public-data-sources

Public Data Sources

Find and acquire source material suited to the political-science question. Dataset choice bears directly on measurement, population, coverage, and comparison, and therefore requires substantive judgment as well as access.

Data source assessment

Prepare a Data Source Assessment, accompanied when appropriate by:

• unchanged copies of the original data releases;
• concise retrieval notes or code when repeated access is needed;
• source documentation and codebooks;
• a concise record of the source, version, date of access, terms, and known revisions; and
• a reasoned comparison explaining why the selected source fits better than plausible alternatives.

Finding and evaluating data

1. Translate the research need into data requirements

Specify the concepts and indicators needed, target population, unit, geography, period, frequency, coverage, linkage keys, and acceptable measurement compromises. Separate essential requirements from features that are merely convenient.

2. Discover candidates

Search the maintained source registry, official statistical and administrative portals, public APIs, research repositories, Dataverse installations, replication archives, discipline-specific projects, journal supplements, and author pages. Search Chinese and international sources when relevant.

Prefer primary data producers and official documentation. Repository copies and aggregators can help discovery, but verify their version, transformations, and relation to the original source.

3. Verify live access and terms

Open current documentation and retrieve a small sample before recommending a source. Record whether access is open, requires free registration or a user-supplied key, or is download-only. Check licensing, redistribution, privacy, attribution, rate limits, revision policy, and file format. Never bypass access controls or print credentials.

4. Evaluate research fit

Inspect the file structure, codebook, sample rows, time and geographic coverage, missingness, revisions, unit definitions, and measurement provenance. Compare candidates on substantive fit, not brand recognition. Pay special attention to changes in definitions, boundary changes, reporting incentives, selection into coverage, and whether nominally similar measures are comparable across places and time.

5. Obtain and document the source

Retain the original release without modification. Record where and when it was obtained, the version, the terms of use, and any selections made during retrieval. When the source must be obtained repeatedly, use concise reproducible code. Do not treat synthetic or literature-imputed rows as observed empirical evidence.

6. Judge limitations and revise the data strategy

Explain what the source can and cannot measure. If no source supports the original population, construct, period, or comparison, propose the smallest defensible change to the question or design rather than pretending the data fit.

Criteria for selection

• Every recommendation has been verified against a current public source.
• The selected data correspond to the research construct and unit, not only the topic keyword.
• Coverage, missingness, revisions, comparability, and provenance are inspected before bulk download.
• Access and licensing claims come from current official documentation.
• Original releases are retained and the means of obtaining them is documented.
• Alternatives and compromises are evaluated explicitly.
• The memo changes or confirms a substantive data decision.

Source notes

• references/source-registry.md — candidate public social-science sources and access notes.
• references/discovery-platforms.md — public repositories and research-data search.

Treat registry entries as leads. Reverify them because access, versions, and terms change.