Brendan O’Leary: Decisive facts and grace are still absent from Professors Coulter and Shirlow…

Brendan O’Leary, Lauder Professor of Political Science, University of Pennsylvania argues that decisive facts remain absent from the arguments of Professors Coulter and O’Leary. This is the final piece in a fascinating exchange.

In their rejoinder to my response Professors Coulter and Shirlow write the academic equivalent of “no surrender.” Namely, they acknowledge no difficulties in their published positions, regret nothing about their past tone, and launch some curious new insults. In short, they take a political rather than a scholarly line.

Their rejoinder repeats at length what they said in their published article about which I had commented in the course of responding to their unjust characterization of the Ireland’s Future conference on June 15 . Let me recapitulate their difficulties and suggest why their ripostes are not convincing.

They maintain that the past “not-stateds” in previous census reports were disproportionally Catholic, and that such respondents have since largely presented themselves as cultural Catholics in the 2021 census. In support of their argument—as I had pointed out—they provide aggregate spatial data showing more “not-stateds” in disproportionally Catholic areas. This supporting argument, I had observed, is vulnerable to the “ecological fallacy.” I had provided a link to Slugger O’Toole readers who may not have heard of the concept. Remarkably, Coulter and Shirlow simply omit this difficulty. Perhaps they think they have worked around it; or perhaps they overlooked the point in their haste to reply.

The “ecological fallacy” is taught in all graduate programs in social science and medicine. Simply put, it means that one cannot infer individual-level behavior from aggregate data, including aggregate spatial data. For example, one cannot infer that because there is a high level of crime in area X that an individual resident of X is more likely to be a criminal.

To address the relevant causal identification problem, Coulter and Shirlow would minimally need survey data of the following kind: from individual 2021 respondents who ten or twenty years previously answered “not-stated” but who now identify as Catholics (or as brought up in Catholic households) in the census. They would also need individual level data from non-Catholics of the same kind. But they don’t have such data. “Proxy” variables don’t cut the mustard. They re-raise the causal identification problem. In the alternative, they might have tried to identify a “natural experiment” to address the difficulty. But they don’t do that. It follows that they have no decisive facts on this subject, merely hypotheses, or speculations. I have, I repeat, no hostility to hypotheses or speculations, I merely object to Coulter and Shirlow’s false claims to have presented decisive “facts” on the subject at hand.

Both previously, and now, I am not claiming to have refuted their speculation about the “not-stateds”; I have, however, suggested that their confidence in their position is unjustified according to contemporary social science (and medical) statistical protocols.

The tone they deployed in their article to explain the transition from “not-stated” to cultural Catholic in the census is, however, objectionable. They did not repeat these explanations before the readers of Slugger O’Toole, so let me quote two passages in their own words:

“Long in advance of the poll, there was widespread speculation that Catholics would, for the first time, represent the largest grouping in Northern Ireland. That change in the standing of a community that had in the past been treated as subordinate might well have been sufficient to persuade some people harbouring an historical mistrust of the Census – and, indeed, some who might not otherwise embrace the term – to disclose themselves as ‘Catholics.’ Furthermore, in the heightened political climate post-Brexit, a cogent campaign has emerged insisting that demographic change is among the critical forces that might prompt a referendum on Northern Ireland’s future. The prospect that the outcome of this particular Census could bring a ‘border poll’ closer might well have convinced some that this was the time to abandon their reticence and declare themselves as Catholics for the first time…”

“Republican politicians, in particular, had been keen to ensure that as many Catholics as possible appear in the Census in order to maximise the labour market opportunities afforded to members of that community.”

These are speculations: there are two palpable “might well haves,” as well as some curious scare quotes.

In my response I had noted that to achieve the required effect, i.e., moving respondents, as they put it, “to disclose” themselves as “Catholics” in their individual or household reports, required extensive coordination. Coulter and Shirlow bluster in reply with an unacknowledged paraphrase from Karl Marx that humans make their own history—while excluding his qualifier “though not as they please.” Absent decisive empirical evidence about campaign-effects on individual form-filling, I suggested that the authors’ speculations displayed a conspiratorial mind-set about cultural Catholics who are nationalists. The above quotations provide some evidence. Differently put, what would one think of a correspondent who wrote to The News Letter claiming that, at the behest of Sinn Féin, the SDLP, and Ireland’s Future, distrusting “Catholics” had come out of the closet, at least in their individual or household 2021 census reports, to advance the cause of a united Ireland and/or to maximize their job opportunities?

Similar cautions about the “not-stateds” must be applied to the “nones.” Responding to my skepticism that Coulter and Shirlow could show that few cultural Catholics were “Nones” they bait and switch. They report “national identity” figures from the census—British, Northern Irish, and Irish. Again, these are proxy variables, but they may be deemed by many to be prima facie plausible. Coulter and Shirlow point out that few among “the “Nones” identify as “Irish.” True, but what they can’t do is eliminate cultural Catholics from the “Northern Irish” nones who include large numbers of cultural Catholics who vote Alliance or SDLP, as we know from individual survey data. They can’t even eliminate them entirely from the British category—some respondents may believe that their nationality or national identity is decided by what citizenship they hold, and yes, some cultural Catholics may identify as British.

On Irish-medium schools Coulter and Shirlow are evasive; their reply is an unacknowledged concession. It was after all they who raised school enrollment data—to describe falling proportions of Catholic children—in their original article, on which I was commenting. I would have preferred had they openly acknowledged this simple but small error in their discussions of the estimates of school pupils. My speculative hypothesis was that most pupils in Irish-medium schools are likely from cultural Catholic households. I suspect they agree.

They are also evasive on schooling data in general. In reply to the point that there was a Catholic majority of 18 and under in the 2021 school census, they reply that there hasn’t been one subsequently. So what? The point for me was to show how the 2021 school census, coterminous with the state census, forecast the possibility—not the certainty—of a future Catholic demographic, and later electoral majority, however brief that might be over the long-run. A school census is not a survey (a survey is a sample). Therefore, it provides harder and more reliable data. Indeed, a school census would usually be expected to have a higher degree of accuracy and completion than the state census because it is organized by highly literate and numerate professionals. It is entirely legitimate to use it for supplementary information. There was no need for Coulter and Shirlow to imply that it is an erroneous or desperate maneuver on the part of nationalists to invoke such data. Here they lost sight of the ball to play the men of their imagination.

The general point is that Coulter and Shirlow exhibit a rhetoric of “just the facts” while providing unproven albeit interesting speculations. Their readings seem intended to deliver the result that there is a stable demographic status quo, and a stable electoral status quo, a world of three minorities. It may also be that such a world would be congenial to their political preferences, but that would be speculative. They use the expression “stalemate” to characterize this world—a world of falling proportions of Protestants, British-identifiers, and Unionist (with a capital U) voters in censuses, surveys, and elections. Time will tell whether they are right.

The most recent local government elections of 2023 delivered an outcome in first-preference votes of 44%, 40%, and 16% for nationalists, unionists, and others respectively. In an ARINS blog on the Royal Irish Academy website (currently down) Jamie Pow and I found that this was the outcome after we had separately coded all the independents as nationalists, republicans, unionists, loyalists, or others as appropriate (see a summary here). It was also the first election in which “all nationalists” outpolled all big U unionists. We treated Alliance votes as other. This 44: 40: 16 outcome is rather different from the 40: 40: 20 configuration of loose local discussion. The upcoming Westminster elections on Thursday July 4 cannot be decisive in assessing the stability or instability of the electoral world because of the strategic withdrawals by several parties which will aid the aggregate Alliance vote—and the SDLP vote. But they will be interesting. Who would have thought two decades ago that Sinn Féin, the largest party in the most recent elections, would engage in strategic withdrawals of candidates to favor Alliance or SDLP candidates against the DUP? Or, as Philip McGuinness has shown, that transfer patterns in STV-elections would reveal that Alliance voters’ ballots now transfer significantly more to nationalist candidates rather than previously, at roughly 50: 50 ratios between the two historically dominant blocs. Or, that it would be conceivable—not certain—that Belfast might have no MPs from a large U unionist party? As these rhetorical questions imply, I do not see stability but rather continuing incremental change.

An odd curiosity in the new slights from Coulter and Shirlow is that I should have delayed the publication of Making Sense of a United Ireland until the census results of 2021 were published. Elsewhere I am accused of being out of date. As John McEnroe might have said: they cannot be serious. The census was published late because of the pandemic. This author was under a scheduled contractual obligation to his publisher, and the publisher had a scheduled contract with the printers. Between final submission of a manuscript and going to press there is always a significant interval though Penguin was faster than any university press with which I have published, and it is always difficult to amend proofs. Should I also have waited until Coulter and Shirlow’s article was published? The whole point of demography is that it enables governments and others to plan, albeit without certainty. And yes, the 2011 census predicted the 2021 census quite well, and it was useful for thinking about 2031 because it provided information on those who would enter the electorate from 2012 onward (18 years’ worth). I used the demography of 2011 to argue that by 2030, barring off-trend migratory movements, or a major shift in differential outbound migration at the expense of cultural Catholics, Northern Ireland’s future would no longer be in cultural Protestant hands. A reasonable judgment, and so far, accurate. I also expressly indicated it was not certain that there would be a cultural Catholic electoral majority, as opposed to a plurality, and that the pivotal voters in a future referendum would likely be among the following overlapping categories: women; recent immigrants with voting rights; the “nones;” current “don’t knows” among Alliance, Green, and SDLP voters; pro-EU liberal Protestants; and “wobblies”—those who have changed their minds more than once on the question of the union or unification. The point of these preliminaries, given the prospect of demographic, electoral, and cultural change, was to prepare the reader and others for the merits of appropriate preparation and planning, to which the bulk of the book is devoted. Readers may judge the book for themselves.

Coulter and Shirlow exploit my reasonable suggestion that I will not use 2024 survey data to predict a referendum outcome in 2030 to suggest that I am weirdly against surveys, and that I am contradictory because I am one of the co-authors of the University College London led project on unification referendums (Renwick, A., et al, Final Report of the Working Group on Unification Referendums on the Island of Ireland. London, Constitution Unit, 2021). In effect, they wonder what I am doing leading the public opinion committee of ARINS, and co-authoring annual surveys—and organizing future deliberative forums. There are simple answers to these almost playful but silly criticisms. The ARINS all-island surveys currently are research-led investigations of what might best facilitate the possible (not certain) peaceful, democratic, and constitutional reunification of Ireland. We are also interested in a reformed Union. We seek to use the best social science standards for surveys, deliberative forums, and for qualitative research. Published at length in the Irish Times, the ARINS surveys so far explore opinion on old and new questions, in depth—all of which may help planning and preparation for future referendums. We do not aim to predict a referendum vote in six to eight years hence, though we do ask Northern respondents about how they would vote in a future referendum and are struck by the number of “don’t knows.”

Future Secretaries of State may use surveys to guide their decision on whether to hold a referendum. For the record, however, as a member of the UCL working group I argued that votes by a majority of MLAs, or a majority of Northern Ireland MPs, provided they represented majorities in the relevant elections, would represent better, and more legitimate evidence, than surveys or internet polls. But as a member of that group, I recognized that others, including unionists in the group, held different opinions—so both election results and surveys are discussed as options for informing the Secretary of State in the report. There was a consensus, which I share, that internet polling, so far, is inferior to surveys, though both have difficulties. Since I was composing my own book, I felt no desire to disrupt the report by demanding that my view on the primacy of election results should be shared by all, which would have been rude—and ineffective. And yes, I am aware, and have often made the point in print, that election results may not predict referendum outcomes because the latter may generate far higher turnouts.

What do Professors Coulter and Shirlow and I agree on? That demography is not destiny, and that politics, including the politics of persuasion, always matters. We disagree somewhat about how to characterize the present, and more deeply about future possibilities. If the SDLP and Sinn Féin, aided by constructive Dublin governments, can prepare adequately for future referendums on unification I believe they can win them. To do so, they will need to persuade an extraordinarily high proportion of cultural Catholics who vote Alliance, Green, SDLP, and Sinn Féin that reunification credibly offers better life-chances for them and their families than the maintenance of the Union. They will need to win a significant share of the new minorities, and they will need to make some inroads among some cultural Protestants (who currently favor the Union, but like the EU). They will also need a credible plan to accommodate hardline Protestant loyalists, British citizens, and British identifiers, and they will need develop a model of a united Ireland that would draw from the best of the North and South. A wider majority would be better, but it may not be within reach. They could win because life-chances under the Union do not look likely to improve significantly while the UK polity and economy remain outside the EU, whereas the future of the Republic looks set to be one of distinct and continuous net improvements.

To return to the original insult, now repackaged as a metaphor: no millenarian suggestion that a united Ireland will be a paradise was or is being made. The Pursuit of the Millennium was the enterprise of medieval religious mystics. “Millenarian” is neither an appropriate metaphor, nor an apt description of those engaged in the democratic, peaceful, and constitutional pursuit of a united Ireland. On July 3, 2024, it looks as if “Things can only get better” in the UK tomorrow night. In the Republic things are much better, and likely to get even better. Of course, unfolding turmoil in the US and the EU could make fools of us all.