The Irish Times/Ipsos MRBI poll published last Tuesday attracted much attention, mostly for its verdict on the Labour Party. At 6 per cent, the poll made tangible the cost to Labour of being the meat in our austerity sandwich.
To Labour’s credit, its response was philosophical. In an interview on RTÉ television, Eamon Gilmore chose to focus on the job that Labour was doing in Government and acknowledged the inevitability of paying a price at the polls for taking tough decisions.
Elsewhere the poll was greeted with some surprise, even shock, as if the result had come as a bolt out of the blue. In reality, recent polling data, from Ipsos MRBI and other polling companies, had been signalling significant voter dissatisfaction with Labour.
In The Irish Times/Ipsos MRBI poll in June, Labour dropped to single digits, having registered a decline in nine of the previous 12 polls. In purely statistical terms, a further descent was more of a probability than a possibility. Polls during the summer showing Labour at 7 per cent (Behaviour & Attitudes) and 8 per cent (Millward Brown) were further ominous signs.
Perhaps the surprise was not that Labour had declined but that a rating of 6 per cent was strikingly low, bringing the party, symbolically and otherwise, back to the 1980s. Certainly low enough to put the Labour leader under pressure.
Methodology
Low enough also for the Ipsos MRBI methodology to be called into question. Writing in the online Daily Business Post, Pat Leahy suggested a Labour vote of 6 per cent might under-represent their true vote, because Ipsos MRBI takes no account of likelihood to vote, and Labour supporters are more likely to turn out to vote.
Over the years Ipsos MRBI’s polling approach has been viewed with some suspicion, despite – as Pat Leahy acknowledged – a track record in delivering accurate polls. This suspicion perhaps stems from a lack of understanding of the approach we adopt. Ironically our methodology could not be more straightforward and transparent.
To say that we do not take into account likelihood to vote is incorrect. We just do it in a different way. To filter out voters who may not vote, Ipsos MRBI excludes those who tell us they do not know how they will vote or plan not to vote. And all the evidence says it has worked remarkably well, delivering an average party deviation of just 0.6 per cent for the most recent general election. Indeed, the Irish Times/Ipsos MRBI 2011 pre-election poll put Labour on 19 per cent, the same level of support they achieved on election day.
Different approaches
There are other approaches to filtering, such as asking a likelihood to vote rating-scale question and removing those who indicate a lower inclination to vote. The argument in favour of this approach is that some of the undecideds may end up voting, while others who nominated a party will not show up on the day, and a likelihood to vote question can provide an extra layer of voter screening. As an approach, it too has proved effective.
Since the beginning of 2013 Ipsos MRBI has been asking a likelihood to vote question on all polls, to better understand how it interacts with other questions and how consistent are the findings we collect. We are constantly experimenting with new techniques and in due course will deliver our verdict on how, if at all, this question can enhance our polling.
In the meantime, in the interests of putting all conspiracy theories to bed, we can confirm that Labour would have registered 6 per cent in the poll, whether or not we incorporated our likelihood to vote question.
Damian Loscher is managing director of Ipsos MRBI