- China: 1,346,197,000
- India: 1,198,372,00
- United States of America: 315,419,000
- Indonesia: 230,452,000
- Brazil: 194,481,000
- Pakistan: 181,507,000
- Bangladesh: 162,531,000
- Nigeria: 155,553,000
- Russian Federation: 141,574,000
- Japan: 127,593,000
- India: 1,614,000,000
- China: 1,417,000,000
- United States of America: 404,000,000
- Pakistan: 335,000,000
- Nigeria: 289,000,000
- Indonesia: 288,000,000
- Bangladesh: 222,000,000
- Brazil: 219,000,000
- Ethiopia: 174,000,000
- Dem. Republic of the Congo: 148,000,000
I understand the UN demographers' predicament. Any more sophisticated model will inevitably make assumptions that (at least to some observers) appear even more arbitrary and wrongheaded, and for political purposes it's safer to go with the crude but simple approach. It's also difficult to imagine what alternative methodology could consistently be applied to the entire world and still produce reasonable results. Caplan's suggestion—that we extrapolate current trends, assuming that existing declines in fertility will continue in the near future—raises a host of potential problems. How do we extrapolate? Presumably we'll need to choose an arbitrary functional form, and unless human biology makes it possible to have negatively many children, we'll have to place some arbitrary floor on each country's fertility. To make matters more complicated, fertility in a lot of developed nations (the US, the UK) is now rising. How long will this increase last, and where will it stop? How do we "extrapolate" it? Certainly fertility in the US is not going to rise above 2.5, but how can we embed this intuition in a model? How can we predict when currently low-fertility countries will take a sudden upward turn, like the US, which has gone from 1.79 to 2.09 children per woman over the last 30 years? This is hard stuff!
On a case-by-case basis, however, I think that it's possible to make informed guesses that improve on the UN's projections. The best example is China. Currently fertility there is stagnant at about 1.77 children per woman. Perhaps it will stay at that level, or nudge upward slightly, making the UN's 1.85 estimate look reasonable. But the closest cultural analogues to what a fully developed China might look like, Hong Kong and Macao, have fertility of around one child per woman, the very lowest in the world. Taiwan is at 1.1. South Korea and Japan, the other large, developed nations in East Asia, are also near the bottom of world fertility, at 1.22 and 1.27 children per woman respectively. This seems to be a regional trend.
Meanwhile, urban China has lower fertility still, and if this holds as the nation rapidly urbanizes, it's hard to imagine that we'll see any increase in fertility. Indeed, the overwhelming weight of circumstantial evidence suggests a decline in the making. True, the fertility level has leveled off over the last decade, but this may be a response to earlier misestimation: as women of the new generation delayed having children, the total fertility statistics measured this as a "decline," and now they're playing catch-up. (This seems to be a common story worldwide.)
You could argue that the "One Child Policy" — something of a misnomer, since it doesn't prevent many women from having more than one child — has artificially deflated fertility, and women will have more children as the government eases. I'm not so sure. The fact that fertility declines were partly coercive doesn't negate their potential to set societal norms. If one or (at most) two child families become the standard in Chinese culture, it doesn't matter how this standard was originally put in place. And again, the regional comparisons are so overwhelming: developed East Asian nations have the lowest fertility in the world, bar none.
With all this in mind, I suspect that the birth rate in China will continue to decline, and by 2050 we will see a China with a population of 1.2 billion (and falling). But there's also a very good chance that I'll be wrong. Hardly anything is more difficult to predict than demographics, and few in 1970 had the slightest idea about the world 40 years hence.