Loose Threads: "Dear Colleagues ..."

Starmer, Free and Frank Advice, and What Three Jurisdictions Reveal About One Constitutional Problem

On 7 May 2026, the night before local elections in which his party faced what most forecasters predicted would be a historic rout, Sir Keir Starmer emailed every civil servant in the United Kingdom. The email was, on its face, an exercise in reassurance. He thanked officials for their service. He affirmed the value of “speaking truth to power.” He insisted that the government was “fixing the processes that have failed” but was “not changing the fundamental value we place on your role” (Starmer, 2026).

The timing, however, told a different story.

Starmer’s email arrived weeks after he had sacked Sir Olly Robbins, the Permanent Under-Secretary at the Foreign Office, over the Lord Mandelson vetting affair. Robbins told the Foreign Affairs Select Committee that he had operated under what he described as an atmosphere of sustained pressure from Downing Street to expedite Mandelson’s appointment as UK Ambassador to the United States. Earlier in the year, Cabinet Secretary Chris Wormald had also been replaced after barely fourteen months in the role. Dave Penman, general secretary of the FDA trade union representing senior civil servants, accused the Prime Minister of sending a “real chill” through Whitehall. Some officials who received the letter reportedly described it as “tone deaf.” One told the BBC they felt exposed not by recent scrutiny but by a Prime Minister who, in their view, had sacrificed senior officials to deflect political criticism.

There is, then, a rather considerable gap between the rhetorical commitment to free and frank advice and the conditions under which it might plausibly be offered. This is not a problem unique to the current UK Government. It is, I would suggest, quite possibly a feature of the Westminster advisory relationship, one that tends to surface most visibly during moments of political stress. And it is the problem I want to explore in this post, across three jurisdictions, because the comparative picture is instructive in ways that a single case cannot be.

This is not the next formal instalment in my Free and Frank series. It is, rather, a live application of the argument I have been developing across this entire Substack project: that the formal commitments routinely outstrip the institutional conditions under which things can be safely and consistently practised.

Why the gap is designed in: it’s not personal

In my doctoral research, recently completed at the University of Canterbury, I argue that free and frank advice operates as what I have termed a bounded plurality (Te Kawa, 2026).

The concept has three parts.

The first is the plurality itself: at least six distinct practitioner postures, each representing a coherent but different way of reasoning about the advisory relationship.

The second is the shared floor that holds those postures together: an expectation, common to almost all practitioners, that advice will always be practical. Not political. Not abstract. Not theoretical. But practical: what I describe as practical statecraft.

The third is a set of four fault lines along which the six postures diverge: whether the official’s duty is to align with the minister’s democratic mandate or to maintain principled independence; whether challenge is a constitutional obligation or a strategic tool; whether free and frank advice is routine practice or an ultimate failsafe; and whether the craft lies in the strategic timing of when to speak or in an unconditional duty to speak regardless.

The notion of practical statecraft borrows, in a sense, from Head’s (2008) work on evidence-based policy. Head argues that genuinely effective policy advice requires three distinct forms of knowledge working together: systematic research evidence and the analytical and technical data about what works on the ground and what happens in the real world when policy settings are changed; political judgment, the nous regarding what is politically feasible and acceptable, including the work of persuasion and coalition-building; and practical wisdom, the operational experience held by frontline professionals regarding how a policy will actually work when implemented. Head emphasises that modern governments increasingly face complex, “wicked” problems involving deep uncertainties and competing values, and that, because purely evidential inquiry cannot adjudicate between conflicting stakeholder interests or ethical concerns, advice must blend technical knowledge, practical stakeholder experience, and evidence. For evidence alone is not enough.

The theoretical lineage of my idea of a bounded plurality arrives from a different direction. In 1956, Gallie established that certain concepts are not merely vague or awaiting better definition but are, by their nature, subject to disputes that no amount of evidence or analytical precision can resolve. The contestation is constitutive: to use an essentially contested concept is to use it against other uses, and to recognise that one’s opponents are doing the same thing with equal sincerity. Connolly (1974/1993) extended Gallie’s idea by arguing that such contestation is not a deficiency but a feature of democratic life. He argued that it is precisely because the meaning of concepts like democracy, justice, fairness, equality, equity and the public interest cannot be fixed that they remain available for democratic argument. Honig (1993), whom Connolly supervised, pressed the point further, insisting that every constitutional settlement generates a remainder: an excess of democratic meaning and energy that the settlement cannot absorb and that persists as a site of ongoing political life. And Mouffe (2005) gave the remainder a political and decolonial form, arguing that the task of democratic institutions is not to eliminate antagonism but to transform it into agonism: the structured and respectful contest between adversaries who share enough common ground to remain active within the same political community while debating and disagreeing about nearly everything the remainder might represent*. Bounded plurality is the form that this remainder takes when practitioners offer free and frank advice: undefined, contested, with at least six different postures, but brought to life and navigated in daily advisory practice between elected and unelected officials.

And it must not be finally defined. Bureaucracy and democracy exist in a state of productive tension. Democracy demands responsiveness to the elected government; bureaucracy demands consistency, due process, and the willingness to say what ministers may not wish to hear. Neither can be permitted to win outright. If free and frank advice were defined with precision, that definition would inevitably privilege one side of the tension over the other: either locking in a version of candour so robust that it obstructed democratic responsiveness, or codifying a version of responsiveness so accommodating that it licensed the suppression of inconvenient truths. The concept of the bounded plurality of free and frank advice captures the importance of a constitutional convention remaining undefined and contested, not because practitioners are confused about what they do, but because settling it would unbalance the détente on which the entire advisory relationship depends.

In my concept of the bounded plurality, free and frank advice is a remainder. It appears in legislation, in codes of conduct, in the rhetorical repertoire of prime ministers and permanent secretaries alike. Yet it resists stable definition. What it means to speak truth to power depends, in practice, on the posture a practitioner adopts, the institutional norms they navigate, and the political context in which advice is offered, received, or withheld. The six practitioner postures identified in my Q-methodology study, designed by 34 practitioners, completed by 58 participants and tested and workshopped with over 132 other practitioners, commentators and academics, revealed not a single shared understanding but a structured plurality: distinct ways of reasoning about the advisory relationship that coexist, sometimes uneasily, within the same institutional space (Te Kawa, 2026). Practitioners hold free and frank advice together not by agreeing on what it means, but by recognising each other’s reasoning as legitimate variants of a shared commitment, depending on the context.

This matters for what is happening in the United Kingdom because it suggests that the problem is not, at root, one of individual ministerial temperament. Starmer is neither the first nor the last Prime Minister to reach for the rhetoric of candour while creating conditions that make candour dangerous. The gap between rhetorical commitment and lived practice is a structural feature of the Westminster model itself: ministers need advice that is candid and independent, yet they also need a public service that is responsive to the government’s political direction. When those imperatives collide, as they have in the Mandelson affair, it is typically the adviser, not the advice, that gives way.

The cipher in the constitution

Alongside the drama in the UK, I have been reading Ben Yong’s recent work on the civil service and the UK constitution. I recommend it to everyone. It certainly helped me explain why this dynamic is so resistant to repair.

In his Current Legal Problems lecture, Bureaucracy and Distrust: The Civil Service in the Constitution, Yong (2025) argues that the civil service has been largely effaced from public law theory. It remains, as he puts it, a cipher: its functions, and the sources and limits of its authority, unclear. The civil service is not seen to have an identity or authority of its own, and, Yong observes, this impression is one that the civil service itself actively encourages. What matters, constitutionally, are ministers; officials merely advise and implement.

This effacement has consequences that bear directly on the Starmer letter. If the civil service possesses no constitutionally recognised identity of its own, then the advisory relationship rests, in practical terms, on convention and ministerial forbearance rather than on any structural arrangement or institutional guarantee. As Martin Stanley has drawn from Yong’s work, while officials can speak truth to power, there is, as Janet McLean has noted, very little constitutional scaffolding to protect them when they do (Stanley, 2025a, 2025b; Yong, 2025). The permanence of the civil service and its capacity to push back depend on norms that an executive under pressure may choose not to observe.

This is, I think, precisely what appears to have occurred in the Robbins affair. However one reads the contested facts of the Mandelson vetting, the outcome is plain: the sacking of a Permanent Secretary in circumstances where his own account suggests he followed proper process, and where former Cabinet Secretaries have publicly called for his reinstatement, sends a signal that convention alone may not be adequate to the weight it is being asked to bear. A letter affirming the value of “speaking truth to power” does not, by itself, restore the conditions under which truth-telling is safe.

Yong’s analysis also illuminates a deeper tension. The traditional account of the Westminster model treats bureaucracy and democracy as in tension: bureaucracy is a means to an end, but it always threatens to become an end in itself. This framing, however, obscures the possibility that the civil service might have a constitutive role in democratic governance, not merely an instrumental one. If free and frank advice is a constitutional remainder, as I argue in my doctorate (Te Kawa, 2026), then the officials who provide it are not simply servants of the executive. They are, in some meaningful sense, bearers of a democratic function that our unwritten constitution itself struggles to name. For those of us working within Aotearoa, there is a resonance here with Te Tiriti o Waitangi, which bears the plurality function while remaining precisely undefined and unsettled, and indeed disputed. Both are cases where the constitutional order depends on something it cannot quite bring itself to codify. I think that is what Yong (2025) is reaching for when he describes the civil service as effaced: it performs a role the constitution depends upon but refuses to recognise.

The Australian warning: Mulgan and the written record

The problem, moreover, is not confined to oral advice or personal relationships. Richard Mulgan’s analysis of the Australian Robodebt Royal Commission reveals a complementary dimension of the same structural vulnerability: the systematic erosion of the written record.

In one of his final published pieces, Mulgan (2024) identified a long-term trend in the Australian Public Service towards leaving important matters unrecorded. The Robodebt Commission found that middle-ranking officials had correctly documented the legal risks the income-averaging scheme would face, and had later reported on significant problems in its implementation. Their superiors, however, deliberately prevented these assessments from reaching ministers. A culture of silence and misrepresentation enabled a succession of ministers to maintain that the scheme was lawful and functioning as intended.

What Mulgan (2024) drew from this is a paradox of modern accountability that deserves to be stated carefully. Under the traditional Westminster system, confidentiality was near-total, and fearless, frank written advice flourished. The introduction of Freedom of Information legislation, and of other accountability mechanisms such as ombudsmen, administrative tribunals, and parliamentary committees, made public servants more directly accessible to public scrutiny. But the principal effect, Mulgan argues, has been to discourage officials from recording any evidence or opinion that might be used to embarrass their political masters. The result is what he describes, with characteristic directness, as the worst of both worlds.

On the one hand, says Mulgan, Australian officials have lost the strong commitment to frank and fearless advice that was nurtured under the old Westminster system, in which all government accountability came through ministers while public servants were largely anonymous. On the other hand, he says, with FOI and other transparency innovations, public servants have become more directly accessible, but the main effect has been to drive candour underground. Not only are public servants muzzled in public, Mulgan (2024) observes, but they are also, in private, being pressured to construct politically favourable advice that shields ministers from awkward matters they ought to know about.

Mulgan’s sharpest observation is perhaps the most sobering. Senior public servants in Australia operated behind an almost impregnable wall of secrecy. Cabinet documents were fully protected from FOI disclosure. Governments regularly invoked public interest immunity to prevent substantial investigation by parliamentary committees. And yet, even within that near-total confidentiality, they still failed to speak truth to power. Confidentiality may be necessary for frank communication between officials and ministers, but it certainly does not guarantee it.

This is the material dimension of the structural problem. Free and frank advice requires not only the courage to speak, but the institutional conditions that make it safe to write things down, to create a record that can be tested, scrutinised, and held to account. When those conditions are eroded, whether by the chilling effects of transparency regimes or by a culture of political risk-aversion at senior levels, the formal obligation to advise freely and frankly becomes, in practice, unenforceable.

Meanwhile, in Aotearoa

The UK and Australian cases both illustrate what happens when the conditions for free and frank advice have already broken down. In Aotearoa New Zealand, the picture is different, not because the tensions are absent but because the institutional response has taken a different form: over many, many years.

On 11 February 2026, the Policy Project within the Department of the Prime Minister and Cabinet convened what it called the inaugural Policy Leaders Summit, held in the Banquet Hall of the Beehive. The event materials, recently released under the Official Information Act in response to a request I made to DPMC, brought together senior policy leaders from across the public service to discuss free and frank advice: what doing it well looks like and why it matters (DPMC, 2026a, 2026b, 2026c). I am grateful to DPMC for their transparency in the release.

The programme was carefully constructed. A panel of chief executives, chaired by DPMC Deputy Chief Executive Janine Smith, explored what free and frank advice looks like in practice, including where it has gone wrong. A keynote from Hon Chris Bishop outlined what he called the “Three Cs” of effective advice: credibility, clarity, and courage. And a session described as “The 9th Floor View” brought together former Chiefs of Staff from both sides of the political aisle, Heather Simpson from the Clark Government and Wayne Eagleson from the Key and English Governments, to offer their reflections on the realities of political decision-making. Prime Minister Luxon gave closing remarks.

Ben King, Secretary of DPMC and Head of the Policy Profession, opened the Summit with remarks that warrant careful attention (DPMC, 2026b). He noted that ministers receive advice from a wide range of sources, much of it well-informed, almost all of it carrying an angle. The distinctive contribution of the public service, King argued, lies in its capacity to take account of all of these views, to assess the evidence, surface trade-offs, and present ministers with a recommendation and a point of view. He cited the public service census finding that only 71 per cent of agencies reported confidence that their organisation was free and frank in its advice to ministers, a figure he described as indicating that there was still work to do.

King also directly addressed the particular pressures of an election year. 2026 marks the end of the first term of New Zealand’s first three-way coalition government, a governing arrangement that has required the public service to develop policy advice that may, as King acknowledged, land unevenly across party lines, or even across factions within parties. The pressures, he suggested, would only intensify as the government focused on delivering its coalition agreement and as coalition parties looked to differentiate themselves to the electorate.

The Summit summary, published internally within DPMC and also released, captured a further theme from the former Chiefs of Staff (DPMC, 2026c). Simpson and Eagleson both emphasised that free and frank advice must focus on execution: ministers need to understand real-world impacts, supported by clear delivery plans and ongoing evaluation. Aotearoa New Zealand, they suggested, has room to improve in this area.

What three jurisdictions reveal about one problem

Each of these episodes, the Starmer letter, the Robodebt inquiry, and the Policy Leaders Summit, is a response to the same underlying problem. What is instructive about their juxtaposition is not their similarity but the difference in institutional register.

In the United Kingdom, the conversation about free and frank advice has been driven by crisis: the Mandelson vetting affair, the sacking of a Permanent Secretary, the forced departure of a Cabinet Secretary, and a Prime Ministerial letter that many of its recipients found unconvincing. The emphasis is on trust broken and relationships to be repaired, after the event.

In Australia, the conversation has been driven by retrospective accountability: the Robodebt Royal Commission’s forensic exposure of a system in which frank advice was not merely absent but actively suppressed. Mulgan’s (2024) analysis takes the further step of asking whether the accountability architecture itself, the very mechanisms designed to ensure candour, may have contributed to its erosion.

In Aotearoa New Zealand, the conversation has been more explicitly prospective and developmental. The Policy Leaders Summit was designed not to respond to a rupture but to strengthen an existing practice: to affirm norms, share practical insights, and create space for senior officials to reflect on what free and frank advice looks like in the context of coalition government and an approaching election. It was, in the language of the event pack, an exercise in capability-building (DPMC, 2026a).

This is not to suggest that the Aotearoa New Zealand system is without its own tensions. It would be naive to draw that conclusion, and my own research would not support it. The Public Service Act 2020 codifies the duty to offer advice in a free and frank manner, giving the obligation a statutory footing that the other jurisdictions lack in any comprehensive sense. Yet codification does not, in itself, resolve the practical difficulties.

Practitioners in the Aotearoa New Zealand system navigate at least four distinct fault lines (Te Kawa, 2026). The first concerns whether the official’s primary duty is to align with the minister’s democratic mandate or to maintain principled independence from it. The second concerns whether challenge is a constitutional obligation that must be discharged regardless of circumstances, or a strategic tool to be deployed when the relationship and the moment permit. The third concerns whether free and frank advice is routine practice, woven into every interaction, or an ultimate failsafe reserved for moments when something has gone seriously wrong. And the fourth concerns whether the craft of advisory work lies in the strategic timing of when to speak, or in an unconditional duty to speak regardless of political cost. These fault lines are not disagreements that better training or clearer guidelines could resolve. They are structured expressions of the same bureaucracy-democracy tension that the bounded plurality holds in place. They are not unique to any single government or political configuration, and the Policy Leaders Summit materials suggest that King and his colleagues are well aware of their persistence.

What is notable about the Summit, however, is the institutional register in which it was conducted. The Aotearoa New Zealand system chose to address free and frank advice as a matter of professional development, system stewardship, and deliberate norm maintenance, and it did so at a point in the electoral cycle when the pressures were real but had not yet produced a crisis. There is something characteristically Aotearoa about the approach: practical, understated, and oriented toward what works rather than toward what sounds good in a communiqué.

The Summit did not produce a grand statement of principles. It produced a room full of Ministers and chief executives comparing notes on what works, what they had got wrong, two former Chiefs of Staff from opposing governments agreeing that execution matters more than strategy, and a Head of the Policy Profession willing to say, on the record, that 71 per cent confidence in free and frank advice across the system was not good enough. That is practical statecraft in action: not a theory about what advice should be, but a professional community working out, together, what it takes to keep doing it under pressure.

Whether this is sufficient to sustain the practice under the intensifying pressures that King himself identified is an open question. But it is, at the very least, a different starting point from a letter written in the wake of a sacking.

The hinge, again

For readers who have been following my Te Rā Whakamana series, the connection to the interpretive hinge may already be apparent. The hinge, as I am developing it, is the space where policy advice meets delivery, and where the work of translation, sense-making, and relational judgment is done. It exists because the state’s frame is never neutral. The weavers are the people who do that work.

Free and frank advice is, in one sense, a thread that runs through the hinge. It is the material that the weaver works with: the assessments, the evidence, the trade-offs, the recommendations that officials bring to the point of decision.

When that thread is weakened, whether by the chilling effects of a sacking, the erosion of the written record, or the slow accretion of political risk-aversion at senior levels, the weave itself is compromised. The hinge does not disappear; the translation work still has to be done. But it is done with thinner material, and the resulting fabric is less able to bear the weight of implementation.

Starmer’s letter, Mulgan’s paradox, and King’s Summit are each, in their own way, responses to the thinning of that thread. They operate at different registers: crisis management, retrospective analysis, and prospective norm-building, respectively. None of them, individually, is adequate to the structural problem. But read together, they suggest something about the nature of the challenge that no single jurisdiction’s experience can fully illuminate.

The formal commitment to free and frank advice outstrips, in all three systems, the institutional conditions under which it can be safely and consistently practised. This is not a counsel of despair. It is a description of the terrain, and it is the terrain on which the rest of this series will now work.

Next week, the free and frank series moves into method: how I designed a Q-methodology study to map the practitioner postures that hold this bounded plurality together, before running through the findings over five parts.

Footnote

*There is a resonance here with Te Tiriti o Waitangi, which bears its own remainder function within the constitutional arrangements of Aotearoa. Te Tiriti o Waitangi concentrates tensions that the settler state requires but cannot resolve: between kāwanatanga and rangatiratanga, between the Crown’s claim to govern and the prior and continuing sovereignty of hapū and iwi, between a constitutional order that presents itself as settled and a founding document that insists it is not. Like free and frank advice, Te Tiriti resists definitional closure not because its meaning is unclear but because closure would require one side of the tension to prevail over the other, and the constitutional order depends on neither side prevailing entirely. The Waitangi Tribunal has spent four decades demonstrating, report by report, that the Crown’s policy failures are not departures from the constitutional settlement but expressions of a settlement that was never, in fact, settled. The bounded plurality of free and frank advice operates at a far smaller scale and with far less at stake. But the structural logic is the same: a concept the system depends on, refuses to define, and that practitioners must navigate through judgment, relationship, and context rather than by reference to a settled rule. For Māori, the consequences of this navigational burden have been, and remain, immeasurably greater. The comparison is offered not to equate the two but to identify a shared constitutional grammar: the remainder is not a flaw in the system. The hypothesis that is offered is what the system runs on.

References

Connolly, W. E. (1993). The terms of political discourse (3rd ed.). Blackwell. (Original work published 1974)

Department of the Prime Minister and Cabinet. (2026a). Policy Leaders Summit: Event pack (Item 2). Released under the Official Information Act 1982, OIA-2025/26-0533.

Department of the Prime Minister and Cabinet. (2026b). Policy Leaders Summit: Chief Executive speech (Item 4). Released under the Official Information Act 1982, OIA-2025/26-0533.

Department of the Prime Minister and Cabinet. (2026c). Building policy capability across the Public Service: Free and frank advice (Item 5). Released under the Official Information Act 1982, OIA-2025/26-0533.

Gallie, W. B. (1956). Essentially contested concepts. Proceedings of the Aristotelian Society, 56(1), 167–198. https://doi.org/10.1093/aristotelian/56.1.167

Head, B. W. (2008). Three lenses of evidence-based policy. Australian Journal of Public Administration, 67(1), 1–11. https://doi.org/10.1111/j.1467-8500.2007.00564.x

Honig, B. (1993). Political theory and the displacement of politics. Cornell University Press.

Mulgan, R. (2024, October 18). Frank and fearless and the written record. The Mandarin. https://www.themandarin.com.au/235055-frank-and-fearless-and-the-written-record/

Public Service Act 2020 (NZ).

Stanley, M. (2025a, November 2). Could the civil service restrain a populist, anti-democratic government? Martin Stanley’s Substack.

Could the Civil Service Restrain a Populist, Anti-Democratic Government?

Stanley, M. (2025b, December 20). ‘Storm and sunshine’: Reform’s plan for the civil service. Martin Stanley’s Substack.

'Storm and Sunshine':- Reform’s Plan for the Civil Service

Starmer, K. (2026, May 7). Letter to civil servants [Letter distributed to UK civil servants]. Office of the Prime Minister.

Te Kawa, D. (2026). The bounded plurality of free and frank advice: Practitioner postures, tensions and the practical state in Aotearoa [Draft Doctoral thesis, University of Canterbury].

Yong, B. (2025, October 16). Bureaucracy and distrust: The civil service in the constitution [Lecture]. Current Legal Problems Lecture Series 2025–26, UCL Faculty of Laws, London. https://www.ucl.ac.uk/laws/events/2025/oct/clp-bureaucracy-and-distrust-civil-service-constitution