ANALYSIS:Unquestioned loyalty to one party is becoming a thing of the past, producing swings and surges that are difficult to register, writes DAMIAN LOSCHER
IT WAS once said of the weather forecast that “it is right too often to ignore and wrong too often to rely on”. Could the same be said of political polls?
The question of accuracy is best addressed in the wake of an election, in this case European and local elections held in June, when poll predictions can be stacked up against actual election results.
Prior to this year's European and local elections, The Irish Timesand TNS mrbi published two polls. The first was in mid-May after the European election candidates had been announced, with a second poll in late May, one week before the elections.
For European election polls, TNS mrbi employs a ballot paper methodology, partly to reflect the importance of candidate profile and personality, and partly because the timing and nature of European election polls facilitate the preparation of mock ballot papers. In the June European election, the ballot paper methodology, combined with a robust 2,000 interviews per poll, proved to be predictive of support at the national level (see Table I).
In European elections, however, most of the attention is focused on voting intentions at the constituency level. Unlike national polls, where constituencies are only pieces of the jigsaw, European polls need to produce reliable measures of candidate popularity for each constituency. To do this, European constituency polls rely on samples of 500 – less than the typical sample of 1,000 used for national polls and, as a consequence, they require more careful analysis before drawing any firm conclusions. European constituency polls can only reliably suggest which seats are safe and which are too close to call.
The 10 candidates that the polls called as “safe” in this year’s elections went on to be elected MEPs. So while on the face of it European election polls appear predictive, you ignore at your peril the Achilles’ heel of European polls – the last-minute explosion or implosion.
In 1999 it was Dana Rosemary Scallon who made a late charge. In 2004 it was the turn of Royston Brady, except his campaign took a dramatic turn for the worse. This year it was Libertas leader Declan Ganley who put his foot on the accelerator and, although he did not win a seat, his late burst nearly upset our prediction that Libertas would not win any seats in Ireland.
Without a doubt, this year’s local elections presented the greatest polling challenge. In the end, local election poll results were shown to be highly predictive for most parties (see Table II), yet were far outside the margin of error for Labour and Fianna Fáil.
Labour, who achieved 14.7 per cent of the vote on election day, were measured at 20.2 per cent in the late May Irish Times/ TNS mrbi poll. Fianna Fáil on the other hand achieved 25.4 per cent of the vote in the local elections despite only registering 18.1 per cent in the late May poll.
The gap between election results and voting intentions for Labour and Fianna Fáil may have a simple explanation. The voting intentions question used in local election polls assumes the ballot paper will contain a credible candidate representing the voter’s party of choice, yet this may not have been the experience of some would-be Labour voters.
As Garret FitzGerald has commented in this paper, many constituencies did not have a Labour candidate on the ballot. In contrast, Fianna Fáil’s local election infrastructure is formidable, and is stocked with candidates that are known in their community and have a track record of delivering for their constituents.
Even if we are satisfied with this explanation for why national polls can struggle to accommodate the very local nature of local elections, there is little that can be done from a polling perspective except to acknowledge these difficulties when interpreting the findings. Unfortunately the use of ballot papers for local elections would not be practical.
Each European and local election poll also asked about party support in the event of a general election, a standard question in all Irish Times/TNS mrbi polls. Our first May poll measured the national Fianna Fáil vote at 21 per cent (after adjustments), falling marginally to 20 per cent in our second May poll – a remarkably low level of support for a party that has been at the forefront of Irish politics and Irish life for the past 83 years.
These results were greeted with some disbelief. How could support for Fianna Fáil have fallen so far? Naturally the poll and, more specifically, the poll adjustments used by TNS mrbi, came under scrutiny.
As readers of The Irish Timesare aware, estimates of party support produced for the newspaper by TNS mrbi are adjusted to more closely align stated voting intentions with actual voting behaviour. For Fianna Fáil, the adjustment is downwards. Whether Fianna Fáil should be adjusted downwards when support for the party has fallen so far was questioned.
Without empirical evidence, as provided by an actual election result, removing the adjustment would be premature. The nearest we have to actual results are the findings from the June 5th exit poll, conducted by Lansdowne Market Research, which asked about voting intentions in a general election.
Party support levels derived from this poll, conducted at polling stations on the day of the election, mirror the adjusted levels produced by the May Irish Times/TNS mrbi polls (see Table III). These findings would appear to support the case for maintaining the adjustment.
A strong theoretical argument for not removing the adjustment can also be made. TNS mrbi research has shown that the Fianna Fáil adjustment is designed to correct for an overstatement of support due to the high level of familiarity and saliency the party enjoys among the public, sometimes called the big brand effect. And while Fianna Fáil’s support has fallen, the party has continued to maintain a very high profile and remains the most widely reported and discussed political party in Ireland.
Returning to the question posed at the beginning of this article, and considering the outcome of the 2009 elections, it seems that polls are predictive within the limits of statistical variance and notwithstanding the potential for voting intentions not to translate into actual voting behaviour as appears to have happened in the local elections. But perhaps the most significant barrier to political polls being consistently reliable is the timing or time limitations of polls.
Political polls are usually conducted no less than seven days before an election and it has been a phenomenon of recent elections that more and more voters are leaving it until the last day, hour or even minute to make up their minds. Unquestioned allegiance to one party is fast becoming a thing of the past, opening the door to late surges and swings that are difficult for polls to register. With voters liable to change their minds as often as the weather, we can safely predict that the shelf life of polls will come into even sharper focus.
Damian Loscher is managing director of TNS mrbi