Northern Ireland has a problem with opinion polls.
For decades, the authoritative Northern Ireland Life and Times Survey – run by Queen’s University Belfast and Ulster University – has found implausibly low support for a united Ireland.
Even after the EU referendum, the latest pro-unity figure stands at 22 per cent, about the same level as always, implying half the North’s nationalist voters are not nationalists.
Other face-to-face polls produce similar results. This problem has come to be explained by the “shy republican” theory, positing that people are reluctant to tell a stranger they back a cause associated with Sinn Féin. That effect would be expected to weaken as the Troubles recede into history, yet results remain stuck at an improbable level.
Commercial polling has moved on to a new methodology of online panels, sometimes supplemented by telephone interviews. This has produced the opposite phenomenon – implausibly high support for a united Ireland.
Hanging a question of this magnitude on polling seen as inaccurate or partisan is clearly a recipe for disaster
Since the EU referendum, such polls have found increasing numbers of unionists becoming nationalist, culminating this week in a campaign poll finding that one unionist in six will switch allegiance due to Brexit, delivering a majority for unification.
Game the results
This is also improbable, to put it mildly. Although online polling has delivered impressive election predictions in Northern Ireland, on the binary question of unification it is only becoming less convincing. There have been complaints that panels are wildly unrepresentative, that adjusting for this inevitably reduces accuracy and that people can game the results by misreporting their community background.
In this week’s poll, one quarter of the sample were Alliance voters, outnumbering DUP and UUP voters. In reality, the latter outnumber the former six to one. Shy republicans have been replaced by numerous neutrals.
Perhaps the fundamental problem is that those who volunteer for online polling are by definition unusually engaged in politics and caught up in its distinct online culture, to an extent beyond statistical correction.
Whatever the cause, the consequence is that polling by all available methods has lost credibility at the moment it is needed most.
Brexit and demographic equalisation have raised expectations of a Border poll but the Belfast Agreement does not specify how that should be triggered – only that nationalist victory must “appear likely” to the Northern Secretary.
The courts have declined to clarify the matter and the balance of the Stormont Assembly is no longer a guide – unionists and nationalists are neck-and-neck, with opinion among the 10 per cent of Alliance and Green voters the deciding factor.
Opinion polls have filled this information gap. All sides, including the British government, have cited them in arguing for or against a Border poll. An assumption has developed that a run of polls favouring Irish unity must lead to a referendum.
Hanging a question of this magnitude on polling seen as inaccurate or partisan is clearly a recipe for disaster.
The potential for opinion polls to cause instability, division and resentment is serious enough to ask if it is worth holding an actual Border poll merely to establish where both sides stand.
The Northern Secretary is entirely free to call a poll outside the terms of the Agreement, meaning it need not be binding or require a parallel vote in the Republic.
DUP leader Arlene Foster appeared open to this idea five years ago, when she believed it would show a clear unionist majority. Today, the DUP’s position is that a Border poll itself would be divisive and destabilising, and on that it may well be correct.
Countdown to unity
Unionists are nervous about a Border poll – even one they would win – because the agreement says one can be held no more than every seven years. This has been interpreted by nationalists as meaning once a poll occurs, it must reoccur every seven years, creating an inexorable countdown to unity.
Even a poll outside the terms of the agreement would bring that expectation into play.
Soothing this seven-year itch has been the subject of two fascinating contributions by former DUP leader Peter Robinson and former Sinn Féin Assembly speaker Mitchel McLaughlin, both made honorary professors at Queen’s this year.
In his professorial lecture in June, Robinson proposed holding Border polls once a generation, with unionists negotiating a "generational settlement" in between.
This was dismissed by nationalists as an attempt to move the goalposts.
In McLaughlin’s lecture later that month, he suggested including a Border poll-type question in the 2021 census – Northern Ireland’s centenary year.
This would have the advantages of avoiding a referendum and leaving the agreement untouched. However, the idea was dismissed as politicising the census, as if that ship had not sailed in 1921.
The enormous media coverage garnered by this week’s opinion poll, despite its obvious shortcomings, shows that McLaughlin’s proposal or something equally imaginative deserves renewed consideration.
Polling has been politicised in Northern Ireland, and the opinion polls will now be coming thick and fast.