Loose Threads: The Prime Minister’s Role

22/04/2026

There is a conversation happening at the moment about who should be Prime Minister. It is, as these conversations tend to be, almost entirely about personality. Who is likable? Who connects? Who looks the part, sounds the part, and seems like someone you could have a beer with or trust in a crisis? I have no quarrel with the instinct to care about these things. But I do think it is worth pausing to ask a preliminary question: what, exactly, are we choosing someone to do? Because the role of Prime Minister, practically speaking, is not what most people imagine. It is not the presidency. The Prime Minister does not command. The Prime Minister chairs. And the difference between those two things is not a technicality. It is one of the organising principles for how Aotearoa governs itself.

Primus Inter Pares

The constitutional position is straightforward, at least on paper. The Prime Minister is the first among equals in Cabinet. The role exists by convention rather than statute, which is itself significant: there is no Act of Parliament that creates the office or defines its powers. The Prime Minister is the person who commands the confidence of the House. That confidence can be withdrawn. The role derives its authority not from an election, strictly speaking, but from the ability to form and sustain a government.

It is worth noting, because it is widely misunderstood, that voters do not elect a Prime Minister. They vote for parties. The party leader is selected internally, by caucus or, under some party constitutions, by the wider membership. The person who becomes Prime Minister is the one who can command the confidence of the House, which, in practice, means assembling and holding together a governing majority. This is not a minor procedural detail. It means the Prime Minister’s authority is not a personal mandate. It is a function of relationships: with caucus, with coalition partners, with Parliament.

In practice, this means the Prime Minister’s job is coordination. The Cabinet is the executive. Decisions are collective. Ministers hold their portfolios and are responsible to Parliament for them. The Prime Minister does not run Health, Education, or Defence. The Prime Minister runs the government as a whole, which means managing relationships among ministers and portfolios, and navigating the tension between political imperatives and administrative realities. It means chairing Cabinet and its committees, managing the coalition, setting the legislative programme, representing the country externally, and bearing ultimate political responsibility for the collective.

From my doctoral fieldwork and from the years I spent working in and around ministers’ offices, I would add something that is rarely discussed but may matter more than anything else: the Prime Minister’s role is to cohere. The public service is structured, by deliberate design, around vertical lines of accountability. Each chief executive is responsible to their portfolio minister. Each agency pursues its own objectives, manages its own workforce, and guards its own institutional knowledge. Vertical accountability, by its nature, fragments. And someone has to hold the whole thing together. Someone has to govern and sometimes enforce work across the horizontals: that is the work of the Prime Minister.

The Ability to Cohere

Readers of this Substack will recognise the theme. I have written at length about the purple zone as the institutional space where political judgement and administrative expertise must be integrated rather than separated. Every minister operates in a purple zone within their portfolio. But the Prime Minister operates in the purple zone on behalf of the entire system. Every tension between what is politically necessary and what is administratively sound, every conflict between one portfolio’s objectives and another’s, every question about whether the government is actually doing what it says it is doing, lands, eventually, on the Prime Minister’s desk.

This is what I mean by the ability to cohere. It is not merely coordination, though coordination is part of it. It is the capacity to see the system as a whole, to identify where its parts are pulling in different directions, and to ask whether the government’s stated priorities are actually being pursued or merely being reported on. The ability to cohere matters because individuals, families, businesses, and communities do not live in government silos. They experience the cumulative effect of what every part of government does at once, and the Prime Minister’s job is to ensure that the silos’ actions do not overwhelm the real world they are supposed to serve. It requires what I have elsewhere described as authoritative critique: the willingness and the capability to look across what government is doing and to say, clearly and without flinching, when something is not working.

The institutional support for the Prime Ministerial role and function has always been thin. The Department of the Prime Minister and Cabinet (DPMC) was established in 1990, but its predecessor, the Prime Minister’s Department, dates from 1976, when Muldoon ordered it into existence over the objections of officials who feared a strong centre would distort the constitutional balance. Bernie Galvin was the first to lead a small group of expert advisors. Jonathan Boston, writing about those early years, described Galvin’s advisory group as vulnerable to what he called “the politics of exhaustion”: a unit deliberately modest in scale that the sheer intensity of working at the centre of government, trying to make sense of all of government, eventually wore its people down. That pattern of deliberate constraint has never entirely disappeared. The main levers available to the centre remain, as they have always been, personality and proximity.

Every Prime Minister Has Done This Differently

If institutional support has been thin, the coherence function has nonetheless been carried out. What is interesting is how differently each modern Prime Minister has chosen to do it. As I traced in my Substack series on the purple zone, each government furnished the space between politics and administration with its own architecture, and each architecture shaped what advice it had to contain, what priorities it could pursue, and what the system paid attention to.

Bolger and Shipley cohered through the Strategic Result Areas and Key Result Areas: whole-of-government mechanisms that crossed portfolio boundaries and included explicit targets, including, crucially, targets to address Māori disparities. The SRAs and KRAs were directive. They told the system what mattered, and they created lines of sight from the Prime Minister’s priorities down through individual agency accountability documents. Whatever one thinks of the targets themselves, the architecture gave the centre a means of asking whether the system was doing what it had been asked to do.

Clark cohered differently. The elaborate cross-portfolio targeting of the Bolger-Shipley era gave way to three organising narratives that shaped the direction of the whole government: National Identity, Economic Transformation, and Families Young and Old. These were not performance targets in the Bolger-Shipley mould. They were governance frames that told the system what the Clark government cared about across portfolios, where to orient the effort, and what concepts and language to use. But the change in method should not be mistaken for a loosening of the central grip. Clark managed the Cabinet process with exceptional discipline. Nothing of consequence moved without her knowledge. Her method was relational and political rather than structural, but it was no less rigorous in its application of central oversight by the Prime Minister and senior ministers. She held the system together through the force of her own attention, through a focus on high-quality policy advice and implementation and a Cabinet process that her office governed closely: the governance narratives functioned as a coherent and powerful set of statements that drove governmental purpose. The coherence was real and enforced.

Key and English returned to explicit cross-government architecture with the Better Public Services result areas: ten measurable targets that required agencies to collaborate across portfolio boundaries. But the BPS targets were only the most visible part of a much more elaborate coherence infrastructure. Ministers were expected to work together with senior officials to deliver against the result areas. The Performance Improvement Framework provided a systematic assessment of the agency’s and system’s capability to achieve results. Input and output costing tools gave the centre a way of interrogating what was actually being spent and on what. Agencies were required to produce four-year plans addressing workforce development. The result areas were published regularly, and new result areas were added as priorities evolved. All of this was integrated into the public service’s accountability architecture. The BPS programme was, in some respects, the most ambitious coherence mechanism any government had attempted. It gave the centre not merely a set of targets but a set of institutional levers for asking hard questions and holding the system to account for its answers.

Ardern jettisoned almost all of it. The BPS result areas were discontinued. The Performance Improvement Framework was wound down by the State Services Commissioner in 2020. The four-year plans, the costing tools, and the integrated accountability architecture that Key and English had refined over eight years were set aside. In their place, the Ardern government cohered around a smaller number of high-profile commitments, most notably the child poverty reduction targets, embedded in legislation that required the government to measure and report on progress. The Wellbeing Budget framework offered a broader organising narrative, though its operational grip on the system was looser than its rhetorical ambition suggested. The coherence was real, but it was concentrated in particular policy domains rather than spanning the whole of government as the BPS architecture had attempted. Whatever was gained in focus may have been lost in reach.

The Institutional Record

But there is a second register to this story, and it is just as revealing. The policy architecture tells you what each Prime Minister said the system should focus on. The institutional architecture of DPMC tells you what each Prime Minister actually reached for when something needed to be held together.

Since its establishment, DPMC has housed more than thirty-five special units, project teams, and dedicated groups, each one created because a Prime Minister decided that something required the centre’s direct attention. The pattern begins almost immediately: the Change Team on Targeting Social Assistance in 1991, the Health Reforms Project from 1991 to 1993, the Crime Prevention Unit through the 1990s. By the 2000s, the units multiply: the Climate Change Programme, the Foreshore and Seabed Group, the Education Sector Review, and the House Prices Unit. Each is a response to a problem that could not be left to a single portfolio, or that the Prime Minister judged too important, too politically sensitive, or the policy problem too complex, to be managed at arm’s length.

The pace accelerates from 2009 onward. The Christchurch earthquakes alone generated a sequence of units: CERA, the Canterbury Earthquake Recovery Policy Team, the Greater Christchurch Group, and the Canterbury Earthquake Recovery Learning and Legacy Programme. COVID-19 produced its own cluster: the COVID-19 Group, the Proactive Releases function, and eventually the Royal Commission of Inquiry coordination. Alongside these are the Child Wellbeing and Poverty Reduction Group, the Health and Disability Review Transition Unit, the Implementation Unit, He Whenua Taurikura, Kāpuia, and the National Emergency Management Agency. Each unit assembled, did its work, and was eventually dissolved or transferred elsewhere.

What this record reveals is the coherence function made visible. Every one of those units represents a moment when a Prime Minister concluded that the existing departmental structure could not hold something together, and that it had to be brought to the centre. And every dissolution or transfer represents the moment when that particular coherence need has passed, or when a new government decides it is no longer a priority.

The pattern is not accidental. It is the institutional expression of a role that has no fixed content. The role of the Prime Minister is to cohere around whatever needs cohering, using whatever instruments are available. Sometimes that means a governance frame. Sometimes, an accountability instrument, like the BPS result areas. Sometimes it means a dedicated unit inside DPMC. Sometimes it means all of them. But the instruments are always temporary, always contingent on the priorities of the Prime Minister of the day, and they almost never outlast the government that created them.

What Is Holding It Together Now?

Which brings us to the present, and to a question I find myself unable to answer with confidence. The coalition agreements that brought the current government into being were, in places, internally contradictory. That is not unusual in coalition politics. Contradictions are managed, traded off, and sequenced. But the agreements also had a finite shelf life, and much of what they prescribed has now been implemented, abandoned, or overtaken by events. If the coalition agreements were the initial mechanism of coherence, what has replaced them?

I do not see an equivalent to the SRAs, the BPS result areas, or the child poverty targets. There is no visible whole-of-government architecture that tells the system what this government is trying to achieve across portfolios, creates lines of accountability for cross-cutting objectives, or gives the centre a basis for assessing whether things are working. Yes, there are signals of cost reductions, but annual budget spending is still climbing. The DPMC special units that remain are largely inherited from previous governments, and several have been wound down or transferred. The Implementation Unit has finished its work. NEMA has moved to Internal Affairs. The Cyclone Recovery Unit has been disestablished.

There is a further difficulty. A coalition agreement structured as a list of deliverables is a contract, not a vision. The agreements specified what would be done mostly as inputs and activities, and that it does not, of itself, tell the system what the government is trying to achieve as a whole. The SRAs created a direction. The BPS result areas created shared accountability. The child poverty targets created a statutory obligation to measure and report on them. A list of agreed actions, once implemented or abandoned, leaves the question of coherence unanswered.

That does not necessarily mean there is no coherence. It may mean the coherence is being held privately, politically, through relationships and conversations that are not visible from the outside. It may mean the Prime Minister is cohering through attention rather than architecture, as Clark did, though the scale and complexity of a three-party coalition makes that harder.

But the comparative evidence suggests we should be cautious about assuming that personal attention is enough. A recent study by O’Malley, Marland, and Palavicini (2026) tested the presidentialisation thesis across twenty-one democracies using forty years of expert survey data, and their most striking finding is what actually predicts prime ministerial power. It is not personality. It is not media dominance. It is the size of the Prime Minister’s party in parliament. A Prime Minister whose party commands a large majority has room to act. A Prime Minister leading a smaller party in a coalition is structurally constrained, regardless of their personal qualities. The appearance of power and the reality of power are diverging: prime ministers look more presidential, but they are not governing more presidentially.

If that is right, then the ability to cohere becomes even more important when the structural conditions are less favourable. A Prime Minister with a large majority can afford to cohere loosely, because the parliamentary arithmetic gives them a margin for error. A Prime Minister in a three-party coalition, where the structural room to act is narrower, needs stronger coherence mechanisms, not weaker ones.

The Job, Not the Person

I raise all of this because the current conversation about political leadership is conducted almost entirely in the register of personality, and I think that register is insufficient. The role of Prime Minister requires the capacity to coordinate a complex system of collective decision-making amid uncertainty, competing demands, and imperfect information. It requires the ability to hold the authorising environment together when different parts of it are pulling in different directions. It requires governance judgement: to know when to intervene and when to leave a minister to manage their own portfolio. And it requires, above all, the ability to cohere: to construct and maintain an architecture, or a set of relationships, or a dedicated institutional capability at the centre, or some combination of all three, that holds the government’s purposes together in a system designed to fragment them.

Thirty-five special units in thirty-five years. Each one represented a Prime Minister reaching for the centre because something needed to be held together. Each one eventually dissolved. The coherence function is real, consequential, and the hardest part of the job. We would do better to spend less time asking who we like and more time asking who understands what the job of Prime Minister actually is, and then who has the skills to do the job well.

References

O’Malley, E., Marland, A., & Palavicini, G. (2026). Assessing the presidentialisation thesis: Prime ministerial authority in an era of rising centralisation and personalisation. European Journal of Political Research.