Forecasting Is A Silly Man’s Game

Once upon a time, when the Viking Dutchman was a younger warthog, he took a class on econometric forecasting. Which is to say forecasting data using econometrics, statistics’ vaguely silly cousin. The thing that really impressed me about the class, was how open the Professor was about how difficult a process forecasting data really is. Most statistically sound forecasting models end up with margins of error so large that its very difficult to tell much useful from them. Never read the precise number (or “point estimate”) given by a forecast: they are largely meaningless.

Remember that the next time you read an economic forecast. 

To my mind, the easiest way to expose the problem with forecasting is to go back in time and read forecasts of the present. If someone says that unemployment will be 5% in 2023, well see if you can go back to 2003 and see what they thought it would be today. 

Ah, but is that fair? After all there have been extenuating circumstances since 2003, that could not have been predicted! True. But there could also be extenuating circumstances between now and 2023. 

Which brings me to this, a UN report brought to my attention by the good folks over at Next City. It’s basically a bunch of 2050 and 2100 projections for population on a global scale, and by country. It has some eye-catching details, like rapid growth in Africa’s share of global population, and in particular the prospect of Nigeria overtaking the US population by 2050 and challenging China by the end of the century for number 2. 

The general thrust of the idea is almost certainly true. Africa is growing faster than the rest of the world, and is likely to continue to do so in the medium term. Nigeria is large and fast growing, and really might challenge the US for the number 3 position in the coming decades. 

But lets think backwards, shall we? These are essentially 40 and 90 year projections. So let’s consider the world in 1970 and 1920. What kind of things might projections from those years have missed? A demographer in 1970 might have projected a much higher 2013 population for a (then still Maoist) China, ignorant as they would be to the coming one-child policy. They also probably would have miss-estimated the impact of women’s lib on family size, failed to predict the post-soviet collapse of Russian fertility rates, and had some silly ideas about Malthusian growth limits in the third world. The 1920’s forecast would have been even worse, especially since it probably would have missed that spot of violence in the 40s that radically shifted the demographic trajectory of a lot of Europe. 

We shouldn’t expect our 40 or 90 year forecasts to be any different. A lot can happen in those kind of time frames. And at least some of it is going to mess with population dynamics. 

The UN’s projection is one future. One possibly future. But remember the margin of error.