This project, my teammate, Casey Nugent, and I designed a model of a the relationships between banks, borrowers, and the Federal Reserve (“Fed”) that maximizes social welfare and minimizes involvement of the Fed all the while avoiding a catastrophe similar to the real-world 2007 subprime mortgage crisis that resulted in the to-date failure of over 130 banks.
I implemented a design for a multiagent system representing a simplified model of the United States banking system, in which a number of bank agents, many borrower agents, and a single Federal Reserve agent interact in ways that resemble their real-world counterparts. Each agent supports a number of parameters that determine its characteristics (e.g., amount of money), and influence its behavior (e.g., ability to make a monthly loan payment). A particular parameterization of an agent can lead to conditions that are different from those of another parameterization.
Since there are many such parameterizations for each of the many agents in our system, it is not feasible to run simulations for all parameter combinations. Using the simulation, this project's investigation was focused on characterizing the conditions under which market bubbles can emerge in loaner bank and borrower relationships. More broadly, this investigation explores how the simulation's parameterizions correlate with the likelihood of bubble emergence, bank loan provision monopoloies, total default situations, and generally, the longevity of each parameterization class. The hypotheses guiding our investigation:
The multiagent model utilizes three types of agents: borrowers, banks, and the Federal Reserve. While borrowers and banks engage in negotiations, banks and the Fed engage in behavior monitoring and regulation. Whether or not to make behavior adjustments will be based upon changes in the moving averages of relevant variables.