[
]
For centuries, humans have used actuarial tables to figure out how long they’re likely to live. Now artificial intelligence is taking up the task – and its answers may well be of interest to economists and money managers.
The recently released Death Clock, an AI-powered longevity app, has proved a hit with paying customers – downloaded some 125,000 times since its launch in July, according to market intelligence firm Sensor Tower.
The AI was trained on a dataset of more than 1,200 life expectancy studies with some 53 million participants. It uses information about diet, exercise, stress levels and sleep to predict a likely date of death. The results are a “pretty significant” improvement on the standard life-table expectations, says its developer, Brent Franson.
Despite its somewhat morbid tone – it displays a “fond farewell” death-day card featuring the Grim Reaper – Death Clock is catching on among people trying to live more healthily. It ranks high in the Health and Fitness category of apps. But the technology potentially has a wider range of uses.
Life expectancy is key to all kinds of financial and economic calculations, by governments, companies and individuals – from retirement income needs, to policy coverage at life insurance and pension funds, and financial planning.
In the US – which has lagged behind other developed economies in the life expectancy of its citizens in recent years – the Social Security Administration has its own table for mortality rates, which features in the annual financial report to trustees.
The government agency currently predicts that an 85-year-old man in the US has a 10% probability of dying within a year, and an average 5.6 years to live. But averages like that can be off by wide margins, says Franson, and the new algorithms can deliver a more tailored measure – a customized death clock.
That such findings are of interest in economics is demonstrated by the publication – in just the past month or so – of two papers around the topic by the National Bureau of Economic Research.
‘Harness The Benefits’
One of them, titled “On the Limits of Chronological Age,” looks at the diverse ways that the aging process affects physiological capacities. It finds that many aspects of economic behavior, like readiness to join the labor force, may not be well captured by people’s calendar age – even though that’s what policies such as statutory retirement are typically based on.
By continuing to rely on chronological age as a proxy for how well people can function, societies may end up failing to “fully harness the benefits of increasing longevity,” the researchers from Harvard and the London Business School conclude.
Another working paper examined the “value per statistical life” or VSL – a callous-sounding measure that’s used for cost-benefit analysis in areas like regulation of pollution or compensation for workplace accidents. It’s typically calculated based on compensation for workers in high-risk jobs.
The researchers behind the NBER study “The value of Statistical Life for Seniors” drew on a different dataset: the propensity of older Americans to spend money on medical services that reduce mortality risk. They found a mean VSL at age 67 of just under $2 million for people reporting their health as “excellent,” compared with $600,000 for those in “good” health.
When it comes to personal finances, better measures of life expectancy will have profound implications for people saving for retirement, according to Ryan Zabrowski, a financial planner with investment advisory firm Krilogy.
“A huge concern for elderly people, our retirees, is outliving their money,” says Zabrowski, who touches on the issue in his soon-to-be-released book, “Time Ahead”.
‘Out The Window’
Decisions such as how much to save and how fast to withdraw assets are often based on broad-brush and unreliable averages for life expectancy. AI-driven tests that can potentially reduce that uncertainty are largely unheard-of now, but likely won’t be such an unusual idea in the future.
What’s more, the AI technology itself along with advances in medicine has the potential to boost life expectancy – and with it the risk of running out of savings. Zabrowski reckons one consequence is clear: longer retirements will mean savers need higher-return investments for their old age, which will push them to allocate more stocks over fixed-income securities.
“The conventional method of measuring demand for equities will be thrown out the window,” he writes his forthcoming book. As people start expecting to live longer, there will be a “massive escalation in demand for equities.”
There are plenty of technologies already out there – like heart-rate monitors and maximal oxygen-consumption gauges from wearables – that have the potential, in tandem with new AI-powered devices, to reduce the uncertainty around personal mortality.
Of course, there’ll always be limits. On top of entirely unpredictable variables, like accidents or even pandemics, there are plenty of intangibles.
Longevity Gap
Loneliness, for instance, is often reckoned to reduce life expectancy. Gratitude may increase it. A Harvard study found that women who reported feeling the most grateful had a 9% lower risk of dying within three years than those who reported feeling the least.
Then there’s the question of inequality. For life expectancy, money matters. Multiple studies – including the work of Nobel prize-winning economist Angus Deaton on “Deaths of Despair” – have found a clear gap between rich and poor Americans.
Research published by the American Medical Association found the longevity gap between the wealthiest and poorest 1%, at age 40, was nearly 15 years for men and 10 for women.
For Death Clock users, who have to pay $40 a year to subscribe, the app suggests lifestyle changes that can hold mortality at bay – along with a second-by-second countdown of estimated time remaining.
“There’s probably not a more important date in your life than the day that you’re going to die,” Franson says.
