A new poll by Millward Brown came out this weekend, with headline figures of 26% for Sinn Féin and 22% for Fine Gael. This let many news outlets to conclude that Sinn Féin is now the biggest party in the Republic. But this conclusion cannot be drawn from the Millward Brown poll, based on 991 respondents.
First, the difference between Sinn Féin and Fine Gael falls within the margin of error of the poll. While the reported margin of error is 3.1%, this is the margin for any single party. If we calculate the margin of error for the difference between two parties, we find that it is 4.3%. That is just slightly bigger than the 4% gap. This means that even if there was no difference between the two parties among all likely voters, there is more than a 5% probability that a poll of 991 people* finds a difference between SF and FG of 4% or more. Just because of the people that randomly end up being surveyed. This is generally considered inadequate to base firm conclusions on.
Second, this calculation assumes the poll is a random sample from the population. In reality, it is very difficult to get a random sample, because (among other reasons) it is difficult to draw up a list of voters and to get them to participate in the polls. Therefore, the error margin might in fact be larger than the reported 3.1%. This is sometimes called ‘Pollster Induced Error’. The Irish Polling Indicator gives an (arguably rough) estimate of the Pollster Induced Error for Irish parties, which it estimates to be between 1.3 and 2.7 for Sinn Féin. That means that the error margin is likely to be at least 1.3 times the calculated margin. So, probably even the 4.3% margin of error is an optimistic estimate.
Third, polling estimates are prone to systematic error. This is visible in the fact that some polling companies have structurally higher estimates for Sinn Féin than others. In the graph below, we can see clearly that Millward Brown (green dots) generally has a higher estimate for Sinn Féin than other pollsters do. This effect can be calculated to be just under 1%. It is helpful to take these ‘house effects’ into account when interpreting polling results.
Therefore, one should be cautious to base sweeping statements on a single poll. The Irish Polling Indicator summarises information available from all Irish pollsters. This provides a more nuanced overview of the state of the parties. The Polling Indicator suggests that Fine Gael, Sinn Féin and Independents are on similar levels of support; based on the available data we cannot draw conclusions about who is the biggest of the three.
So, is support for Sinn Féin currently bigger than for Fine Gael? Perhaps, perhaps, perhaps**
* And bear in mind that 24% of those 991 gave a ‘Don’t know’ answer, so the effective sample size is in fact 753, which would yield an error margin of 4.9% for the difference between SF and FG.
** Mind the difference with the title of this blog post: the title reflects the fact that we cannot say that SF is bigger based on the Millward Brown poll, but that does not mean that SF is certainly not bigger. It might be. We don’t know.
11 thoughts on “Is Sinn Féin the biggest party according to new poll? No, no, no”
Have you calculated the effect of weightings for respondent’s recall of vote in last election?
Millward Brown, to my knowledge, does not employ such weighting procedures. That might partly explain why it has a 1% positive house effect for Sinn Fein (compared to, for example, Red C, which does).
polls are always open to a margin of error. Sinn Fein will poll the highest number of votes in 2016
Haver you taken into account that a lot of people that swear they would never vote for SF will vote for SF this time round because two of the most important things in their lives their water and the roof over their heads will be tax free under SF. In truth it does not take much to get people in Ireland to sell their votes but this water charges scam has drawn a truly viceral reaction. I am surprised O’Connor can walk down the street. Then I have been told he uses the back door.
Property tax is a lethal concoction water charges will bring SF to power. Kenny in opposition said “taxing the family home would be immoral….a vampire tax”! Has there ever been a more cynical betrayal?
Pingback: The Squiggle Of Doom | Broadsheet.ie
I am not surprised by this effort to downplay the apparent extent of popular support for SF, but I’m a little surprised that the steady flow of polling evidence which suggests that a majority of voters have rejected the mainstream parties (that traditionally have governed either in combination or in rotation) is not attracting more attention. The extent of this apparent rejection is more extreme in Ireland than in all other EU member-states. And the strength of this rejection was emphasised by the mass protests about water charges on the weekend. Unlike voters in many other EU member-states Irish voters so not do mass protests. (They tend to form a settled view and patiently bide their time until they can cast their judgement at the polling booths at the net time of asking.) But they have been gouged to pay for the banking fiasco. They continue to be gouged by powerful special interests in the sheltered private, semi-state and public sectors – even if they don’t know precisely how or by whom they are being gouged. It appears that water charging is a gouging too far.
Some form of direct water-charging was inevitable (whether the Troika demanded it or not), but the Government would have much preferred to kick this can down the road. But it got too clever by half when it was forced to confront this issue. It crafted a cunning and complex package tht would (1) keep direct charges as low as possible until after the next election, (2) exploit Eurostat rules to create fiscal space for some budgetary gymnastics, (3) compensate Bord Gais for the sale of its non-network activities by establishing Irish Water within it and trasnferring c €12 billion assets to it for free, (4) fund an underemployed army of local authority water sector workers under a 12-year Service Level Agreement and (5) pass the poisoned chalice of setting water charges to the CER.
All of this is now in disarray. The credibility of the CER as an ‘independent’ economic regulator (which was always an optical illusion) has been totally shredded. This will have serious implications for the financing of energy network and infrastructure investment. The Government may retrieve some ground by setting a basic capped charge (as it now appears to be considering) to keep its cunning package broadly intact. But the problem won’t go away and unless it makes a serious effort to tackle the funding and financing of the semi-states and the role and function of economic regulation it is difficult to see the Government surviving until early 2016.
I am not downplaying anything, but merely pointing out that a poll is given an incorrect interpretation. Polls are powerful instruments, but must be interpreted with caution. I could have written exactly the same piece with the positions of FG and SF being reversed.
In fact, the second post about the Irish Polling Indicator focused very much on the rise of Independents, who at that point were the largest group in the polls. In that post I made a very similar remark about the difference between SF and FG:
To be sure: both SF and Independents are making huge gains compared to the last elections. One doesn’t have to draw incorrect conclusions, however, to make that point.
Thank you. I understand your precise and narrow focus.
I will never give these guys a red cent. They are worse than FF and I did not think that was possible. Giving away 12bn of infrastructure to these rogues is treason plain and simple.
RedC Feb 23 Undecided 14, Others 26, FF 15, FG 21, Lab 5, SF 19
MB Feb 15 Undecideds 29, Others 17, FF 14, FG 18, Lab 4, SF 19
The Dates of the polls here are so close together that no actual significant change in what is being measured can have occurred. The votes for parties are within the margin of error but the differences in “others” and “undecideds” are HUGE. The explanation can only be a major difference in the system of measurement. The application of the Likelihood to vote filter cannot be the explanation because it has not yet been applied if your(Liberius, Cedar Lounge) understanding is correct. Recall of past voting cannot provide the explanation either as it has not been applied yet either. The huge variations both between the two Red C polls and between the RED Cs and MB is centred on Others and Undecides Maybe Tom Louwerse can explain it?