Checking publication dates

Policy relationships: latest evidence and sources

Withdrawn evidenceDerived survey analysis

The former matrix displayed correlation coefficients without committed respondent-level inputs, named variables, survey coverage, weighting, exclusions or a reproducible calculation. No relationship strength, ideological cluster or causal link is being inferred or carried forward.

What must be true before this evidence returns

  • Publish the respondent-level or aggregate input dataset with a lawful and reusable licensing basis.
  • Name every source variable, survey wave, sampled population, geography, weighting scheme and missing-data rule.
  • Define the correlation or association statistic, coding, exclusions, multiple-comparison treatment and uncertainty before calculation.
  • Keep statistical association separate from causal or political interpretation and publish sensitivity checks.
  • Test the full transformation against fixed fixtures and expose the method and limitations beside any returned matrix.
Read the evidence and withdrawal policy