The first step in their argument is to make plausible the claim that government and private retirement plans may well be understating how much longevity is likely to rise (citations and references to exhibits omitted): "The main source of longevity risk is therefore the discrepancy between actual and expected lifespans, which has been large and one-sided: forecasters, regardless of the techniques they use, have consistently underestimated how long people will live. These forecast errors have been systematic over time and across populations. ... In fact, underestimation is widespread across countries: 20-year forecasts of longevity made in recent decades in Australia, Canada, Japan, New Zealand, and the United States have been too low by an average of 3 years. The systematic errors appear to arise from the assumption that currently observed rates of longevity improvement would slow down in the future. In reality, they have not slowed down, partly because medical advances, such as better treatments for cancer and HIV-AIDS, have continued to raise life expectancy ..."
Here are a couple of illustrative figures. The first shows projected life expectancies for the United Kingdom. Starting at the bottom left, the lines show the projected rise in life expectancies at that time. Notice that the projected increases consistently underestimate the actual rise, which is the black line on top.
The table below shows the typical life expectancy at age 65 used for pension funds in a number of countries in the first column. The second columns shows the currently estimates of life expectancy at age 65. Notice that in each case, the number used for future life expectancy in the first column is above the current estimates of life expectancy, which is wise. But also notice that for a number of countries, including the United States, the actual increase in life expectancy since 1990 is substantially larger than the difference between columns 1 and 2.
The IMF asks what would happen if life expectancy by 2050 turns out to be three years longer than current projected in government and private retirement plans: "[I]f individuals live three years longer than expected--in line with underestimations in the past--the already large costs of aging could increase by another 50 percent, representing an additional cost of 50 percent of 2010 GDP in advanced economies and 25 percent of 2010 GDP in emerging economies. ... [F]or private pension plans in the United States, such an increase in longevity could add 9 percent to their pension liabilities. Because the stock of pension liabilities is large, corporate pension sponsors would need to make many multiples of typical annual pension contributions to match these extra liabilities."
What's to be done? One step could be to build into the benefit formal for public pensions, like Social Security, provisions for an automatic decline in expected benefit levels for future retirees as life expectancy rises. The enormous change that has taken place in private retirement plans, switching from "defined benefit" plans in which the employer promises a stream of future payments to "defined contribution plans where the employee has a retirement account with a certain amount in that account acts as a way of transferring longevity risk from employers to workers. Of course, retirees can then protect themselves from longevity risk and outliving their assets by putting a substantial share of their retirement funds in annuities.
There are also proposals for innovative financial assets like "longevity bonds." Say that a pension plan worried that it has underestimated longevity risk, and thus will have to make higher payments than it expects. The company buys a longevity bond, where the payments that the company receives from that bond would rise if longevity exceeds certain benchmarks. Because the return on the bond would offset some longevity risk, the buyers of the bond would be willing to accept a lower interest rate than they otherwise would demand. However, no longevity bonds have yet been successfully issued.
It's not enough just to set an expectation for how much the population will age, and to plan accordingly. We also have to plan for the historically likely possibility that life expectancies may grow faster than we expect.