Opinion poll cannot be absolute forecast of results of election

When I first saw the findings of the Irish Times/MRBI opinion poll of June 2nd/3rd, I was conscious of three factors, which had the capacity to influence the outcome of the European election.

The first was that there were still nine days to go, and a lot could happen in that time. I would be surprised if some candidates, on the basis of an unfavourable position in the poll, did not make a determined effort during the final week to improve his or her impact. In fact, I have observed this happening in at least one instance.

On the other hand, the opinion poll also showed a number of candidates sitting comfortably, and it is also possible that some, in this situation, assumed that the day was already won. In any event, I do not accept that every candidate, on reading the poll findings, sat back and accepted the figures as a foregone conclusion. It is a logical reaction that those unfavourably positioned would endeavour to improve the situation over the final days.

With so much activity taking place, as election day approached, movements and changes in voting intentions were the order of the day, particularly in such a personalised environment as a European election.

The second factor is not unrelated to this point. When assessing the MRBI figures, I had knowledge of the recent movements and trends relating to all candidates in the Dublin constituency, from private research conducted by the company. These indicated a measurable swing over the previous two weeks in favour of two particular candidates, and away from two others. For professional and ethical reasons, I did not comment on this, but I was very much aware that movements in ongoing voting intentions were taking place.

It can be also assumed that similar movements and swings arose during the final weeks of the campaign in the three other constituencies.

To assume therefore - as some commentators have done - that the findings of an opinion poll conducted nine days before the election represented an absolute forecast of the actual outcome is neither realistic nor valid, and implies that opinions and intentions remained unchanged during the final weeks.

The third factor was the anticipated low turnout. As it happened, these anticipations were correct, and the 51 per cent figure represents a market researcher's nightmare, when one considers that while 1,600 electors participated in the survey, it appears that only half actually turned out to vote.

Before addressing the findings further, relative to the election outcome, a redefinition of the parameters is appropriate. The statistical variation of national figures in the opinion poll is calculated at approximately plus or minus 2.5 per cent, and the corresponding variations for those for each constituency is plus or minus approximately 5 per cent.

My primary post-election interpretation is that there was a swing away from Fianna Fail across all four European constituencies during the final week or so of the campaigns. The table above of cumulative candidates cum party figures confirms this.

Excluding Fianna Fail - which is discussed below - no opinion poll figure for any party (except for the two crucial Independent candidates in Connacht/Ulster) is outside the statistical variation of the samples. In fact, in most instances - particularly in regard to Fine Gael and Labour - the survey and election figures are remarkably close. A partial clue to this Fianna Fail performance can be identified if we refer to previous election and campaign opinion polls.

In a number of elections in the 1980s, opinion polls charted a downturn in Fianna Fail support as soon as the election campaign commenced. In the 1990s the downturn in poll figures was not as dramatic, but the actual outcome was much more startling: 39 per cent in the two most recent elections.

The phenomenon is again evident in the European election, and one hypothesis is that the downturn in Fianna Fail support is in some way relative to the low turnout. Understanding this scenario represents a challenge not only to politicians, but also to opinion polling companies. It needs further research.

Taking the interpretation a stage further down to candidate level, it is clear that support figures for a relatively large number of candidates were stable and consistent between the Irish Times/MRBI poll and the election.

These included Mary Banotti, Avril Doyle, Alan Gillis, John Cushnahan, Jim Corr and Joe McCartin (all Fine Gael); Brian Crowley and Ben Briscoe (Fianna Fail); Proinsias De Rossa, Bernie Malone, Sean Butler, Paula Desmond, Gerard Gibbons (Labour) and Pat Cox (Independent) - a mix of 14 candidates, of whom seven were successful.

The Fianna Fail candidates whose election levels reflected a late drop in support compared to the MRBI figures were: Niall Andrews, Liam Hyland, Jim Fitzsimons, Gerard Collins, Pat the Cope Gallagher, and Noel Treacy. Against this background the interpretative commentary relative to probable successful candidates was very accurate, although the sequence of election was not spot-on in all instances.

In Dublin, the MRBI indications were Niall Andrews, Mary Banotti and Proinsias De Rossa, and the final seat was seen to be between Patricia McKenna and Ben Briscoe. The actual outcome was: Banotti, Andrews, McKenna and De Rossa. Against a turnout of 36 per cent, the survey indications were remarkably accurate.

In Leinster, MRBI showed Liam Hyland, Avril Doyle, Jim Fitzsimons and either Alan Gillis or Nuala Ahern as likely to take seats. The outcome was Doyle, Fitzsimons, Hyland and Ahern.

In Munster, the opinion poll had Brian Crowley in the low 30s topping the poll, followed by Gerard Collins, Pat Cox, and the final seat between John Cushnahan and Jim Corr - with the odds favouring the former. This is precisely what happened.

In Connacht/Ulster the survey figures saw Fianna Fail retaining two seats and Joe McCartin holding the Fine Gael seat. The opinion poll also identified the fact that both Independent candidates were likely to poll well, and had almost a quota between them - nine days before the election. There was already some evidence therefore of the spark with which both eventually set the heather blazing in the west. The result is now history and it is clear that both of these candidates campaigned very successfully right up to polling day, and maximised their respective support levels to the utmost.

The primary overview comment is that an opinion poll conducted nine days out from an election cannot conceivably identify movements that materialise during the final week of the campaign - i.e., after the survey was conducted.

A comparison with the tally-based predictions is interesting. Until the actual result was known, all commentators maintained over the weekend that, in party terms, there would be no change in Connacht/ Ulster - and this was the situation up to two days after the election.

A final comment. Sinn Fein was the only party in the election which increased its vote in all four constituencies relative to 1994, and when the survey was undertaken this was not evident. National opinion polls here have shown support as high as 4 per cent on occasions, and in the 1997 election the party obtained 3 per cent. Traditionally, in Northern Ireland, however, opinion polls have tended to understate support for the party relative to election figures.

Jack Jones is chairman, MRBI Ltd.