In November 2023, a database administrator named Barry Young walked out of Health New Zealand with a slice of the country’s vaccination records and handed them to the tech entrepreneur Steve Kirsch. Within weeks Kirsch was at a lectern at MIT, telling an audience the data proved roughly one death for every thousand doses, a number he then scaled up to thirteen million dead worldwide. It was a spectacular figure built on almost nothing: four million doses out of twelve million with no proof they were a random sample, an “excess deaths” baseline set arbitrarily three weeks after each shot, and no unvaccinated group to compare anyone against. Strip those props away and the thirteen million evaporates.
We have seen this move before, and it is worth naming the move rather than just the man who made it. Excess mortality is unusually easy to weaponize, because “excess deaths” is not a body count. It is the gap between how many people died and how many a statistical model says should have died, and whoever builds the model builds the answer. Feed it a confident story and it hands you a confident number. That is what Kirsch did in 2023. It is also worth holding onto when you read the study now being used to bury him, because that study leans on the very same kind of modelling, and both sides of this fight are asking it to carry more weight than it can.
The number that is supposed to settle it
In June 2025 the International Journal of Epidemiology published a full accounting of New Zealand’s pandemic dead by Michael Plank, Pubudu Senanayake, and Richard Lyon. The year-by-year figures are the heart of it, and they do genuine damage to the crude vaccine-timing theory. In 2020, New Zealand recorded 2,276 fewer deaths than expected. In 2021, the year it put more than ten million doses into a population of five million, it again ran under baseline, 654 deaths short of expected. The excess did not show up until 2022, with 2,735 more deaths than expected, and 2023, with 1,235 more.
| Year | Deaths vs expected |
|---|---|
| 2020 | 2,276 below |
| 2021 | 654 below |
| 2022 | 2,735 above |
| 2023 | 1,235 above |
Here is the problem for the “the jab did it” crowd, stated plainly because the evidence supports it plainly. The shots went into arms in 2021, when deaths were below normal, and the dying spiked a year later, when the country abandoned its elimination strategy and Omicron finally got in. The bulk of the rollout ran from early 2021 into early 2022; the elevated mortality ran from March 2022 through 2023, a stretch during which comparatively few first and second doses were given. You cannot make a poison arrive twelve months before the bodies. Skeptics who staked everything on the 2022 curve picked the wrong hill, and the honest thing is to say so.
That is where the modellers and the fact-checkers want the story to end. It shouldn’t.
What the study cannot do
Look at the cumulative line, the one the press releases skated past. Add the four years together and New Zealand’s total pandemic excess comes to 1,040 deaths, with a 95% confidence interval running from 1,134 fewer deaths than expected to 2,927 more. Read that range again. It crosses zero. The authors’ own arithmetic cannot rule out that New Zealand, across the entire pandemic, suffered no net excess deaths at all, or even finished slightly ahead of where it would otherwise have been. This is the result being waved around as proof that the vaccine question is closed, and it is statistically compatible with nothing having happened.
Then there is how the excess gets pinned on the virus. The paper reports that COVID-attributed deaths accounted for 102 percent of the 2022 excess and 82 percent of 2023, which reads like a clean match until you ask what “COVID-attributed” actually captured in the Omicron era. A positive test near the time of death, and died-with folded into died-of, are the known soft spots of pandemic death coding everywhere, and an all-cause model cannot resolve them one way or the other. Matching your excess-death curve to your COVID-death curve is not independent confirmation when both are drawn from the same ledger. It is the model nodding along with itself.
Now notice what a national all-cause mortality model in a country of five million simply cannot see. With a confidence band already swinging by thousands of deaths, it has no hope of detecting a rare adverse event buried in that noise. The documented safety signal from the mRNA products was never a mass-casualty wave. It was myocarditis in young men, documented well enough that the FDA ordered it onto the warning labels and, in 2025, expanded the language to spell out the risk in 16-to-25-year-old males. A study whose error bars are wider than any such signal could ever be will, of course, come back empty. “We saw no vaccine mortality wave” and “the vaccines carry no safety cost” are different sentences, and the second does not follow from the first.
The modellers are grading their own pandemic
The last thing worth saying is who wrote this. Michael Plank is not a neutral referee who wandered in after the fact. He was one of the country’s most visible pandemic modellers, co-lead of Covid-19 Modelling Aotearoa, the group that supplied the government with the projections that fed New Zealand’s pandemic response. Now that same modelling world has come back to tally the toll, and it has concluded that the toll was among the lowest on earth and almost entirely the virus’s doing. The paper declares no funding and no competing interests, and I take that at face value. But a competing interest is not only a check that clears. It is a structural stake in the answer, and people who modelled the response have an obvious one in finding that the response worked and cost few lives.
None of this means Kirsch was right. He wasn’t, and TrialSite, a publication no one accuses of pro-pharma sympathies, put the honest question in its own headline: does the timing tell us anything about vaccination? On the crude version of the theory, no. But the public-health establishment has taken a narrow, uncertainty-soaked result and laundered it into a blanket exoneration, the same way it laundered “flatten the curve” into two years of mandates and “safe and effective” into a decree that ended debate. That reflex, the leap from a modest finding to a closed case, is the exact habit that torched public trust the first time. A confidence interval that runs through zero is not a verdict. It is a confession of how little these numbers can carry.
The thing that would actually settle this is not modelled, it is counted. Cause-specific mortality, the granular by-cause records, is what would show whether anything besides COVID moved in 2022, and this study never opens that ledger; it is an all-cause estimate by construction. So put the by-cause data on the table. The actuaries and statisticians who work in this exact space, including one of this paper’s own authors, are the people who could force it into daylight, and they are the ones to watch. Until Stats NZ’s by-cause breakdown is out where everyone can read it, both camps, Kirsch and the modellers alike, are reading tea leaves in a five-million-person cup. The honest position is not that the vaccines killed New Zealanders. It is that the people telling you the matter is closed are working from a number they know cannot bear the weight, and they are betting you won’t check the interval.
Sources
- International Journal of Epidemiology – Plank, Senanayake & Lyon, excess mortality in Aotearoa New Zealand (2025)
- PMC full text – Estimating excess mortality during the COVID-19 pandemic in New Zealand
- FDA – Required updated warning labeling for mRNA COVID-19 vaccines regarding myocarditis and pericarditis
- Covid-19 Modelling Aotearoa – about the modelling group co-led by Michael Plank
- FactCheck.org – Flawed analysis of New Zealand data doesn’t show COVID-19 vaccines killed millions (2023)
- AAP FactCheck – NZ data does not show a spike in vaccine-related deaths
- TrialSite News – New Zealand excess-death study finds 2022 spike
- arXiv – preprint version of the New Zealand excess-mortality study