The New York Times had an article recently about academics and financial advisers who want to bring back a Baroque-era investment vehicle—the tontine—where an annual dividend is split among surviving investors (the Washington Post had a similar story two years ago). The present-day appeal of the tontine is partly based on its supposed transparency. It’s unlikely, however, that potential investors would be able to accurately predict the payouts they might receive, which would depend on their health relative to that of others in the pool, among other variables. Still, it’s a morbidly interesting excuse to think about insurance markets and innovative retirement schemes.

Gambling on other people’s death isn’t unique to tontines. The AIDS epidemic created a secondary market in life insurance policies, allowing ill policyholders to tap some of their benefits to pay for health care and living expenses. Though this may have served a useful function in a country with inadequate social insurance—especially pre-Obamacare—it’s hard to feel sorry for investors who lost big after the discovery of antiretroviral drugs.

Tontines, like Social Security, traditional pensions, and life annuities, insure against the risk of living longer than expected in retirement. The problem of outliving one’s savings has gotten worse as Social Security benefits have been trimmed back and private sector employers have replaced traditional pensions with 401(k)-style savings plans. In theory, 401(k) savers can insure against longevity risk by purchasing life annuities, but few actually do. There are several reasons for this, starting with the fact that few have significant savings to begin with—a problem exacerbated by current low interest rates that lock annuitants into low annual payments. In addition, potential buyers must navigate complex and tricky insurance markets and face prices driven up by adverse selection and asymmetric information, the classic problem of markets for individual insurance whereby people at greater risk (of living longer, in this case) are more likely to purchase insurance and have an incentive to conceal information to avoid higher risk-adjusted premiums, leading to higher prices for all consumers and a shrinking market

Potential annuity buyers also behave in ways are hard to square with fully-informed and rational behavior, such as overvaluing lump sums relative to their equivalent in annuitized benefits and exhibiting loss aversion—in this case, the tendency to dwell on the potential financial losses associated with dying prematurely rather than the potential gains from living a long life. Could tontines at least counter these behavioral challenges? One psychological hurdle for would-be annuity buyers is the fact that insurance companies profit from annuitants’ early death, which puts people in a pessimistic and suspicious frame of mind. Advocates say tontines could be structured so that only investors—not issuers—would benefit from the deaths of others in the pool, which might or might not alleviate these concerns. (Tontine murders were once a common melodramatic plot device in plays and murder mysteries).

To see how a tontine might work in a modern-day context, imagine a pool of a thousand 65-year-old investors sharing the 3 percent yield on $100 million in 30-year U.S. Treasury bonds. The first year, each investor would receive close to $3,000 (slightly more, assuming a few investors die before the end of the year). By the 30th year, if only 127 investors remained, the payment would be around $23,600. These estimates are based on life tables for the Social Security population, though actual amounts would likely be lower because investors in tontines are likely to be healthier than average. Regardless of the exact amounts, we would expect payments to spike significantly in the out years, even assuming there’s a predetermined payment period and the issuer, not the last survivors, recoup the principal. (This variation on the classic tontine model would also lower the cost of participation below the $100,000 face value of each participant’s share of bonds as well as reduce the incentive for murder.)

Who needs an annual income stream that grows from $3,000 at age 65 to $23,600 at age 94? Even if you consider that other income sources tend to be eroded or exhausted in old age and that anyone who makes it to 94 is expected to live another three years and should prudently plan on living even longer, this is a strangely back-loaded investment vehicle. Though enthusiasts are right that tontines should cost less than annuities with income streams that are equivalent in value, they’re inefficient at producing income when people need it. They also have the marketing disadvantage of uncertain payouts, which could be mitigated with experience and larger risk pools. At best, therefore, tontines might be a specialized product for sophisticated investors who have other income sources and want a backup in case they live to a very ripe old age.

It’s also questionable whether tontines would overcome people’s aversion to investing in products that conjure up the double whammy of facing premature death and realizing you made a bad investment decision. Tontines’ popularity back in the day was the appeal of a macabre lottery, not that of a retirement plan. And people tended to buy into them when they were young and optimistic, not after getting banged around by life. Pessimism about our own life expectancy is another reason people tend to under-save and under-annuitize. Past a certain age, most of us are aware of having less-than-perfect health and typically expect to live about as long as our parents or as the average life expectancy when we were born. This is a mistake. Not only are we likely to live longer than our parents, but each year of survival extends our expected lifespan. Baby boomers born in 1950, for example, had a life expectancy at birth of around 75 years, but half of those alive today will live past their 84th birthday, while a quarter will live past 90, and a tenth will live past 95.

Tontines (at least a modernized version) do have a useful feature that explains their appeal to retirement wonks. They can provide some longevity insurance to individuals without saddling issuers with systemic risk—hence their lower cost compared to conventional life annuities. In other words, they insure individuals against longer-than-average lifespans while adjusting payments for changes in average life expectancy. This feature is worth thinking about, because pooling individual risk is costless in large groups, but insuring against systemic risk has a cost that is passed on to consumers in the form of higher premiums.

To some extent, insurers can hedge longevity risk by selling both life insurance and life annuities, so in the event of higher-than-expected mortality, say, larger-than-expected payments to life insurance beneficiaries are offset by lower-than-expected payments to annuitants. This hedging is limited, however, by the fact that different types of people may purchase life insurance and annuities, and the two products typically cover people at different life stages while mortality trends can vary by age and other demographic characteristics. Recent alarms about rising mortality among whites, for example, really reflect more narrow increases among middle-aged women in southern and rust belt states.

So, though some of us might theoretically prefer cheaper annuities that can be adjusted to offset trends in life expectancy, the challenge is consumers’ ability to understand how this works and government’s ability to police insurance companies so they don’t game the system. Moreover, since the experience of even large insurance pools will differ from that of the overall population, insurers would still retain some longevity risk if adjustments are based on population trends—a problem tontines avoid.

These and other insurance market failures can be minimized with social insurance, which, in the case of Social Security has the underappreciated benefit of insuring the same people against both premature death and longer-than-expected lifespans. Thus, the dependents of many of the working-class white women who are now dying at increasing rates will receive Social Security benefits that the women themselves will never get in retirement. Historically, Social Security contributions have increased to offset increases in life expectancy and other cost drivers, though the last ad hoc rate increase legislated by Congress was over a quarter century ago. Going forward, it might be a good idea to link contributions to life expectancy, though this should be coupled with higher—that is, proportional—taxes on the wealthy to avoid making low-income people pay more for what in recent years has been mostly high-income people’s longevity gains.

The question of how to provide retirees with secure lifetime incomes in a way that is robust to changes in life expectancy has also come up in discussions about how state and local governments can help workers save for retirement. In the absence of federal action to address a looming retirement crisis, some state and local governments are in the process of setting up low-cost plans aimed at the roughly half of prime-age private sector workers who aren’t covered by employer plans. While policymakers backing these initiatives like the idea of annuitized benefits, they’re leery of asking taxpayers to take on additional pension liabilities at a time when many public sector employee pensions are underfunded and under attack. The challenge, therefore, is devising ways to encourage or require participants in these voluntary plans to annuitize some or all of their savings without discouraging people from participating in the first place.

The simplest solution is to offer an annuity from an insurance provider as the default payout option, presumably at a lower cost than annuities purchased outside the plan. But it’s likely that participants who face a choice between lump sums or annuities will still tend to undervalue annuities, exacerbating the unavoidable adverse selection problem. An alternative worth considering is based on hybrid DB-DC plans in Europe in which benefits can take the form of traditional “defined benefit” (DB) pension income streams but employers can be shielded from long-term liabilities as in 401(k)-style “defined contribution” (DC) plans. These hybrid DB-DC plans can take different forms, but typically have annuitized benefits that can be adjusted for gradual changes in life expectancy without being as sensitive to investment returns as DC account balances annuitized at retirement based on market interest rates.

Unlike a tontine scheme, where payments simply increase in inverse proportion to the share of surviving investors, such longevity and return-smoothing adjustments are complex and require trust in the system, so may be better suited to government-sponsored plans than private sector ones. The simplest solution, of course, is simply to expand Social Security, an increasingly mainstream idea among Democrats but not one that is likely to fly in the current Congress.