Numbers Game #3: Oh No, Not Me

On Tuesday, 23rd September I attended a presentation by Prof. John Curtice, the doyen of TV pollsterism, on the subject of “How Scotland’s Citizens Voted”.[1]

Warning us that everything was so hot off the press that there were bound to be typos and glitches, Prof. Curtice took us through his analysis of the referendum result. The data used included not just the details from the breakdown of the votes as actually counted, but also information from the various polls published before the referendum date and from a few taken afterwards.

Much of what was covered has since been commented on elsewhere, and is pretty obvious stuff in any case. I won’t dwell on those aspects here, therefore, but instead concentrate on the things that stood out for me and use them as the launch pad for some of my own thoughts on the topic.[2]


The now infamous Ashcroft post-referendum poll results seemed to be saying that the 16/17 year-olds voted overwhelmingly for Yes, there had been a curious dip backwards for 18-24 and then a rise until age 54 at which point support slipped back again, becoming overwhelmingly No for age 65 and over. The headline was that the wrinklies had sold out the youngsters, but the hope might be that time would eventually overturn this position.

The sharper-eyed had noticed, however, that the widely circulated initial presentation of the results was misleading. For one thing, the age bands were not all the same size. The category of 65+ does of course cover a significant section of the population, spanning a very wide age range indeed. The age 16/17 category is exactly the opposite, covering only two years of age. Yet, they had been given identical prominence in the charts because the column width used was constant throughout.

I notice that the one-page summary currently available from Ashcroft shows the first age band as 16-24, which corrects things to a degree by combining those youngest age groups as first reported on. That new age band shows only a slight majority for Yes of 51%.

Even more of an issue, however, is the low sample numbers for these youngest age groups. For the 16/17 age band only 14 people had been interviewed and for 18-24 the number was 84. Even as a combined age band of 16-24, the total of 98 is only 24% of the number of people in the 35-44 age grouping and 20% of the number in the 65+ category. The reason that matters is to do with the margin of error that is created as a result; based on the sample sizes, the reliability of the findings seems to be far lower for those youngest ages than for the other, older ones.[3]

The standard error rate quoted for such surveys, which they are purposely designed to fit within, is 3% +/-. On that basis, with one exception, every age group either definitely voted No or shows a Yes lead which falls within the margin of error of 3% and must be treated, therefore, with caution.

The only age band that we can reliably say voted for Yes in a majority was the single one of 25-34.

Change in intention during the campaign

Curtice’s graphs told the by now familiar story of a very large advantage for No at the start of the campaign, followed by a progressively narrowing lead, then a flip to give a very small margin for Yes with a few weeks to go and finally a late surge for No in the last week or so, ending up with the actual result of a 10-point win.

Although it wasn’t covered in Curtice’s presentation, there’s a very important feature of Ashcroft’s findings which should be discussed here. What one of his tables says is that 62% of those voting No had always had that intention. As regards the other Nos, 28% of them had decided to vote that way during this year and 19% within the last month before the referendum.

On the other hand, 52% of the Yes voters had made up their mind during 2014 and 39% during the final month. If nothing else, that has to say something about the effectiveness of the Yes campaign overall: most of those saying Yes decided to do so during 2014.[4]

The Yes campaign lost, though. In particular, there was a softening of support during the final week which may well have turned out to be critical. As Lesley Riddoch has suggested, the key learning has to be that you must keep up the pressure right till the very end: play for the full ninety minutes.

My own worry throughout the final months was the persistent size of the Don’t Knows. I couldn’t help feeling that anyone still reporting being uncertain then was probably really a No. Curtice appeared to say that most of the DK group did indeed vote No when it came to it, but I may not have picked up his complete point about that.

Reasons for voting No

One of the appeals made was to vote Yes because of the opportunity to create a more equal society via an independent Scotland. Sadly, this aspect of things never did feature in the principal reasons given for voting intention, by either side’s supporters. The economy is what dominated throughout, if the polls are to be believed.

A phenomenon much commented on recently is that of the “Shy No”. This refers to those who intend to vote No, but decline to reveal that when asked. It’s something that happens with all surveys: for whatever reason, the person has a reticence about going public on the matter (even if the poll is guaranteed to be confidential). It may be that they don’t want to voice what they think will be an unpopular view (e.g. a racist one), that they would feel embarrassed about opting for narrow self-interest over the common good, that they fear retribution, or any combination of these.

A good example is what happened to Neil Kinnock. In the run-up to the General Election of 1992, the polls suggested that he was within an inch of leading the Labour Party to victory and the exit poll on the day indicated that he had just managed it. Instead, he lost to the Tories under John Major, who recorded the highest ever share of the popular vote. The problem seems to have been that there was a significant reluctance to admit to intending to vote Tory; many people simply don’t like to report having an unpopular opinion.

It’s clear, I think, that many of those who were Don’t Know until late in the day regarding the 2014 referendum fell into this Shy No camp. I’d go further, however, by suggesting that there’s another important category: the Dissembling No.

The Ashcroft post-referendum poll asks a sample of No respondents for their main reason for voting that way, which it does by inviting them to choose the most important to them of three candidates. The results are as follows:

  • 47% – Unacceptable risk associated with economy, currency, EU, jobs & prices
  • 27% – Strong attachment to UK’s shared history, culture & traditions
  • 25% – ” No” still means additional devolution of powers (“best of both worlds”)

Now, there’s a difficulty in interpreting these results: the respondents have to choose from the limited set of ready-made reasons rather than being able to volunteer their own ones. There will be other reasons not included here, which means that the percentages above are inflated as a result. It wasn’t possible to pick “none of these”, for example.

As I alluded to earlier, another Ashcroft table tells us that 19% of Nos made up their minds within the final month of campaigning and 28% during the rest of this year. The impact of the margin of error complicates things, but these two sets of results taken together seem to be suggesting either

  • that effectively all of those who moved to No during the final period did so because of the promise of more powers, or
  • that, as the final movement therefore included some folk swayed by economic or Britishness arguments, some of those who voted for the “best of both worlds” had decided to do so before the start of 2014 (which begs the question: how did they know at that point that there would be more powers promised later on?).

I don’t buy either of those explanations, because I don’t believe in taking these responses at face value. I didn’t accept the argument that the 1979 Devolution vote was significantly affected by the intervention by Lord Home (who was already by then a superannuated old buffer most folk had long since forgotten about) and I’m wary of attaching much weight to the equivalent argument being made now.

What I mean by “Dissembling No” is the person who was never going to vote any other way, irrespective of any campaign message in either direction, but can’t bring themselves to say that. Instead, they rationalise things by latching on to a “reason”. Give them a choice of three possible reasons and they’ll pick one. Give them a choice of a different set and they’ll still pick one of them. You would be able to match the current justification to the headlines in the papers that week: currency then the NHS then supermarket prices and so on.

After all, 62% of Nos said that they had always known how they would vote, which does suggest that the 27% for those who claimed to have voted No because of an innate sense of Britishness is too low a figure. Another 10% claimed not always to have been Nos, but to have made up their minds more than a year ago, which was well before Osborne told us we were not going to get to share the Pound Sterling. It’s all complete tosh. I’d like to have seen Ashcroft including the question: do you just feel comfortable as you are and don’t like the idea of change at all?

Of course, you can make the same kind of observations about those who voted Yes.[5] The difference is that only 38% were dyed-in-the-wool Yes supporters, with another 10% being converts of more than a year’s standing. A slight majority claims to have moved to Yes within this year, which is very interesting.

As a final observation, let me refer to the repeated assertion that “people need answers to the questions, Alex”. Well, if 72% had made up their minds already, Alex really needn’t have bothered his shirt at all.

Party loyalty

The result that’s perhaps the most surprising at first is that a significant number of those who have voted SNP before were Nos in the referendum: 14% of SNP voters in the 2010 UK General Election and fully 20% of those in the 2011 Holyrood contest.

Many of those in Dundee East, Angus and Aberdeenshire who have previously voted SNP did not support the party in this referendum.

I don’t think there’s too much mystery here. The SNP’s electoral success in recent years has been built on attracting voters from other parties in an effort to squeeze out Labour. Several of its seats were previously Tory or Lib Dem. Many of these supporters were never natural nationalists at heart.

Other observations

Prof Curtice was at pains to point out that, right from the beginning of the campaign, those who favoured Devo Max (however defined) were always going to vote No and in fact ended up doing exactly that. The notion that the removal of the third question favoured No seems to have been borne out in practice, therefore.[6]

There was one observation made by Curtice which did surprise me greatly. As many have commented on, there’s a clear positive correlation overall between deprivation and voting Yes, and likewise between economic well-being and voting No. Irrespective of that overall result, what seems to be the case is that, if the district as a whole is comfortably off, then those within it who are at the bottom of the ladder did not vote Yes in great number. The reverse is also true: if the district as a whole is dominated by areas of high deprivation, then the more privileged people within that district tended to vote Yes along with their less fortunate neighbours.

Could that be the underlying reason for the difference between Edinburgh’s 61% for No and the 51% for Greater Glasgow?[7] Perhaps the Weegie middle classes are more aware of, and affected by, the deprivation in their midst.




1. One of the The University of Glasgow’s Stevenson Lectures in Citizenship.

2. The extended commentary is mine. Prof Curtice is not to blame.

3. To assess the difference in error rate would require looking at the breakdown by these age bands of the total population registered to vote. There are of course far fewer 18-24 year-olds than there are those aged 65+, which means you do not need as great a sample size; a smaller one will do, but the question concerns how much smaller.

I’m struggling to believe, however, that the differences in total population between these age bands are anywhere near justifying the differences in the sample sizes here.

4. To be rigorous about things, it would better to talk about relative effectiveness. Some of those voting Yes will have done so because Better Together pissed them off.

5. You might reasonably want to ask Yes supporters this: do you feel so disenchanted with, and detached from, the UK State and its economy that you would vote for any change, of whatever kind?

6. Of course, that’s to ignore the argument that opting for No has since mutated into Devo Max in any case. I’ve made the point previously here.

7. “Greater Glasgow” is the notional aggregation of Glasgow City, East Renfrewshire and East Dunbartonshire. See my earlier post for more on this.

Leave a Reply

Your email address will not be published. Required fields are marked *