An Algorithmic Framework for Modeling Institutional Processes

Abstract

Human behavior in a given moment is simultaneously influenced by a variety entangled institutional structures. Economic, political, legal, cultural rules and norms all contribute to this process in explicit and implicit ways. This dissertation details a framework to model institutions as a set of computational algorithms with the goal of isolating and determining the impact of particular methods of interaction. This formal analysis allows for a better understanding of the mechanisms agents require to properly engage with one another and the types of economic outcomes made possible by the presence of certain institutions. To demonstrate the utility of this approach, I apply this toolset to a variety of contexts, including O-Ring production, HOA governance, and macroeconomic policy.In the first chapter, I program an agent-based version of the O-Ring theory of development, allowing workers to endogenously match and select which countries or firms to inhabit within the model. Through this process, I demonstrate that in a world with heterogenous technology, a previously unknown Nash equilibrium exists as agents sort in the opposite manner prescribed by the original model. Further, I determine that the predicted equilibrium cannot be recovered through the adjustment of any parameters within the original piece. In other words, a market institution cannot in isolation sort workers in the predicted manner, showcasing the need for complementary mechanisms. In the second chapter, I argue that government and market institutions, when interacting with one another, lead to the generation of a network effect with respect to property value. Using data on all home values in 17 counties from the Florida Department of Revenue, I demonstrate that living within a basic governing structure, an HOA, increases the market value of an average home by nearly 20%. I then estimate the size of the network effect and demonstrate using an agent-based model that the only way to take advantage of this value increase is through interaction between market and political institutional structures as these arrangements in isolation cannot generate the desired impact. In the third chapter, I propose the initiation of a new economic cybernetics and apply this approach to the traditional problem of determining monetary policy and pursuing macroeconomic stability. Using a New Keynesian macroeconomic model as the initial framework, I remove the Taylor rule equation and give an artificial intelligence control of interest rate policy. I then outline three distinct institutional approaches to central banking as alternate reward functions in a reinforcement learning algorithm. Under this approach, the AI, with no knowledge of the underlying model and access to only a subset of the data generated by the economy, can capably achieve near optimal monetary policy under a variety of parameter values due to the algorithmic institutional structure. Further, I find that different types of central banks generate superior policy outcomes dependent upon the behavior of agents in the macroeconomy.