There is some unreality to the current clamour for reform of Northern Ireland’s devolved government. The three major non-unionist parties, as well as the British and Irish governments, have proposed or are open to substantial changes in the teeth of unionist opposition. That debate too often overlooks cross-community consent as a principle shaping constitutional form, focusing only on its function as a rule governing decisions within the political institutions. Little is therefore said about what would legitimate deep reform and make it durable.
Fulfilling the principle of cross-community consent means that structural reform of the terms of participation in Northern Irish powersharing and cross-community protections requires renewed majority support within each of the nationalist and unionist political blocs. But also overlooked is the dearth of unionist support for removing the government-formation veto, as well as the, at best, partial support for replacing community designation-based voting rules.
This post explains the origin and significance of the principle of cross-community consent in the modern Northern Irish constitution and how it is reflected in institutional arrangements. It outlines recent reform proposals and looks to history to distinguish legality, legitimacy and durability in constitutional reform.
WHAT IS THE PRINCIPLE OF CROSS-COMMUNITY CONSENT?
The modern Northern Irish constitution is rooted in the Good Friday (Belfast) Agreement (“1998 Agreement”). It takes domestic legal form in the Northern Ireland Act 1998 (“NIA”), as amended many times by Westminster to implement reforms negotiated by the local parties and British and Irish governments. The constitution’s central principle is the requirement that nationalists and unionists govern together through institutions that command popular support from both communities. The principle precedes the 1998 Agreement. The 1996-98 multi-party talks leading to it operated under a ‘sufficient consensus’ rule: where agreement was lacking on devolved institutions or north-south cooperation, a proposal required support from parties that together won a clear majority of the vote in the 1996 Forum election and represented clear majorities in both main traditions. Brendan O’Leary describes “reinvented consociational practices [used] to conduct the talks”, with sufficient consensus determining what became part of the Agreement. The Agreement was emphatically endorsed in May 1998 by voters on the island of Ireland, though surveys suggest that only a narrow majority of Northern Irish unionists voted in favour. Cross-community consent therefore regulated the making of the institutional settlement, while later structural changes were also, for the most part, negotiated with broad cross-community agreement.
Strand One of the 1998 Agreement safeguards participation in, and protections from, government for both main traditions. Members of the Legislative Assembly (MLAs) designate themselves as unionist, nationalist or ‘Other’. The main safeguard requires cross-community agreement for key governmental decisions. Under section 4(5) of the NIA, that support means either ‘parallel consent’ –– a majority of MLAs voting, including majorities of nationalists and unionists voting — or a ‘weighted majority’ of 60% of MLAs voting, including at least 40% of MLAs voting within each of the two main designations. The 1998 Agreement required cross-community consent for electing the Speaker and First and deputy First Ministers, adopting standing orders, and budget allocation. It also applies to transfers of reserved powers, Assembly size, the Ministerial Code, ministerial exclusion or censure, and other financial and procedural matters.
The St Andrews reforms, implemented in 2007, replaced joint election of the First and deputy First Ministers with separate party nomination from the two largest parties from separate community designations: if either minister resigns, both offices become vacant and the Executive cannot meet. Since 2022, other ministers may remain as caretakers for up to 48 weeks after a breakdown but cannot make contentious or cross-cutting decisions. Refusal to nominate thereby hands both parties an effective government-formation veto. The withdrawals from government of Sinn Féin in 2017 and the Democratic Unionist Party (DUP) in 2022, followed by refusals to nominate, left Stormont largely inoperative for five of the seven years to 2024.
Under section 42 of the NIA, a petition of concern presented by 30 MLAs from at least two parties, and confirmed again after a 14-day consideration period, can, subject to statutory exclusions, require cross-community support for an Assembly vote. Section 28A(8) of the NIA requires the Ministerial Code to allow three ministers to trigger cross-community voting in the Executive.
Taken together, the 1998 Agreement’s founding principle, the statutory safeguards in which it is embodied, and the practice of institutional reform, support a legitimating requirement of renewed sufficient consensus when altering the terms on which the two traditions share power. It can hardly be the case that routine decision-making is subject to the principle of cross-community consent while constitutional reform is immune from that constraint. Different ways of gauging such support exist. But a vote by the Assembly is a practical measure, reflecting voter preference and MLAs’ alertness to shifting voter attitudes on constitutional preference. This would necessitate a two-part test: a majority of all MLAs, and majorities of all nationalist-designated MLAs and all unionist-designated MLAs. That measure warrants ongoing review with the electoral growth of constitutionally unaligned parties, whose first preference vote has doubled from 8% to 17% since 1998. But the touchstone of legitimacy must remain the basis on which the 1998 Agreement was negotiated and popularly endorsed.
CALLS FOR REFORM
Probably the most striking contribution to the reform debate arrived in June 2026 when Sinn Féin published Building Better Politics, calling for an effective end to the government formation veto. Sinn Féin would preserve the initial right of the largest nationalist and unionist parties to nominate the First and deputy First Ministers, but proposes that, if either declines, “the opportunity to nominate should pass to other parties”. The paper is silent on whether this means the next largest party overall, the next within the same designation, or the largest party designated ‘Other’. Sinn Féin also proposes removing the veto over electing the Assembly Speaker but leaves the proposed replacement rule unspecified. Its similarly vague nod towards equality for ‘Other’ MLAs remains subject, somewhat contrarily, to protecting the 1998 Agreement’s designation safeguards.
The constitutionally agnostic Alliance Party’s longstanding calls for reform intensified after its 2022 electoral breakthrough. Its equivalent proposal on government formation is clearer, with the entitlement passing to the next largest party, as in the D’Hondt allocation of other ministries, should an entitled party decline to use it. An immediate proposal for change is replacement of the parallel consent standard with a weighted majority of around two-thirds of the Assembly recalculated after each election. The party also proposes restricting petitions of concern, and the three-minister Executive trigger, to questions of identity, conflict legacy, and the constitutional structures created in 1998. More radically, Alliance proposes abolishing community designation in the medium term, paving the way for a voluntary coalition formed by willing parties.
The Social Democratic and Labour Party (SDLP), the second largest nationalist party, supports staged reform: removing the formation veto, weighted majority election of the Speaker, equal titles for the two heads of government, and a pledge by Executive parties not to collapse the institutions during the mandate. Changes to government formation would form part of a broader discussion.
The appetite for reform is very different on the unionist side of the political divide. The DUP’s position has sharply changed: in the early 2010s (when in a much stronger electoral position than it is now) the party supported eventual voluntary coalition, voting by a majority of around 65%, and ending community designation. The party no longer supports those reforms. The DUP’s leader, Gavin Robinson MP, reiterated that position on 19 September 2026 when he said that the party “will never support reforms that are designed to strip unionists of our consent for our governing arrangements”. The Ulster Unionist Party (UUP), the second largest unionist party, supports limited reform, including a supermajority vote for Assembly Speaker. Its leader has criticised single-party collapse but warned against nomination reform that would place the First and deputy First Ministers in the hands of one tradition, while not yet specifying a preferred replacement mechanism.
The most noteworthy proposal endorsed by the Northern Ireland Affairs Select Committee in its 2023 report on the 1998 Agreement’s institutions is the use of a supermajority of two-thirds to elect two ‘Joint First Ministers’, with any two MLAs from different parties eligible to stand jointly for the offices. It also recommended electing the Speaker by a two-thirds majority: this could allow Assembly debate and scrutiny during an Executive boycott. In response, the Sunak government stressed in 2024 that constitutional reform in Northern Ireland required widespread consent, including cross-community support for changing existing protections. Labour has since continued that approach.
LEGALITY, LEGITIMACY, AND DURABILITY
Alliance’s 2026 “Democratic Renewal” document grounds its argument for structural reform in an Assembly vote in December 2025. By 41 to 32, MLAs approved an SDLP amendment criticising Executive parties’ failure to prioritise institutional reform. Every unionist MLA voting opposed it, while every nationalist and ‘Other’ MLA voting supported it. That division prompts two distinct questions about the limits of constitutional reform. First, are there limits to Westminster’s power to effect structural change at Stormont? And could the reformed institutions endure if parties representing a majority within either tradition refused to participate?
Cross-community consensus is not a legal precondition to constitutional reform from Westminster. Further, s 5(6) of the NIA preserves Westminster’s legislative power over Northern Ireland. But legality is a separate question from legitimacy and durability. Legitimacy turns on the existence, or otherwise, of sufficient consensus for adoption, while durability depends on whether the reformed institutions can sustain or, after a crisis, recover broad participation of both traditions. On current Assembly numbers, reform imposed over Sinn Féin’s or the DUP’s heads would fail that legitimacy test. A subsequent boycott by parties commanding a majority within either bloc would also threaten durability. The British government therefore has clear practical and principled reasons to seek sufficient consensus before enacting statutory reform. Transferring nomination rights within the same designation would preserve representation of both main traditions in joint leadership, but this alone would not reflect majority consent in both blocs to a new rule. Nor would keeping the offices filled necessarily secure broad participation if the dominant party in one tradition maintained a boycott.
Richard Fallon’s account of sociological legitimacy helps explain the connection between cross-community authorisation and continuing participation. In its ‘stronger’ sense, sociological legitimacy concerns the body politic’s belief that public authority deserves acceptance. The sufficient consensus test advanced here is a normative standard for authorising reform, with an Assembly vote providing a clear measure of such consent expressed by elected representatives. A new rule lacking cross-community authorisation may nevertheless prove durable because alternatives are unattractive or institutional withdrawal too costly. In that sense, Fallon’s ‘weaker’ concept of sociological legitimacy captures acquiescence, which may reflect habit or self-interest. Such acceptance can develop while the rule continues to fail the sufficient consensus test advocated here.
History offers credible candidates for all four combinations on the legitimacy-durability matrix. The overarching 1998 settlement, with its philosophy of cross-community partnership, and mutual protections, was legitimate and has proved durable through repeated crises and renegotiations. The original 1998 arrangements were legitimate at adoption in 1998, but repeated suspensions and significant renegotiation of government-formation rules show the limits of their durability. Yet those revisions also helped the wider settlement to endure.
The Article 18 democratic consent process under the Northern Ireland Protocol offers a possible example of illegitimate yet durable reform. It provides for an Assembly vote on continuing certain post-Brexit trading arrangements in Northern Ireland. The petition of concern was disapplied, a change upheld by the Supreme Court in Allister. The December 2024 vote lawfully passed by 48 to 36 despite every unionist MLA voting against. That reform lacked cross-community legitimacy under the test advanced here. Yet almost two years on, the mechanism remains in place and Sinn Féin and the DUP remain in government. Ongoing participation is consistent with acquiescence in a rule to which unionists remain fiercely opposed.
Brendan O’Leary’s account of the Northern Ireland Act 2000 offers an example of an illegitimate reform that not only proved non-durable but dented powersharing. The Act authorised the Secretary of State to unilaterally suspend devolution. O’Leary writes that it lacked legitimacy because “[n]o suspensory power had been endorsed with cross-community consent” and criticises its enactment outside the Agreement’s review processes. He also regards it as destabilising, leaving “every aspect of the Agreement… vulnerable to Westminster’s sovereignty” and obstructing the development of stable institutions. The power was repealed in 2007 as part of the St Andrews reforms, thus showing its illegitimacy was matched by non-durability.
Even if reform underpinned by sufficient consensus were to cross Northern Ireland’s constitutional Rubicon and remove the government formation veto, that would not by itself replace the legitimacy standard for subsequent structural reform. The 1998 referendums offered popular endorsement for consensus-based mutual vetoes. Subsequent negotiated changes were attempts to renew that consent to the shared arrangements. St Andrews, for example, responded to the existential challenge that, by the mid-2000s, the 1998 arrangements no longer enjoyed majority unionist support.
Any reform of the ultimate rule of legitimacy, centring on designation-based voting, would, in my view, require fresh popular authorisation as it would alter the very basis of the 1998 Agreement-imposed constitutional settlement: while the exact threshold would require prior demographic analysis, a referendum endorsed by a supermajority of Northern Irish electors should be required. Any such change could remove a single party’s ability to collapse government but not compel cross-community participation. Northern Ireland’s entire existence warns against constitutional change that leaves a third or more of the population distrustful of its governing arrangements.
WHERE NEXT?
The government-formation veto and most core designation safeguards are, for now, subject to a form of what Albert calls ‘constructive unamendability’. That is, statutory reform of those provisions is possible, but the absence of sufficient consensus and the prospect of yet another boycott make that path politically prohibitive. On current Assembly strength, the only viable reform of government formation –– the transfer of nomination rights to the next-largest party within the same designation –– collides with political reality: collapse of government has historically been an act of last resort during crises, which suggests it may be implausible that, in a future period of political instability, the relevant designation’s next largest party would opt to remain in government. For so long as Sinn Féin and the DUP sustain electoral dominance within their respective traditions, the transfer of nomination rights is unlikely to deter withdrawal. And even if that measure achieved legitimacy, such a reform would be of uncertain durability.
The petition of concern, for example, was significantly altered where sufficient consensus to do so existed. Sinn Féin’s proposal to surrender its effective veto shows that political incentives can shift and open new ground. The utility almost three decades after the 1998 Agreement of handing a veto over institutional reform to a majority within a minority bloc is plainly problematic. But that power stems ultimately from how the devolved institutions acquired legitimacy. With four in five voters still opting for unionist or nationalist parties, and sectarian mistrust remaining a societal blight, a consociational settlement with designation-based safeguards remains the only legitimate and durable form of devolved government. Substantial cross-community support for powersharing and some degree of institutional reform exists. But the exact design of constitutional reform supported by the public is unclear, while one in six voters favour moving away from, or replacing, powersharing. It would also appear that public support is limited to “reforms that could be accommodated within the consociational framework”.
Structural amendment, including removal of the formation veto, can acquire legitimacy through sufficient consensus and its durability would depend on sustained, cross-community participation. But reform cannot magic away the function performed by what Mark Durkan famously described in 2008 as the 1998 Agreement’s ‘ugly scaffolding’: the lumbering mechanisms that sustain participation in government by a majority within both traditions. The present lack of sufficient consensus for deep reform, such as removal of the government formation veto and a dilution, or even removal, of community designations, reflects a societal dysfunction that structural change alone cannot repair.
I would like to thank the Editors for their very helpful advice and comments on this post.
Dr Patrick Graham is a Lecturer in Law at the University of New England (Australia).
(Suggested citation: P. Graham, ‘Legality, Legitimacy, and Durability – The Limits of Constitutional Reform in Northern Ireland’, U.K. Const. L. Blog (28th September 2026) (available at https://ukconstitutionallaw.org/))
Leave a Reply