Investigating claims and using public records
Drawn from the Governing for Life discussion paper. Linked references open the supporting sources.
How to investigate a claim
Instead of relying on the label “fake news,” investigate the specific statement. An outlet may publish an accurate report, a misleading headline or unsupported commentary. A generally reliable publication can make mistakes, and an unfamiliar source may point to a genuine document. The useful questions concern evidence, context and verification.
- Write down the precise claim. Separate statements such as “voted against a bill,” “opposes education” and “caused poor educational outcomes.” They require different evidence.
- Find the original material. Seek the full bill and vote, complete speech, dataset, research paper, financial filing or court decision. A short clip or screenshot may omit crucial context.
- Check identity and date. Confirm who produced it, the jurisdiction, the event date, the publication date and whether an old item is being presented as current.
- Examine the surrounding context. Check qualifications, the exact question answered, bill amendments and whether a vote concerned procedure or final passage.
- Investigate the source outside its own website. Look for ownership, relevant expertise, corrections and independent descriptions of its record. Professional-looking design establishes little.
- Seek genuinely independent corroboration. Several articles repeating one allegation or press release do not provide several independent confirmations.
- Match confidence to the evidence. Distinguish documented facts, disputed interpretations, allegations, forecasts and matters that remain unknown.
- Record the link and explanation. Before sharing, be able to tell another person what the evidence establishes and what it does not.
Library of Congress guidance emphasizes examining a primary source’s creator, purpose and perspective, and considering additional evidence. An original document can faithfully record an allegation without proving the allegation. A full interview can establish that words were spoken without establishing that their content was true. [V1]
Research also needs scrutiny. Ask whether the sample represents the people discussed, whether the finding is a correlation or a causal estimate, whether uncertainty is reported and whether later research contradicts it. A claim that a policy “works” needs a specified outcome and comparison, not simply an approving quotation.
Public records readers can use
For U.S. federal offices, the FEC provides candidate and committee financial disclosures and guidance for researching candidates. These records reveal reported contributions and expenditure; they do not expose every motive or all political influence. [V2]
The U.S. Senate’s vote guidance links readers to recorded congressional votes. Read the associated legislative text and the question voted on before accepting an advertisement’s characterization of a vote. For state and local offices, consult the relevant legislature, election authority and public meeting records. [V3]
In New Hampshire, the Secretary of State provides an online campaign-finance system and candidate financial-interest filings. Other jurisdictions maintain their own disclosure arrangements. These are useful starting points for investigation, with the completeness and relevance of each record evaluated on its merits. [V4]
For claimed outcomes, look for original statistics and independent audit or evaluation reports. Government publication is useful provenance but does not make an interpretation infallible. Check definitions and revisions; budgets, appropriations and actual spending are different figures. Court filings, charges, findings and final judgments also have different meanings.
Full interviews, debates and candidate policy documents help establish positions and commitments. Independent journalism can investigate facts those materials omit. Fact-checking articles are useful when they make evidence and reasoning visible; inspect their underlying sources instead of relying solely on a rating.