Access
This article is part of Nature's premium content.
Published online 5 August 2009 | Nature 460, 680-682 (2009) | doi:10.1038/460680a
News Feature
Economics: Meltdown modelling
Could agent-based computer models prevent another financial crisis? Mark Buchanan reports.
It's 2016, and experts at a US government facility have detected a threat to national security. A screen on the wall maps the world's largest financial players — banks, governments and hedge funds — as well as the web of loans, ownership stakes and other legal claims that link them.
To read this story in full you will need to login or make a payment (see right).
Comments
Reader comments are usually moderated after posting. If you find something offensive or inappropriate, you can speed this process by clicking 'Report this comment' (or, if that doesn't work for you, email webadmin@nature.com). For more controversial topics, we reserve the right to moderate before comments are published.
Reading between the lines, the term 'agent based' seems to represent TWO things. First it is a modeling technique that produces emergent behavior in ways we cannot produce analytically. Second, the underlying complaint seems to be that rational economics models use only a fraction of the actual independent variables and interactions that collectively produce real world behavior. Agents create domains of similar behavior, greater or lesser interactions over a network, friction, latency, and so on. A simulation with differential equations could be made to do this as well, though perhaps not so elegantly. But the heart of the problem is the second characteristic, which rational economists have ignored for far too long, armoring themselves with a shell-like material that goes by the name of 'dismal.' How about we make economics a science, using behavoral approaches to define true independent variables, and agent-based models that are more than fantasy matches for reality?
Great article, but: we do not need better models, but better maps (and mappings). As long as I do not possess perfect online maps (in real time) of all (digital!) financial transactions happening world-wide (e.g., by tracking all mobile phones), there will be no much hope for a "better" steered Economy. And please note that even if I had such perfect online maps in future (2020), there will still be the nasty possibility of "dark nets"... (i.e., Mafia circles building up their own financial webs, i.e., intranet economies that are not mapped by THE official and transparent economic webs, economists, and scientists). And those who will have the "best" maps...
Agent-based modeling may be a great technology for financial and economic reseach, especially when the market is far-from-equilibrium. Frankly speaking, there may be some questiones need to be talked about: ·How can the agent-based models simulate the real marekt? |-The virtual markst's complexity and the agents' intelligece need to be talked about. |-As well as the relationshipes between these agents. ·How can the agents make their decisions? |-There maybe different patternes. Recient papers show that there may be simply patternes and more intelligent patternes in a nutshell. |-Simples patternes may be all the agents are one type or several types, just like fundamentalists, chartists and noisy traders, but these decision making patternes are not agility. In other words, these patternes can not be evolutionary. |-Intelligent decision making means evolutionary patternes, which can be evolution depend on the agents' decision records and performance,just like genetic algorithm, genetic programming, neural network,fuzzy algorithm, etc. ·How about the network between these agents? |-Different decisignes of networks, different relationshipes between agents, different coplexity of the market. ·How many assets in the market? |-This need to disscuss the complexity of the market or world which is the environment for agents living. The regulations in the virtual market just as the regulations in the real world, they are the constraints of agents' behavior. In a word, agent-based modelling is a good tool for the solution of the world. Also, there are some areas of this methodology need to be explored deeply.
Modelling (Simulations) commes in three levels: Process, Product and Economic. The main issue is change. The main problem is what are the rules that define this change. Process simulations are the easiest. The mathematics involved are often well known and the systems can easily be tested. Product(ion) simulations are more difficult. They often include human behavior. The smaller the area studied the easier. Economic models are the most difficult because financial issues are involved. In the past most economic systems were small (between production unit and bank) and country based, currently they are global (between banks world wide, pension funds etc) It is true that "the major chalenge is spycifying how the agents behave and in choosing the rules the agents use to make decisions". On the other hand IMO in many cases this is impossible. To use common sense as part of a simulation is even "wrong". IMO creating a carefully crafted agent-based model of the whole economy is impossible accurately and not worthwhile the effort. One of the major problems is time delays and all internal money transfers. On the other hand to continue to study world production as done by the Club van Rome I think is a must. Nicolaas Vroom Schilde