How accurate can we be about Population Growth?

There was an interesting article in the Economist last week (July 23rd edition) about the growing dependency ratio in the developed world and whether this is sustainable. (the dependency ratio is the ratio of the number of working aged people to number of non-working age people).

It was definitely an interesting read, and I will post a link below, it raises interesting questions about the transfer of wealth from one generation to the next and the fairness of this transfer in a system where there are fewer working age people paying in.

One thing about the article that struck me though was that the UN Population Division publishes estimates of the working age population all the way out to 2100! I can't even predict what I will have for lunch tomorrow. Is this a case of the UN being wildly optimistic about their modelling skills? Or perhaps over such a large data sample (the entire human population) the law of large numbers prevails and we can pick out some clear trends which we expect to continue in the future.

Let's try to look into how good these population models are.

My first step was to try to find estimates of population growth published by the UN Population Division in the past and see how accurate they had been.

Every couple of years the UN Population Division publishes a report on World Population Prospects, I managed to find report going back to 2000, link below for the 2000 addition:

http://www.un.org/esa/population/publications/wpp2000/highlights.pdf

Previous Estimates

I found as many of these reports as I could and chucked the data into a table. Luckily the reports are fairly standardised and it is possible to look at the sequence of estimates.

We see that the best estimate of the World Population in 2050 has actually been relatively stable since the year 2000. The standard deviation of the medium estimate is only in the range of 200m, which to me seems like a reasonable volatility for such a far reaching statistic.

An interesting aspect of this data is the drop in the last column between 2000 and 2004. The final column represents the estimate of the population in 2050 assuming that birth rates do not decline. Since we do not have a large shift in the medium estimate between these two dates, this drop in birth rates must have been factored in for the 2000 analysis.

So to answer the original question - how accurate can we be about Population Growth - so far pretty accurate. If the UN is able to accurately model the change in birth rates, then perhaps there is hope that they will be able to model population as a whole accurately.

Link to the original article from the economist:

http://www.economist.com/news/finance-and-economics/21702477-can-debt-fuelled-model-growth-cope-ageing-populations-vanishing