Fact-First Governance
-
Lead with facts, not vibes. Ask: What’s the claim? What’s the evidence? How strong is it?
-
Prefer primary over punditry. Laws, datasets, audits, and studies beat hot takes.
-
Name the tradeoffs. Real solutions have costs and limits; acknowledge them.
How to spot good evidence (vs. vibes)
Good evidence tends to be:
-
Checkable: Links to data, statutes, audits, or studies you can open.
-
Transparent: Methods and limits are stated; sample sizes shown; uncertainty acknowledged.
-
Replicable/Corroborated: More than one credible source points the same way.
-
Proportional: The strength of the claim matches the strength of the data (no sweeping certainty from a skimpy survey).
Vibes look like:
-
Anecdotes as proof: “My cousin’s district did X, so…”
-
Cherry-picked charts: Selective endpoints, no denominators, misleading axes.
-
Certainty theater: Absolute language, zero limits, lots of adjectives, few receipts.
-
Authority swap: “Trust me, I’m an expert” without method or citations.
Red flags: No link to sources, screenshots of stats with no origin, tiny or unbalanced samples, polls with leading wording, “study” press releases without a paper.
The 3-Question Source Test
-
Who produced it, and what are their incentives?
Government auditor? University lab? Advocacy group? Commercial vendor? What’s their track record?
2. How was the evidence collected?
Random sample? Full population? Administrative records? Self-reports? Is the method fit for the claim?
3. Can an informed critic reproduce or falsify it?
Are data and methods open? Are limitations and uncertainty quantified? Has anyone independent replicated it?
Verdict rubric:
-
Green (use with confidence): Transparent methods/data, credible publisher, replicable.
-
Yellow (use with caution): Partial transparency, minor limitations, no replication yet.
-
Red (do not rely): Opaque methods, conflicts not disclosed, non-replicable.
Common traps (and fixes)
-
Trap: Confusing correlation with causation.
Fix: Ask what else could explain the pattern; look for experimental or quasi-experimental designs. -
Trap: Overweighting dramatic outliers.
-
Fix: Ask for medians, interquartile ranges, and trend lines.
-
Trap: One-study certainty.
Fix: Seek meta-analyses or at least two independent studies.
Quick checks you can do in under a minute
-
Open the link trail: Does the headline trace to a study/report…or to another headline?
-
Scan the methods box: Randomization? Sample size (n)? Margin of error? Time frame?
-
Look for denominators: “+50%” of what baseline? Per 100,000 people or total counts?
-
Compare across sources: Do CRS/GAO/CBO/state fiscal notes align or diverge? Why?
