It’s that time of year again – British Crime Survey data are released, and the media misinterpret them. This time the problem is that old chestnut ’statistical significance’. Journalists have looked at the figure for all crime, seen that the survey reports a rise of 4%, and then written their copy. A typical line, this one from the BBC is ‘ Separate figures from the British Crime Survey (BCS) – measuring people’s experiences of crime – suggest crime went up by 4%. Not wanting to single out the BBC, the Telegraph did no better, reporting that ‘the British Crime Survey showed an increase’ [in crime]. The Telegraph went one better though. Having failed to interpret the BCS data correctly, their article goes on to cast doubt on the survey data (collected via face-to-face interviews with over 40,000 people), preferring instead to rely on what we can only characterise as ‘common sense’: ‘People are not stupid. They know that it stretches credulity for official figures to suggest that 2011 was a good year for crime-fighting.’. Who needs data when we can just rely on people’s impressions?
Top marks to the Guardian on this occasion though, who correctly reported that the survey figures show ‘no statistically significant change’ in overall crime.
Is the concept of statistical significance really so difficult to grasp that it eludes the cream of British journalism?