This month, the High Court of Australia confronted for the first time an issue that has vexed courts around the world for decades. How can the compelling need for an impartial bureaucracy be reconciled with the fact that public servants are not “second class citizens” (Finn 1993, 61) and therefore cannot be “silent members of society”?

Superior courts in Canada, the United States, the United Kingdom, Europe and elsewhere have long been puzzled by this question. As early as the 1950s, American scholars were warning that its resolution would “continue to perplex democratic statesmen”.

The need to balance an impartial public service with the democratic rights of government employees, and the public interest in their political participation, dates back in the Anglo context to at least 1854. The Northcote-Trevelyan report of that year planted the initial seeds of the Whitehall model of bureaucracy, a model that would become central to the public administration of Commonwealth countries and persist to this day.

At the core of the Northcote-Trevelyan paradigm of a professional public service, free from patronage and ineptitude, was impartiality. A neutral bureaucratic class chosen on merit would provide stability as political fortunes ebbed and flowed; in the words of one early observer, “[t]he commander may be often changed … but the crew remains the same, equally prompt to obey every varying order” (Hearn 1867, 238–9).

This principled vision soon found explication in the public service regulation of Australian colonies. Just two years after the Northcote-Trevelyan report, Victorian government employees were warned off “political partisanship” (Plehwe 1983, 365). The first regulations governing Commonwealth public servants following federation, promulgated in 1902, provided that government employees could not “publicly discuss or in any way promote political movements”.

While the scope of this rule was eventually eased, it remains the case today that employees of the Australian Public Service (APS) – and, to varying extents, their state, territory and local government counterparts – are expected to uphold the political neutrality of the bureaucracy. The Australian Public Service Commission (APSC) has warned public servants that even liking anti-government posts on social media could breach the APS Code of Conduct. Several years ago, the Department of Prime Minister and Cabinet urged staff to “dob in” colleagues breaching these obligations.

Although similar limitations on political engagement have been the subject of extensive litigation in other jurisdictions, it was not until the early 2000s that the constitutional validity of these regulations was challenged in Australia. In the first such case, Bennett v President, Human Rights and Equal Opportunity Commission (2003) 134 FCR 334, Finn J of the Federal Court struck down one restriction on the basis of the implied freedom of political communication: “Though I am mindful that [the regulation] has been a threatening presence … for over 100 years I am satisfied that it is invalid” (at [108]).

The spectre of invalidity was again raised in Starr v Department of Human Services [2013] FCA 1052, when the Fair Work Commission reinstated a Centrelink employee – dismissed for anonymous online comments – on free speech grounds. In a cognate context, Buchanan J of the Federal Court found in Gaynor v Chief of the Defence Force (No 3) (2015) 237 FCA 188 that the termination of an army reservist’s commission for homophobic comments was ultra vires for contravening the implied freedom. But the Australian Defence Force (ADF) was successful on appeal: Chief of Defence Force v Gaynor (2017) 246 FCR 298, with the Full Federal Court emphasising that the implied freedom is not a personal right. While the reservist’s application for special leave to the High Court was denied, the ADF’s counsel admitted in oral argument that “there are some interesting issues here just over the horizon”, relating to “social media and members of the public service”. When Comcare v Banerji was removed to the High Court in 2018, the horizon came into view.


In 2011, Department of Immigration and Border Protection employee Michaela Banerji joined Twitter, under the pseudonym “LaLegale”. Banerji proceeded to tweet extensively, often critical of her Department’s policies regarding refugees. Her tweets were sometimes expressed in vitriolic language. With the possible exception of a lone tweet, her social media activity was undertaken in Banerji’s personal time, on her own personal communication devices, and disclosed no confidential information.

In 2012, a colleague came to suspect that Banerji was responsible for the @LaLegale Twitter handle and made a complaint. Under the Public Service Act 1999 (Cth), Banerji was required to “at all times behave in a way that upholds the APS Values” (s 13(11)). Among the APS Values was a declaration that “the APS is apolitical, performing its functions in an impartial and professional manner” (s 10(1)). These sections – collectively the APS Code of Conduct – have subsequently been amended, but the changes are not material.

Following an investigation, the Department determined that Banerji had breached her obligations under the Code of Conduct. She commenced proceedings in the Federal Circuit Court, seeking an injunction to prevent the termination of her employment. In Banerji v Bowles [2013] FCA 1052, Judge Neville refused the application: “[t]he unbridled [free speech] right championed … does not exist” (at [101]). Banerji’s employment was terminated.

She then lodged a workers’ compensation application with Comcare. Her claim was denied on the ground that the termination constituted reasonable administrative action taken in a reasonable manner, pursuant to the Safety, Rehabilitation and Compensation Act 1988 (Cth). After the decision was affirmed by an internal reviewer, Banerji appealed to the Administrative Appeals Tribunal (AAT).

In Re Banerji and Comcare (Compensation) [2018] AATA 892, the AAT found for Banerji on the basis that her termination was contrary to the implied freedom. The AAT focused on the anonymous nature of her tweets, holding that the interest in an impartial public service which might otherwise justify restrictions fell away when there was no connection between the comments and the public service. To the contrary, the AAT mused, “restrictions in such circumstances bear a discomforting resemblance to George Orwell’s thoughtcrime” (at [116]). Comcare appealed to the Federal Court, and the Commonwealth Attorney-General removed the dispute to the High Court.

Comcare v Banerji

On 7 August, the High Court delivered its first opinion on balancing the competing policy concerns that have troubled other courts globally in this context. Across a majority joint judgment (Kiefel CJ, Bell, Keane and Nettle JJ) and three concurring sets of sole-authored reasons, the entirety of the bench upheld the appeal.

Comcare had conceded in the AAT that the APS Code of Conduct burdened political communication. The joint judgment indicated that this “concession was rightly made” (at [29]). Justice Edelman described the burden as “deep and wide” (at [197]).

Once an effective burden is established, the relevant inquiry – following Lange v Australian Broadcasting Corporation (1997) 189 CLR 520 and its progeny – is to ask:

  1. Does the impugned law have a legitimate purpose consistent with the constitutionally-mandated system of democratic government; and
  2. If so, is the law reasonably appropriate and adapted to the achievement of that purpose. On the view of a majority of the current High Court, that process involves a structured proportionality exercise with three stages: suitability, necessity and adequacy in balance.

The High Court unanimously held that the APS Code of Conduct met each requirement.

On the first, the joint majority held that there “can be no doubt” that the “maintenance and protection of an apolitical and professional public service is a “significant” and consistent purpose (at [31]). They did not elaborate on the difference between a “significant” and a “legitimate” purpose. Justice Gageler described the Public Service Act’s aims as “unquestionably legitimate” (at [100]) while Gordon J indicated that public service impartiality “is essential to upholding the constitutionally prescribed system of representative and responsible government” (at [155]). Justice Edelman made similar observations.

On the second question, the joint majority offered a relatively cursory analysis. They accepted that the APS Code of Conduct was “suitable in the necessary sense” and found no “obvious and compelling alternative” (at [34]–[35]). After highlighting the limitations on the imposition of sanctions on a public servant – the requirements for procedural fairness, proportionality and internal, external and judicial review options – they concluded that the provisions “present as a plainly reasoned and focussed response” to the pursuit of the legitimate purpose. As such, they held that the law “trespasses no further upon the implied freedom than is reasonably justified” (at [42]).

The three concurrences came to the same result, albeit via varying methods. Justice Gageler expressed the need for “close scrutiny” in cases such as this of the correspondence between the “direct and substantial burden” and the “compelling justification” (at [97]). Nevertheless, his Honour was satisfied that this inquiry was met: Gageler J found that the APS Code of Conduct was “reasonably appropriate and adapted to achieve the identified object of establishing an apolitical public service in a manner that involves minimal impairment of freedom of political communication” (at [102]).

For Gordon J, in light of the “tailored and limited” scope of the law’s application and that the provisions are not “self-executing” but rather entail a transparent, “just and appropriate” sanction process, “the closeness of the means to the legitimate end” meant that these factors “conclude the issue” (at [157]–[159]).

Initial Observations

There will, no doubt, be much rigorous analysis of Comcare v Banerji in the coming months. As an initial contribution to that dialogue, I offer the following six observations.

1. An icy wind

“[P]roperly interpreted,” admitted Edelman J, the APS Code of Conduct “casts a powerful chill over political communication” (at [164]). Much the same might be said of the impact of Comcare v Banerji.

Several of the Judges were keen to highlight that the Public Service Act does not equate to a blanket ban on political speech by public servants. Justices Gageler, Gordon and Edelman each, to greater or lesser extents, explored the factors that would influence the acceptability of that speech, including seniority, the relation between work area and subject matter, the tone of the comment (“from vitriolic criticism to objective and informative policy” (Edelman J at [183])) and breadth of dissemination.

As such, Edelman J argued, the APS Code of Conduct “does not impose behavioural obligations that preclude a public servant from making political comment on social media.” Instead, it:

creates a boundary, albeit ill-defined, between acceptable expression of political opinions and unacceptable expression of political opinions. Taking into account that a public servant is intended to be able to take part in their political community, that boundary will only be crossed when comments sufficiently imperil the trust between, on the one hand, the APS and, on the other, Parliament, the executive government, or the public. An assessment of when that trust will be sufficiently imperilled will depend upon all the circumstances (at [182]).

This is nice in theory. But what rational public servant would wander anywhere near that “ill-defined” line? The practical consequence is a significant chilling effect on political discourse – why would a government employee risk sanction when the determination of the acceptability of speech will involve a complex factorial analysis? The safer option is to say nothing at all. The American First Amendment chilling effect doctrine anticipates this negative effect and seeks to partially mitigate it through various procedural advantages for plaintiffs, to avoid the possibility that “people will self-censor”. While the doctrine may not be apposite in the Australian implied freedom context, it does amply illustrate the consequences of the High Court’s approach.

2. Constraining executive power

An issue left unaddressed in Comcare v Banerji was the interplay between the implied freedom and executive power. Banerji and the Australian Human Rights Commission (intervening) had sought to agitate this point, but the Court adopted the approach outlined by Brennan J in Miller v TCN Channel Nine (1986) 161 CLR 556 and utilised in the present context in Wotton v Queensland (2012) 246 CLR 1.

That approach provides that where a burden on political communication has its source in statute (the Public Service Act), the constitutional analysis is undertaken solely at that level. Any particular application of the statute (such as Banerji’s termination) raises questions of statutory power, not distinct constitutional questions. As Edelman J explained, “[t]he constitutional constraint does not operate directly upon the exercise of executive power. It invalidates the executive act only by operating upon the legislation, disapplying the legislative authority for the executive act” (at [209]).

 It remains to be seen how the implied freedom would operate on “pure” executive power exercised in the absence of statutory authorisation, or inform administrative (rather than constitutional) review of decisions made under enactment. The joint majority mused: “So to conclude does not mean that the implied freedom may not be a relevant consideration in the exercise of different discretions under other legislation” (at [45]). On the other hand, Gageler J quoted Basten JA’s comments in A v Independent Commission Against Corruption (2014) 88 NSWLR 240 that characterising the implied freedom as a mandatory consideration introduces “an element of conceptual confusion”. There were several other relevant tea leaves in the judgment, but until the issue squarely arises the proper understanding of the interaction will remain unclear.

3. Criticism, without more, is not a reputational hazard

One of the more alarming claims from the bench in Comcare v Banerji came in response to Banerji’s contention (accepted by the AAT) that the anonymous nature of her tweets meant the persuasiveness of the relevant provision’s neutrality purpose fell away. The joint majority observed:

where the employee broadcasts tweets commenting on policies and programmes of the employee’s Department or which are critical of the Department’s administration, damage to the good reputation of the APS is apt to occur even if the author’s identity and employment are never discovered (at [24]).

There are good reasons why, in the joint judgment’s words, Banerji’s attempt to draw a bright line between attributable and anonymous comment might be considered “facile”.

Justice Gageler explained with more nuanced analysis:

Confidence cannot exist without trust, and trust cannot exist without assurance that partisan political positions incapable of being communicated with attribution will not be communicated anyhow under the cloak of anonymity. That is so irrespective of whether a particular comment made anonymously might or might not end up being attributed to its maker (at [105]).

That may be so. But the suggestion that the public service’s reputation is “apt” to be damaged by criticism that is unattributable to a public servant is troubling. If I, as a non-public servant, engage in criticism of a government department, I am exercising a democratic freedom. One might accept that a public servant does not enjoy the same freedom because their comments, in light of the nexus with the public service, could imperil community or government confidence in the public service.

But criticism is not problematic per se – the risk that justifies limitations on public servant political expression only arises where that nexus exists. Acknowledging the potential that anonymity might be dissolved is one thing. Suggesting that even absent that dissolution the reputation impact exists is another entirely.

There is a not insignificant irony in the fact that it was the Department of Immigration itself that made public the identity of LaLegale as Banerji, a public servant, only to then argue (via Comcare) that the reputational damage (real or potential) of the tweets in light of her public service role justified termination.

4. Substance or semantics?

In the Gaynor appeal, the Full Federal Court seized on the “rights-based analysis” they said Buchanan J had erroneously adopted (at [55]). This was despite the fact that his Honour had initially stressed that the implied freedom is a freedom and not an individual right.

The High Court adopted the same approach in Comcare v Banerji. The joint judgment highlighted that the AAT had approached the issue, “wrongly, as if the implied freedom of political communication were a personal right” (at [19]). The joint majority leapt on the AAT’s language concerning the “serious impingement on Ms Banerji’s implied freedom” (emphasis in Comcare v Banerji) as illustrative of the misstep. This was despite the fact that, like Buchanan J in Gaynor, the AAT had emphasised the right/freedom distinction.

Much has been written about the right/freedom dichotomy in the present context (perhaps most prominently by Professor Adrienne Stone). However, the distinction remains blurry. It may be that the objections of the Full Federal Court and the High Court in Gaynor and Comcare v Banerji respectively are astute identifications of a substantive analytical failing. But the risk is that this claim of misconception provides a convenient excuse to substitute an alternative value judgment, without a robust engagement with the underlying reasons for that initial position. In other words, one judge’s (implied) freedom might be another’s right.

5. You’re fired

Comcare v Banerji raised interesting issues surrounding the relevance of the severity of sanction to the assessment of the statutory burden on the implied freedom and its adequacy in balance. That is a complex topic for another day. But it is worth highlighting two points.

First, the High Court hung much hope on the unfair dismissal regime in the Fair Work Act 2009 (Cth) to ameliorate the risk of disproportionate penalties for breaches of the APS Code of Conduct. The joint judgment, for example, highlighted that “a penalty of dismissal must not be ‘harsh, unjust or unreasonable’” (at [40]). This gloss misstates the law. It is not unlawful per se for an employer to dismiss someone harshly, unjustly or unreasonably. The right that exists for terminated employees is to apply to the Fair Work Commission for an unfair dismissal remedy (which is not automatically granted even where the termination is harsh, unjust or unreasonable) – and that right does not extend to senior public servants, nor those with less than six months’ employment.

Secondly, several comments in the decision seem to downplay the significance of the termination of a public servant’s employment. Justice Gageler, for example, writes “non-observance of the requirement [of neutrality] can lead only to administrative action, the most extreme outcome of which is that the person ceases to be an APS employee” (at [103]).

That is quite the extreme outcome, and it is a shame that only Edelman J confronted this reality. His Honour observed: “A person’s employment can be fundamental to him or her. The person’s entire life might be built around it. The consequences of a loss of employment, particularly as a disciplinary penalty, could be catastrophic” (at [195]).

6. Neutrality cuts both ways

In intervening submissions in Comcare v Banerji, the Solicitor-General for Western Australia struck on an important point that has largely been overlooked. The Public Service Act, he submitted, requires public servants to “not promote or criticise” (emphasis added). Justice Gageler also emphasised that neutrality has dual dimensions when he referred to “praise for or criticism of” government policy in his reasons (at [105]).

It is striking that each of the cases referenced above has involved criticism of government policy or other negative comments. Indeed, the APSC’s official guidance provides that, while “[c]riticising the work, or the administration, of your agency is almost always going to be seen as a breach of the Code [of Conduct]”, this “doesn’t stop you making a positive comment on social media about your agency.”

If Banerji had vociferously defended the government’s border protection policy, even making “vituperative” comments about critics, would she have ended up in the High Court? It seems not. Yet, particularly on an issue as polarising as refugee policy, neutrality must cut both ways. Other than Gageler J’s passing comments, the bench did not seem alert to the double standard infecting the public service’s approach to impartiality.

Concluding Remarks

It is unlikely Comcare v Banerji will be the last time the political expression of public servants is litigated. Certainly the American experience, where such issues have troubled the Supreme Court of the United States repeatedly over the past half-century, foreshadows that this will not be the last word.

The central failing of the High Court’s first foray into this area is a failure to grapple with complexity – or at the very least identify the tensions that arise and explicitly defer the resolution thereof to the legislature on the basis of the implied freedom’s limitations. Instead, the Court has effectively given Parliament and the executive a blank cheque, holding that the legitimate purpose pursued by the Public Service Act justifies sweeping intrusions into the private lives of public servants, whatever the adverse consequences.

Comcare v Banerji might not have been the optimal test case to explore these issues – the Court would perhaps have been more concerned by the limitations had Banerji tweeted on climate change policy, in moderate language. But Comcare v Banerji will make it very difficult for a public servant whose case sits closer to that “ill-defined” boundary between “acceptable” and “unacceptable” expressions of political opinion (Edelman J at [182]) to prosecute a constitutional claim.

That is a pity. Comcare v Banerji compares poorly to more nuanced jurisprudence on these issues in other jurisdictions. Instead, it comes closer to the antiquated view of Wendell Holmes Jr. J in the late 1800s: “The petitioner may have a constitutional right to talk politics, but he has no constitutional right to be a policeman” (McAuliffe v Mayor of New Bedford, 29 NE 517, 517 (Mass. 1892)).

Kieran Pender is a visiting fellow of the Centre for International and Public Law at The Australian National University. He acknowledges the helpful comments of Associate Professor Heather Roberts, and the blog’s editors, on an earlier draft.

Suggested Citation: Kieran Pender, ‘“A powerful chill”? Comcare v Banerji [2019] HCA 23 and the political expression of public servants’ on AUSPUBLAW (28 August 2019) <“a-powerful-chill”?-comcare-v-banerji-[2019]-hca-23/(opens in a new tab)>