Opinion poll evidence is increasingly used by politicians and interest groups to argue their favoured policies are in line with public opinion – that they ‘get it’ and their opponents don’t. But what do opinion polls really tell us about public attitudes and beliefs? Polling evidence from earlier this week raises some interesting questions.

Institute-of-Economic-Affairs-zombiesPollsters ComRes were commissioned by zombie right wing think-tank the  (IEA) to get views on their latest proposals to ‘cut public expenditure and slash taxes’.

ComRes press released the findings thus:

“Opinion poll shows overwhelming support for new IEA plan to cut public spending and slash taxes…

“Given a straight choice between the coalition’s plans of government spending of 40% of national income and the IEA’s more radical plan of reducing spending to about 30% of GDP and implementing average tax cuts of £7,500 per household, the overwhelming majority (70%) favour the IEA’s proposal and only 30% favour the coalition plan.”

This would have suited the client fine had ComRes not also released the full polling data. The findings are truly amazing, but you’d want to be pretty desperate to lean on them as evidence to support any policy platform.


Here’s the most striking finding (hat-tip to Mark Gettleson at PoliticsHome) – public expenditure for 2010-11 is projected to be 47% of GDP. When asked to estimate what proportion of national income the government currently spends, only 11% of respondents landed in the right range (41% to 50%). See Graph 1 below.

This percentage varies a little by age, social class, region and voting intention but never gets above 16% for any group.

Averaging all the responses gives a figure of 61%, which overestimates government spending by some 14 percentage points. But it gets worse – much worse – because most respondents were nowhere near the average. Some 12% gave answers in the range 91%-100% (this rises to 28% for UKIP voters, by the way, but no party’s voters come out of this exercise looking good).

Graph 1:

Pushing interpretative charity to the limit, we might say that those who estimated public expenditure as between 30% and 60% of GDP gave answers with some relation to reality (on the basis that the figure has in fact ranged between 36% and 50% over the last forty years – see Graph 2 below).

But even this represents only a third of all respondents; the remaining two thirds appear to be living on a different fiscal planet altogether. And the errors are not random. Responses are massively biased towards high estimates, with more than half of respondents appearing to believe that public expenditure is more than 60% of GDP and nearly two fifths believing it is more than 70%.

Graph 2:

In the context of these findings, which are not mentioned in the press release, what can we make of the claim there is overwhelming public support for the IEA’s proposals? The third (of four) questions put to respondents was:

“By the time of the next election, the current government’s plan is to spend about 40% of the UK’s national income each year. What proportion of the nation’s income do you think the government should be aiming to spend?”

The average response was 33%, in line with the IEA’s proposals.

But this result is vitiated by the fact that in asking the question, ComRes made it plain to respondents just how far out their previous answers had been by giving them new information (‘government is planning to spend 40% of GDP’).

Rather than expressing their attitude to public spending, we would suggest that having had the scale of error in their previous estimates demonstrated to them, respondents were looking to give the ‘right’ answer this time around and over-corrected for their previous bias.

In other words, there is reason to believe that that the sequence of questions itself generated the results, which happened to correspond to what the client was looking for.

The ComRes poll can probably be discounted as evidence of support for the IEA’s proposals, but the startling overestimation of current public expenditure is an important finding which raises questions about public beliefs as opposed to public attitudes. Is this evidence that the public is just irrational on the subject of public expenditure?

The idea that voters are basically irrational has recently been in vogue on both left and right. But common sense suggests we should exhaust rational explanations for false beliefs before appealing to irrationality.

Take the finding that 12% of respondents thought that government was spending more than 90% of GDP. That sounds crazy, but a plausible interpretation is that some people thought ‘national income’ meant tax revenue. So even this apparently bizarre response can be interpreted in terms of explicable error rather than irrationality.

More generally, the pattern of overestimation of spending might mean that respondents were not generally answering the same question the pollsters were asking.

The real lesson may just be that it’s very easy to generate gross errors when people are being asked to quantify things which they have no reason to put figures on in their daily lives. The public may just not have any view on the ratio between Total Managed Expenditure and Gross Domestic Product, because most people neither know nor care how these abstract economic and fiscal aggregates are related.

If you ask the wrong question, the answer won’t tell you much. You may get results that appear to support cutting expenditure to 30% of GDP – but if people cared enough about the public spending share of GDP to make that result politically significant, we would expect them to have a better idea of how much the government spends than they apparently do.