Below I take up a challenge from Matthew Yglesias to describe an economic model in which increased numbers of scientists and engineers in China and India (from now on chindia) imply that the US should increase public spending on science and engineering education. He notes that , in the simplest model, scientists and engineers in Chindia are substitutes for scientists and engineers in the USA so the optimal response would be to lower US spending the more scientists and engineers there are elsewhere.
I had two semi joke models. Now I think of a model which is almost not totally ridiculous (still ridiculous). One can argue that the services supplied by scientists and engineers in Chindia complement the services supplied by scientists and engineers in the USA even if they all have the same knowledge and skills.
The argument is that the most profitable use of engineers in poor countries is to assist the adoption (imitation if you want to be rude) of rich country technology -- that they are needed to design factories and to explain to untrained workers what to do with such factories. In rich countries, these needs are met and, besides, it takes fewer engineers to expand a factory than to design a new one (one can copy the old part) and many fewer to train new workers if the firm employs experienced workers who have learned on the job and can show the new workers what to do. Also chindian engineers might profitably spend their time adapting the technology appropriate to capital rich labor poor countries to their own labor rich capital poor countries. No sane manager would pay an engineer to do any such thing in rich countries.
Thus engineers in the USA are likely to work on developing new technology -- new products and processes. The increased supply of old products from Chindia increases the relative price of new inventions increasing the marginal product of engineers in rich countries.
Thus, even if the engineers are identical except for their location (and language skills) they may provide their employers with services which are different and, indeed, complementary.
Can such a model be written ? Sure. Do I believe it is relevant to the real world. Certainly not.