Jeremy Singer-Vine at Slate noticed a new Pew study predicting that the global Muslim population will grow twice as fast as the non-Muslim population over the next twenty years. The report, released today, is simply one example of population forecasting. At the core of most reports on demography is not just the make-up of the current population, but also predictions about the future population. Rather than taking them for granted, Singer-Vine investigated how these predictions are made and how accurate they are. In short, he found that demographers use three major variables: the fertility rate, the morality rate and the immigration rate. These three variables may appear obvious choices, but predicting them for years or decades into the future remains challenging. The baby boom of the middle of the twentieth-century, for instance, defied predictions by demographers.

Failure to accurately predict these variables has held back improvements in population forecasting which continue to be wrong at the national level by about six percent on average. In fact, estimates for developed countries have barely improved in a century (in developing countries, where record-keeping a century ago was quite poor, forecasting ability has improved) and the UN's worldwide predictions haven't gotten any more accurate since the 1950s.