New Zealand employers are expected to manage psychosocial risk. But the tool most organisations use to find hazards was never built to see them
Try a simple experiment. Ask your incident register how much bullying your organisation has. For most, the answer will be a handful of entries, maybe none.
Now ask your people. When WorkSafe surveyed 3,612 New Zealand workers using a validated psychosocial questionnaire, 23 per cent reported being bullied in the previous year. For Māori workers it was 28 per cent. More than a third had experienced at least one form of offensive behaviour.
Both numbers describe the same workplaces. One of them is wrong, and it is not the survey.
That gap gets read as a resourcing problem or a will problem. I think it is an instrument problem. Incident reporting is not the right tool for detecting psychosocial risk. It does not sample harm. It samples a particular kind of event, and the empty column is a blind spot, not safety.
Three hurdles before harm becomes a report
It has to be an event. A report needs a moment: observable, immediate, traceable to a cause. Paul Slovic's risk perception research found hazards divide sharply on exactly these qualities. Some are visible and immediate. Others are unobservable, delayed and hard to attribute. Burnout has no onset date, so there is nothing to file, and the person carrying it rarely connects it to a workload set months earlier.
It has to be safe to say. Reporting tracks psychological safety, the shared belief that speaking up will not be punished. Studies of accident reporting show under-reporting climbs steeply where the climate is poor. That suppresses physical data too, but it bites hardest here.
Something has to happen next. The employee voice literature calls the alternative organisational silence: people stay quiet not only from fear but from futility. In one large survey of nurses who did report violence, 40 per cent said no action followed. A report that changes nothing is a risk with no return.
Nobody reports the hazard to the hazard
A large share of psychosocial risk comes from how people are managed. Workload is set by someone. So are role clarity, the pace of change, and how a team is spoken to. To file a report about those things is to record that the hazard is your manager, in a system that usually sits inside the reporting line the complaint is about.
Research on workplace bullying finds most targets never formally complain, and the explanation is not confusion about the process. It is that the power of the perpetrator or the organisation is perceived to be too great. American task force data on harassment found roughly 70 per cent of those affected never told a manager at all. That is structural, not cultural, and it will not be solved by encouraging people to speak up.
Reported is not the same as actioned
There is a second filter. Consider two risks a community-facing worker carries: a vehicle crash and an assault. For the sake of argument, both cause minor injuries, both are visible, and both land reliably in the data. They part company on what happens next. Each gets investigated and closed out, but the assault is far more likely to trigger a security review, a change to how the work is done, a paper to the board. Slovic found demand for action rises when a risk feels imposed rather than chosen, and dreaded rather than familiar. An assault is both. The crash gets closed as an individual event, reframed as "just driving", chosen and familiar.
Most chronic psychosocial hazards sit where the crash sits, or further down. The workload does not feel imposed. No single duty was visibly breached. And they are common enough to read as normal rather than as harm.
So a hazard can fail at either filter. Some never enter the data. Some enter and are tolerated. Chronic psychosocial risk fails both. An incident system samples hazards by how detectable they are and how much they feel like they warrant action, and neither of those tracks actual harm.
Our compensation system makes the same assumption. ACC covers work-related mental injury only where it flows from a single sudden event, a physical injury or sexual assault, so harm that accumulates has no claimable moment.
Measure exposure, not incidents
The alternative is not difficult or costly. It is what WorkSafe itself did. The New Zealand Psychosocial Survey asked workers directly about their exposure to demands, control, support and offensive behaviour. That is how we know bullying sits near one in four.
The same information is available to any organisation, and it changes the question from what happened to what is it like to work here. Run a validated psychosocial assessment on a cycle: anonymous, analysed at group level so nobody has to name anyone to be counted, through a channel outside the management line it measures. Treat the output as hazard data, not engagement scores. Sustained high workload in one team is an exposure reading, and it goes on the risk register with an owner, a control and a review date. Keep the incident register for assaults, traumatic exposures and near misses.
This is not an engagement or wellbeing survey. The same finding in a staff feedback report is received as an HR update. On the risk register it is received as risk. Asking people to describe harm and then doing nothing visible teaches, very efficiently, that reporting has no return.
The things that damage people most at work rarely arrive as an incident. They build slowly, until they are just how things are. If your only instrument counts incidents, you will keep finding silence and calling it safety. You will not find these hazards by waiting for something to happen.
Sean Versteegh is a registered psychologist and co-CEO of 3 Big Things
