Moshe Milevsky on Tontines

I’m writing from a great conference at the Stanford Center on Longevity put together by Steve Vernon and William Sharpe. As always, Moshe Milevsky provided a great and thought provoking presentation. It’s about tontines and their similarities and differences with traditional income annuities (single-premium immediate annuities). I must add that this is my interpretation of the presentation and any errors are my own. 

He began the presentation discussing all of the groups who have basically come to the conclusion that they don’t like income annuities, despite their theoretical advantages as a retirement income tool. These groups include: 

• Consumers: payout rates are too low relative to subjective yield expectations (i.e. people think they can do better investing assets on their own). Consumers also end up killing the mortality credits component by adding self-defeating features like period certainty, cash refunds, etc. These features just leave them with an expensive bond portfolio.
• Advisors: Income annuities offer low commissions relative to other annuities and also result in a loss of assets under management and loss of the need to provide further advice to clients.
• Employer plan sponsors: fiduciary concerns, default risk of insurer possibly leaving the employer on the hook, newness, regulatory matters, etc.
• Media and press: confuse income annuities with other expensive and complex annuities
• Solvency II: capital costs for systematic longevity risk are increasing in Europe, which will reduce annuity payments by 20% there after this happens. More generally, income annuity payouts are reduced to hedge systematic longevity risk (which is that, on average, people start living longer than expected)

How is a tontine different from an immediate annuity?

In a basic way, they are similar. A single premium is added to a pool with other annuitant contributions. These assets earn interest and payouts are made contingent on remaining alive. 

The difference is that with tontines, the payouts are not necessarily fixed. In the most basic form, the interest earned on the pooled assets is divided among those still alive. As members of the pool die, income to survivors increases. The big difference is that there is no need for an actuary, or for holding reserves against systematic longevity risk, etc. [Though there is a need for someone to put together the pools of anonymous tontine subscribers sharing similar mortality characteristics and to keep track of who remains alive and to distribute the payments. This will cost something].

With tontines, the participants are the ones taking systematic longevity risk (payouts will be less as more of the pool stays alive than planned) rather than the insurance company. This allows contributed dollars to go further in providing income to the pool of participants. With income annuities, systematic longevity risk stays with the insurance provider and reserves must be set aside, resulting in a lower payout.

A fair life annuity is preferable to a classic tontine for those seeking constant spending, because the classic annuity has a constant payout pattern. The payout pattern for the tontine is less ideal because the actual pattern of realized mortality will result in some uncertainty about the annual income provided by the tontine. No matter how it is designed, the tontine leaves some residual risk about what payouts are going to be. 

But after accounting for price loading with the income annuity in terms of capital and reserve costs, the tontine (with some longevity risk sharing regarding income received) will be preferable.  At some point the charges for the conservatism of the income annuity will lead people to prefer the tontine.  With a larger tontine pool, the tontine more easily becomes attractive. 

Prof. Milevsky provided a historical example of King William’s Tontine of 1693. Subscribers could choose between an income annuity with a 14% payout rate, and a tontine in which the underlying pool of assets would be credited with 10% interest per year for years, and then 7% thereafter. Actual income to tontine owners would be higher, then, as based on the number of remaining survivors in the pool. 

To choose among these (or whether to participate at all), factors to consider include subjective health status, the health and characteristics of the tontine pool, sponsor default risk for the income annuity, the term structure of interest rates, and longevity risk aversion (how much you care about possibility of what happens if live a very long time).

At the time, interest rates were about 6%. Edmund Halley (of Halley’s Comet fame), analyzed the choice and modernized the idea about pricing annuities in terms of mortality and interest rates. He thought that the income annuity was a better deal. About 70% of participants chose the life annuity and 30% chose the tontine.

Prof. Milevsky concluded by asking some questions. Would tontines be an attractive option for employer defined contribution plans? Would they be attractive to retirees as part of the menu of retirement income tools? Could adding a tontine choice possibly increase the appeal of a life annuity by giving them an opportunity to think more about life contingent payments? What regulatory and legal burdens would need to be overcome, such as the matter that tontine insurance (though not tontine schemes) are currently illegal? 

On a fun note, Prof. Milevsky also mentioned the Bogleheads in his presentation as a population of sophisticated consumers who might be the first to jump on the tontine bandwagon as an income tool in the coming years.