ThinkCritical.eu · Manipulation atlas
Numbers and chartsBiased sample
Only those who wanted to answered, or too few did – the result can't be generalised.
A survey only tells you about the people who answered it. If they chose to take part – the happiest, the angriest, the most active – or if there were only a few of them, the result can't be applied to everyone.
A cycling fans' group on social media asked: “Do you want more cycle lanes in town?” 94% of voters said yes. The result shows what the group's followers think – not what all the town's residents think.
How to spot it
- An online, social media or TV poll where anyone who wants to can vote.
- The number of respondents is missing or very small (a few dozen people).
- The question was asked by someone followed mainly by people with one point of view.
- The result is suspiciously one-sided – 95% or 100% for one answer.
How to check and what to do
- Find out how many people answered and how they were chosen – at random, or did they sign up themselves?
- Look for a survey with a representative sample; a reputable pollster states the sample size, dates and method.
- Allow for uncertainty: even with 1,000 well-chosen people, the result can differ from reality by around 3 percentage points.
- Ask who is missing from the survey – and whether they would have answered differently.
Tip: An internet poll measures who felt like voting, not what everyone thinks.
Full explanation and exercises:
thinkcritical.eu/en/manipulation-atlas/biased-sample