Since I read Social Atom by Mark Buchanan, I have been investigating agent-based models applied to economics and finance. Andrew Lo at MIT is working on behavioral finance models based on adaptive agents that evolve and are selected in a “survival of the fittest” environment. Rajiv Sethi refers to a Nature article applying agent-based models to macroeconomics.
What is interesting in agent-based models is that very simple rules at the individual agent level can create extremely complex system through interactions and networks of agents. Such networks would be able to model non-linear effects, by introducing stocks, flows and delays as suggested by System Dynamics theory.
That reminds me of a the Mandelbrot set fractals, where the iteration of an extremely simple equation generates an infinitely detailed image.
I think that a behavior rule for agents as simple as “do what the most successful agent did and avoid what failed/bankrupt agents did” would be enough to generate interesting results. As I develop my simple agent-based model, I will keep you posted with the results.
One of the discussions that I am following closely is the one related to the crisis of financial modeling. A side effect of the financial meltdown of August 2007 (subprime) - September 2008 (Lehman, AIG) and the following credit crunch, is that existing models, mostly based on the Efficient Market Hypothesis, showed all their shortcomings.
My take is that all pre-meltdown models were representing static snapshots of the markets based on equilibrium hypotheses. New models will have to add a temporal axis and describe markets as dynamic systems, at the cost of abandoning formal elegance, nice equations and closed-form solutions.
The bare existence of “stocks” (i.e. pools of money, investment funds, house inventories), each with its own incoming and outgoing “flows” (i.e. investment flows, equity commitments and exits, construction of new houses and house sales) and related delays (i.e. time between investment decision and time to measure investment results, especially in illiquid markets) creates a dynamic system regulated by differential equations whose solutions are exponential and periodic functions over time.
Accumulating and depleting stocks of vast size allow the financial markets to deviate from equilibrium positions for long periods of time. System unbalances can build up and grow exponentially over years before collapsing. Social interactions among investors (aka herd behavior) can inflate bubbles that are sustainable for much longer than any static model can predict, hence reinforcing herd behavior.
The presence of stocks and delays in the system makes it possible for sources of instability to accumulate until finally some random event triggers a chain effect reversing the same self-reinforcing loops into self-destructive loops.
Some more food for thought:
