State and Country Evidence Guide

Compare social conditions by examining evidence, definitions, and time periods. This guide helps readers investigate a state or country without treating a composite number as a verdict on its residents.

Start with one question

Ask a question that a dataset can answer: how did a state’s age-adjusted overdose death rate change between two specified years? Avoid starting with “Which state is most moral?” No dataset below can answer that question.

A comparison you can reproduce

1. Choose the measure. Record its definition, unit, population, and geographic coverage.
2. Match the period. Use the same reference years for both places. The year a report appears may differ from the year it measures.
3. Choose the denominator. Raw counts reflect population size. Use an appropriate rate when comparing populations, and explain any age adjustment.
4. Check coverage. A missing observation is not zero. Changes in reporting agencies or survey methods can disrupt comparisons.
5. Keep the uncertainty. A small difference between estimates may not establish a meaningful difference.
6. Separate interpretation. State what the evidence shows before explaining what you think it means.

Choose an evidence source

Crime and violence

FBI Crime Data Explorer. Use a rate and a stated reporting year. Check agency participation and distinguish reported offenses from victimization surveys.

Overdose mortality

CDC final overdose mortality, 2022–2023. Final counts and provisional estimates are different releases. Compare the same series; record whether the rate is age-adjusted.

Households and families

Census families and households. Household composition is descriptive. It does not measure parental care, family worth, or an individual child’s future.

Mental health and substance use

SAMHSA NSDUH. Check the age group, survey year, uncertainty interval, and whether state estimates combine multiple years.

Religious participation

Pew Religious Landscape Study. Distinguish affiliation, attendance, and belief. Participation is not a measure of a person’s integrity.

International comparisons

World Bank Open Data. Use matching indicator definitions and reference years. Missing observations must remain missing rather than being assigned an invented score.

Civic knowledge

Annenberg 2026 civics survey. A question about naming branches of government is not a full citizenship test. Check sampling uncertainty and changes in survey panels.

Media trust

Gallup media trust, 2025. A survey of public trust measures attitudes toward media, not the factual accuracy of every news organization.

Why this guide does not assign a grade

A single grade requires defensible normalization rules, weights, treatment of missing data, and a reproducible calculation. Those choices are judgments that must be visible to readers. Giving an undocumented number a “provisional” label does not supply the missing evidence.

Compare the same measure across places before combining unrelated measures. A place can improve on one outcome while worsening on another. Presenting both findings is more informative than forcing them into a predetermined story of decline.

Keep a record

For each comparison, save the dataset link, table or variable name, retrieval date, reference years, values, units, and limitations. Include a brief explanation of what would change your interpretation. Readers can send a documented correction through our contact page.

Articles · Methodology · Sources · Submit a correction