|August 3rd, 2013
People who want to do whatever most improves the world still have varying preferences and talents, so I suspect that fewer than 99 out of 100 prospective effective altruists would like going into earning to give. This suggests that people on the fence should lean towards the "make money, give money" approach.
(This is very rough and ignores a lot of issues, but is an attempt to get the scale of things.)
Update 2015-08-20: This overlooks that people earning to give aren't the only source of funds. Foundations etc have a lot of money. It also ignores that GiveWell's recommendations are based off the work of a lot of development researchers, which I'm not counting.
 This is my estimate from reading their "room for more funding" analysis on each of their top charities. They tend not to give a specific number, but an upper bound. They also don't mention annual vs total numbers, but a lot of their limits are based on how much money the charity currently has experience handling and not the size of the problem. (AMF, GiveDirectly, SCI)