Heterogenous Agent Modelling for Central Banks

[akshayshanker]

Project

Econ-ARK

Summary

We will build a lightweight PolicyForecast module that lets central-bank analysts apply interest-rate or fiscal-transfer shocks to a calibrated heterogeneous-agent models in HARK and instantly view impulse-response results. Companion Jupyter notebook will demonstrate the workflow and serve as a template for the development of transparent central bank modelling practices and tools.

submitter

Akshay Shanker

project lead

@llorracc

Community benefit

Central-bank forecasting tools still assume all households behave identically, but policy effects hinge on differences in income, wealth, and spending. Heterogeneous-agent (HA) models capture these differences, yet they remain difficult to run with the software used by policy teams in government agencies.

To start addressing this challenge, this project seeks to bring central-bank scenario analysis into the open-source scientific-computing community. By extending Econ-ARK —and, where helpful, related toolkits such as the SSJ Toolkit, Open-Source Economics, and QuantEcon —we will prototype how policy analysts can run the same heterogeneous-agent simulations academics use, but inside the reproducible, community-maintained Python ecosystem.

• Impact for central banks: plug-and-play functions for interest-rate or fiscal-transfer shocks.
• Impact for the open-source community: a concrete applied policy use-case of open source tools that invites contributions. The proposed will also tighten links among existing computational communities in economics.
• Impact for researchers, students, and the public: openly shared notebooks lower the barrier to entry to understand and replicate policy simulations.

Ultimately, our project aims to help close the gap between a) cutting-edge HA modelling research and advances in computational methods on the one hand, and b) the practice of economic policy decisions on the other hand. The significant potential for the long-term social gain from more transparent policy analysis underscores how this proposal directly supports NumFOCUS’s mission: advancing open, trustworthy software for science and research and improving decision-making on matters of broad public interest.

Amount requested

9500

Execution plan

Project completion date: Sep 2025.

Month	Deliverable	Output format
1	Policy interface prototype - a PolicySimulator class that: • loads any calibrated HARK HA model • applies user-chosen monetary / fiscal shocks • returns impulse responses and counterfactual paths	Merged Python module + unit tests
2	Demonstration notebooks 1. Monetary-policy shock (HANK model) 2. Fiscal transfer scenario (RBC + transfers) Both include interactive dashboards (ipywidgets / voilà)	Two Jupyter notebooks
3	Outreach & training - Slide deck + live demo for a virtual workshop	Slides (PDF)

The project will be delivered by Matthew White and Akshay Shanker—computational economists with extensive experience in frontier heterogeneous-agent modelling, open-source development (HARK, Econ-ARK), and also use of computational methods in government policy applications.

Task	Hours	Rate	Cost
Build PolicySimulator class	66	$75	$5 000
Write notebooks + docs + outreach collateral	60	$75	$4 500
Total			$9 500