Poll methodology: data no longer adjusted

IN TODAY’S Irish Times /Ipsos MRBI poll, for the first time in more than 10 years, we report only unadjusted party support data…

IN TODAY'S Irish Times/Ipsos MRBI poll, for the first time in more than 10 years, we report only unadjusted party support data. By unadjusted support data we mean exactly how voters told us they would vote when polled in their homes, without making any compensation for the fact that this may be different to how they would actually vote on election day.

Where trend data is shown, like-for-like comparisons are made with unadjusted data from previous polls.

Before explaining why Ipsos MRBI decided to switch off our poll adjustments, it helps if we remind ourselves why the adjustments were first introduced.

In the 1990s, a phenomenon was observed whereby some voters who told us on the doorstep that they intended to vote for Fianna Fáil did not actually vote for that party on the day. At the time Fianna Fáil were by far the most prominent political party and subsequent research indicated the party benefited from a “big brand” effect, which was diluted when voters were asked on election day to choose an actual candidate and not just a party.

READ MORE

The poll adjustments were introduced in the late 1990s, calibrated on actual general election results. At its most basic level, the adjustment systematically downweighted Fianna Fáil party support data, with support for the other parties upweighted as a consequence. For the following 10 years this adjustment proved remarkably effective when put to the test at election time. For example, in the most recent general election in 2007, the final Irish Times/Ipsos MRBI poll put Fianna Fáil at 41 per cent, Fine Gael at 27 per cent and Labour at 10 per cent. This was an exact prediction of the actual result, right up until the last count in Laois-Offaly when Brian Cowen won 19,102 votes and edged Fianna Fáil up to 42 per cent of the national vote. In recent years, however, the political landscape in Ireland has undergone quite a dramatic transformation, prompting Ipsos MRBI to examine our poll adjustments and to assess their relevance to the current political situation.

Our review unearthed, on the one hand, strong evidence to suggest Fianna Fáil are still enjoying a “big brand” effect, albeit to a lesser extent perhaps. This evidence comes from the 2009 local elections, which measured, amongst other things, voting intentions if a general election were called. In response to this question, voters gave Fianna Fáil 21 per cent support, which was two points lower than the Irish Times/Ipsos MRBI pre-election poll measure of 23 per cent – unadjusted, indicating some level of downweighting was still desirable.

But on the other hand, internal and external telephone tracking polls have shown a more shallow decline in the Fianna Fáil vote than that evidenced by the Irish Times/Ipsos MRBI series of face-to-face polls, suggesting a natural form of realignment has occurred over the past three years, in which case there may be no need for an artificial adjustment.

There are also qualitative reasons for questioning how likely we are to see a Fianna Fáil “big brand” effect at the next election. Bearing in mind how the landscape has shifted, it may now be reasonable to argue that there could be a “big candidate” effect instead, if disgruntled voters nonetheless choose a Fianna Fáil candidate on the day because of name recognition and general familiarity.

Following our review, Ipsos MRBI concluded the available evidence does not conclusively support the continued use of the current adjustments. It was our ambition to build a dynamic model, one which would draw upon the current adjustments should Fianna Fáil support return to historic levels, but would have the flexibility to accommodate existing and future party support scenarios.

To build such a model, empirical measures of party support (such as those provided by a general election result) are needed at least, at two points in time. The 2007 general election result provides a first measure, but it is difficult to find a second, more recent measure.

We had hoped the 2009 local election exit poll data would provide that second measure, but as robust as these poll findings are, they are not fully empirical (relying on “claimed” voting intentions). Also, the political landscape has shifted further in the interim, which raises questions about the stability of any model built on pre-2010 data. Ultimately, we concluded a dynamic model would not be reliable, and perhaps not relevant at current party support levels.

Irish Times/Ipsos MRBI poll figures are now shown unadjusted.