Mortality pooling raises safe income, and almost nobody buys it
The claim
Pooling longevity risk lets a given pot support more lifetime income than any self-managed withdrawal rule can safely promise, yet voluntary annuitisation rates are a small fraction of what standard economic models predict.
Why we rate it moderate
The welfare result is a theorem under stated assumptions and survives most attempts to relax them. The empirical gap between theory and behaviour is well documented; the explanations for it are still argued over.
Yaari's result is the starting point: a consumer with no bequest motive and access to fairly priced annuities should annuitise everything. The intuition is that self-funding an unknown lifespan forces you to plan for the tail — you must underspend for forty years in case you live for forty years — while a pool only has to fund the average. The difference between those two is the mortality credit, and it is large at older ages.
Davidoff, Brown and Diamond showed the conclusion is much more robust than it first looks. Full annuitisation depends on strong assumptions, but partial annuitisation survives relaxing almost all of them: incomplete markets, bequest motives, and pricing loads reduce the optimal share without ever driving it to zero.
Observed behaviour looks nothing like that. Voluntary annuity purchase is rare almost everywhere it is not compulsory. Benartzi, Previtero and Thaler catalogue the reasons — the decision is framed as an irreversible gamble on death rather than as insurance, the payout is compared against a wealth number rather than an income need, and the option to keep the money is highly salient while the risk of outliving it is not.
The practical version is not all-or-nothing. Covering essential spending with life-contingent income — Social Security first, then a simple immediate annuity if there is a gap — converts the retirement problem from 'will this last' into 'how much discretionary spending do I get', which is a much easier problem to hold in your head and a much easier one to survive a bad decade in.
The products to be careful with are the complicated ones. A single-premium immediate annuity is a bond-like promise you can compare on price across insurers. Variable and indexed annuities with riders are opaque enough that comparison is difficult, which is usually a sign about who the complexity is serving.
Where this breaks down
- An annuity is only as good as the insurer. State guaranty association coverage is real but capped, and diversifying across insurers costs efficiency.
- A nominal annuity is destroyed by sustained inflation — the 1970s halved the real value of fixed pensions — and inflation-linked annuities are expensive and often unavailable.
- Annuitising forecloses bequests and liquidity. Both are genuine goods, and a household that values them is not making a mistake by holding back.
- Pricing depends on interest rates at the moment of purchase, which makes the timing of an irreversible decision uncomfortably consequential.
Sources
Follow these rather than taking our word for the summary.
Menahem E. Yaari (1965). Uncertain Lifetime, Life Insurance, and the Theory of the Consumer
The Review of Economic Studies, 32(2), 137-150
Finding: A consumer with no bequest motive facing an uncertain lifespan and actuarially fair annuities should hold their entire wealth in annuitised form.
Thomas Davidoff, Jeffrey R. Brown and Peter A. Diamond (2005). Annuities and Individual Welfare
American Economic Review, 95(5), 1573-1590
Finding: Substantial partial annuitisation remains optimal after relaxing Yaari's assumptions on complete markets, bequests and actuarially fair pricing.
Shlomo Benartzi, Alessandro Previtero and Richard H. Thaler (2011). Annuitization Puzzles
Journal of Economic Perspectives, 25(4), 143-164
Finding: Annuitisation rates respond strongly to how the choice is framed and defaulted, indicating the low take-up is not a considered preference.