The Community Library is an open-source library of baseline models in computational economics, implemented and documented across the toolkits that can solve them. For each model the library documents the model statement and the calibration, so that implementations in different toolkits can be compared side by side. Each solution is authored and maintained by its submitters, whether a toolkit's project or individual contributors, and is open to review by the community.
The library is hosted by QuantEcon and published at https://quantecon.github.io/community-library/. It is run in collaboration with the SCE Working Group 1 on Language and Semantics. This repository holds the community website, the shared model descriptions and the solutions submitted by toolkit projects and individual contributors.
Read the buffer stock model description or its PDF, which covers model theory, calibration, numerical solution and the results each implementation reports. The HARK notebook provides a runnable example with short explanations; its instructions give the installation and run commands.
Keep the model and calibration fixed, and choose your own prose, solution
method and runnable format. The submission guide
explains where to put a solution, which results to report and how to
document its use. Every solution includes a short metadata.yml
recording its authors, the description version it follows and its AI
disclosure; the contributing guide explains these, and the
guide's CSV formats and draft comparison tolerances are optional.
The working group has proposed five baseline models. Projects choose one or more that their toolkit can solve and also contribute a tutorial on a model of their own choosing.
| Baseline model | Model class | Description |
|---|---|---|
| Buffer stock saving | Partial-equilibrium consumption–saving under income risk | Draft available |
| Aiyagari · Krusell–Smith | Incomplete-markets heterogeneous agents | In preparation |
| A two-asset HANK model | Heterogeneous-agent New Keynesian | In preparation |
| A small New Keynesian DSGE model | Representative-agent DSGE | In preparation |
| A baseline agent-based macro model | Macro from interacting heterogeneous agents | In preparation |
| Directory | Purpose |
|---|---|
models/<model>/ |
The shared model description, its PDF and the submission guide. |
solutions/<model>/<NNN>-<toolkit>-<method>/ |
A runnable solution in its submitters' chosen format, with its metadata.yml, supporting code and results. |
docs/ |
The library website. |
templates/ |
The page layout used to produce model PDFs. |
schemas/ |
The schema each solution's metadata.yml follows. |
tools/ |
Scripts that build the model description pages and check versions and disclosure records. |
A model has the same directory name in both places, and its solutions are
numbered in submission order: the buffer-stock description is in
models/buffer-stock/ and the HARK solution in
solutions/buffer-stock/001-hark-egm/.
Questions about a shared model description belong in
the working group issue tracker.
The website consists of static HTML, CSS and JavaScript in docs/. Its
appearance follows the QuantEcon book theme.
Edit docs/index.html for text, docs/site.css for appearance and
docs/site.js for the contents navigation. docs/contributing/index.html is
the contributing guide for people; CONTRIBUTING.md holds the full rules, for
contributors and coding agents, so a change to a rule there updates the page
too. docs/sce-wg1-2026-10-01.html
is the Project B talk to SCE Working Group 1 (1 October 2026), a
self-contained reveal.js deck.
Each model description also gets a page at docs/models/<model>/, built from
its model-description.md by tools/build_pages.py with the PDF alongside.
The pages are built when the site is published, not committed. To preview
them:
python -m pip install -r tools/requirements.txt
python tools/build_pages.pyThe publishing workflow builds those pages
and runs tools/check_metadata.py, which checks that the descriptions, each
solution's metadata.yml and the library table agree on versions and
disclosure. It then checks local file links, links to sections and the
contents navigation on every push and pull request. It publishes docs/ to
GitHub Pages from main; pull requests provide a downloadable site preview.
Notebook execution is not yet automated.
Publishing requires Settings → Pages → Build and deployment → Source: GitHub Actions. The buffer stock guide gives the commands for generating the PDF from the model description.
Text is CC-BY; code carries an OSI-approved licence. Contributions remain authored and maintained by their submitters — the library curates and publishes; it does not own anyone's work or rank the toolkits.