Comments and Questions

My impression is that the referee is right that the authors have added the same thing to both sides of the equation. If he is right, then it seems that the reported results tell us little. The question would be settled if the authors would add a brief mathematical discussion of the national income accounting issues involved. Here is how I see the problem. The issue does not concern foreign borrowing, even though that is what ultimately is being investigated, so let's drop it and work with a closed economy. The same point emerges when openness is included. We write the national income identity C+I+G = C+S+T. The left side is measured by expenditures. Education expenditure is carried out by households, private investors, and government, so it is parceled out among all three of C, I, and G. On the other side, we have total income, which don't measure by adding C, S, and T but rather by measuring income directly. We then argue that C is the same on both sides and cancel it. We can measure T directly, leaving S as a residual. This is standard stuff, but it is important to be clear about it. Now let's consider expenditure on education. The authors are right that such expenditure is human capital investment that should be included in I. They therefore want to add total expenditure on education to I on the left side, which is correct. On the right side, they want to move household expenditure on education from C to S, which also is correct. They emphasize in their reply to the referee that they are adding total expenditure on the left and only private expenditure on the right. However, that seems in fact *not* to be what they are doing. They explain in their reply to the referee that they add OECD total expenditure and then subtract government expenditure because it *already is included in I*. In other words, the new item they are adding on the left is indeed private expenditure, which is exactly what they add on the right. If my understanding is correct, then the referee's third concern is also correct and the authors have a serious problem. If my understanding is incorrect, then the authors must explain carefully in the paper what is going on with the accounting so that readers like me don't get confused. In any case, it would be useful for the authors to include a table showing the following: I and S as originally measured by F-H, the net series the authors add to the left side (that is, OECD numbers less government expenditure on education), and the numbers the authors add to the right side. That will allow the reader to see two things. First, how the numbers added to the two sides compare (which is what the referee and I are concerned about), and second, how the total values of I and S are changed by these additions. If the numbers added are small relative to the original values of I and S, then it is no surprise (though perhaps still interesting) that the regression results are not changed.