Contribute
nstate is built by people who believe the evidence should be public and the methods should be checkable. You don't need to be an economist. If you can find a primary source and write a sentence, you can contribute.
Scorecard
Find a government programme with a measurable objective and public data. Write a scorecard following the methodology template. Submit a PR.
Data pipeline
Python or R script that fetches, cleans, and versions a government dataset to CSV/Parquet. Even a one-source pipeline is useful.
Source verification
Check a published scorecard's data sources. Confirm the figure matches the original release. Report discrepancies via GitHub issue.
Country chapter
Adapt the methodology for your country's government data. Become a chapter maintainer.
Methodology challenge
Disagree with a counterfactual approach or evidence quality rating? Open a PR with a counter-analysis meeting the same evidence standards.
Shadow agent (Phase 3)
Build a shadow automation of a government administrative process. Must run in read-only mode initially. Agents recommend; they do not act.
how to submit a scorecard
This is the core contribution type. Follow these steps exactly. A PR that skips a step will not be merged.
State your claim
Open a GitHub issue: "Scorecard: [policy name] [country]". State the policy, its stated objective (verbatim from official source), and the primary data source you will use. Wait for a maintainer to assign it to you — this prevents duplicate work.
Find the data
Every figure must come from a primary official source. If a figure only appears in a news article, it does not qualify. If you cannot find primary data, close the issue and explain why — that is a valid outcome.
Write the pipeline
Write a script in country/pipelines/ that fetches the data from the primary source and outputs to country/data/. The pipeline must be reproducible from scratch with a single command.
Complete the template
Copy uk/scorecards/TEMPLATE.md. Fill in all seven sections. Every section is mandatory. If you cannot complete a section with evidence-quality HIGH or MEDIUM, the evidence quality for that dimension is LOW or NOT MEASURABLE — write that.
Open a PR
PR title: [SCORECARD] [Country]: [Policy name]. Include: the scorecard file, the data pipeline, and the raw data output. A maintainer will review source citations and evidence quality ratings. Expect back-and-forth — that is normal and expected.
start a national chapter
Title: "Country chapter: [country name]". Provide your name (or handle), country, primary data source for your first scorecard, and confirmation you have read the methodology.
Copy COUNTRY-TEMPLATE.md → your-country/README.md. List your country's primary statistical releases, the equivalent of OGL, and your first-wave scorecard priorities.
Submit at least one scorecard via PR following the full methodology. The chapter is not listed on the countries page until the first scorecard is merged.
Chapter maintainers must disclose political affiliation or material conflicts relevant to the policies they audit. Independence is not optional — it is the methodology.
what we will not merge
- Figures without a primary source URL and data release date
- Scorecard verdicts that pre-determine the outcome before seeing the data
- Content that targets named individuals rather than programmes and policies
- Counterfactuals at Level 5 (descriptive only) presented as causal evidence
- Evidence quality ratings that overstate certainty
- Analysis of policies below £100m/year without demonstrated high public interest
- Scorecards where the contributing maintainer has an undisclosed material conflict
The repo is at github.com/nstate-project/nstate. Issues are the first step.