Lies, damned lies… let’s talk about stats

‘Lies, damned lies and statistics’ may be a cliché, but it highlights a cynicism about how statistics can be manipulated to mislead or exaggerate. And in today’s climate of mistrust in institutions, ‘official’ truth and experts, are statistics fit for purpose?

Shockingly, the Office for National Statistics (ONS), once a trusted institution, is no longer believed to be producing reliable data by a majority of the public (according to ‘some’ statistical evidence).  While ONS data is still used by the government and the Bank of England, popular confidence is falling.

For many, the accelerated use of statistics by government during the Covid period created a mood of scepticism about the reliability of the figures. National briefings – quoting stats derived from forecasts – were often used to hector the public with figures to legitimise government policies and to persuade us to comply with harsh restrictions on liberties, often with little relation between the evidence and the policies in question.

‘Facts and figures’ have become increasingly relied upon by politicians to depoliticise debates around everything from economic growth to healthcare. Crime is falling, we are told, so why is everyone complaining about Broken Britain? How can you argue with the facts and figures? With politicians bamboozling us with percentage rises and falls to prove their success or highlight their opponents’ failures, statistics themselves seem have become increasingly politicised.

On contentious issues – like climate change or the trans debate – both survey data and scientific studies can give wildly varying results.  Methodology is often the root of this, due to factors like selection bias, badly worded survey questions or even an inability to get enough respondents. But why does this matter? Good statistical practice is honest about its methodology – how you define a category, how you word a question, who responds, what’s excluded – and then adjusting when the methodology is shown to be flawed. But when it comes to highly charged ideological issues, such good practice seems to have become strained.

One ONS scandal in 2021 showed a trend away from this. ONS guidance told respondents they could answer the Census’s binary ‘sex’ question according to their self-identified gender rather than their biological sex, without outlining or understanding that this changed what the variable meant. In 2023, migration statistics in the UK became so politically charged that the ONS suspended their reporting, in order to change their methodology.

And then there’s the plethora of statistics created by ever more polling organisations treated as ‘truth’ by commentators and as the basis of countless news stories. But how reliable are these surveys? The Quiet Revival report by YouGov in 2025 claimed that there had been a growth in Christianity across Gen Z, leading to headlines across mainstream media outlets that young people were flocking back to church. However, there was very little evidence for this: the data was misleading. YouGov later retracted its report as a large amount of the responses to the survey were fraudulent.

But isn’t a dismissal of statistics the road to ruin and misinformation? Is it healthy that when reputable researchers produce reliable data, the public often refuse to believe the numbers? While it may be fair enough to note that statistics don’t show the whole picture, and that cherry-picked stats can be used as  ‘evidence’ for a one-sided argument, how can we debate the challenges we face if there is no way to measure how big these issues really are? If the numbers are suspected as being shaped by political bias, how can society believe in any objective truths?

Are stats and data always provisional as human creations, with categories drawn up within social and historical contexts? Or should we focus on depoliticising an important resource for society?