Mark Delaney: Michael, tell me honestly — when you hear 'record-low death rates,' is your gut reaction good news or bad news?
Michael C. Vincent: Depends entirely on whose balance sheet you're reading.
Mark Delaney: Ha — yeah, okay, that's kind of the whole episode right there. Because 2025 England and Wales data just recorded the lowest all-age death rates ever, two to six percent below 2024, and the response from pension consultancies is not champagne — it's a longevity report with the word 'warning' in the subtext.
Michael C. Vincent: Longevity risk — the liability-side of a long life. If your members live three years longer than the model assumed, the money isn't there.
Mark Delaney: Right, and LCP — Lane Clark and Peacock — their 2026 report is literally titled 'Are life expectancies bouncing back?' Stuart McDonald, their longevity partner, is calling it a potential inflection point. And I want to put a slightly heretical take on the table: the biological fact of people living longer is almost the secondary story here.
Michael C. Vincent: Go on.
Mark Delaney: The primary story is that CMI_2024 — the Continuous Mortality Investigation's big June 2025 model revision — introduced a new half-life parameter to deal with pandemic distortion, and then CMI_2025 landed in March 2026 as the follow-up, and actuaries at Isio publicly said: we like the structure, we hate the CMI's own calibration choice on that parameter. Two credentialed teams, same framework, different life expectancy outputs. That's the thing that should make a trustee sweat — it's not the biology, it's who gets to turn the dial.
Michael C. Vincent: You see, that framing is sharper than the way most of this gets reported — and it's also the uncomfortable one, because the Institute and Faculty of Actuaries owns the CMI. This is the profession's own tool, disagreeing with itself.
Mark Delaney: Wait — but that's the part I want to push on, because 'the profession disagreeing with itself' makes it sound almost chaotic, and I don't think it's quite that, is it?
Michael C. Vincent: No, and that's exactly where I'd pump the brakes on my own framing. Think of it this way — the CMI model is like a satnav that every single driver on this particular road is legally required to use. Same map, same routing algorithm. But two co-drivers are arguing about which speed limit to trust on one stretch of road. The car still gets there. The question is whether it arrives in ten minutes or twelve.
Mark Delaney: Okay — that actually lands. So what's the speed limit they're fighting over?
Michael C. Vincent: The half-life parameter. That one dial controls how fast the model stops treating pandemic-era death data as noise and starts treating it as signal — as real long-run evidence about where mortality is heading. And Isio, in their July 2025 update, said publicly: we back the structural redesign in CMI_2024, we think the IFoA got the architecture right — but their Core Model's actual choice of value for that parameter? We disagree.
Mark Delaney: Huh — so they're not rejecting the model, they're rejecting one number inside it.
Michael C. Vincent: One number that produces materially different life expectancy outputs. And here's what makes that uncomfortable — the Institute and Faculty of Actuaries has owned this model for a hundred years. They published a centenary history in August 2025 to mark it. A century of being the profession's authoritative baseline. Public disagreement isn't just a technical footnote in that context. It's institutionally awkward in a way that doesn't happen often.
Mark Delaney: A hundred years — and now uh, the first real public break on calibration lands right when the underlying data is, I mean, the most ambiguous it's ever been because of the pandemic. That timing is not great.
Michael C. Vincent: Now that is the crux of it. Nearly every UK defined benefit pension scheme runs its liability calculations through this model. Same data going in. Depending on which half-life value your actuary selects — the CMI's or Isio's — your scheme's projected life expectancy shifts. And that shift is the difference between telling your CFO the funding gap is manageable and telling them it isn't.
Mark Delaney: So the science isn't broken — the judgment call inside the science is where the money lives.
Michael C. Vincent: And a real company just made that visible. HSBC UK Bank plc — their 2025 interim report, published figures: male life expectancy at 60 moved from 26.1 years to 26.4 years. December 2024 to June 2025. Six months. Three tenths of a year.
Mark Delaney: Wait — that's in a corporate filing? Like, disclosed to shareholders?
Michael C. Vincent: Interim report. And the reason that number moved is itself a judgment call buried inside it — HSBC weighted 2020–21 pandemic data at zero percent. 2022–23 at fifteen percent. Those aren't scientific constants. Someone chose those.
Mark Delaney: Wait, no — that's the point, isn't it. The calibration disagreement I was describing isn't theoretical anymore. It's already sitting in annual reports affecting what liabilities look like on a balance sheet.
Michael C. Vincent: That is where the hot take holds. And early 2026 mortality data — pensioner mortality specifically — is tracking back toward pre-pandemic trajectories. Which means the next CMI calibration could push life expectancy assumptions higher than either of the two preceding model releases.
Mark Delaney: Larger than the last two. Combined?
Michael C. Vincent: Larger than each of the two preceding releases individually, the way LCP reads the trajectory. And increased assumptions translate directly — not abstractly, directly — into higher scheme liabilities, different bulk annuity pricing, sponsor balance sheets taking a hit.
Mark Delaney: And what we haven't even gotten to yet — the part that I think makes this whole dilemma genuinely worse — is whether a trustee acts now on imperfect numbers or waits for the next CMI update, and those two bets don't have symmetric downside. That's uh, that's the thing.
Michael C. Vincent: Now that is the question that deserves its own moment.
Mark Delaney: Because if you wait — I mean, waiting sounds prudent, right, like you're being careful — but if the inflection point is real and you've sat on your hands, your scheme is underfunded. And that's not a modeling error anymore, that's a governance failure.
Michael C. Vincent: And the cost of being wrong is not symmetric. Act early, assume higher longevity — your sponsor funds more than necessary, for now. Wait, and the mortality inflection proves real — you're chasing a gap on behalf of people's actual retirements.
Mark Delaney: One of those mistakes is recoverable.
Michael C. Vincent: The other one isn't. Now — here's where it gets genuinely complicated, because the NHS angle cuts against the clean version of this. The BBC investigation found UK residents are living fewer years in good health. So longer raw lifespan — yes. But more of those years spent waiting for a hip replacement, or caught in a cancer backlog. LCP flags A&E wait times specifically as a headwind on healthy life expectancy.
Mark Delaney: Which means — wait, actually, does that complicate the liability picture or make it worse? Because a pension scheme pays out regardless of whether you're healthy.
Michael C. Vincent: It makes it morally complicated and financially unrelieved. The liability rises either way — the member's experience is just worse. Those aren't the same problem, and conflating them lets sponsors frame this purely as a balance-sheet crisis when it's also a health-system failure.
Mark Delaney: And then LCP throws GLP-1 drugs into this — the anti-obesity drugs — as a potential longevity upside, and uh, I dunno, I'm skeptical. We can't even agree on how to weight pandemic data and we're modeling Ozempic's population-level mortality impact?
Michael C. Vincent: The Pensions Policy Institute would say the problem runs deeper than any single model update. They've called on MPs to act urgently — the UK pension system, structurally, has not kept pace with a changing economy. And the IMF has flagged longevity risk as a systemic threat to global financial stability. This isn't one Thursday morning in a trustee meeting. It's the whole architecture.
Mark Delaney: So the calibrated take is: the model isn't broken, the data is genuinely moving, acting now and waiting are both bets — and at the system level, nobody's actually in charge of making sure the right bet gets made.
Michael C. Vincent: That's where I'd leave it too, actually. Nobody in charge of making sure the right bet gets made — that's the honest summary. And the part that stays with me is this: Stuart McDonald and the Isio team are looking at the same CMI framework, the same underlying data, and arriving at different conclusions on the half-life parameter. The 'standard' was never fully standard. It just looked that way until the numbers got hard.
Mark Delaney: Yeah — and I mean, that's the thing that I keep turning over. It's not a modelling crisis and it's not purely a demographic one either. It's that whoever sets that dial answers two questions at once: what does longevity look like, and who's on the hook when they get it wrong. And the scheme members never knew there was a dial to turn.
Michael C. Vincent: No, they didn't. And they still don't. That's an uncomfortable place to land, but it's the right one.
Mark Delaney: Uh — yeah. Good conversation, genuinely.
Michael C. Vincent: Likewise. I mean that.
