In his new book, “King William's Tontine: Why the Retirement Annuity of the Future Should Resemble Its Past,” Moshe Milevsky argues convincingly that retirees have a product problem. Today's financial marketplace of investment and insurance products doesn't offer the perfect solution. To find that solution, he goes on an archeological dig in Renaissance Europe and unearths the tontine. Cast away years ago, the tontine may hold secrets to solving the income puzzle faced by retirees looking for a private sector solution to replace employer pensions.
King William's tontine was originally dreamed up as a lottery-like bond scheme meant to fund the king's war spending issues. The tontine bond provided bond interest payments until death. There was also the possibility of huge payments to those who lived a long time since the total interest amount was shared among the survivors until the last one died.
Although the king's tontine might not sound like the perfect way to fund retirement income, it contains a number of features that could be resurrected to provide retirees with greater financial security.
As it turns out, everything we need to know about building a perfect retirement product we learned in kindergarten. Life is better when you share. When retirees share the risks of living longer than expected, or of facing low portfolio returns, they are better off than if they face these risks alone. As we learned in kindergarten, the idea of sharing can be painful at first. But sharing retirement wealth with others underlies what Milevsky terms “tontine thinking.”
Tontine thinking throws out some of the promises employers and insurance companies make to workers. That's because making promises means predicting how long retirees will live and what investment returns will be in the future. Unfortunately, actuaries don't have a crystal ball. They have data on past mortality rates and asset returns. But what if there's a cure for cancer? Or a 21st century plague? Or a sustained low interest environment? In each case, the estimates could be wildly off.
Institutions making future income promises will either face a shortfall, or they’ll be flush with cash that could have been used to pay higher incomes. Neither is ideal.
At a recent Wharton Pension Research Council meeting, a European panelist argued that he “didn't want to live in a world where we couldn't guarantee an income to retirees” in response to a question (by me) about whether workers shouldn't face an income haircut if bond and stock returns didn't turn out as well as the pension anticipated. The problem with his response is that somebody will have to pay when the assumptions used to estimate the guarantees don't pan out.
Milevsky deals head on with the moral implications of a pension scheme that doesn't involve risk sharing among peers. When pensions or insurance companies make income promises based on optimistic assumptions, someone has to pay when reality doesn't measure up. If you’re an insurance company, this means the very real risk of insolvency that many faced after making generous income guarantees of GMWB policies before the financial crisis. If you’re a private pension, this means increasing contributions or reducing benefits of younger workers, which involves a wealth transfer from younger to older participants. If you’re a government, then you’re tempted to promise big guarantees to older citizens at the expense of younger taxpayers who either don't, or can't, vote.
The reasonable response of insurance companies is to face these risks by increasing the cost of annuities. This involves increasing expenses, holding back greater reserves or providing far more modest guarantees. While a fair market price is more likely to provide a sustainable pension system, it also means that retirees will need to pay more to transfer this risk to an institution. Like any insurance product, it sacrifices expected wealth for security.
The Jared Tontine
Milevsky seeks to address what he calls “the plight of Jared”—referring to the Bible's second-oldest person (his lifespan was exceeded by that of his better-known grandson Methuselah). Jared, he explains, is his euphemism for the many retirees who will live to advanced ages without breaking records or even being considered all that unusual—and who will have to pay for their extended retirements.
“Yet,” Milevsky writes, “no insurance company, retirement system, or pension plan can afford to support all these Jareds, especially considering how little we have collectively saved for retirement.”
Milevsky's solution pulls together some of the best ideas from existing products using economic theory and past experience to weed out ideas that don't work. These include complexity and a lack of risk sharing in annuities, historical tontine features that don't make much economic sense such as those in King William's tontine, and the lack of mortality pooling in investments. This means starting with a shell of a tontine and tweaking the rules to create a more perfect retirement product.
To understand why “tontine thinking” makes sense, it's important to revisit the benefit of pooling longevity risk among retirees. Today's retirees often draw from a lump sum of money in a defined contribution plan to fund retirement income. Even if they place their savings in safe investments and withdraw their money prudently so it will last, they are taking two very important risks that can be reduced by pooling.
The first risk is that they’ll live too long. We know that a proportion of today's retirees will live to 100. If you carefully ladder a bond portfolio to cover expenses up to age 100, what happens if you live to 105? It's not easy to plan when you don't know your time horizon. This is longevity risk, or more accurately idiosyncratic longevity risk (or retiree-specific longevity risk).
All retirees who don't pool their savings face this idiosyncratic risk and must deal with it by spending very little to avoid running out of money, or by spending more and taking the risk that they’ll have to cut back if they beat the odds.
I’ve just alluded to the second risk. A retiree who wants to avoid running out of money can assume an extended lifespan, say tailoring an investment portfolio and withdrawal strategy to last to age 105. How much, though, can retirees spend if they plan to live to 105? Not very much at all. If we assume an investment in safe bonds, they’d need to save perhaps 35 times their spending goal or more if they retire at age 65 and withdraw a tiny slice each year to cover living expenses.
Most retirees will die long before Methuselah or Jared. Dying early is great news for the (not emotionally attached) heirs, but it isn't so great for the retiree. This creates the great tradeoff of pooling retirement savings. Pooling means more money goes to the retirees, and accepting idiosyncratic risk means more money goes to the children or charities of retirees (or means that the retiree runs out of money early). Since Milevsky is an economist, he believes that pooling is a better deal.
I’ll give a quick example of why pooling works. Let's say we have two 95-year old retirees. One will die after one year and the other will die after two years. Neither knows which retiree will die first. They each have $90,000 in savings. If the retirees are conservative, they’ll want to make sure they don't run out of money early and they’ll each spend $45,000 a year. The unfortunate retiree who dies after the first year will give $45,000 to their heirs. The other retiree will run out of money after the second year and then dutifully pass away.
But what if they instead pool their $180,000? This means that together they need to fund three total years of spending. They can now each spend $60,000 in the first year, leaving the remaining $60,000 for the fortunate retiree in the second year. The heirs gave up $45,000 to fund a 50% increase in spending for each retiree. This is the magic of mortality pooling.
That's why an annuity with safe investments will pay more than a bond ladder that provides income up to a Jared-like 105 years of age. Money from dead annuitants goes back into the pool to fund the spending of the living. So why not just buy an annuity? What's the advantage of a tontine?
Milevsky provides a colorful history of the first tontine schemes, which makes his book a treat for the pension geek who happens to be a history buff (like me). Tontines provide a guaranteed rate of return (say 8%), which is less than the going market rate for bonds when they are first issued on the total capital borrowed by the tontine issuer (say the British government). If the government borrows $1 million and there are 10 investors who each write a check for $100,000, then each would get $8,000 in the first year.
The difference between a tontine and a bond is that if two of the original investors die, then the surviving ones are still paid their share of the $80,000 interest payment. That means that instead of getting $8,000, through the magic of mortality pooling the eight survivors now get $10,000 each. After five investors die, each is now getting $20,000. The income keeps rising until the last investor gets $80,000 a year until dying. Unlike a bond, however, none of the heirs is entitled to the principal or interest payments after the original investor dies. The dead retirees and their heirs provide the vigorish that juices up returns for the survivors.
Mystery novel readers might be intrigued by the sinister possibility of doubling one's income when there are only two remaining participants. Being an economist, Milevsky sees the main problem with the traditional King William tontine as this lottery-like increasing payments path. An optimal path would smooth spending even after participants started dying off.
Milevsky's ingenious proposal is to level the expected payments by reducing the tontine rate each year. Instead of starting at 8% of the original $1 million, why not start at 10% giving all retirees a 25% higher income in the first year and then reduce the rate over time so that each surviving retiree gets about the same amount as they age? The average retiree will be able to spend more but lose the fun opportunity to blow fat stacks at the casino in your scooter.
To an actuary, this is actually not that hard. We can use historical survival rates to estimate how much we’ll need to reduce the total tontine payout rate over time in order to roughly smooth income. Of course, investing in very safe inflation-protected assets will ensure a constant spending stream. Or one can accept greater investment risk, which will raise or lower the remaining assets from which the fixed percentage is drawn.
For an insurance company, the appeal of a tontine-style product is that you get rid of two big risks of selling an annuity. If there is a cure for cancer, you won't go out of business making payments to a much longer-lived pool of annuitants (systematic longevity risk). If everyone lives longer, there will be more of them to share a fixed amount of remaining assets and they’ll have to spend a little less. They also won't be on the hook if investment returns are lower than the insurance company expected when the company made the income guarantees. Poor returns simply mean that the pool is smaller and the participants will receive an income that varies with the performance of the investment portfolio.
That's the good news for an insurance company. The bad news is that if done correctly the only comparative advantage they have versus an investment company is access to good actuaries. But actuarial tables aren't that difficult to find and some companies are already selling similar products, such as managed payout funds, that could be tweaked to do a better job of pooling mortality risk. Once the investment companies are able to compete for instruments that effectively pool mortality risk, will insurance companies still love tontine thinking?
Milevsky's income smoothing tontine could be actuarially simplified by allowing sales to only a specified gender and age. Investors could gain a little bit of mortality pooling benefit by initiating the tontine before retirement so those who die before retirement essentially chip in to boost the pool (which is no different than workers who have contributed to a pension).
Tontines also don't lose much efficiency with a modest number of participants, which can allow medium-sized employers with, say, 75 male employees all aged 45 to have their own tontine arrangement. Of course, the employer would love the prospect of handing off longevity and market pension risk to employees through the tontine.
Groups of like-minded individuals could also band together to form a tontine. For example, a church could sponsor a tontine in which participants may be more likely to join with less fear that their proceeds from an early death would go to a complete stranger.
Perhaps the greatest benefit of a tontine is that participants could select the one with an investment policy that best suits their risk tolerance in a manner similar to target date funds and they can select the tontine with the most attractive expense ratio. Highly salient, easily compared mutual fund expenses have pushed down average fees within the industry. Tontine thinking could be a game changer in an insurance industry that often suffers from opaque pricing.
What may be the most surprising piece of history in Milevsky's book is that tontines were very popular only a century ago. But poor regulation, and in particular the abuse of more vulnerable investors by insurance companies, gave tontines a black eye and resulted in legislation that eliminated some features of these early tontine schemes. The book argues that there are no real legal barriers to an ideal tontine scheme—indeed TIAA-CREF currently sells an instrument that employs quite a bit of tontine thinking.
With some luck, tontines could make a comeback. Are tontines the killer app of the retirement industry? Milevsky makes a strong argument that they’re better than many products on the market today, and certainly better than a retirement strategy where retirees don't share the risk. Sharing and participation might hold just the key to solving the great retirement income puzzle.
--- Related on ThinkAdvisor: