In between two models in Air Traffic Management, I have also helped designing and running a huge Agent-Based Model from the CRISIS project, led in particular by Doyne Farmer.
The project was aiming at bringing together a macro-economic model and a financial model in order to see their interaction and to predict new financial crisis. The project started in the aftermath of the 2007/2008 crisis, where it became clear that the “Shadow Banking System” was one of the source of the huge speculation bubble around derivative products, more specifically mortgage-back securities.
For a short (and simplified) story of the crisis itself, you can have a look here.
The CRISIS-economics model was designed as an Agent-Based Model, with many different types of agents, including:
- banks of different types,
- pension funds,
- central government.
As a consequence is was able to catch to feed-back effect in the economy, see the official github page for more details.
More specifically, I was involved in the core mechanisms of market clearing. With one of my colleagues, we designed in a new approach which is able to clear multiple inter-related markets (networks between buyer and seller). We found a proof that given some pretty general assumptions on the demand and supply functions of the agents, there would be always a solution (i.e. cleared prices or interest rates). We also found an algorithm for this, with is quite powerful and which is able to clear high numbers of edges (pairs buyer-seller) at the same time. We are in the process of finishing a paper on the subject.