Perceptron inspired macroeconomic ABM- A path to multiple equilibria?


I will present a simple agent-based model of the macroeconomy to demonstrate that a large number of stable equilibria can exist in macroeconomic models. The starting point is a simple budgetary constraint mapped to the perceptron, a well-known model in machine learning. We generalise the perceptron problem to allow for the introduction of debt. We then proceed with the addition of some basic dynamical rules to make the model economically complete. Two distinct regimes are found called “flexible” and “disciplined”, respectively characterised by the level of indebtedness allowed for each agent. We will then present differences between these two regimes by looking at various macro-level indicators such as the distributions of supply and demand, and their dynamics, Finally, we will show that the transition between these two phases is robust over a wide range of parameters permitting us to have a parsimonious model.

Jan 21, 2020 12:00 PM
2 rue Andre Pascal, Paris, Ile-de-France 75016

This is a recent talk I gave at the OECD for the New Approaches to Economic Challenges (NAEC) initiative. This brief talk distills the essence of the approach that I have adopted for my doctoral work. Given the time constraints, it’s necessarily short on details but all the big ideas are there.

A word on NAEC. This is a recent initiative undertaken by the OECD. Its members seek to understand the challenges facing the world economy by using the latest research and identifying the best analytical tools to build a systemic perspective. The people I have met here are very open-minded and eager to learn, understand, and question diverse approaches to macroeconomic modelling.

Dhruv Sharma
Dhruv Sharma
Researcher - Portfolio Construction, Risk Management, Agent-Based Models, Statistical Physics.

Physicist turned economic modeler trying to make sense of the world through ABMs.