behavior in a complex micro-meso-macro system to generate artificial societies, it is particularly important to ground ABM in population reality. This paper seeks to add to the methodological literature for systematically grounding an ABM in empirical data, demographic data in particular. Starting with a philosophy of science view on grounding ABM in population reality and providing discussions on the unique rationales underlying empirical calibration of initial conditions and model parameters and empirical validation of model results, this paper contributes to the overall methodological foundation of ABM. Our methods and procedures and the illustration using our ongoing project provide a practical guide for implementing empirical calibration and validation of ABM for a wide range of social scientists.

Author

• Lingxin Hao, Johns Hopkins University