Long-term causal effects via behavioral game theory

Random experiments are the gold standard in reliably comparing the causal effect of switching from a baseline policy to a new policy on socio-economic platforms. One critical shortcoming of classical methods, however, is that they do not take into account the dynamic nature of response to policy changes and may fail to capture long-term effects. We formalize a framework to define and estimate long-term causal effects of policy changes in multiagent economies, using behavioral game theory and a latent space approach, where a model of how agents act conditional on latent behaviors is combined with a temporal model of how behaviors evolve over time.

To appear in NIPS 2016. Joint work with Panos Toulis, Econometrics and Statistics, University of Chicago, Booth School.

https://arxiv.org/abs/1501.02315