How to use mistakes to improve credibility

Abstract

The article argues that errors in statistical publications should not be seen solely as a threat to the credibility of the statistical institution, but that they could, with careful handling, actually contribute to public confidence in the institution. Hiding errors or disguising them as “revisions” puts credibility at risk. The author calls for a policy for dealing with errors in statistical institutions. Errors should be identified and corrected figures published loud and clear as soon as possible. The policy must include systematic collection of information on the occurrence, causes and consequences of errors. This information should be used internally to prevent future errors, and published externally as part of the institutions quality declaration of statistics. Preliminary experience from Statistics Denmark seems to indicate that serious or minor errors occur in 1–2% of statistical publications.