Almost thirty years ago I was a Watson Fellow. In between recreational diversions in various corners of the world I made a serious study of how one might try to change fertility behaviors.
Even then we'd known for a long time that as nations "industrialized" (became technocentric) families became smaller (Mormons are an interesting exception).
This disturbed me. Back then what's now "evolutionary psychology" was "sociobiology" (name change for branding reasons), and within that framework I couldn't understand how the heck wealth produced infertility. I thought the usual theories  were nonsensical in the long term, but they were widely accepted.
My best guess was that technocentric wealth was so different from anything in human evolutionary history that we were "outside system bounds" and that women (in particular) were behaving "irrationally" . In time we'd adapt to our new technocentric world, fertility would climb and we'd be back to the core business of churning out copies of our genes (children). 
I've been waiting thirty years for biologists and behavioral economists to recognize that this was an unsolved mystery. Now, at last, some new thinking has come along ...
...Now yet another explanation has been added to the pot. This is that the mixing-up of people caused by the urbanisation which normally accompanies development is, itself, partly responsible. That is because it breaks up optimal mating patterns. The demographic transition is thus, in part, a pure accident...
This is a variation of my "out of bounds" theory. It implies that we'll eventually adapt to urbanisation and that family size will again increase.
It's good to see this topic is in play. Took long enough!
 I was looking for ways to reduce fertility, so this suggested that we shouldn't rely on the wealth effect to shrink family size forever.
 Lowered infant mortality, economic benefit of education and employment of women, etc, etc.
 In a Darwinian sense.
Update 2/14/08: Another new theory. It's nice to see this paradox getting a bit of attention!