This week we saw the first candidate poll for this month’s European Parliament elections. These are polls that present the respondent with a mock-ballot, listing the candidates in their constituency and asking them to indicate how they will vote.

These have been an occasional feature of European Parliament elections since 1989, with the 2014 election campaign seeing four such polls, as well as an exit poll. More are on the way.

The IPSOS MRBI poll for the Irish Times published on 10 May was conducted earlier in the week, and came with two major health warnings. First, it was conducted on only 500 people in each of the three constituencies. This means that, in each constituency, the degree of random error expected in a sample is larger than in a typical national opinion poll.

The estimate given for any party or candidate is being within a band, with the size of the band dependent on the margin of error, as well as on the degree of certainly one is looking for.

The margin of error in this poll was declared as plus/minus 4.4%. This would apply to a candidate getting around 50% of the vote; it would be somewhat smaller for a candidate on 20, at plus/minus 3.5, and smaller again for a candidate estimated at 10%: plus/minus 2.6. In other words, a candidate given 10% in the poll is likely to have support in the range 7.5-12.5, with a 95% level of certainty. We could be 50% certain that support is between 9 and 11. With several candidates often separated by small percentages, this suggests that poll rankings could easily be upset in the real election.

The second health warning concerns the timing. The campaign is only just getting going, and while posters are up and many flyers delivered, most of the main television events have yet to take place, and lots of voters have arguably yet to tune in. This could have the effect of inflating the popularity of the better-known candidates.

Perhaps more importantly, events yet to take place could also sway voters towards or away from a candidate. A post-election poll carried out as part of an academic study of the elections across Europe recorded almost 20% of voters as saying they decided on their vote in the final week. These were more inclined to support independent candidates, but that is not to say a different pattern of change might not be seen this time.

It is worth looking back at the various candidate polls carried out in 2014 to see how effective they were at indicating eventual results, and whether the later polls were more accurate? (Pollsters will say that their estimates are snapshots, not predictions, and of course that is true, but people will base predictions on these polls nonetheless.)

Difference between polls and outcomes for candidates in the 2014 European elections.

Date of polling

Polling company

Dublin

Midlands NW

South

Average difference

26–28 April

Millward Brown

13%

16%

17%

15%

1–2 May

RED C

12%

11%

12%

11%

3–15 May

Behaviour and Attitudes

8%

9%

15%

11%

13–14 May

Millward Brown

11%

6%

12%

10%

23-May

B&A EXIT POLL

4%

8%

6%

6%

Note: Difference is defined as half of the sum of the absolute differences between the poll and the results for each candidate.

There were four candidate polls in 2014. There was also an exit poll. This is included for comparative purposes, and it should be noted that the sample size (3,500) in the exit poll is much larger than in any of the campaign polls. It should be closer to the actual results for that reason, as well as the fact that it records votes and not intentions. The table above shows the difference between each of these and the actual outcome, by constituency and on average.

Difference here is measured by calculating how many voters we would need to reallocate to different candidates to get the same result. For instance, if a poll gave a referendum results as 45% yes and 55% no, and the outcome was in fact the other way around, the distance between the two could be bridged if 10% of voters were moved from yes to no.

The average difference (across all three constituencies) was 15% in the first poll, conducted over three weeks from election day, but 10-11% in the next three. It was just 6% in the exit poll. Arguably opinion was settling by the start of May, but it may be that real change also happened in the final week.

The patterns are a little different across the three constituencies, but in all cases the last two polls were closer to the result than the first two. It also seems that Dublin saw less movement that the other two, with the South seeing most change.

Some of that movement could be within parties, as voters settle on which of their party’s offerings are preferable but in fact, we see little evidence of any trend. Where Fianna Fáil and Fine Gael ran two candidates, polls and outcomes are close in most of the cases with no obvious sign that later ones were more ‘accurate’.

"With more fragmented voting patterns the last seats may well be in doubt until very late on Sunday"

Whatever the differences with respect to vote percentages, the polls did come pretty close to telling us the top candidates in each constituency.

In Dublin, all polls suggested correctly that Hayes (Fine Gael) and Boylan (Sinn Féin) would be elected. Only two of the four polls placed Fitzpatrick (Fianna Fáil) in third place, but the margins were tight and in fact she was not elected.

In the Midlands/North-West constituency, Carthy (Sinn Féin), McGuinness (Fine Gael), Harkin (Ind) and Flanagan (Ind) were the top four in three of the four polls, the earliest one placing the Fianna Fáil candidate ahead of McGuinness.

In the South constituency, all polls put Crowley well in front with Ní Riada (SF) and Kelly (Fine Gael) next, but most saw Clune out-polling Harris in the race for the second Fine Gael seat. She did get the seat, but Harris won more first preference votes.

Evidence from a poll carried out for the Irish Times three weeks before the 2004 election by MRBI also suggests that, although we can expect change between now and election day, most of that movement will have little significance for who wins most first preferences.

That 2004 MRBI poll ‘predicted’ three of the top four candidates in Dublin, two of the three in the East, two of the three in the South and all three in the North-West. Differences between polls and outcomes ranged from 11-16% across the four constituencies, with South at 11 and East at 16, where Mairead McGuinness’s vote seems to have been vastly underestimated in the poll.

Of course, first preferences do not always tell us who will win. In 2014 Childers squeezed out the Fianna Fáil candidate in Dublin and, as said above, Clune beat Harris in South, despite fewer first preferences.

With more fragmented voting patterns the last seats may well be in doubt until very late on Sunday, but these polls are only measuring first preferences, and it is generally the case that the seats go to those with most first preferences.

While the focus in these polls may be on candidates, commentary is sensitive to how the parties are doing.

In this week’s IPSOS MRBI poll for the Irish Times, Fine Gael seemed to be performing very strongly while Sinn Féin looked to be falling behind its 2014 performance and Fianna Fáil candidates were lagging well behind those of Fine Gael.  

Aggregating the results across the three constituencies, and weighting by the number of seats that are supposedly at stake (post Brexit) underlines the very strong performance of Fine Gael and the relatively poor showing of Sinn Féin and Fianna Fáil.

Fine Gael is at 35%, Fianna Fáil at 18 and Sinn Féin at 14. This is a crude way to assess overall performance, as turnout would vary across the three in a general election and seat numbers are a rough approximation to constituency size, but this does underline what a good poll this was for Fine Gael.

It is notable this figure is well above the 29% voting intention for FG in a general election, a difference that again underlines the fact that the results at these European Parliament elections cannot be translated directly into votes at the next Dáil election.