As the results rolled in on Saturday morning, it was apparent within five or six minutes of boxes opening that the Government had suffered a crushing defeat – in the case of the referendum on “care” or “mothers in the home”, a defeat which we can say is literally without precedent. No Government has ever had a referendum rejected by a greater margin.
One of the most frequent questions directed to me, and to other journalists, on social media and in copious emails, was how on earth had the opinion pollsters gotten it so wrong?
No opinion poll published in the campaign gave the “no” side a lead. Indeed, many such polls, including the Red C poll two weeks ago in the Sunday Business Post, had the “yes” vote at the commanding heights of 60% or so, data which led this reporter to write a piece suggesting that the referendums would be very hard to defeat indeed.
They were not alone. The Irish Times’ final data on the referendum, published on February 9th, almost a month before polling day, showed “a clear majority” in favour of yes, with 53% supporting the family referendum and a whopping 60% supporting the deletion of mothers.
On the face of it then, the pollsters got it badly wrong. In reality, they should be given something of a pass for a few reasons.
First, referenda in Ireland are notoriously hard to poll for the simple reason that voters have a long and established habit of telling pollsters that they are undecided when in truth they are likely to vote no. For example, the final Sunday Independent Ireland Thinks poll, published last weekend, was clearly and obviously horrendously bad news for the “yes” side, even though it showed them in the lead. As I wrote just one week ago today:
Last week, the Sunday Business Post/Red C poll found, with two weeks to go, the “yes” vote at 52% in the “durable relationships” referendum, and 55% in the “women in the home” referendum. Yesterday, by contrast, the Sunday Independent/Ireland Thinks poll found those figures at 42% and 39%, respectively. In both cases, if you allocate 80% of the undecided voters to the “no” side, then the “no” side would win.
By way of illustration, in the last “major” referendum defeat for the Government, the first Lisbon Treaty vote, the final polls from the Irish Times and the Sunday Business Post had the yes side on 35% and 43% respectively. That’s the territory we’re suddenly in, if you’re a tea-leaf reader. This poll is the first published during the campaign that is consistent with a “no” side win on Friday.
Without wishing to sound elitist, the data in the final poll was very consistent with the result we saw on Saturday: The “yes” vote falling, and “undecided” increasing.
The other factor here is that Irish referenda generally – not always, but generally – take shape in the final week. Most people who are politically motivated know how they are voting, sometimes months in advance. The average person, however, will often only really begin to engage with a campaign in the final few days. Again, as I wrote in that same piece last week:
Since the last Sunday Independent/Ireland Thinks poll on this question, the “Yes” vote has fallen by five per cent in the “durable relationships” vote and ten per cent in the “women in the home” vote. This suggests that the more voters hear about the proposed changes, the less sure about them they are.
This evidence was there in the polls: As we went into the final week, it was clear that the more people heard about the proposals, the less sure they were about them.
In defence of the earlier polls, people have an understandable tendency to look at the headline data more than they look at what’s driving those figures. For example, in the Irish Times poll on February 9th showing 60% yes, if you looked under the hood you would have found that only 8% of the electorate said they knew “a lot” about what they were being asked to vote on, and 53% – more than half – admitted to “knowing barely anything at all”. That suggested that the “yes” vote lead was very soft, and could be changed. It’s also consistent with what I wrote last week: the more people learned, the less they liked. And they tuned in late.
No opinion pollster wants to put out incorrect information – that would be an act of gross irrationality. Their entire business model, after all, rests on the idea that companies can hire them to get an accurate sense of public opinion. What the polling companies get out of election and referendum polling is, largely, free marketing: If the result matches their prediction, that’s a good sign that they’re good at their jobs.
Getting it wrong, though, is not always a sign that they’re bad at their jobs.
What it does show, though, is that you should never get too up or too down about any opinion poll result. Public opinion can swing sharply, and especially so when a particular idea catches the public mood. It’s not just in referendums, either: Sean Gallagher infamously saw a strong lead in the 2011 Presidential election evaporate in a week after RTE’s botching of the final debate in that campaign. Polls dramatically over-estimated the “green surge” in Ireland in the last local elections in 2019. The final polling in the abortion referendum was wrong in 2018, though in the opposite direction, as it drastically under-estimated the scale of the “yes” vote.
In the context of fast-moving public opinion, pollsters have a very hard job. The lesson of this result is not to discount what they say – the data is always valuable. The lesson is to look underneath the headline result, and see if there are factors driving that result which can be changed. At the end of the day, Gript readers who read my piece last Monday will have seen that the polls showed a “yes” lead but great news for the “no” side. Those who just looked at the topline number may have missed that.