Predicting population change: How can a graph be so very wrong?

Different ways of seeing stats
BBC News 17 Apr 09;

Sir David Attenborough is now patron of the Optimum Population Trust, which lobbies to cut the number of people on Earth. But predicting population change is tricky, says Michael Blastland in his regular column.

A child watching Star Trek, open mouthed. The scene, a hideously overpopulated planet. A stirring caldron of people at the window. A privileged few who close the curtains.

But I grew up, not noticeably squashed.

The idea of overpopulation has been around a while. Let's not go there. Let's stick to a simpler part of the problem. How many of us will there be, a decade or two hence? We can decide later if so many is too many.

The answer, crudely, is that the track record of population projection is abysmal. It borders on being a statistical lottery.

People are not good at seeing the future, particularly the future of people. The problem is that people are changeable, they don't always do in the future what they do now. People, being people, know this - or should - but often carry on regardless.

Total population change in a country comprises three elements: migration, lifespan and fertility. In no case have people done what they were meant to. The numbers came out and then, drat them, people had the nerve to live even longer, just like that.

They had lots of babies and then stopped having lots of babies. Loads suddenly decided to turn up on the doorstep just when it was thought more people might leave.

It's hardly surprising that the UK projections for all three elements have repeatedly failed to hit a barn door. Here, just for fun (they serve little other purpose) are some past attempts at forecasting.

First, total UK population.

In 1965, the population at the turn of the millennium was projected to be about 76 million (the yellow line sticking up). That's about 16 million too many, or more than two extra Greater Londons.

The 1994 projection had UK population falling after about 2020. Ten years later, this was revised to point us above 70 million again. In general, the figures are all over the shop.

The next chart shows part of the explanation for at least some of the ups and downs: children.

In 1965, the total fertility rate (TFR) was about three - that is, women were having about three children each. Therefore, women in the future would do the same…

Except that in the past decade, the rate of fertility for UK-born women fell briefly to about 1.5, half the projected rate. The number of births in one year in the UK was at one point nearly a million fewer than projected. There's almost no need to study this chart - just marvel at the way it flails around like a cat o' nine tails - and you'll get the general idea.

Coming and going

The next component is migration. A good example of where migration goes wrong is that few people expected half a million Poles to arrive in the UK in the space of about two years in the mid-2000s. Nor that almost all would travel through ports and minor airports, where far less effort goes into monitoring migration.

So picture the official survey teams, mostly stationed at Heathrow, Gatwick and Manchester, as about 500,000 people tiptoe behind them on some other route. "Seen any Poles, mate?" "No mate, no Poles here."

Whether the authorities should have seen this change in the pattern of travel coming is another question. The fact is they didn't.

Not only did we not expect them, nor project meaningful figures for them, we didn't know they were actually arriving.
PROJECTED LIFESPAN IN 2001
# 1971: men 71.9, women 78.6
# 1975: men 70.4, women 76.7
# 1979: men 70.6, women 76.9
# 1983: men 72.3, women 79.3
# 1987: men 73.8, women 79.9
# 1992: men 75.4, women 80.6
# 1996: men 75.3, women 80.2
# 2000: men 75.7, women 80.4
# Actual: men 75.8, women 80.5

Source: ONS and Government Actuary

Finally, there is how long we live. In the past 30-odd years, projections of the typical lifespan in 2001 has varied considerably - see factbox on right. These got close in the end. Though by the end, of course, there's not much future left.

As the Office for National Statistics (ONS) reported a few years ago, projections for total population have been better recently but only because the errors on each of the components evened out.

The report said: "Demographic behaviour is inherently uncertain." Which is a restrained way of putting it.

It then continued: "The number of children we have, how long we live and the number of people who migrate from one country to another are variables that have changed continually in the past and will continue to do so in the future. Even if we understood perfectly the factors that have brought about past changes (which we clearly do not), our ability to predict the demographic future would inevitably remain limited."

What's more, not only do we not know where we are going, we don't really know where we are.

"We do not even know, with complete certainty, the size and age structure of the current population at the time a projection is made."

To be fair, the ONS beats itself up (and the government actuary's department, which used to do these projections) unnecessarily by reporting on the projections' accuracy. They were mostly not intended to be a prediction of the future. Projections are not predictions, but simply describe current behaviour and ask what the world would look like if this behaviour continued.

Here's a prediction: it won't.

All-seeing eye

To go back to where we started, people change, and they change not least in response to changes already occurring. If the birth rate rises fast, the resulting housing pressures and costs alone of 78 million people in the UK would probably do at least something to temper the attraction of large families.

As for population on a global scale, it does seem likely that it will go up, maybe even faster than the steepest projections, but here's a telling story. In the 1970s, a woman in Bangladesh typically had about seven children, of whom two died in infancy. Today, she typically has about three, who tend to live. Family size has fallen dramatically. In not much more than a generation, Bangladesh changed in a way that took the rich world 200 years.

One image of population is that people clock in and clock out, by birth or death or migration, that they do so in reliable, easily measured ways, and that we should be able to take the swirl of political, cultural, economic, scientific, social and personal influences, and somehow divine the mass human reaction.

Are we talking about the ONS, or God?