How can I question official data responsibly?

Official data deserves examination, not automatic belief or automatic dismissal. Begin with what the producer says it measured and how that relates to the claim being made.

A department's press release, an administrative count, a sample survey and a court record serve different purposes. Identify the source type before deciding what it can establish.

Ask about the measurement

Find the population or cases covered, period, definitions, collection method and missing information. Ask whether the table counts people, households, events, applications or payments.

Then ask about the conclusion. A programme may correctly report applications received while a commentator incorrectly describes that number as benefits delivered. The source and the interpretation require separate assessment.

Read the producer's own limitations

MoSPI's release for PLFS Annual Report 2025 explains that the sampling design changed from January 2025 and that annual reporting moved to a calendar-year cycle. Its comparability note warns that the new estimates may not be strictly comparable with previous annual estimates. A chart that ignores this qualification can overstate what a trend shows. 1

The useful response is not “therefore the data is fake”. It is to make the methodological change visible and avoid claiming a precise trend the comparison cannot support.

Do not average incompatible official numbers

A 2026 parliamentary answer gives three different 2024 road-death figures and explains differences in reporting channels and categories. It distinguishes total road fatalities from deaths in a penal-law negligence category. These are not automatically three interchangeable estimates of precisely the same measure. 2 3

First determine what each series counts. If the explanation leaves a gap, name that gap instead of picking the number that suits your argument.

Bring in other evidence

An individual's observation may reveal something a national measure does not capture. Independent reporting may identify omissions or inconsistencies. Evaluate whether those sources address the same place, period and definition before treating them as a contradiction.

Criticism should be specific: “This excludes these cases”, “this definition changed”, or “the underlying table is unavailable”. An unsupported claim that every number is manipulated is not a method of checking data.

The aim is a conclusion you can explain and revise. Sometimes that conclusion is that a statistic is useful for one question and unsuitable for another.