By Roger Mortimore
For more than a decade now, whenever Ipsos Mori conducts a poll of voting intentions, we published two sets of figures: which party everybody who expresses an opinion says they would support if there were a general election tomorrow, and which parties those people who are certain they would vote would support.
Once, when election turnouts were much higher, we could be pretty sure that almost everybody who knew which party they preferred would get to the polling station to vote for them, and the figures based on everybody's answers were an accurate political indicator. Today that is no longer the case, and we use the 'certain to vote' measurement as our headline statement of voting intentions.
Over the years, one effect of relying on the 'certain to vote' figures has usually been that our polls have shown Conservative support at a higher level, and Labour support at a lower level, than would have been the case if we had used the unfiltered figures. Nothing surprising in that – older and more middle-class people are more likely to vote these days and are also more likely to be Conservatives rather than Labour supporters.
The proportion of each party's supporters who are certain to vote goes up and down from one poll to another and tends to rise for both parties as an election approaches, but, despite the fluctuation, up to 2010 the Conservative advantage never entirely disappeared.
However, as the chart shows, things have changed dramatically since the general election. Labour supporters have tended to be much more sure they would vote than was the case when the party was still in office, and for the last year-and-a-half or so the Conservatives have also been noticeably less certain of voting than they used to be. As a result, the gap between the two lines has narrowed substantially, and there have even been a few of our monthly polls when those who said they would vote Labour have been more likely than Conservatives to say they are certain they would vote.
There are various possible explanations for this but by far the likeliest is that it is simply a reflection of the difference between being in government and being in opposition. Nobody should ever believe that being in government is easy. Putting policies into practice will inevitably alienate or disappoint some supporters, even if not to the extent of deciding to vote for a different party altogether; meanwhile the opposition may get a free ride, able to rally their own side to an enthusiasm for getting the current government out without (yet) having to prove in practice that their own stewardship would be better.
When the Conservatives were able to combine the advantages of opposition with their own natural advantage of drawing their core support from the heavily-voting groups of the public, they had a big lead over Labour in their ability to turn their raw support into votes; now, with Labour reaping the opposition benefits, the two forces work in opposite directions and things are more finely balanced.
What the change means is that filtering the voting intentions by certainty of voting will have less of a pro-Conservative effect on our headline figures, and sometimes (when the red line in the chart edges above the blue one) may even favour Labour instead. That was the case this month, when the certainty of voting correction turned a five-point Labour lead into a nine-point Labour lead.
One watcher of our polls noticed this movement in our January findings, and tweeted to ask whether we thought this was a trend or just random variation. In fact, neither may be quite the right description. Certainly, as the chart makes clear, there has already been a long-term change, from a wide gap to near parity, but it is too early to tell if the most recent shift will prove to be permanent or only a blip – we have seen this sort of movement before, with the Labour line edging above the Tory one, but on each occasion after a couple of months it has slipped back again.
But that does not mean that the changes are simply random variation. The lines do move up and down a lot, but this seems to reflect something more concrete than purely fluctuations within the survey margin of error. As can be seen on closer examination of the chart, when Labour's vote certainty has moved ahead of the Tory score, it has tended to be for more than one month at a time, suggesting that the figures reflect real movements in public opinion, if small and short-lived.
Of course, these movements in themselves may be effectively random, driven by trivialities rather than by any real appraisal of the political situation. But if on general election day Conservatives were less likely to vote than Labour supporters, that would damage the Conservative party's electoral prospects regardless of why it had happened or whether the same would have been true a week earlier or a week later. Turnout matters.
It is because we are convinced that these variations in certainty of voting are real, and that our respondents' answers genuinely reflect in some degree their likely behaviour if there were an election tomorrow, that we make so much allowance for them in the calculation of our headline voting intention measure.
The thing that determines election outcomes is not how many people support a party but how many vote for it, so likelihood of voting is just as much a part of what we are trying to measure as party loyalty. In fact, modern elections are often as much about turning out the vote as about winning over floating voters - maybe more so.
It is true that, because we are including likelihood of voting in our voting intention calculations and other polling companies are not (or are giving it less weight), this can make our measurements more volatile. But this is because our method makes our measurements more sensitive, picking up the reactions that are strong enough to weaken or strengthen the enthusiasm of one side's supporters but not strong enough to swing their loyalties altogether.
We don't apologise for that: it's something that is real and genuinely a factor that affects voting behaviour and the relationship between actual votes and polling numbers. (But, of course, we will carry on publishing all the figures so readers of our polls can see exactly how much difference the turnout factor is making and what the trends are in underlying party support, if they prefer to take that as a political indicator.)
However, concentrating on a single month's figures and worrying whether the Conservatives or Labour will gain a slight advantage from the turnout filter in the next poll is probably missing the point. What is more significant is that there has already been a huge change in the long-term equilibrium.
Unless the revitalised enthusiasm of Labour's supporters proves to be all talk and no substance, or unless the Conservatives can find some way to reverse it, this change will make the Conservatives' task at the next election even harder than it looks.
Roger Mortimore is a professor at King’s College London and a research director at Ipsos Mori.
The opinions in Politics.co.uk's Comment and Analysis section are those of the author and are no reflection of the views of the website or its owners.