Workshop on Economic Science with Heterogeneous Interacting Agents (WEHIA) 2019

Last week we attended WEHIA 2019 at City, University of London. WEHIA is an annual conference bringing together researchers, mainly in academia, from a number of disciplines, including economics (e.g. econophysics; complexity economics; experimental economists; behavioural economics), finance, physics, computing – all interested in understanding the economy as a complex system made up of heterogeneous interacting agents. The conference took place over three days and involved plenary talks, sessions and roundtables.

The conference started on Monday with a wide variety of sessions, including topics such as: innovation and technological evolution and adoption; financial markets; agent-based macroeconomics; market microstructure; and even crypto and distributed ledger technology.

We attended the agent-based macroeconomics track and while many talks were not directly relevant, it was interesting to see how formal the models were when compared to, say, models in social simulation. There was an interesting talk by Jakob Grazzini, who presented work showing Bayesian estimation of an macroeconomic ABM.

After lunch, there was a plenary talk given by Daniela Puzzello looking at evaluating monetary policy in the laboratory (experimental economics). It was a great example of experimental economics: comparing theoretical results with experiments done with actual people. More info here.

The day finished with an interesting roundtable discussion on ABMs and market surveillance moderated by Paul Davies (from the Wall Street Journal) with a panel that included Doyne Farmer, Rama Cont, Andrei Kirilenko and Sheri Marakose. They provided a number of insights about the current state of use of ABMs, in particular in finance. The noted that how compared with 10 years ago, there is more data available to calibrate models at the micro-level, but that it can be difficult to get access to the data, in particular financial data.

During the roundtable, Doyne Farmer also posed a challenged to the community: we’re seeing some adoption of ABMs (e.g. Bank of England) but to get to the next level ABMs need to predict and to beat the competition (i.e. DSGE models in macroeconomics). When will this happen?

On Tuesday, there were a number of interesting sessions, including: ABM estimation and calibration; spatial economics; behavioural economics; game theory; empirical financial analysis; and market microstructure.

We attended the ABM estimation and calibration track. A talk of particular interest was given by Donovan Platt, who presented his work comparing different approaches to estimating economic agent-based models. He found Bayesian approaches to show some promise compared to frequentist approaches. For his work, Donovan won best methodology paper at the conference. More info here.

More generally, Donovan and the community are working to produce a set of benchmark models for evaluating estimation/calibration approaches. Something to keep an eye on.

In the afternoon the Bank of England organised a track focussed on policy and gave out an award for the best policy paper. The track included papers from a diverse set of methodologies, including DSGE models (the “competition” in macroeconomics), as the Bank is trying to encourage a more pluralistic, interdisciplinary approach.

The final day’s session were also diverse, including: growth and development; financial stability; mathematical finance; ABMs for financial markets; systemic risk; information aggregation and shocks; inequality; and of course machine learning and artificial intelligence. It’s unfortunate that we were not able to attend more sessions!

In the morning we attended a great talk by Omar Guerrero (UCL and Turing institute). Omar presented a really interesting approach for policy priority inference to be used by governments in accessing and planning economic development. For more info, see the brochure.

Another talk we found interesting, by Olivier Simard-Casanova, looked at management style linked to turnover/collective action in organisations. Olivier wrote his model in Mesa – an open source Python ABM framework (integrated with Sandman) – go Python!

Friederike Mengel gave an interesting plenary looking at opinion dynamics (belief formation) in networks (real and virtual). Another great example of experimental economics. More info here.

Overall, we found the conference really interesting, particularly the variety of sessions, plenaries and rountables. Thanks to the organisers.

Finally, as for ABMs and their wider adoption, it was fascinating to see the presence of a number of central banks. It appears the joint work of the Bank of England and Doyne Farmer’s INET group on a UK housing market models has sparked wider interest and a number of variants. We look forward to seeing how this progresses…

Comments are closed.