What the law was for

THE FINDINGA well-intentioned reform that made itself impossible to judge

The question was whether the intent behind worker-protection legislation matches its outcome. The answer has three parts, and the middle one is the surprise: for most of this legislation there is no outcome measure, because nobody ever set one.

01 · THE INTENTWhat the authors said they were doing

Four instruments, four stated purposes. All are on the record, and none is about siding with anybody.

Psychosocial WHS regulation — clarify duties and shift the narrative

Boland's Recommendation 2 asked that the model Regulations "deal with how to identify the psychosocial risks associated with psychological injury and the appropriate control measures to manage those risks." Her own account of what it would achieve is cultural, not statistical:

Recommendation 2 is "the recommendation that I see as being critical to supporting greater emphasis on the culture of organisations — on increasing the focus on a systematic safety culture and shifting the narrative of discussions about inappropriate behaviours from being an individual's problem to an organisation's challenge."

Marie Boland, on the model WHS laws review, June 2019 High

The instruments themselves are more modest still. The Commonwealth Explanatory Statement says the amendments "will ensure that physical and psychosocial risks are managed using the same framework" and will "assist duty holders to meet their obligations." Queensland's Explanatory Notes are blunter: the regulation "will provide a principles-based approach that raises the profile and awareness of existing duties," achieving its objective "by providing clarity and certainty for PCBUs about existing obligations." High

Read carefully, this is a declaratory reform. On its own account it created no new substantive duty — it made an existing one legible. That framing matters when judging the outcome: a law whose stated objective is "awareness" has succeeded when people are aware, regardless of whether anyone is safer.

The Respect@Work positive duty — move the burden off the victim

This is the clearest prevention intent in the whole set, and the diagnosis is worth quoting at length because it is the same diagnosis this project's first report reached from the customer's side:

"Overwhelmingly, the Commission heard that the current system for addressing workplace sexual harassment in Australia is complex and confusing for victims and employers to understand and navigate. It also places a heavy burden on individuals to make a complaint."

"Throughout the Inquiry, the Commission heard of the need to shift from the current reactive, complaints-based approach, to one which requires positive actions from employers and a focus on prevention."

Respect@Work: Sexual Harassment National Inquiry Report, AHRC, March 2020, pp. 14–15 High

Recommendation 17 proposed the positive duty; 18 and 19 proposed AHRC assessment, compliance-notice and systemic-inquiry powers. The baseline was measured precisely — 33% prevalence, 53% among Aboriginal and Torres Strait Islander workers, $3.8 billion cost in 2018, fewer than one in five seeking support. No target was attached to any of it. High

Harmonisation — uniformity, lower burden, fewer injuries

The 2008 COAG Inter-Governmental Agreement stated four objectives: "uniform, equitable and effective safety standards"; addressing "the compliance and regulatory burdens for employers with operations in more than one jurisdiction"; "efficiencies for governments"; and "significant and continual reductions in the incidence of death, injury and disease." Jurisdictions were bound to implement "no later than December 2011." The 2011 Decision RIS, modelled by Deloitte Access Economics, predicted about $250 million per annum in economy-wide net benefit, roughly $80 million of it to multi-state businesses — while conceding a net cost of $3.27 per worker per annum to single-state and small firms, and declining to quantify any safety improvement. High

Frontline offence laws — deterrence, then measurement

NSW's 2023 retail-worker assault offences were sold on deterrence. Minister Sophie Cotsis: "These changes will provide a deterrent for those who would behave in this manner." Attorney-General Michael Daley: the new offences "send a strong message that this behaviour is unacceptable." High

The UK Home Office was markedly more candid. Its economic note for the Crime and Policing Bill's retail-crime measures states the aim is to show offenders that abuse "will not be tolerated and will be punished, which the government believes — but there is no definitive evidence to suggest — may have a deterrent effect." It sets a breakeven at "between 380 and 475 retail assaults a year" prevented, reports a negative net present social value driven by prison costs and "the lack of monetised benefits," and monetises no benefits at all. The achievable objective it actually claims is measurement: ensuring assaults are "separately recorded, so we know the true scale of the problem." High

02 · THE MEASUREMENT GAPThe one thing that was never legislated was a number

Across the entire Australian chain, exactly one document produced a genuine cost-benefit analysis of psychosocial regulation — and its own government then discarded the options it justified.

Where a quantified predicted outcome should be, and what is there instead

Document	Date	Quantified predicted outcome
Boland Review, Recommendation 2	Dec 2018	None. Predicted effect is a cultural shift in narrative
Safe Work Australia Consultation RIS	Jun 2019	None. "provided no quantified figures regarding benefits or costs associated with either option" Med
Safe Work Australia Decision RIS	Dec 2019	None locatable. Document body not retrievable Unresolved
Commonwealth Explanatory Statement F2023L00012	Dec 2022	None. "The Office of Impact Analysis has advised that a Regulation Impact Statement is not required"
Queensland Explanatory Notes SL 2022 No. 141	2022	Borrowed generic figures only — $11bn annual cost, $2.30 return per $1 spent. Neither is a prediction about this regulation
Australian WHS Strategy 2023–2033	Feb 2023	Six numeric targets, none for psychological injury. Psychosocial risk gets an activity target: Members will "build the capability of PCBUs, regulators, and workers"
Respect at Work Act 2022 — Explanatory Memorandum	Sep 2022	A cost, not a benefit. $226.4m average annual regulatory cost. "The Bill will not have a financial impact on the Commonwealth"
Victoria — Regulatory Impact Statement	Mar 2022	The exception. Four options costed; preferred option net benefit +$132.10m, BCR 1.06. Breakeven requires improvement "in approximately 24 per cent of Victorian workplaces"

Victoria's RIS deserves credit and a caveat. It is the only serious attempt at quantification in the wave, and it was candid about its own limits — the analysis "should be regarded as a test of whether … the benefits are likely to exceed the costs, rather than a point estimate," acknowledging "an evolving state of knowledge in relation to mental health and the establishment of appropriate workplace controls." Even so, it never predicted a number of claims avoided.

03 · SCORECARDPromise by promise

Only where a promise was specific enough to score. Everything else appears as "no measure set," which is itself the most common result.

Stated promise against measured outcome

Promise	Source & year	Outcome	Verdict
≥20% reduction in worker fatalities from injury	AWHS Strategy 2012–2022	Baseline 268 (2007–2010 avg) → 188 (2020–2022 avg), about −30%	Met
≥30% reduction in the serious workers' compensation claims rate	AWHS Strategy 2012–2022	Never officially closed out. Safe Work Australia's own Key Statistics 2024 shows claim volumes up 20.5% and the frequency rate up 0.8% across 2012-13 to 2022-23; one secondary source reports −24% on the incidence-rate measure. The two readings use different metrics and cannot be reconciled from the public record	Missed
≥30% reduction in musculoskeletal claims rate	AWHS Strategy 2012–2022	No closing measurement published by Safe Work Australia	No result
Reduce psychological injury	The entire psychosocial wave, 2021–2025	No target was ever set. Serious mental-health claims 11,700 → 17,600 across 2021-22 to 2023-24; +161% over the decade; median compensation $67,400 vs $16,300 for all serious claims	No measure
Uniform WHS laws in all jurisdictions by December 2011	COAG IGA 2008	WA adopted a modified Act commencing 31 March 2022 — ten years late. Victoria never adopted and still operates the OHS Act 2004. Queensland excluded mining. Safe Work Australia's own September 2025 review paper concedes jurisdictions "have increasingly made variations within their own versions of the model laws, which poses a real challenge to the maintenance of harmonisation"	Missed
~$250m per annum net benefit from harmonisation	Decision RIS 2011	No verification of delivery exists in the public record. One 2021 study found a 0.9 percentage-point reduction in claims in harmonised states	Unverified
Shift from a reactive complaints model to employer-led prevention	Respect@Work 2020 / positive duty from Dec 2022	AHRC enforcement powers commenced 12 Dec 2023. In 2023–24 no enforcement counts were published. In 2024–25 — the first full year — the AHRC "launched 4 formal inquiries and engaged with 35 other employers in a voluntary capacity," covering roughly 7,500 workers. No compliance notices, enforceable undertakings or court applications reported. And no sixth national prevalence survey has been published since Time for respect in November 2022 — the single most direct test of the reform does not exist	Early
Industrial manslaughter (Boland Rec 23b)	Boland 2018	Rejected for the model Act by WHS Ministers in May 2021, then added to the model Act in July 2023; legislated in all nine jurisdictions by September 2024	Delivered late
Deterrence of assaults on retail workers	NSW 2023 Act	Ministerial releases report 192 individuals charged with 242 offences and 41 custodial sentences since July 2023. BOCSAR publishes no retail-worker assault series and no evaluation exists, so deterrence is untested	Untested
Deterrence and measurement of retail-worker assault	Scotland 2021 Act	Recorded serious assaults of retail workers 15 (2022-23) → 13 → 11. But the Scottish Government states the Act "led to changes in how crimes were recorded but did not change the volume of crime being recorded." Five years on, no evaluation has been commissioned by anyone	Unmeasurable
Standalone offence protecting retail workers	UK, campaigned from 2020	Aggravating factor only (2022) → standalone offence defeated in committee Jan 2024 → Bill fell at dissolution May 2024 → enacted as s.45 Crime and Policing Act 2026 on 29 April 2026 → still not in force as at July 2026, omitted from the first commencement regulations. Covers assault only, not threats or abuse as Scotland's does	Not in force
Federal US healthcare workplace-violence standard	OSHA, petition granted Jan 2017	Request for Information Dec 2016; SBREFA panel report May 2023; proposed rule targeted June 2025 and missed; moved to Long-Term Action in the Spring 2025 agenda. Nine and a half years, no proposed rule ever published. Enforcement remains via the General Duty Clause	Abandoned

04 · ENFORCEMENTA duty on paper, thinly tested in court

Every concluded psychosocial conviction in Australia, as far as the public record shows

Defendant	Sentenced	Penalty	Hazard
Court Services Victoria (Coroners Court)	19 Oct 2023	$379,157	Traumatic material, workload, role conflict, bullying — conduct 2015–2018, one worker suicide. Prosecuted under the general OHS duty, two years before Victoria's psychosocial regulations existed
Marist Youth Care Ltd	11 Mar 2024	$300,000	Risk of sexual violence and assault by clients
State of NSW (Corrective Services)	20 Dec 2024	$600,000	Two custodial officers ambushed by inmates
Department of Defence	19 Dec 2025	$188,000	RAAF technician suicide — first Commonwealth psychosocial conviction; maximum available was $500,000

The pattern matters more than the count. The one prosecution that squarely tested an organisational psychosocial failure — Court Services Victoria — was brought under a general duty that had existed since 2004, which means it did not need the new regulations. The two cases that did follow them concerned third-party violence, also long prosecutable. And the two attempts to prosecute novel organisational psychosocial failures both collapsed: SafeWork NSW withdrew against Western Sydney Local Health District after three weeks of hearing and twenty witnesses, following a successful objection to its expert evidence.

Activity is rising, though. SafeWork NSW conducted 228 psychosocial checks in a single three-day statewide blitz in October 2025, appointed 20 dedicated psychosocial inspectors in March 2026, grew its inspectorate from 370 to 469, and stood up a Psychosocial Advisory Service. Its 2026–27 regulatory statement prioritises "higher-order controls" and "preventing violence and aggression toward frontline and customer service workers." The inputs are real. Medium But note what the October 2025 blitz release did: it itemised non-compliance notices for falls (209) and mobile plant (175), and gave no psychosocial notice count at all.

05 · DID WORKERS GET SAFER?The prevalence line has not moved

Nine years, two regulatory reforms, no change in verbal abuse

Retail and fast-food workers reporting abuse in the previous 12 months, SDA No One Deserves a Serve survey waves

Verbal abuse Physical violence

The SDA's own reports say it plainly. The 2021 wave: "the nature and prevalence of abuse and violence by customers has not improved much since our last survey at the end of 2016." The 2023 wave: "the prevalence of abuse and violence by customers has not improved since the previous surveys." Verbal abuse: 88% (2016) → 88% (2021) → 87% (2023) → 88% (2025). Reported physical violence roughly doubled between 2023 and 2025, from 12.5% to 25%. Workers who do not feel safe at work: 26% (2023) → 34% (2025). Workers abused repeatedly by the same customer: 38.5% → 52% → 59%. High

The UK, running a parallel reform arc, shows the same. Usdaw's Freedom From Fear series: verbally abused 68% (2019) → 77% (2024) → 78% (2025); threatened 43% → 53% → 54%; assaulted 5% → 10% → 11%. The 2025 report's own framing: "the growing levels of abuse and violence recorded in 2019 have continued to increase year on year since the pandemic … they remain higher than before the pandemic — and are growing." High

And the hardest-edged Australian number, from official hospital data rather than a union survey: assault hospitalisations of health care workers rose from 49 in 2015–16 to 126 in 2024–25 — 4.3 to 8.1 per 100,000 — moving assault from the fourth to the second leading cause of injury hospitalisation in that workforce. Medium-High

06 · DILUTIONWhat fell out between the draft and the instrument

• Victoria dropped both distinctive obligations. Written prevention plans for five prescribed hazards, and six-monthly reporting of aggression, bullying and sexual harassment complaints by employers with 50+ staff — both excluded from the 2025 regulations by ministerial decision, replaced by an optional template. High
• Victoria did keep the strongest control rule in the country. Where a combination of controls is used, information, instruction and training must not be the predominant measure, and may be used alone only if nothing else is reasonably practicable. This is more prescriptive than any other Australian jurisdiction and directly targets the substitution problem in §8. Credit where due. High
• The "harmonised" duty was not uniform at birth. The Commonwealth extended the model hierarchy-of-controls provision to psychosocial risks, expressly so that "physical and psychosocial risks are managed using the same framework." NSW and Tasmania excluded it. Same recommendation, opposite implementations, in the same year. Medium — NSW's position after its 2025 regulation remake is genuinely unclear; secondary trackers conflict
• New York weakened its hard control. The Retail Worker Safety Act's panic-button mandate for employers with 500+ employees nationwide became a "silent response button" for employers with 500+ statewide, deferred to 1 January 2027 — a narrower population, a weaker device, and 28 months after signature. The policy-and-training obligation shipped on time. High
• The UK's flagship offence is on the statute book and not in force. Section 45 was omitted from the first commencement regulations (in force 29 June 2026), which commenced s.44 and then skipped to s.97. No commencement date has been announced. High

The important qualifier: no Australian jurisdiction has narrowed or delayed a psychosocial prevention duty. On the prevention side the duties have only expanded. The retreat happened somewhere else.

07 · REVERSALNSW narrowed the injury three years after regulating the hazard

This is the single sharpest intent-versus-outcome event in the whole domain. NSW inserted psychosocial duties into its WHS Regulation on 1 October 2022. Three and a bit years later it made the resulting injuries substantially harder to claim for — and it said why, openly.

Two Acts: the Workers Compensation Legislation Amendment Act 2025 (assented 24 November 2025) and the Workers Compensation Legislation Amendment (Reform and Modernisation) Act 2026 (assented 11 February 2026), with the main tranche commencing 1 July 2026.

What changed

• A "relevant event" gate. Primary psychological injury is compensable only if caused by a listed event — violence or threats, indictable criminal conduct, witnessing trauma, vicarious trauma narrowly defined, sexual or racial harassment or bullying, or "excessive work demands."
• "Excessive work demands" defined cumulatively: demands beyond the requirements of the role and repeated or persistent and not reasonable in all the circumstances. All three limbs. This excludes chronic-but-in-role overload — which is the most commonly reported psychosocial hazard in Australia.
• "Reasonable management action" expanded to expressly include performance appraisal and feedback, counselling, suspension or stand-down, disciplinary action, demotion, transfer, training, investigation of alleged misconduct, and dismissal.
• 130-week cap on weekly payments for primary psychological injury, with continuation requiring whole person impairment above 25%.
• Lump sum and common law threshold rising from 15% WPI to above 25% (1 July 2026), 26% (2027) and 28% (2029).
• Medical treatment test tightened from "reasonably necessary" to "reasonable and necessary" — with SIRA guidance on the new test not due until October 2026, months after the test began to bite.

Why, in the Government's own words

"Without the reform, employers were facing average premium increases of 36 per cent over the next three years."

"The average cost of these claims has jumped $100,000 in five years to almost $300,000."

NSW Government ministerial releases, 2025–2026 High

The cost pressure was real: psychological claims are 12% of NSW claims and 38% of scheme cost; average claim cost rose from $146,000 (2019–20) to $288,542 (2024–25); the Nominal Insurer held 85 cents in assets per dollar owed; 40% of psychological-injury workers return to work within a year against 88% for physical injuries.

What the parliamentary inquiry found instead

The Legislative Council's Public Accountability and Works Committee reported on 3 November 2025 and located the cost driver somewhere other than eligibility. Large employers account for 47% of the workforce, 58% of claims, but only 49% of claim costs; small employers 25% of the workforce, 15% of claims, and 27% of costs. The committee's inference was that what matters is how claims are handled. It recommended withdrawing the impairment-threshold increase, a full claims-management review, and capping insurer investigation expenses at 50% of claim cost. The Law Society warned that the then-proposed 31% threshold "will conceivably exclude nearly all workers with psychological injury from making a claim," requiring proof of inability to "live alone, care for dependants, or to function in society." Unions NSW called the package "callous and draconian." High

In fairness: the 31% proposal was defeated and the final landing is materially lower; premiums are frozen to 30 June 2028; a stop-bullying jurisdiction and an intensive return-to-work program were added. This is calibration as well as retreat. But the shape is unmistakable — the identified problem was claim cost, the evidence located the driver in claims handling, and the legislated remedy operated on eligibility.

08 · SUBSTITUTIONWhat employers actually did with the duty

The best available Australian dataset is the Australian HR Institute's 2025 report, surveying 609 senior decision-makers in October 2024. It is the analytic core of this report.

Psychosocial measures Australian organisations reported implementing (AHRI, n=609)

Measure	%	What kind of control this is
Flexible working options	42	Administrative / individual
Work–life balance initiatives	39	Individual
Activities and regular conversations with workers	36	Administrative
Training in resilience and stress management	32	Individual, worker-side
Employee Assistance Program	29	Individual, post-hoc
Leadership and management capability investment	28	Administrative
Stress management programs	25	Individual
Staff surveys and focus groups	25	Assessment
Regular psychosocial risk audits	20	Assessment — the lowest-reported measure

Not one reported measure is a structural control. No staffing-level change, no workload redesign, no queue management, no barriers, no job redesign, no expansion of worker discretion to resolve a customer's problem appears anywhere in the reported list. The top responses are worker-adaptation measures — flexibility, resilience training, employee assistance, stress management. Meanwhile the most-reported cause of claims and complaints in the same survey is job demands, at 30% — precisely the hazard resilience training cannot control. High

Two further findings from the same data sharpen it. 38% of organisations reported psychosocial claims increasing over the previous twelve months against 13% reporting a decrease — and yet 80% rated themselves as managing psychosocial hazards "very" or "somewhat" effectively, with only 7% rating themselves poorly. That confidence-versus-outcome gap is what compliance-by-documentation feels like from the inside.

Industry guidance reproduces the same substitution. The Australian Customer Experience Professionals Association's guidance on customer abuse recommends backup support, security precautions, zero-tolerance posters, post-incident support, and "soft skills and stress management" training — with no reference to WHS or psychosocial duties at all, and no queue-management, staffing or job-design measure. High

Queensland's Code says the opposite of what is happening, clearly: control measures "should predominantly be considered at an organisational, work and system design level … rather than at an individual level," and administrative controls "are less reliable at minimising risk because they rely on human behaviour." Victoria's regulations go further and make training-as-predominant-control unlawful. No evidence was found that either rule is being enforced, and no regulator statement criticising training-only compliance was located. The rule exists; the practice runs the other way; nothing in the public record connects them. Medium — absence of evidence

09 · THE ABSENT PARTYThe law regulates a two-party interaction and analysed one side

This is where this report meets the first one. Nothing in the legislative intent concerns customers or complainants — and that is not a criticism of the drafters' motives. It is the finding.

The impact statements do not mention them

Searched document by document for customers, clients, patients, students, complainants, consumers, service users, third parties and the public:

• Commonwealth Explanatory Statement F2023L00012 — clean negative. No mention of customers, clients, patients, students, complainants or service users. The only non-worker reference is definitional: reg 55B's "a worker or other person." Impact analysis is confined to duty holders and workers. High
• Queensland Explanatory Notes SL 2022 No. 141 — clean negative. None of those terms appears in consultation or impact assessment. Consultation comprised "worker representatives, industry representatives, employer representatives, relevant government agencies and Commissions, and technical experts and academics." No service-user or consumer body was consulted. High
• Victoria's 2022 RIS and Safe Work Australia's Decision RIS — unverified, not negative. Neither document body was retrievable. Notably, Victoria's RIS is no longer publicly available four years after publication, which is its own small comment on evaluability. Gap

Where a regulator did engage, it prescribed service refusal

The AHRC's Guidelines for Complying with the Positive Duty (August 2023) is the exception, and it cuts the other way. It correctly recognises customers as perpetrators — "more than one third of sexual harassment incidents in the fast-food and retail industries were perpetrated by a customer (36%)" — and then, at page 43, offers this as a practical compliance example under the Culture standard:

"Workers (including leaders and managers) refuse service or entry to customers or clients who engage in relevant unlawful conduct."

AHRC, Guidelines for Complying with the Positive Duty under the Sex Discrimination Act 1984 (Cth), August 2023, p.43 High

The Guidelines do not address the consequences of refusing service for the person refused. It is presented as a unidirectional control.

Queensland's Code contains the contradiction on its own pages

Verified against the primary PDF, three passages in the same statutory Code of Practice:

Page	What it says	Function
p.21	"Service methods that cause or escalate frustration, anger, misunderstanding or conflict (e.g. long waiting times, denying someone service, an absence of queue management leading to queue jumping)"	Listed as a risk factor to be avoided
p.27	"refusal of service to customers with higher risk of violence or abuse"	Listed as an elimination control — top of the hierarchy
p.32	"implementing procedures to restrict or terminate service in the event of aggression"	Listed as a combined control

The same Code names denial of service both as an aggression-escalating hazard and as a top-of-hierarchy control. This is the cleanest documentary artefact of the second-order effect in Australian regulatory material. It also resolves an item flagged as unverified in the first report: the "restrict or terminate service" wording is genuine and is in the Code. High

And the mechanism travels downstream, citing WHS as its authority

• NSW Ombudsman's UCC manual lists as an objective: "Comply with work health and safety and duty of care obligations by identifying the potential risks posed by unreasonable conduct to staff health, safety and wellbeing." Restrictions on access to staff, premises and services follow in Part 5. To its credit the manual cautions that special care is needed where disability or disorder may affect behaviour.
• Tasmania's model council policy imposes restrictions "in accordance with our duty of care and work health and safety responsibilities under the Tasmanian Work Health and Safety Act 2012," extending to prohibiting premises access and "terminating provision of services altogether."
• AFSA's model policy: conduct is unreasonable where "it unreasonably compromises the health, safety and security of our staff"; the agency "has a zero tolerance policy," managed "under our duty of care and work health and safety responsibilities."
• NSW Health PD2015_001 cites the WHS Act as the legal basis for a zero-tolerance approach, providing for conditional treatment agreements and, as longer-term responses, "conditional visiting rights or exclusion from the premises."

None of this was legislated. It is a chain of inference — WHS duty, therefore staff protection, therefore restrict the service user — built downstream of instruments whose impact statements never considered the service user at all. High

10 · DOES ANY OF IT WORK?The evidence on the two most-adopted controls

Employers overwhelmingly chose training and zero-tolerance policies. Both have been evaluated, and neither performs.

De-escalation training does not reduce incidents

"individual staff training programmes, such as in de-escalation, do not reduce violent incidents" — though there is "some evidence that de-escalation training can help staff to manage patient aggression" once it occurs. "very few rigorous evaluation studies … include a robust comparison and counterfactual."

Leach, Gloinson, Sutherland & Whitmore, Reviewing the Evidence Base for De-escalation Training: A Rapid Evidence Assessment, RAND Europe for NHS Improvement, 2019 High

Zero tolerance has never been shown to reduce aggression — and has documented side effects

The strongest retrievable finding is from UK general practice. Paniagua, Bond and Sallah (2009) record that the zero-tolerance campaign was endorsed "despite any evidence of the effectiveness of this approach"; that a 70% rise in recorded violent incidents followed implementation; and that "staff training programmes, which emphasise de-escalation strategies, have not proven to be effective and can even increase violent incidences." Their unintended-consequence finding is directly on point for this project: patients were removed from GP registers on "subjective behavioural interpretations," in decisions that were "unregulated and subject to ambiguity," producing care gaps that fell disproportionately on already socially excluded groups. High on the source; UK not Australia

Wand and Coulson's Australian paper — "Zero tolerance: A policy in conflict with current opinion on aggression and violence management in health care" (2006) — carries the claim in its title, but its abstract was paywalled and it should be cited as a position paper rather than as empirical evidence. And on the broader intervention literature, the Cochrane review by Gillen et al. (2017, five studies, 4,116 participants) concludes there is "very low quality evidence that organisational and individual interventions may prevent bullying behaviours in the workplace," calling for "large well-designed controlled trials." High

11 · THE FAIR VERDICTWhat this is, and what it is not

What genuinely worked

• Fatalities. The 2012–2022 Strategy's target of a ≥20% reduction was met and exceeded, from a baseline of 268 to about 188 — roughly a 30% fall. This is the domain's clearest success and it deserves to be said first.
• Industrial manslaughter. Rejected for the model Act in May 2021, added in July 2023, legislated in all nine jurisdictions by September 2024. Slow, but delivered.
• Some harmonisation dividend. A 2021 study found a 0.9 percentage-point reduction in workers' compensation claims in harmonised states, stronger in construction.
• Victoria's control-hierarchy rule is a genuine, well-targeted response to the substitution problem — the only one in the country.
• Enforcement capacity is being built, visibly, in NSW: dedicated psychosocial inspectors, an advisory service, and a stated 2026–27 priority on higher-order controls and frontline aggression.

What this report is not saying

• Not that the intent was bad faith. Every stated objective in this domain is defensible and evidence-linked. The Respect@Work diagnosis in particular is one of the better pieces of Australian public policy analysis of the last decade.
• Not that it is too early to judge, in the way that is usually meant. It is too early for claims data to turn. It is not too early to observe that no target was set, no leading indicator was baselined, the one reporting mechanism that would have produced a dataset was deleted, and the one national strategy that sets numbers set none for this hazard.
• Not that regulating psychosocial hazards was a mistake. Customer aggression against workers is real and, on official Victorian and AIHW measures, rising. The duty is warranted. The critique is of instrument design and measurement, not of purpose.
• Not that employers are villains. Given a duty with no specified outcome, a hierarchy of controls whose top tiers cost money and whose bottom tier costs a training licence, and no enforcement signal distinguishing the two, documentation is the rational response. The incentive was built that way.

12 · CONFIDENCECorrections, conflicts and gaps

Corrections — including to the first report

• Industrial manslaughter WAS adopted. It was rejected for the model Act by WHS Ministers in May 2021, then added in July 2023, and legislated in all nine jurisdictions by September 2024. Any framing of Boland Recommendation 23b as "never implemented" is wrong; the correct finding is a roughly four-and-a-half-year delay. High
• The Queensland Code's "restrict or terminate service" wording is genuine. The first report flagged this as an unresolved source conflict. It is verified against the primary PDF: p.27 lists "refusal of service to customers with higher risk of violence or abuse" as an elimination control and p.32 lists "implementing procedures to restrict or terminate service in the event of aggression." Both sit in the same document as the p.21 risk factor. High
• Safe Work Australia's "action on all of the 34 recommendations" is not accurate as plain English. Recommendation 23b was expressly rejected for the model Act in 2021 — a decision that failed the required two-thirds majority, with NSW, South Australia, Tasmania and the Commonwealth opposed.
• The SDA quote "has not materially changed since 2016" could not be verified and should not be attributed. The verified equivalents are the SDA's own report language: prevalence "has not improved much since our last survey at the end of 2016" (2021) and "has not improved since the previous surveys" (2023).
• The "475 psychosocial checks and 41 notices" figure could not be verified in any SafeWork NSW primary document and should not be used. The verified figures are 228 psychosocial checks in the October 2025 blitz and roughly 2,600 psychosocial-related requests for service across 2019-20 to 2021-22.

Unresolved conflicts

• The 2012–2022 serious-claims target result. One secondary source reports −24% against a −30% target; Safe Work Australia's own Key Statistics 2024 shows claim volumes +20.5% and the frequency rate +0.8% over 2012-13 to 2022-23. Different metrics, different baselines, and Safe Work Australia never published a closing measurement. Report the target as missed and the magnitude as contested.
• Whether NSW now applies the hierarchy of controls to psychosocial risks. It did not in 2022; secondary trackers disagree about the position after the 2025 regulation remake. Do not assert either way without the primary text.
• NSW's 130-week exception threshold is stated as 25% WPI by SIRA and 21% by one major law firm. Treat SIRA as authoritative.

Documents that could not be retrieved

• A Roadmap for Respect (April 2021) — so the exact response category applied to Recommendation 17 rests on the Parliamentary Library and the Government's own 2022 consultation paper. Use "deferred for further consideration on duplication grounds," not "rejected."
• Victoria's 2022 Regulatory Impact Statement — no longer publicly available. The §9 negative on customers is therefore based on two of five documents fully searched; the other three are unverified, not negative.
• Safe Work Australia's Decision RIS body; the Boland final report; the Scottish Bill's Policy and Financial Memoranda; California SB 553's official analyses; and the Safety Science policy evaluation "Australia's national laws for worker psychological health," which is the highest-value unread document in this research and needs institutional access.
• The Dreyfus second reading speech was not retrieved verbatim. The "reactive, complaints-based → preventative" formulation is verbatim from the AHRC's 2020 report and should be attributed there, not to the Attorney-General.

Worth commissioning

1. An Australian evaluation of zero-tolerance policy effectiveness. None exists after more than two decades of operation.
2. A published pre- and post-2022 psychosocial safety climate series. The leading indicator the field nominated has no before-and-after reading.
3. The sixth national workplace sexual harassment survey. Without it the positive duty cannot be evaluated at all.
4. Whether any regulator has ever enforced the "training must not be the predominant control" rule.

SOURCESPrimary references