Start with what we knew. For years, the 'harm or endanger persons' category — the bucket police use for offences like endangering life, reckless acts, and threats to safety — moved in one direction. Up. Not every quarter, but the trend line pointed higher. More incidents. More reports. More people put at risk.

Then something shifted. The latest data shows 33 recorded offences in this category. The period before? 45 offences. That's a 26.7% drop (Source: NZ Police, ANZSOC Tab_Full Data_data).

To understand why that matters, you need the context. This isn't shoplifting or car theft — categories where numbers bounce around based on reporting rates and police priorities. This is harm. Endangerment. The kind of offending that puts someone in hospital or makes them fear for their safety. It doesn't usually fall by a quarter in a single reporting period.

So what happened? The timeline doesn't give us a press release explanation. It gives us a pattern. Over the past few years, harm offences climbed during periods of social stress — lockdowns, economic anxiety, housing instability. When people were crammed into bad situations with no escape valves, the numbers reflected it. More volatile households. More confrontations. More incidents that crossed the line into criminal offending.

But the latest period tells a different story. The drop isn't marginal. It's steep. It suggests something in the environment changed — whether that's better intervention, fewer flashpoints, or people finding ways out of dangerous situations before they escalate.

Here's what the timeline shows: harm offences don't drop by accident. They drop when pressure eases. When housing becomes slightly more stable. When families aren't as financially desperate. When support services have capacity to step in before things explode.

This data won't make headlines the way a crime spike does. A 27% drop doesn't feed the narrative that everything's getting worse. But if you're one of the 12 people who didn't become a victim this period because that trend reversed, the data isn't abstract. It's the difference between safe and endangered.

The question now is whether this holds. One period of data is a datapoint. Two is a pattern. Three is a trend. If harm offences stay at this lower level — or keep falling — it's evidence that something structural improved. If they bounce back up, it tells us the factors driving harm are still there, just temporarily dormant.

Either way, the timeline matters. Because harm doesn't happen in a vacuum. It happens when systems fail, when pressure builds, when people run out of options. And when it drops by a quarter, that's worth watching.