Macrocosm

Macrocosm builds agent-based models of the economy: simulations in which each firm, household, bank and government is represented on its own, with its own constraints and rules, and figures such as output, prices and employment come out of their interactions instead of being assumed.^1,3 This approach is known as complexity economics.³ It was founded in 2023 to turn into products more than a decade of work by researchers at Oxford's Institute for New Economic Thinking, and it is a public benefit corporation.² Doyne Farmer is its founder and chief scientist.²

It offers three simulators.¹ The supply-chain simulator takes a shock such as an earthquake, a pandemic or a tariff and traces it week by week along supply links, giving output losses by firm, sector, country or region.⁶ As an example the company has published a scenario of a 15-week closure of the Strait of Hormuz across 80 countries and 49 sectors.⁶ The energy simulator represents generators, transmission lines and market participants, so a user can shut down a plant or add a battery and follow the effect on dispatch, prices and investment.⁵ The macroeconomic simulator forecasts GDP, inflation, employment and movement of workers between occupations from simulated firm and household behaviour, and is aimed at central banks, asset managers and policymakers.⁷

By the company's own account, its models start from a mix of proprietary data and public sources such as OECD input-output tables (accounts of what each industry buys from every other), national inventory statistics, bilateral trade flows and infrastructure data.³ It says it calibrates them with Bayesian inference, keeps them stock-flow consistent so that the accounts balance at every step, and has backtested its macroeconomic simulator across 38 countries and its supply-chain models against the 2011 and 2016 Japan earthquakes.^3,5 The published work it points to includes a 2022 paper by Anton Pichler, Marco Pangallo, R. Maria del Rio-Chanona, François Lafond and Farmer in the Journal of Economic Dynamics and Control.^4,9 That paper describes a disequilibrium input-output model that made real-time forecasts of the fall in UK economic activity by sector in the second quarter of 2020.¹⁰

Access is by contacting the company; its design-partner form offers partnerships, paid pilots and research or data collaborations.^1,8 For someone modelling a place, Macrocosm shows what a whole-economy agent-based model looks like; its home page zooms from one city's households, firms and banks out to the globe.¹ A team asking how a supply or energy shock would reach a region's industries and jobs could commission such a study, and the 2022 paper shows the method in the open literature.^6,10