November 11, 2015

Mastering Agent-Based Economics

My master’s thesis in economics (pdf) touches many topics that I plan to research more deeply in the coming years.

Abstract: An agent-based production economy with saving consumers and a stock market is incrementally built, applying methods from modern software engineering and benchmarking outcomes with classic equilibrium results. A suite of complementary tools is key to attaining high accuracy and robustness: exponential search for price finding, sensor prices to separate information exploitation from information exploration, a stability ranking of seemingly identical firm decision heuristics, a decentralized variant of Walras’ tâtonnement process to accelerate convergence, and reincarnating agents to find selfconfirming equilibria in configurations with anticipated shocks. The production economy is extended by listing the dividend-bearing shares of firms on a stock market, where overlapping generations of mortal consumers save for retirement. Market makers provide liquidity and are shown to qualitatively impact price formation. Further adding fundamentalist traders that are listed companies themselves can lead to self-reinforcing, chaotic dynamics driven by circular ownership structures.

“In using systems of adaptive agents to create a ’guess’, we are counting on the tendencies of these systems of adaptive agents, as plodding as they are, eventually to find their ways to equilibria that we economists (who, after all, made up the model!) have difficulty finding.”
Thomas Sargent in 'Bounded Rationality in Macroeconomics'