Economists win prize for agent-based forecast model

Authors say theirs is the first agent-based model to produce competitive economic forecasts
forecasting

Economists Sebastian Poledna, Michael Gregor Miess and Cars Hommes have won a prize for a paper detailing what they say is the first agent-based model that can compete with more conventional economic forecast methods.

Agent-based models (ABMs) model the economy as a system of agents that interact in complex and unpredictable ways. Unlike the dominant forecast methods in use at central banks, they are run as a simulation, with results emerging from the interactions of agents. More conventional methods tend to involve solving equations or extrapolating data to reach a forecast.

The authors note ABMs have faced major hurdles in economics. They lack a “commonly accepted basis” for modelling the boundedly rational behaviour of agents. Furthermore, simulations can be designed to fit almost any outcome, so a model that fits historical patterns is not necessarily any use.

Poledna, Miess and Hommes develop a model with a large number of agents across financial firms, non-financial firms, households and the government. Each agent in the model represents a “natural or legal person in reality”. Each agent has its own balance sheet, and flows between them are based on national accounts data.

The authors test their model’s out-of-sample forecast performance against a standard dynamic stochastic general equilibrium model and a vector autoregressive model. They find the ABM produces comparable forecasts and outperforms the other models in some cases.

The paper, Economic forecasting with an agent-based model, won a competition organised by Rebuilding Macroeconomics, a four-year project to develop interdisciplinary approaches to macroeconomics.

  • LinkedIn  
  • Save this article
  • Print this page  

Only users who have a paid subscription or are part of a corporate subscription are able to print or copy content.

To access these options, along with all other subscription benefits, please contact [email protected] or view our subscription options here: http://subscriptions.centralbanking.com/subscribe

You are currently unable to copy this content. Please contact [email protected] to find out more.

You need to sign in to use this feature. If you don’t have a Central Banking account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an individual account here: