Warren Sanderson & Sergei Scherbov had a very interesting article in Nature earlier this year (you can find the full article reproduced here on page 5). The article title really tells the story in itself: average remaining lifetimes can increase as human populations age. Put differently, we may be facing the interesting enigma that the longer we live, the longer we have left to live.
But, riddles aside, what Sanderson and Scherbov actually propose is a new metric: the median age of the population standardized for expected remaining years of life. Now why would that be interesting?
Well, as regular readers will already know, I tend to use median age as a rough and ready – rule of thumb type – guide for looking at all kinds of issues, from savings and investment behaviour, to social and political phenomena. Yet I have long been aware that there was something problematic about this way of doing things. Median age is, as Sanderson and Scherbov argue, a moveable feast.
To get an idea of what I am talking about here, let’s take a simple analogy from economics. Economists often think about apples and pears, but sometimes they want to just think about apples, and compare them, from one year to another. Now one comparison we make is in terms of price, how this changes from one year to the next. But often this rather crude way of doing things just isn’t sufficient, since the procedure ignores changes in quality. So to try and get round this economists have invented (well better put Zvi Griliches invented) a thing called hedonics.
Well, in a nutshell, this is what Sanderson and Scherbov are proposing: a hedonics of human age value, using remaining life expectancy as a crude proxy. What do they actually argue?
“Population ageing differs from the ageing of an individual. People who survive grow older with each year they live. Populations, on the other hand, can grow younger. Because a wide variety of matters such as the cost of medical care, retirement, bequests, consumption and the accumulation of human and tangible capital depend not only on age but also on time left to live, our understanding of population ageing must also reï¬‚ect both of these factors. Because conventionally measured old-age dependency ratios (the ratio of the number of people at the retirement age and above divided by the number of people in the working ages) have caused worry about the sustainability of pensions, it is important to recognize that these ratios, rescaled for life expectancy increases, are forecasted to change comparatively little over the century, suggesting caution in our assessment of long-term pension problems.”
Now one idea that Sanderson and Scherbov implicitly advance (although they don’t in my view make this sufficiently explicit) is that this “getting younger” is a re-evaluation of the prime age working life of the human individual. I think this is the point they are making when they say:
“Medical care expenditures provide an example where calculating the median remaining life expectancy in a population is useful. Health care costs rise rapidly in the last years of a personâ€™s life. The change in the median remaining life expectancy between years is equal to the change in the median time to the onset of that phase of rapidly rising costs.”
That is to say, we shift the more feeble and fragile years up through the age course. They also use the idea of “proportional life cycle rescaling”, which amounts to the same thing.
“Proportional life cycle rescaling is a heuristic not a predictive concept. It provides one simple way of thinking about a complex future in which the lengths of life cycle phases will be inï¬‚uenced by social policies and demographic constraints not modelled here. We use proportional life cycle rescaling by adjusting the conventional start of the working age phase (assumed to be age 20 in the year 2000) and the conventional end of that phase (assumed to be age 65 in 2000) proportionally to changes in life expectancy from 2000 onward.”
So the basic idea is that we need a benchmark (say the year 2000) and a moving age index calibrated in terms of the benchmark values. This could give us a measure of how the productive potential of a population was changing through time (and thus some sort of indication of the growth potential of the economy in question).
Two further implications of this approach would be that life rescaling implies a later average start date for work (this is what is implied by moving up the value chain, with more education continuously required) , and of course (as suggested by me in this post) an upwardly flexible retirement age.