Volatility And Public Opinion

August Political Commentary:

Some commentators have noted in recent months that Ipsos's voting intention figures are "more volatile" than those of the other companies, which in one sense is true; but they have also assumed that this implies they are less accurate, which is not necessarily the case, and some of them have clearly not understood why our figures sometimes move more dramatically than those in other polls.

The difference comes entirely because we report as our "headline" figure only the voting intentions of those who are "absolutely certain to vote", whereas the other companies include the views of those who say they would be less likely to turn out. That makes for a more volatile measure, because it takes less to persuade a voter not to turn out than to switch altogether from one party to another; but it is a real volatility, measuring real fluctuation in public opinion, and since the proportion of a party's supporters that turn out is as much a part of the equation that makes up a final election result as the number of supporters the party has in the first place, we believe it is also a more meaningful measure.

But our samples are no more inherently unstable than those of the other conventional pollsters, as can be seen by comparing the Ipsos results before filtering (which are always published together with the headline figure on our website) for the monthly polls so far published in 2006 with those from Populus and ICM, looking at the poll-to-poll change for each party every month.

Ipsos was somewhat more "volatile" on Conservative share than ICM or Populus, less so than both on Labour and Liberal Democrat share, and overall there is not a hair's breadth between us and Populus, with ICM marginally the most volatile of the three. The method of weighting by reported past vote which ICM and Populus use does not make their samples fundamentally more stable than ours -- and nor, presumably, can it therefore be supposed to make them more accurate. (YouGov's figures move somewhat less from month to month, but their respondents are drawn from an online panel, which raises different questions.)

We filter by certainty of voting because we believe that changes in this factor reflect real volatility in British public opinion, and our job is to report public opinion as it is at the moment. That means our figures are based on the responses of (on average this year) 57% to 58% of the sample; turnout at the last two general elections, we might remember, has been 59% and 61%, with no signs that a dramatic increase is likely next time. Indeed, at last year's general election (when, you will recall, MORI successfully predicted the shares of the three major parties with an average error of two-thirds of a percentage point) far from being too restrictive a filter it proved not restrictive enough.

What this means, though, is that what we are measuring is not exactly the same as what ICM, Populus or YouGov are measuring, so little wonder that the findings are not always in line. Nor, for that matter, are they necessarily measuring exactly the same as each other, either. (And, incidentally, it makes a nonsense of "polls of polls", which in these circumstances are akin to measuring England's sporting performance by averaging the most recent scores of the football, rugby and cricket teams.)

We will continue to provide two sets of figures. The headline results based on those who say they are certain to vote at a General Election; and a set based on no filtering for turnout so as to provide the public, politicians and commentators with powerful long-term trends of voting intentions going back to 1979.