Sci. Aging Knowl. Environ., 26 November 2003
Over the last century, life expectancy has soared in the United States and other industrialized countries. Some researchers argue that we can sustain that progress, but others think we've stretched our time nearly to its limit
Abstract: During the 1900s, life expectancy climbed steeply and mortality fell precipitously. In trying to predict whether these trends will continue during this century, demographers come up with widely varying life expectancy increases, from a few years to more than a decade. The differences among estimates often come down to basic assumptions. For example, some researchers favor extrapolation, which projects past conditions into the future; however, that approach provides solid answers only if trends continue. Other researchers and institutions, such as the Social Security Administration, hold that experts' judgment about future medical advances, economic conditions, and other factors should take precedence, although such methods often give the greatest weight to subjective assessments.
When newborn Humphrey Bogart howled his first protests against this crazy world on 25 December 1899, his prospects for reaching old age were almost as poor as those of a private investigator who turns his back on a blonde carrying a grudge and a .38. Even for a future tough guy, life expectancy was a paltry 48 years. Would-be femmes fatales could count on living only 3 years more. Before Bogie could sink a German gunboat with jury-rigged torpedoes, heed Lauren Bacall's seductive whistle, or set hearts a-thudding with famous lines like "Here's looking at you, kid," he'd have to survive beyond his first birthday. Ten percent of babies born at the time didn't.
By the time Bogie's youngest child, Leslie, was born in 1952, these dark statistics had brightened in the United States. Life expectancy, the average number of years a person has left to live, had shot up more than 25% for newborns, and infant death rates had plummeted nearly 75%. The medical and public health breakthroughs that spurred these improvements couldn't save Bogart, who died from smoking-induced cancer in 1957, but the demographic revolution raced on. Today, infant mortality has dipped to 6.5 deaths per 1000 births, a decrease of about 94% since the turn of the last century. Meanwhile, life expectancy at birth has ballooned to 74 years for men and nearly 80 years for women, a rise of more than 50% over 1900 values.
From astute 30-year-olds planning how much money they'll need for early retirement to senators trying to keep Social Security solvent, plenty of people are clamoring to know how much higher life expectancy will climb. Trouble is, demographers disagree about the answer. For instance, a recent study projects that women in at least one developed country will achieve a life expectancy of 100 years shortly after the middle of this century, with women in the United States not far behind. Another demographer, arguing that we can't squeeze out many extra years without attacking aging itself, speculates that life expectancy probably won't even reach 85 years by 2050. Navigating between these extremes, the Social Security Administration (SSA) projects a modest upswing of about 4 to 5 years over the next 50.
Although demographic models can be mathematically subtle, the differences among forecasts stem largely from scientists' surmises about the future of medicine, the environment, the economy, and other factors. Or as Harvard statistician Gary King puts it, what dictates the answers to questions about future life expectancy is "some evidence but many assumptions." Analyzing these assumptions can demystify how researchers arrive at disparate visions of the next 100 years.
Taking Life's Measure
Predicting the future is always chancy--just ask any investor whose portfolio was gutted when the dot-com bubble popped in 2000. However, pension managers, life insurance companies, bureaucrats, sociologists, and others need firm numbers for life expectancy. Although the projections listed in Figure 2 might not seem to differ much, even small miscalculations can be expensive. For example, SSA requires accurate projections of how many people will be drawing benefits and for how long. Because of the way the agency adjusts payments for inflation, an 80-year-old costs the government slightly less than a 65-year-old, says Steve Goss, chief actuary for SSA in Baltimore, Maryland. But every 1-year rise in life expectancy hikes the cost of Social Security by roughly 5%. So estimates that are off by a measly 2 years could lead to a 10% funding shortfall. If that happened this year, for instance, the agency would need to rustle up an extra $47 billion to cover its obligations. Looking beyond the usual issues of pensions and retirement age, rising life expectancy could shake up society, changing the structure of families and how we plan our lives, says James Carey, a biodemographer at the University of California, Davis (see "Extending Life or Compounding Misery?" and "Life Extension--Our Salvation or Our Ruin?"). If people regularly live to 100, for instance, society will have to figure out what they can do during those extra decades--even if we postpone retirement. And if your relatives irk you today, imagine living at a time when families typically contain five or more generations.
Calculating life expectancies for recent years is much easier than making prognostications. To determine last year's value for the United States, for example, researchers need to know the age of every person who died during the year and the number of people in each age range--data that are available from the National Center for Health Statistics, SSA, and the Census Bureau. Looking far into the past is trickier because of the paucity of data. The United States began gathering national figures only in the early 1930s, although other developed countries boast much longer records. For instance, in Sweden, tabulations started in 1650 and reliable statistics date back to 1750, says James Vaupel, director of the Max Planck Institute for Demographic Research in Rostock, Germany.
Although they might bicker over estimates of future life expectancy, demographers have a pretty clear picture of what has happened over the last 100 years. They've identified a host of advances that have helped drive down infant mortality in the United States. Vaccines and antibiotics, along with improvements in sanitation, routed child-killers such as typhoid, measles, and pneumonia. Better nutrition boosted resistance to disease. Even pasteurization of milk, introduced at the turn of the 20th century, helped thwart potentially lethal digestive bugs. Children weren't the only beneficiaries. Death rates among the elderly plunged 84% after 1900. The reasons for this decline aren't certain, but researchers give credit to several changes in addition to better health during childhood, including the introduction of Medicare and more effective treatments for "old age" ailments such as diabetes and cancer.
Projections of future life expectancy and mortality are what get demographers riled, but the reasons for disparities don't lie in the data themselves. If the number of people who die annually remained stable, this year's life expectancy would hold true for next year and for 2100, and demographers could apply their statistical know-how to concocting surefire systems for winning in Las Vegas. But fluctuating death rates force them to use imperfect methods to peer into the future. Some researchers, convinced that current trends will continue, extrapolate future values from past changes, whereas other scientists use their judgment to adjust the numbers. Either way, the baseline assumptions shape the results.
Reading the Future in the Past
Ronald Lee, a demographer and economist at the University of California (UC), Berkeley, used extrapolation to forecast U.S. life expectancy in a 1992 paper and in a 2001 follow-up. "It's hard to believe it's the best we can do when we've got all this knowledge" about genetics and the molecular mechanisms of diseases, he says. But attempts to incorporate that information haven't succeeded, he says: "It's clear that going on hunches, intuition, and subjective sense is not reliable."
By analyzing U.S. death rates from 1900 to 1989, Lee and colleague Lawrence Carter of the University of Oregon in Eugene developed a model to describe the decline, then applied the model to project future mortality and life expectancy. This long-term approach provides an accurate guide to the future, Lee says, because it smoothes out ups and downs in death rates caused by factors such as changes in the rate of smoking and the advent of AIDS, allowing researchers to tease out a general trend. In the year 2050, according to the 1992 analysis, the method foresaw a life expectancy of 84 years for men and women combined (versus 77 years today). Two years ago, Lee and Timothy Miller, also of UC Berkeley, revisited the predictions using updated data. The 1992 model predicted a 1998 life expectancy of 76.5 years, only 0.2 year below the actual figure, Lee notes. But 9 years of additional information spurred Lee and Miller to hike the forecast life expectancy for 2050 to 85 years.
An even more optimistic prognosis through extrapolation comes from Vaupel and his colleague Jim Oeppen of the University of Cambridge, U.K., whose analysis hinges on the country with the highest life expectancy for women (see Vaupel Perspective). In 1840, that country was Sweden, where women lived an average of 45 years. New Zealand, Iceland, Denmark, and other nations have each taken their turn at the top, but today Japan leads the world, with life expectancies of 78 years for men and 85 for women. A plot of the record-setting value for every year since 1840 shows that it rises about 2.5 years per decade, Vaupel and Oeppen noted in a 2002 Science article. "Our idea is that if life expectancy has been going up linearly at a rate of 2.5 years, it will continue," says Vaupel. Extrapolating with that rate means that women's life expectancy in the top-ranking country--perhaps Japan, perhaps another nation--will hit 97.5 around 2050 and 100 around 2060.
The method also provides a way to gauge the U.S.'s performance. Flag-waving patriots might be dismayed to learn that the United States has never been life expectancy champ--it has always lagged 1 to 10 years behind. If the differential remains about the same, the U.S. value in 2050 would fall between 87.5 and 96.5 years, an increase of between 8 and 17 years over today's numbers.
Lee's and Vaupel's methods assume that the past foretells the future--that prior declines in mortality will continue apace. That supposition, in turn, requires that medical research continue to deliver life-extending advances at a substantial rate. Given the bounty of data pouring out of genetic and protein studies and given our booming knowledge about the molecular underpinnings of killers such as cancer and heart disease, it's a reasonable assumption, says Vaupel.
Hitting the Wall?
Bunk, says S. Jay Olshansky, a biodemographer at the University of Illinois, Chicago. Scoffing at projections of large increases in life span, he diagnoses the failings of these studies as "too much math, too little biology"--leading to unrealistic assumptions about future declines in mortality. For instance, he says, industrialized countries stretched life expectancy in the 1900s mainly by chopping the death rate of children. If you spare a child, he notes, you add decades to life expectancy. But "you can only save the young once." The benefit from saving a 60- or 70-year-old is only a few years, so it's not possible to eke much extra time from countries with low youthful mortality.
Olshansky says that Americans could add about a decade to their lives if they stopped smoking, ate healthful diets, and exercised more. Curing cancer, heart disease, and diabetes--the illnesses that result from shunning these good habits--would prolong life by about 15 years. But people would still get old and die. "The only way we are going to see another quantum leap [in mortality reduction] is by learning to modify aging," he says. Still, Olshansky rejects the label "pessimist," because he's willing to accept that mortality rates among the elderly could fall by another 50% even without such a breakthrough. In a 2001 Science paper, he and colleagues Bruce Carnes and Aline D�sesquelles calculated that if the decline in mortality between 1985 and 1995 continued at the same pace, U.S. life expectancy for men and women could creep up to 85, but not until 2182--more than a century after Vaupel's and Lee's timetables.
If Olshansky is right, life expectancies should be leveling off in the richest nations. Whether they are is a matter of dispute. In the United States, women's life expectancy has barely budged from the 1990 value of 79 years. Olshansky sees this constancy as evidence of stagnation. But Lee says it could be a temporary slowdown, caused by an increase in the prevalence of women's smoking, which peaked in the United States in 1975. Vaupel adds that longevity increases in Japan and other industrialized countries show no signs of slowing. For example, by 1990 life expectancy for women in Japan had already shot past 80 years, and it has risen by nearly 2 years since. What's more, he notes, for the last 70 years researchers have been glumly prophesying that life expectancy was about to plateau--and they've always been wrong.
Consulting the Oracles
When academic demographers publish projections, only their reputations are at stake. But forecasts from actuaries at SSA can make a practical impact that lasts decades. Their figures can drive debate on controversies such as rejiggering payroll taxes, Medicare reform, and the Bush Administration's proposal to partly privatize the retirement savings system. Congress mandates that the Social Security trustees--a group that includes high-ranking federal officials and other worthies--issue annual reports on the financial health of SSA's trust fund. These Trustees Reports contain a demographic snapshot of the United States in that year, including projections of life expectancy and mortality up to 75 years into the future.
To derive those figures, the trustees confer with economists, demographers, medical researchers, and other experts and tweak the projections according to their input. For example, says Goss, life expectancy increased unusually rapidly during two bursts. The first came right after World War II, when medical advances that had been developed to help soldiers moved from the field hospital to the local hospital and antibiotics came into widespread use. The second boom started in the late 1960s and continued through the early 1980s, when Medicare and Medicaid gave more of the elderly and disabled access to health care. This year's panel of experts doubted that future medical advances would replicate the gains from these improvements. They saw other disturbing trends that might check life span, from antibiotic resistance and the resurgence of infectious diseases to the exploding cost of medical care. But the experts were also bullish on the prospects of new medicines based on information from the recently deciphered human genome. In the end, Goss says, the report set the projected decline in mortality at about 0.7% per year--lower than the historical rate of 0.75% but above the 0.5% chosen in 1999. The promise that genetic medicine would bring more effective treatments for diabetes, cancer, heart disease, and other ailments was the motivation for boosting the rate, he says. Those numbers give a life expectancy in 2050 of 79 for men and 83 for women.
Vaupel and Olshansky dispute many things, but they agree that SSA's "pundits know best" approach is faulty. They note that SSA has chronically underestimated the increase in life expectancy and the fall in mortality. Vaupel sees the current values as dangerously conservative, because the government will be making policy based on figures that could be more than a decade too low. Olshansky, on the other hand, judges this year's predictions as recklessly liberal. However, Goss argues that SSA projections have started to come quite close. For instance, in 1983 the trustees' report estimated that by the year 2000, a 65-year-old man could expect to live another 15.7 years; a 65-year-old woman another 20.8. The actual figures were 15.8 and 18.9.
At Harvard, statistician King also defends SSA's approach and the reliance on "gurus," but he sees a way to improve it. Instead of adjusting estimates afterward, he says, the organization as well as other researchers should debrief the experts beforehand and build their knowledge into models. Using that strategy, his team is constructing a megamodel that will project mortality for 191 countries, 17 age groups, and 24 causes of death--churning out 155,000 estimates in all, such as how many 50-year-old men in the United States will die from heart disease in a particular year. These projections can also yield life expectancy forecasts.
Working with Federico Girosi, a former postdoc now at the RAND Corp., King and colleagues are weaving in variables such as the effect of water quality on life span, which is negligible in the United States but massive in African countries, and tobacco consumption. Forecasts often fall short, he says, because they rely on so little data--often a mere 20 or 30 years' worth, not enough to ensure accurate projections 3 or 4 decades down the road. Because people who are close in age, say, 5- and 10-year-olds, often die of the same things and at similar rates, the researchers combined neighboring age groups--using some mathematical adjustments--to derive what King says seem to be superior predictions, at least so far. Although he's not ready to unveil the results, King has posted a manuscript on his Web site that describes the new methods. "So far, our approach usually outperforms existing methods in large-scale comparisons," he says
Reversing a Century of Progress?
Although their vision of the future is blurry, demographers usually see a rising curve of life expectancy. But even the most optimistic among them can rattle off a list of possible events that might send life span plummeting: environmental degradation, war, economic collapse, a spate of terrorist attacks. Although the SARS and West Nile viruses haven't killed that many people, other new pathogens might, says Lee. HIV has already reduced life expectancy in some parts of southern Africa to Medieval levels. However, the biggest threat to future longevity could be growing corpulence and indolence in many of the industrialized countries, particularly the United States. "We're going to need an awful lot of improvement from genetic interventions to overcome the effects of sedentary lifestyles and bad diet," says Goss. Olshansky agrees. If Americans don't curtail their bulging bellies and break out of their lethargy, "we are predicting a decrease in life expectancy," he says.
Demographers don't know whether citizens of developed nations will continue to bloat and start dying sooner or will vanquish their paunches and cottage cheese thighs, but King sees progress toward more accurate forecasts. "The available methods have gotten much better," he says, citing Lee's work as an example, "so I think there will be a lot of changes in the field." Improved models will bring demographers closer to the solution, but they'll have to be patient. To paraphrase Humphrey Bogart and Ingrid Bergman's theme song from Casablanca, we'll finally know the truth as time goes by.
November 26, 2003
Mitch Leslie, a science writer in Albuquerque, New Mexico, hopes to buy Humphrey Bogart's toupee on e-Bay.
Suggested ReadingBack to Top
Science of Aging Knowledge Environment. ISSN 1539-6150