During the course of research my co-author Maggie McMillan and I were doing on broad patterns of structural change, our research assistant Inigo Verduzco stumbled on an interesting stylized fact that is captured in the figure below.
What it shows is that the relative productivity of agriculture exhibits a U-shaped pattern over the course of development. It first falls and then rises, as countries get richer. (By productivity, I am referring to labor productivity.)
You can see the same pattern within countries as well. Here is a chart that collates the trends in three countries at different levels of development: India, Peru, and France:
In India, which is the poorest country, as overall productivity has increased, the productivity of agriculture (relative to other activities in the economy) has come down steadily. In France, the richest country, agricultural productivity has converged with that of the rest of the economy. In Peru, which is somewhere in between, the relative productivity of agriculture has fluctuated around the minimum point of the U-curve.
What these results point to are two critical dynamics that take place during the course of development. First, you need some new, high-productivity activities to emerge in order for development to happen. These typically arise outside of agriculture. So as these new activities propel the economy forward, agriculture falls behind in terms of relative productivity. Second, labor tends to move from low- to high-productivity activities as an economy gets richer. This tends to reduce the gap between productivity in agriculture and non-agriculture (think of diminishing marginal returns). At some point, the second dynamic overwhelms the first, and the curve begins to slope outward.
Lesson: economic development requires both new activities (diversification) and ongoing transfer of resources from traditional to modern activities. Some countries are stuck with no new industries, so they never grow. Others get a few new industries (e.g. mining and other natural resource-based industries), but these do not expand sufficiently and absorb much labor, so development gets stuck at an intermediate level of income. The real successful countries are those that pull off both tricks.