Property rights play a big part here. Why work to improve something when there is not much guarantee it will be yours once improved?
Many areas of the world where yields are low suffer not only form lack of money and technology but property rights are likely as significant.
Property rights are key. One stat I read (that I can't remember where) said that about 70% of the people in the world don't have a formal title to the land they live on.
A buddy of mine actually has a start-up that focuses on that called Suyo: http://suyo.co/
Sure you could get _an_ African farmer to American level yields, but that's different than what essentially is economic reorganization for the _majority_ of workers in African economies.
Also, remember this isn't an agile software project that can go through an iteration a day. You plant once a year, you harvest once a year; an entire continent doubling yields in only 15 iterations is pretty good.