Hacker News Reader

Hacker News

Comment by PaulHoule | Hacker News Reader

Full thread

PaulHoule

It's a general problem in retirement that if you live a long time you could outlive your savings. Some answers are

https://en.wikipedia.org/wiki/Life_annuity

and

https://en.wikipedia.org/wiki/Tontine

View on HN

The OECD Pensions team has recently confirmed that in this current environment, tontines offer better value for consumers than annuities.

Sounds like an arbitrage opportunity - shouldn't they pay out roughly the same?

Insurers make a double profit by charging management fees and by mispricing the longevity risk which essentially means that those dying earlier generate excess profits for the insurer.

In a tontine, those dying earlier generate excess income for the retirees. It's explained here: https://tontine.com/explainer

I used to price annuities (but no longer have any affiliation or financial interest in the industry) and this isn’t true IME - annuities are typically priced using best-estimate mortality rates. Outside of immediate annuities, which are a pretty small chunk of the business, mortality rates just wouldn’t be a useful lever to increase profits, since the predominant decrement is generally lapse/surrender, not death. Annuities are profitable because insurers price to high target IRRs or equivalent metrics, build in enough fees or spread to achieve that IRR, and build in enough management levers (e.g. the ability to increase those fees/spread) that they can compensate for any mispricing after the fact.

Yes, there’s no reason for the expected total value of your payments to inherently be higher or lower for a tontine vs an annuity, after adjusting for time value of money.

Typically a tontine would have lower payments in early years with much faster acceleration later on compared to an annuity. That means that healthier people would generally do better with a tontine than an annuity, and less healthy people would do better with an annuity. It also means that the nominal amount of payments from a tontine is higher than those from an equivalently priced annuity, but that difference is entirely due to the tontine having more time to earn interest on your money before it pays out.

Tontines have no liabilities so do not need guarantees. Therefore incomes do not suffer from the underwriting fees of insurers.

This means that tontine payments typically start meaningfully higher than annuities at the outset even before the faster acceleration kicks in.

I guess I don't understand what you mean when you say "tontines have no liabilities." Mechanically, the way that a tontine works (or at least the way that they worked historically) is that I give you a fixed amount of money today, and in exchange you promise me a series of payments that are contingent on my life and the lives of the others in the risk pool. That promised series of payments is a liability, as a matter of accounting but also for all other practical purposes. Maybe you have a different structure in mind, but I don't see a way to operate a tontine-like product without a balance sheet.

The liabilities of the insurer are typically fixed amounts regardless of investment performance or changes in mortality. If and when the insurer miscalculates they will be wiped out if their assets don't match their liabilities.

Modern tontines are structured more like the Dutch/Swedish/Danish state pensions (the safest in the world) which have the ability to adjust the ongoing payments to members based upon the investment returns and mortality experience.

Asides from saving on the cost of guarantees, the fact that the trustees of the tontine don't have to cover their liabilities by only investing in low-yield bonds means that the trustees are free to invest in a much broader set of asset classes which in the OECD's opinion will generate higher returns resulting in the tontines being able to provide meaningfully higher levels of retirement income to the members.

> The liabilities of the insurer are typically fixed amounts regardless of investment performance or changes in mortality. If and when the insurer miscalculates they will be wiped out if their assets don't match their liabilities.

Again, for deferred annuities (and immediate-election GLWB annuities), which make up the vast majority of annuities that are sold and in force today, that's generally not true. The guaranteed credited rates on fixed and indexed annuities sold today are typically very close to zero (often 0.1% to 0.25%), and the guaranteed option budgets on modern variable annuities are often negative, which gives insurers a ton of leeway to reprice inforce policies as needed.

> Modern tontines are structured more like the Dutch/Swedish/Danish state pensions (the safest in the world) which have the ability to adjust the ongoing payments to members based upon the investment returns and mortality experience.

So pay-as-you-go? Does volume risk get passed along to the members, i.e., your payment in a given period is proportional to the amount of new money that gets put in during that period?

I'm with you on the asset issue, it would be great to have a less capital-heavy way to hedge longevity risk so that people aren't stuck funding their retirements with low-yield interest-bearing assets. But I don't think pay-as-you-go is a great workaround. Volume risk is significant even if you're a state-level actor and can make participation compulsory; I imagine it would be much worse as an individual player in the private sector.

The risk profile is different. Yes, both have a payout that depends on your own mortality, but there are other risk factors to consider.

An annuity is also a credit risk: if the counter-party goes under, then your payments will stop. Since most of the time the counter-party is a well-capitalized insurance company (potentially with an implicit government backstop), this risk is pretty small.

On the other hand, a tontine has a risk profile that depends on the mortality of the other nominees who participate: your payout in a given year depends on the number of nominees still alive. Also, depending on how it's structured, there might be some market risk as well.

From another perspective, for there to be an arbitrage opportunity, you'd need a way to create a synthetic annuity using a tontine (or vice-versa). But the risk profile of the tontine, which depends on the mortality of the other nominees, is hard to come by unless you are an insurance company. You can view a participant in a tontine as owning a certain life annuity as well as having sold the other participants a smaller life annuity. So each time one of them dies, you no longer have to pay that life annuity and can keep more of the income from the annuity you own.

I don't understand why tontines are illegal in the U.S.

Too many murders?

Probably judges got tired convicting geezers who committed murder to win the tontine.

it was mostly because of fraud. if you are a banker seems quite a nice premise to sell your rich friends a tontine with them and a bunch of Bering sea crab fishermen in it.

Tontines are perfectly legal. See: https://tontine.com/news/tontines-can-a-pension-this-good-be...

I also wanted to ask this but realized I only "knew" they were illegal from the Simpsons[0], which is maybe out of date and might have been inaccurate at the time.

[0]: https://en.m.wikipedia.org/wiki/Raging_Abe_Simpson_and_His_G...

Reply on news.ycombinator.com