Jesus Felipe of ADB gave an interesting presentation yesterday. His paper debunks the theories of total factor productivity, and hence the well-known Solow growth model. His main objections are 1) Aggregate production function is nonexistent, 2) Solow's Cobb-Douglas' growth function is not a growth function. It's a rewriting of an accounting identity at the aggregate level -- and hence, not a behavioral model, but an identity. I agree with him on the danger of aggregation. I agree that Solow proceeded his elegant 1956 paper realizing the growth accounting. But I don't see the connection of aggregation biases to his proof on identity to behavioral equations. As I raised in the discussion session, if he has problem on Solow's Cobb-Douglas function such that to him it is simply an accounting identity, then try doing it the other way around. Go from the Cobb-Douglas function and do the math back to reach the accounting identity. You WILL get there, too. I'm sure the same thing applies to CES and other functions, given the approriate assumptions on factor shares (by the way, he never discussed the main difference between Solow's growth model and the endogenous growth model: diminishing- vs constant marginal returns). And this reverse proof should also apply to micro behavior. So, if you say Solow was wrong (at macro level), you have to say that Deaton-Muelbauer was also wrong (at micro level).