Four young people were arrested in Dunedin this week after a series of unprovoked group assaults, as reported by RNZ. A 65-year-old man was ordered to appear in court over a hammer brandished in the Sylvia Park car park. Arrests are the part of the justice system we hear about. What happens in the months afterwards sits in a Stats NZ spreadsheet that almost nobody opens.

Open it, and one line stands out. For sexual assault and related offences, the remand count hit 1,029 in 2025, up from 654 the year before (Source: Stats NZ, remand-prisoners). That is 375 more people, a rise of 57% in twelve months.

To see how unusual that is, look at the rest of the series. It runs back to 1998, when the figure was 267. The previous high was 750 in 2023. Every year from 2001 to 2024 sat somewhere between 549 and 750: a slow-moving band, drifting up and down by tens, not hundreds. Then 2025 blew straight through the top of it.

Here is the comparison that stopped me. Between 1998 and 2024, across 26 years, the number grew by 387. The single jump from 2024 to 2025 was 375. One year did almost as much as the previous quarter-century.

Now the part that matters most, and the part that gets skipped whenever remand numbers are used as evidence of anything. These are people on remand. They have been charged. They have not been convicted. Remand counts are driven by three things at once: how many people police charge, whether judges grant bail, and how long courts take to get to trial. A rise can come from any of them. It is not a measurement of how much offending is happening in New Zealand, and anyone using it that way is either confused or hoping you are.

The strongest honest alternative reading of my own finding: a jump this size in one year is more likely to reflect how the system is processing cases than a sudden change in behaviour. If remand durations lengthen, the count rises even with the same number of people charged, because more of them are inside at any given moment. The series also has a suspicious early wobble, from 267 in 1998 to 450 in 1999, which hints that collection or definitions changed in the early years. And a single latest data point, in a series that has never moved like this before, deserves a second year of evidence before anyone builds a policy on it.

What I will say plainly is this. If you want to argue that New Zealand is getting more dangerous, remand data is the worst possible place to look, because it moves when courts get slower and bail gets tighter, not just when crime changes. And if you want to argue the opposite, it is no better. The honest reading of 1,029 is that far more people are being held in custody, unconvicted, on these charges than at any point since 1998, and that the reason sits inside the courts, not on the front page.

Worth watching what the 2026 figure does. A one-year spike that holds is a system change. A one-year spike that falls back was a queue.