The House Opinion Article | The Professor Will See You Now: dunno

4 min read1 hr

The first (from Punch) showed a schoolteacher addressing his pupils: “Some of you students have urged me to teach that bourgeois society is corrupt, so here goes. Bourgeois society is corrupt. Returning now to the question of congruent triangles…” That one comes back to me occasionally in seminars: “Anyway, let’s get back to the subject of Early Day Motions…”

The second (by the great Tony Husband, maybe?) featured a man being questioned by a clipboard-armed pollster. “I am,” says the man, “less a don’t know, and more a couldn’t give a toss.”

That one came back to me recently while reading a fascinating new project examining the ‘Don’t Knows’. The chaff of opinion poll responses, most public polls simply discard them – along with the ‘Won’t Says’ and the ‘Couldn’t Give A Monkey’s’ – and report findings based on those who cough up a response. Yet those who don’t answer can often be considerable in number and they are not random.

It has, for example, long been known that the Don’t Knows are much more likely to be female. This new research shows just how much. In an impressive piece of work (which joins the growing list of projects I have often thought of doing, never got around to, and which are now, thankfully, being done by people much more able), researchers analysed every single question asked by the British Election Study (BES) over the last 10 years. That’s more than 2,000 questions, asked of almost 120,000 unique respondents.

In all 29 waves of the BES, women were more likely to say dunno, at roughly twice the rate of men, and around three quarters of those with a high proportion of don’t know responses were female. There were other differences too – education increases the likelihood of offering an opinion, for example – but sex appears to be the most significant factor.

Preliminary results appear to show this effect varied by both focus and format of questions. It was most pronounced when asking about people’s knowledge and/or about European politics. But although its scale varied, the effect remained, regardless of the topic, type or format being examined.

As so often with these sorts of findings, it’s important to remember the differences are probabilistic and at the margins. Women answer plenty of questions in surveys; plenty of men frequently say they don’t know. But one group is clearly more likely to do it than the other.

So, when we casually drop the Don’t Knows from a survey result, we are disproportionately dropping women. That might matter less if they genuinely don’t know – but, given that the project also found significant differences based on question formatting and wording, some of these differences aren’t genuine. Plus, there is almost no gender gap in eventual electoral turnout, so we are almost certainly dropping people who are still participating. 

Some of these gaps are already known to be caused by men’s tendency to give answers based on less certainty – and sometimes just to guess. Several years ago, there was an experiment in which respondents were set a series of unanswerable political knowledge questions – in that every one of the proffered responses was false.

Who said: “We shall fight them on the beaches?” A) John Lennon B) Boudica C) Rastamouse D) Don’t Know. (The real ones were more subtle than that, but you get the idea). The good news is that most people responded by picking option D. The less good news is that men were much more likely to give an answer regardless. Men were basically more likely to think they were right, even when they had to be wrong. Women never seem to be surprised by this finding.