You are here: Home Seminar 2018/19 Winter Term Applied Micro Workshop – Felix Mauersberger, “Thompson Sampling: Endogenously Random Behavior in Games and Markets”

Applied Micro Workshop – Felix Mauersberger, “Thompson Sampling: Endogenously Random Behavior in Games and Markets”

— filed under:

Felix Mauersberger: “Thompson Sampling: Endogenously Random Behavior in Games and Markets” Economists often consider human behavior as stochastic, being reflected by the wide use of random shocks. However, random shocks have important limitations in explaining or forecasting behavior, if shock sizes differ systematically across environments or over time. This paper proposes Thompson Sampling, a learning heuristic from the bandit literature, as a positive theory of human behavior across very different situations in economics. The distinctive feature of Thompson Sampling is an endogenous distribution of random shocks. Agents, having limited information about their environments, update their subjective belief distributions in a Bayesian way and subsequently make a random draw from the posterior. Conditional on that random draw, agents optimize. The empirical value of Thompson Sampling is demonstrated using data on experimental games. These datasets are chosen, because (i) strategic interaction is an important ingredient of economics, (ii) the controlled laboratory environment allows observing the cognitively driven randomness. The empirical fit and the out-of-sample predictive ability of Thompson Sampling are compared to other hypotheses of noisy decision-making: mixed strategy Nash equilibrium, Bayesian learning with exogenous shocks and quantal response equilibrium (QRE). The applications show that Thompson Sampling has superior empirical properties to the mixed strategy equilibrium and the other benchmarks.

Event details
What
  • Presentation
When Nov 06, 2018
from 16:00 to 17:30
Where briq, Seminar Room 9/1.1, Schaumburg-Lippe-Straße 9
Add event to calendar vCal
iCal

More information about this event…

Document Actions