This week, a dramatic dip in the number of victims of violent crime was revealed – a remarkable turnaround in just eight months that the government was quick to take credit for. But, as Alice Neville explains, crime data is far from clear-cut. 

In September last year, the government announced a concerning development: the number of violent crime victims in the 12 months to June 2024 had risen to 215,000. This put its reduced violent crime target – 165,000 victims by 2029 – officially “at risk”.

Fast forward seven months, and this week, a more pleasing development was revealed: the number of violent crime victims for the year to February 2025 had dropped to just 157,000.

Wow, 58,000 fewer victims in less than a year – that’s great, and the 2029 target has been surpassed four years early!

Well, kind of.

You don’t sound convinced?

A more reliable number was quietly released six weeks earlier, in late February: 191,000 victims, which was for the 12 months to October 2024. 

You’ve lost me.

Let’s start from the beginning. All these datapoints come from the New Zealand Crime and Victims Survey (NZCVS), an annual survey of around 8,000 people that provides a nationally representative picture of New Zealanders’ experiences of all crime, including crime that is not reported to police. The Ministry of Justice has been running it since 2018, and there have now been seven annual cycles, as it calls them.

How exactly does it work?

According to the official NZCVS methodology report, “victim screener questions are used to ask respondents if they have experienced any crime within the last 12 months of the interview date and subsequent victim forms are used to get detailed information on those incidents. Trained offence coders use the incident description to determine if an offence has occurred and the offence type.”

Hold on. If it’s an annual survey, why have you just thrown three quite different numbers at me that all came out in the last year?

Great question. Because to fit in with the current government’s corporate-inspired quarterly reporting requirements, in addition to its usual once-a-year “cycle” release, the NZCVS now publishes quarterly updates that give us the number of violent crime victims in the 12 months prior to each quarterly point.

Still confused, to be honest.

That’s understandable. Basically, data for the NZCVS is collected on a near-continuous basis, meaning at any point, stats boffins can provide an estimate for the 12 months to that point. But the quarterly datapoints are not as reliable as the annual ones, and the NZCVS team does not recommend using them to suggest there has been any statistically significant change. 

How come?

Because the survey is designed to be carried out across set annual periods, with the number of people interviewed in each region across the year proportional to the population of that region. While there are targets per quarter for the number of interviews to be carried out per region, “in reality, these are often not met for each quarter”, says the methodology report, so the quarterly datapoints might not be weighted correctly against the population. This makes the data more “volatile”, particularly when looking at regional-level data or other subgroups like ages. 

An increase or decrease between quarterly datapoints “does not necessarily mean that the difference does not reflect a true change in the population”,  according to the NZCVS methodology report. “However, when the difference is statistically significant, we are more confident the difference is ‘real’.”

Comparing the annual data points to the quarterly ones is not recommended, according to the report – something government ministers do all the time.

Hmm. So what was the latest annual datapoint, if those are so dang reliable? 

That one, for the 12 months to October 2024, was released on February 27. It was 191,000, up a bit on the 2023 figure – the baseline of 185,000 violent crime victims the government is using for its violent crime target. 

Hold on, isn’t late February when the National Party’s social media accounts were trumpeting about the first drop in violent crime since 2018? 

Yep. In what I can only presume was remarkably coincidental timing, two days prior to the release of the NZCVS Cycle 7 results, justice minister Paul Goldsmith and police minister Mark Mitchell announced some great news about a drop in violent crime. The stats came from a tweet.

Despite it being the most gold-standard, statistically sound, highly reliable, tick-all-the-boxes violent crime stat we have, the government did not put out a press release or post anything about the official NZCVS Cycle 7 results. 

But the number had increased?

Yes, though it was not an increase considered statistically significant. (In fact, there has not been a statistically significant change in this data since the survey began: the number of victims of violent crime has remained steady each year.)

Here are the annual violent crime victim totals for each cycle:

• Cycle 1 (2018): 177,075
• Cycle 2 (2019): 183,931
• Cycle 3 (2020): 178,789
• Cycle 4 (2021): 174,486
• Cycle 5 (2022): 162,403
• Cycle 6 (2023): 184,815
• Cycle 7 (2024): 190,806 

None of these year-on-year changes was considered statistically significant, even the Cycle 5 dip (the sample size was smaller and response rate lower that year because of the impact of Covid-19, which “increased the level of statistical uncertainty”, according to the official report of the findings). 

The quarterly releases, meanwhile, have looked like this (there was no December release):

• June 2024:  214,737
• August 2024: 201,230
• October 2024: 190,806 (the same as the annual release for 2024)
• February 2025: 157,000 (presumably a rounded number; the exact total doesn’t appear to have been released yet)

As you can see, these numbers are certainly decreasing, but only because that June 2024 spike was so damn high. 

Here is a graph (which quite possibly doesn’t adhere to the rules of good data journalism but please go easy on me, I tried.)

Looks OK to me, but what would I know? Back to this latest update, the 157,000 number they’ve just released. What did the government have to say about all this?

The justice and police ministers, Paul Goldsmith and Mark Mitchell respectively, put out a press release saying the data showed “Kiwis are becoming safer”. Goldsmith said the quarterly results for the other government targets weren’t being released until June, but “given the significance of these results we are releasing them early, as we believe they are robust and in the public interest”.

Robust, eh. Remind me, what did he say about the statistically more robust figure of 191,000, released just six weeks prior?

Not a damn thing.

Right. Did they say anything else?

They did. “This is very encouraging and shows our work to restore law and order is paying off,” reckoned Goldsmith, pointing to new gang laws, the reintroduced three strikes regime and sentencing reforms, and the scrapping of funding for section 27 reports. 

When the Herald’s Derek Cheng pointed out that three strikes and the sentencing reforms haven’t actually come into force yet, Goldsmith, channelling his inner Dennis Denuto, pointed to “the vibe”.

Mitchell, ignoring the caveats about the unreliability of regional-level data, lauded a decrease of 12,000 in Auckland and 5,000 in Canterbury, to which he credited the local police. 

Goldsmith did warn that the data was likely to “remain volatile”, and also pointed out it could be relating to crime that had occurred over a 24-month period (because the interviews are conducted over a 12-month period, and interviewees are asked about experiences of crime from the year up until the point they are interviewed), “so we will continue to see the results of Labour’s soft on crime approach filter through at points”.