Ireland: The Machinery That Sat Before the Country Voted

Ireland is usually told as a story about a mind that changed. A Catholic country, the account runs, became a secular one, and two referendums three years apart recorded the moment it happened. The figures are quoted as though they were the finding: sixty-two per cent for marriage in May 2015, sixty-six per cent for repeal in May 2018.

This Library is not chiefly interested in those figures. It is interested in what happened before them, because Ireland did something with those two questions that very few countries do with any question. Before the country voted, a body of ordinary people — selected at random, holding no office, with no standing beyond having been picked — sat down with doctors, lawyers, submissions from organisations on every side of the argument, and worked for months. Then they voted among themselves and published what they had concluded and by what margin, along with the transcripts and the papers they had been given. Only afterwards did the electorate go to the polls.

The claim this essay wants to make is narrow, and the narrowness is the point. It is not that deliberation produces better answers. Ireland’s own record refuses that reading, and refuses it decisively, as a later section will show at the essay’s own expense. The claim is that how a society decides a question about private life changes what the decision can mean to the people living inside it afterwards — including, and especially, to the people who lost. A settlement that was argued leaves a record of the argument. A settlement that was announced leaves only the announcement.

There is a second country in this chamber this month where the terms of private life changed and nobody voted. The two pieces are meant to be read together. But the comparison has a hard limit, and it is stated here at the start so that nothing later can be read past it: nothing in the Irish record shows that the Polish outcome would have been different under a different procedure. That is unknowable. This Library does not know it and will not imply it.

Precision first, because this is the part that is most often reported loosely, in Ireland and outside it.

The Constitution of Ireland can be changed in exactly one way. Article 46 requires that a proposal be introduced in the lower house as a Bill, passed by both Houses of the Oireachtas, and then submitted to the people; if a majority of the votes cast is in favour, the Bill is signed by the President and the Constitution is amended accordingly. There is no parliamentary shortcut and no judicial one. This is not evidence of a national appetite for direct democracy. It is a structural fact about one document, and it produces a country that has been obliged to vote on matters other states settle by statute: on divorce, on the death penalty, on the age at which a person may stand for the presidency, on treaties, and on the definition of a family.

In 1983 that machinery was used to put a position on pregnancy into the constitutional text itself. The Eighth Amendment of the Constitution Act 1983 added a new subsection to Article 40.3, in the Irish text and in the English. The English reads that the State acknowledges the right to life of the unborn and, with due regard to the equal right to life of the mother, guarantees in its laws to respect, and, as far as practicable, by its laws to defend and vindicate that right. It was approved at a poll on 7 September 1983, by 841,233 votes to 416,136, on a turnout of 53.7 per cent, and signed into law on 7 October 1983.

Two features of that wording are worth holding onto, because everything afterwards runs through them. The first is that the subsection did not by itself prohibit anything. It created a constitutional right against which every statute had to be measured, which meant the legislature could not legislate its way out of the consequences; the lock was above the level at which ordinary politics operates. The second is that the clause was symmetrical in form — two lives, equal rights — while the work of deciding what that symmetry meant in a particular hospital on a particular night fell to courts and to clinicians, not to the people who had voted for it. In 1992 the country’s highest court read the subsection as permitting termination in defined circumstances of risk to the woman’s life; twenty-one years later a statute was passed to give that reading a procedure. This essay refers to that sequence structurally and does not narrate the case, for reasons set out at the end.

The subsection did not remain as drafted either. In 1992 two further amendments were attached to the same subsection: one providing that it did not limit freedom to travel between the State and another state, and one that it did not limit freedom to obtain or make available, within the State and subject to conditions laid down by law, information about services lawfully available elsewhere. A country had voted to place a right in its constitution, and then voted, nine years later, to make explicit in the same subsection that the right did not reach as far as the airport.

The deliberative machinery was not invented for the abortion question, and getting the order right matters, because the two bodies are routinely confused with each other.

The first was the Convention on the Constitution, approved by resolution of both Houses in July 2012. Its membership was sixty-six randomly selected citizens, thirty-three politicians drawn from both Houses of the Oireachtas and from the Northern Ireland Assembly, and an independent chair, the economist Tom Arnold. It considered a list of topics set for it and two it chose itself. In April 2013 it recommended that the Constitution be amended to provide for marriage without distinction as to sex, on a vote of seventy-nine per cent in favour. The Government responded in December 2013, and the referendum followed in May 2015. So the marriage question was prepared by a mixed body of citizens and politicians, not by the later all-citizen assembly.

The second was the Citizens’ Assembly, established by resolution of the Houses in July 2016 and chaired by Ms Justice Mary Laffoy, a judge of the Supreme Court. Its membership was ninety-nine citizens and the chair, with ninety-nine substitutes held in reserve. It was given five topics: the Eighth Amendment, the challenges and opportunities of an ageing population, fixed-term parliaments, the manner in which referendums are held, and how the State might respond to climate change. It took the Eighth Amendment first.

The selection deserves to be described exactly rather than admired vaguely, because the difference between the two is where most of the argument about this model actually lives. A market research company was engaged after a public tender. It used stratified random sampling: sampling points chosen on a random basis from census data, and recruiters calling at addresses within electoral divisions, working to quotas for gender, age and social class derived from the census. Members had to be on the register of electors. People who were, or had been, members of advocacy organisations campaigning on the Assembly’s topics, or who intended to act in an advocacy role, were excluded and screened for at interview. Direct applications from the public were refused, and recruiters were forbidden to enrol friends and family.

That is a serious procedure and it is not sortition in the pure sense. Nobody was conscripted. Recruiters called; people agreed or declined; and those who agree to give five weekends to the most divisive subject in the country are not themselves a random draw from those who were asked. The Assembly also had attrition, which is why the substitutes existed, and in February 2018 it emerged that seven replacement members present and voting at the January meeting had been recruited outside the agreed methodology, by telephone and through a recruiter’s acquaintances. They were removed and took no further part; the chair called the episode regrettable and stated that it was isolated and did not affect the Assembly’s earlier work. That is a defensible position on the timing, because the ballots on the Eighth Amendment had been taken ten months before. It is also, precisely, the kind of thing an essay admiring a procedure is obliged to print.

The Assembly met on the Eighth Amendment across five weekends between November 2016 and April 2017. Its own report records twenty-five experts, more than eighty hours of listening and discussion, and over twelve thousand submissions from the public. The public sessions were streamed live, the papers were published as they were delivered, and the transcripts were appended to the report. Voting was conducted with the assistance of a former returning officer.

The recommendations, published on 29 June 2017, were these. Eighty-seven per cent voted that Article 40.3.3 should not be retained in full. Fifty-six per cent voted that it should be replaced or amended rather than simply deleted. Fifty-seven per cent voted that it should be replaced with a provision explicitly authorising the Oireachtas to legislate on termination of pregnancy, on any rights of the unborn and on any rights of the pregnant woman. In other words, the Assembly did not recommend repeal on its own. It recommended handing the question to the legislature, in terms, and writing that handover into the constitutional text.

On what the legislature should then do, sixty-four per cent recommended that termination without restriction as to reason should be lawful. Of those who voted on that ballot, forty-eight per cent would have confined it to twelve weeks, forty-four per cent to twenty-two weeks and eight per cent set no gestational limit. Separately, a majority of members recommended twelve specific grounds on which termination should be lawful, ranging from risk to the woman’s life through risk to her health to socio-economic reasons, and seventy-two per cent recommended that no distinction be drawn between physical and mental health.

Two things follow that cut against a sentimental reading of this body. The first is that the Assembly landed, on grounds and on limits, some distance beyond what the legislature would eventually enact, which is awkward for the argument that a mini-public simply mirrors settled opinion back at the country. The second is that what the 2018 referendum did was almost exactly what the Assembly had asked for in structure: not a bare deletion, but a substitution handing the matter to the Oireachtas. A joint committee of the Oireachtas took the Assembly’s report, held its own hearings and reported in December 2017 with recommendations of its own, and the Government announced its decision to proceed in January 2018. The chain from the room to the ballot paper is documented at every link, which is itself the unusual thing.

The 2015 vote is widely misdescribed, usually generously. On 22 May 2015 the country approved the Thirty-fourth Amendment of the Constitution (Marriage Equality) Bill by 1,201,607 votes to 734,300, on a turnout of 60.52 per cent. The amendment inserted a new section into Article 41: marriage may be contracted in accordance with law by two persons without distinction as to their sex. It was signed into law in August 2015, and the Marriage Act 2015 followed on 29 October, commencing in November of that year. Ireland was the first state to extend marriage to same-sex couples by a nationwide popular vote rather than by legislation or by a court.

What it did not do is the more useful list. It did not touch Article 41.2, the provision about a woman’s life within the home, which remained in the text for another nine years. It did not remove the State’s pledge to guard the institution of marriage. It did not oblige any religious body to recognise a form of ceremony or any religious solemniser to conduct one, a point written expressly into section 7 of the Marriage Act 2015. And it did not itself construct the surrounding law on parentage or recognition; separate statutes of the same year, including the Gender Recognition Act 2015, dealt with matters the referendum did not decide. A constitutional amendment of one sentence is a narrow instrument, and treating it as a general change of national disposition is where most of the misreporting starts.

The 2018 vote was narrower still in what it authorised, and this is the fact most often lost. On 25 May 2018 the country approved the Thirty-sixth Amendment of the Constitution Bill by 1,429,981 votes to 723,632, on a turnout of 64.13 per cent, and it was signed on 18 September 2018. The polling card put the question as a deletion and a substitution: delete the 1983 subsection together with the travel and information clauses attached to it in 1992, and substitute a single sentence — provision may be made by law for the regulation of termination of pregnancy. That sentence permits nothing on its own. It removes a lock and returns the question to the legislature, exactly as the Assembly had recommended.

The legislature then did the work. The Health (Regulation of Termination of Pregnancy) Act 2018 was passed on 13 December and signed on 20 December 2018, and the service was introduced on 1 January 2019. Its structure is worth stating because the debate abroad rarely does. Termination is lawful where two practitioners form the opinion that there is a risk to the life or of serious harm to the health of the woman, with a separate section for emergencies; where two practitioners form the opinion that a condition is present that is likely to lead to the death of the foetus before or within twenty-eight days of birth; and, under a separate section, where a single practitioner forms the opinion that the pregnancy has not exceeded twelve weeks, with a required interval of not less than three days between certification and the procedure. The Act’s own notification returns show how that distributes in practice: of 8,156 terminations notified for 2022, 8,042 were carried out under the early-pregnancy section.

The settlement is not finished, and honesty requires saying so rather than closing the story at the referendum. A statutory review of the Act’s operation, published in 2023, recommended changes including the removal of the mandatory three-day interval and an end to the criminalisation of practitioners, and reported barriers in how the service is delivered. Some of those recommendations remain unimplemented. A country can decide a constitutional question cleanly and still be arguing about the statute six years later, which is what ordinary politics looks like once the lock is off.

Four arguments run against the reading above. They are set out here at their strongest, because a version weak enough to knock down would prove nothing.

The first is that referendums on minority rights are a dangerous instrument, and that this essay risks admiring one because the result was congenial. That objection is correct on its own terms and this chamber has already published the evidence for it. The Australian essay records what a national campaign on a minority’s family life cost the people the question was about, and states plainly that winning a vote on your own legitimacy is not the same as never having been asked to. The Taiwanese essay records a referendum on the same subject that went the other way, and a legislature obliged to work around it. Ireland’s 2015 campaign was a campaign about particular people’s families, conducted in their hearing, and the fact that it ended well does not retire the objection. It is also worth separating the two Irish questions, which are usually flattened together. In 2018 a majority undid a clause that a majority had itself inserted in 1983; that is a population correcting its own act. In 2015 a majority was asked to extend something to a minority; that is a different kind of question with a different cost, and Article 46 is the reason it could not have been done any other way.

The second is that the deliberation may have been decoration. Peer-reviewed work makes this case carefully. Courant, writing in 2021, argues that there is no clean transferable Irish model: the assemblies were not institutionalised and each differed in composition and design, politicians chose the topics, the budgets and which recommendations reached a ballot, an expert group had considerable influence over how ballot questions were framed, and public awareness of the process was limited. He also notes the arithmetic that sits uncomfortably with the heroic account — the Assembly’s recommendation for termination without restriction carried sixty-four per cent inside the room, and the referendum returned sixty-six point four per cent outside it, which is consistent with a country that had already moved. On this reading the Assembly is a legitimating device that allowed politicians to go where they already wished to go, with cover.

The third is that the causes lay elsewhere. Elkink, Farrell, Marien, Reidy and Suiter, analysing exit-poll data from 2018, find that age and church attendance were the key drivers of the vote, with a large difference between those attending weekly and those not, and read the age pattern as a cohort effect: one generation replacing another. They also find that voters aware of the Citizens’ Assembly were more likely to vote yes, which is a correlation and not a demonstration that the Assembly moved them. Around that vote sits a decade of state and church inquiries into institutional abuse that did lasting damage to a particular kind of moral authority, and a measurable change in affiliation: the census of 2022 recorded sixty-nine per cent of the usually resident population as Roman Catholic against seventy-nine per cent in 2016, and 736,210 people recording no religion, an increase of sixty-three per cent on 2016, though the statistics office itself notes the question wording changed between the two censuses. A purely procedural reading of Ireland under-weights all of this, and it should be said rather than managed.

The fourth is about standing. A publication in Japan writing admiringly about another country’s constitutional process is exposed to a specific vanity: praising a procedure whose costs it did not carry and whose losers it will never have to meet. Roughly a third of the people who voted in 2018 lost, and they lost a provision they had been told for thirty-five years was a settled moral commitment of their country. An essay that cannot hold that in view has not understood what it is describing.

The strongest evidence against this essay is not an argument. It is a date.

A Citizens’ Assembly on Gender Equality, convened on the same model, reported in June 2021. Among its recommendations was that Article 41.2 — the provision about a woman’s life within the home and about mothers and economic necessity — should be deleted and replaced with language that was not gender specific and that obliged the State to support care. A parliamentary committee took it up. The Government drafted, and on 8 March 2024 two proposals went to the country: one to recognise the family whether founded on marriage or on other durable relationships and to delete the pledge to guard the institution of marriage, and one to delete Article 41.2 and insert a new article recognising care within families and committing the State to strive to support it.

Both were rejected. The family proposal was defeated by 1,021,546 votes to 487,564, and the care proposal by 1,114,404 to 393,053, on a turnout of about forty-four per cent in each case. Those are wider margins of rejection than either of the majorities of 2015 and 2018 were margins of approval. The same country, the same instrument, the same preparatory machinery, and the answer was no, twice, decisively.

The defeats were argued about on their wording as much as on their substance, including by people who wanted Article 41.2 gone and thought these were the wrong words for removing it; that characterisation is a reading of the public argument rather than a count, and this essay does not adjudicate it. What matters here is simpler. A deliberative assembly is not a machine for producing liberal outcomes. It is a machine for producing a recommendation, which a government may or may not draft well, and which a country may or may not accept. Anyone who admires the Irish procedure only in the years it delivers the answer they wanted is not admiring a procedure. They are admiring a result and crediting the method.

It follows that this essay’s claim has to be stated at the size it can actually bear. Deliberation before a vote did not make Ireland liberal and cannot be shown to have changed any of the three outcomes. What it did, reliably, in all three, was leave a public record of the reasoning — who heard what, from whom, in what order, and how the room divided. In 2024 that record is what allows the country to say what it rejected, and why, without either side having to invent an account of the other’s motives.

The other country in this chamber this month arrived at the opposite settlement by the opposite route. A constitutional court ruled on 22 October 2020 that the ground on which the great majority of lawful terminations in Poland had been carried out was incompatible with the constitution; the judgment entered into force when it was published in January 2021. No one voted at any stage, and no body of randomly selected citizens sat with the evidence beforehand.

Three things this essay does not say about that. It does not say the Polish outcome would have been different under another procedure; that is unknowable, opinion polling is not a referendum and a referendum is not opinion polling, and any claim in that direction would be a partisan guess wearing a comparative costume. It does not say that a vote is legitimating merely by being a vote — the preceding section on referendums and minority rights stands, and stands in both directions. And it does not say that Irish institutions are better than Polish ones, which is not a judgment a foreign publication has the standing to make.

The comparison it does make is narrower and, this Library thinks, defensible. The two countries are differently placed to say afterwards what their own settlement rests on. Ireland’s rests on documents anyone can read without permission: ballot papers, transcripts, expert submissions, recorded margins, and a returning officer’s count published constituency by constituency. A person who disagrees with the outcome can see precisely what she is disagreeing with and who else disagreed alongside her. Poland’s rests on a judgment. That is a difference in what is available to a citizen who wants to know why her private life is arranged as it is. It is not a difference in whether the arrangement is right, and this essay does not convert the first into the second.

The Constitution of Japan can also be amended only with the people’s agreement. Article 96 requires a concurring vote of two-thirds of all the members of each House of the Diet, followed by ratification by a majority of votes cast at a special referendum or at an election the Diet specifies. That procedure has never been used. The document has not been amended since it came into force, and no such referendum has ever been held. The instrument exists on paper and has no practice attached to it, while the deliberative machinery this essay has described has no counterpart here at all.

The wrong lesson is therefore the obvious one, and it should be named before anybody draws it. Nothing here recommends putting Article 24, or the question of who may marry, or any other question about particular people’s families, to a national campaign in Japan. On this chamber’s own evidence from Australia and Taiwan, that would be a decision to spend the well-being of the people being voted on, and it would be spent whatever the result. Recognition here has moved instead by statute, by municipal certificate and by litigation carried, as this chamber has recorded, by a small number of named plaintiffs. That route has its own costs and they are not smaller; they are differently distributed.

What travels is smaller than a constitution and it is not a matter of public policy at all. It is a distinction between a settlement you were talked into and one you reasoned your way to, and an observation about how differently the two behave when new information arrives. Ireland held its 1983 position the way a person holds an inherited conviction: as something to be defended, because the reasoning behind it was not hers and could not be inspected. A position arrived at by sitting with the evidence, including the evidence against, is held differently. It can be revised without collapsing, because the reasoning is still there to be revisited, and it can be explained to somebody who disagrees without the explanation turning into a fight about loyalty.

This chamber has argued that what religion was doing was never mainly rule-giving, and that the societies which removed the rules have not replaced the three things that sat underneath them. Ireland is a case where the rules went and the question of the replacement is visibly still open; the votes of 2024 are the evidence that it is open. It has argued, in the Spanish essay, that when a Catholic moral order gives way to a public language of desire the older structures often go quiet rather than disappearing. And it has argued, in the essay on the vacuum where a law should be, that an absence of law protects nobody. Ireland is the opposite failure and it makes the same point from the other side: a presence of law, set so high in the hierarchy that no legislature could reach it, also protected nobody, because it took thirty-five years and a national campaign to alter one sentence. What distinguishes a workable settlement about intimate life is not mainly its content. It is whether the people bound by it can revise it, and can find out why it says what it says.

For a reader here the operative form of that is private and unglamorous. It is the difference between knowing what you want and knowing why — between a preference you would have to defend and one you could explain. The second survives a disagreement, a change of circumstances and an ageing body. The first has to be protected from all three.

This house sells evenings, and it should say first what that has to do with an essay about constitutional procedure, because the honest answer is uncomfortable.

What the house actually sells is a deliberated arrangement. Before anything begins there is a form, an explicit statement of what is wanted and what is not, and a word that ends things without explanation; the terms are agreed in advance by the person they bind, and nobody on the house’s side has standing to reinterpret them afterwards. In the vocabulary of this essay, the house sells a private settlement whose reasoning is on the record and which the woman who made it can revise. That is the same good the Irish procedure produced at national scale, sold in a room, for money.

The house therefore benefits directly from the scarcity this essay has been describing. A woman who has mostly encountered arrangements she was talked into — by a family, by a partner, by an employer, by a script she never agreed to — is a woman for whom a room with written terms and a guaranteed exit has a price. The house profits from the absence of that elsewhere in her life. It should say so plainly rather than let the value appear to come from warmth, or skill, or anything else it would prefer to be credited for.

What follows from that is a list of things it must not claim. Nothing here is an argument for any political position in Ireland, in Poland or in Japan, and the house takes none. Nothing here suggests that an ordinary relationship should be conducted by form and agreed word; a courtship run like a constitutional convention would be an absurdity, and the essay’s own point is that the procedure is what makes a commercial arrangement safe, not what makes an intimate one good. Nothing here teaches a woman how to obtain terms in a room where nobody offered them. And one evening cannot rehearse a life. What it can honestly offer is a single experience of having decided something about her own body in advance, with the reasoning intact and the exit real — and whether that experience travels out of the room is not in the house’s gift to promise.

It does not claim a position on abortion or on marriage. The house has none and would not be a competent holder of one. Everything above is about how decisions were made and what the making left behind, and it has been written so that a reader on either side of either question can check the account against the record and find it fair. If it reads as advocacy in either direction, it has failed at the thing it was attempting.

It does not claim that deliberation produced the outcomes. The evidence does not support that and one section of this essay is given over to the evidence against it. It does not claim that the Citizens’ Assembly was representative in the strict sense; it was quota-recruited from randomly chosen addresses among people who agreed to serve, it lost members, and its recruitment failed once in a documented way. It does not claim that referendums are a safe instrument for questions about minorities; this chamber has published the opposite finding and that finding is not suspended because Ireland’s vote went the way it did.

It does not claim anything about the Polish outcome under a different procedure. That is the one sentence in this essay that is load-bearing for the whole comparison, and it is a refusal.

It does not narrate any individual case. Not the death that is commonly cited as the turning point of the 2018 campaign, not the judgment of 1992, not any named plaintiff or patient. The omission is deliberate and there are two reasons for it. The first is that the subject here is procedure, and a single story, however true, does the argumentative work of a hundred pages without submitting to the same checking; a reader moved by it has not been shown anything about how decisions get made. The second is a matter of standing. A commercial house that sells intimacy does not get to use a named woman’s worst day as an illustration in its own copy. The reader who wants that account will find it named in every history of this period and does not need this Library to supply it.

It does not claim completeness on the law. The constitutional texts, the referendum results, the Assembly’s ballots and the principal sections of the 2018 Act have been read in the original and are cited as such. The practical operation of the service, the unimplemented recommendations of the statutory review, and the fate of Article 41.2 after 2024 are live and moving, and any figure here should be checked against a current official return. And the essay is written from outside both of the societies it compares, in a third language, about a country whose argument with itself is not finished.