Agent-based Computational Economics â€“ The start of rational forecasting for good
Historically, economic forecast â€“ on which most countries taxation, interest and spend policy is based - has used Dynamic Stochastic General Equilibrium (DSGE) models (1). In 2008, these models failed and a financial crisis of gargantuan proportions followed. The echoes of that crisis can be still heard in the economic, financial and political systems of the western world.
Human resourcefulness knows no bounds and shortly after, a number of academic institutions look for solutions to the forecasting problem. And they found a promising one. Not too dissimilar to weather forecasting â€“ economics can also be forecast bottom up. In this approach, an economy (or a sector or a region) is modelled by modelling each one of the economic atoms (an â€œagentâ€) - be a family or a firm - and each one is giving a behaviour, which today can span from basic to quasi-AI, and thus provide with nuances not possible in DSGE models â€“ such as adaptive, irrational or even criminal behaviour. Such agent models can then be run under a number of scenarios and give us valuable data on the future.
This is referred as Agent-based Computational Economics (ACE for short). And this is not a dream, some small scale exercises in the EU and USA have already been done (2). Technology is here that allows to simulate systems in the region of 500 to 1000 million families and firms â€“ we just need to get going to code the behaviours of interest.
And while the benefits to mankind are vast â€“ there are also practical upsides. Firms will then be able predict the impact of new marketing campaigns, product launches, new competitor entry, regulatory changes or new supply chains.
We live in a complex world â€“ of the kind people could not have imagined 50 years ago, and which is getting more complex by the pass of each day. We cannot leave the future to random guessing, intellectually flawed arguments or politicians motivated by short term goals.
Juan E Amador
Global Head Financial Crime Risk Technology
HSBC Bank PLC
(1) For an eloquent, detailed, introduction see: Andre Haldane, The Dapple World, 10 November 2016.
(2) A good summary of a EU-sponsored initiate can be found here:
Gencer, M; Ozel, B. Agent-Based Modelling of Economic Systems: The EURACE Project Experience; 2010. http://www.ecomod.net/sites/default/files/document-conference/ecomod2010/1316.pdf
And for the US here:
Atxwell, RL; 120 Million Agents Self-Organize into 6 Million Firms: A Model of the U.S. Private Sector; Proceedings of the 15th International Conference on Autonomous Agents and Multiagent Systems, 2016. http://www.ifaamas.org/Proceedings/aamas2016/pdfs/p806.pdf