Professional longevity research has turned squarely to climate - including work on climate-mortality channels in longevity bulletins and recent research on wildfire smoke mortality. The life side of climate risk has been under-modelled relative to the property side; this article maps what belongs in the models.
Climate change mortality modelling has lagged its property-catastrophe sibling by a full professional generation. P&C insurers have priced physical climate peril for decades; life insurers and pension schemes have mostly treated mortality improvement as a socioeconomic and medical story in which the physical environment barely features. That is no longer tenable, and the profession's own research agenda - longevity bulletins addressing climate, mortality studies of wildfire smoke, ageing-population workstreams - says so.
The modelling question is not "does climate affect mortality?" - it plainly does - but which channels are material, at which ages, over which horizons, and how they should enter models built around cohort improvement rates.
The direct channels
Heat. Excess mortality during heat episodes is the best-evidenced direct channel, concentrated at older ages and in cardiovascular and respiratory causes - precisely the ages and causes that dominate annuity and pension exposure. Two features matter for modelling: the effect is event-shaped (mortality spikes with episodes, rather than drifting smoothly), and it is strongly modified by adaptation - housing stock, cooling prevalence, urban form, care-home practice. A UK book and a southern-European book with identical age structures carry different heat sensitivities.
Cold, and the asymmetry trap. Warming reduces cold-related mortality, and in cool countries dying of cold currently outnumbers dying of heat. Some analyses therefore net the two and conclude near-neutrality. The trap is that the netting is portfolio-specific and horizon-specific: the cold benefit accrues broadly and gradually, while the heat cost arrives in episodes, escalates non-linearly with temperature, and concentrates in vulnerable subgroups. Net-zero-effect conclusions computed at population level can be materially wrong for a specific annuity book.
Air quality and wildfire smoke. Recent mortality research has sharpened this channel: wildfire smoke exposure carries mortality effects at long range from the fires themselves, meaning the exposed population is far larger than the evacuated one. As fire seasons lengthen, smoke becomes a recurring respiratory-mortality load on regions that will never see a flame - a channel property books ignore and life books have barely begun to price.
The indirect channels
Indirect effects are slower, harder to attribute and plausibly larger over annuity horizons: vector-borne disease ranges shifting; food price and nutrition effects; mental-health and substance-related mortality following displacement and economic disruption - an area where mortality research has been increasingly active; and healthcare-system strain during compound events. None of these fits an event-loss framework; all of them fit the framework life actuaries already have - as modifiers of the trend in mortality improvement, varying by cause, age and socioeconomic group.
How this should enter longevity models
Decompose by cause and age, then reassemble. Climate channels act on specific causes at specific ages. Improvement models operating on all-cause aggregate rates smear the signal; cause-of-death-informed analysis, for all its known difficulties, is the natural resolution at which climate enters.
Treat climate as a scenario overlay on improvements, not a new base table. The honest representation of uncertainty here is scenario-conditional improvement paths - warming pathways mapped to cause-specific mortality adjustments - sitting alongside the existing socioeconomic and medical scenarios, with the same discipline I have argued for elsewhere on separating observed from projected components.
Respect the tails and the asymmetry. Episode-driven mortality belongs in stochastic mortality models as jump or excess components, not absorbed into smooth trend. For annuity books the direction of danger reverses - the risk is lighter mortality if adaptation and cold-benefit dominate - which is why climate belongs in longevity risk assessments in both directions, not only as a mortality-cost story.
Socioeconomic differentiation is not optional. Every direct channel - heat exposure, housing quality, cooling access, smoke exposure, care quality - is socioeconomically graded. Books concentrated in particular socioeconomic segments (as most pension schemes are) will experience climate-mortality effects very differently from population averages. This is the life-side analogue of the geographic fairness questions I have written about in pricing: the physical effects and the demographic gradients arrive together.
And date-stamp the science. Climate-health research is moving quickly; assumptions should carry vintages and review triggers, exactly as vendor hazard models do on the property side.
The practical starting point
For most life offices and schemes the first deliverable is not a new model. It is a channel-by-channel materiality assessment against the book's actual age, geography and socioeconomic mix, an explicit statement of which channels are in and out of current improvement assumptions, and one stress - an episodic heat-plus-smoke scenario appropriate to the book's geography - run through to liability impact. That is a quarter's work, and it converts climate-mortality from a literature review into a governed assumption set.
The property side learned to model climate because losses forced it to. The life side has the rarer opportunity to build the models slightly ahead of the experience - and longevity risk transfer pricing will reward whoever does.
Key Takeaways
- Climate-mortality channels are real, evidenced and under-modelled: heat episodes, cold offset, wildfire smoke at range, and slower indirect effects on improvement trends.
- Heat-versus-cold netting is portfolio- and horizon-specific; population-level near-neutrality can be materially wrong for a given annuity book.
- Climate enters longevity models naturally as cause- and age-specific scenario overlays on improvement assumptions, with episodic components in stochastic models.
- Socioeconomic gradients in every channel mean scheme- and book-specific effects diverge from population averages - differentiation is mandatory, not refinement.
- Start with a channel materiality assessment against the book's mix plus one episodic stress to liability impact; date-stamp the underlying science.
Frequently Asked Questions
How does climate change affect mortality assumptions? Through direct channels - heat-episode excess mortality at older ages, reduced cold mortality, wildfire smoke and air-quality effects - and indirect ones including disease range shifts, nutrition, mental-health and healthcare-strain effects. These enter most naturally as cause- and age-specific adjustments to mortality improvement assumptions under climate scenarios.
Does climate change increase or decrease longevity risk? Both directions are live. Heat and smoke channels raise mortality (reducing annuity liabilities but hurting protection books), while cold-mortality reduction and adaptation can lighten mortality, increasing longevity risk for annuity writers and pension schemes. The net depends on the book's ages, geography and socioeconomic mix - which is why portfolio-specific assessment matters.
Should pension schemes include climate in longevity assumptions? Yes, at minimum as a documented materiality assessment and scenario overlay. Scheme demographics are socioeconomically concentrated, so population-average conclusions do not transfer; a channel-by-channel review against the scheme's actual membership, plus an episodic stress test, is the proportionate starting point.