In Understanding Business Fluctuations Not all GDP is Equal

In standard macroeconomic models, GDP is GDP and an industry is important only to the extent that it produces GDP. In other words, there are big industries and small industries but there are no special industries. The standard view implies that the structure of production can be ignored. It doesn’t matter, for example, whether an industry sells to final consumers or to other businesses. It doesn’t matter whether an industry sells an easy-to-substitute product or a hard-to-substitute product. And it doesn’t matter whether an industry is a weak-tie link or a strong-tie link. Thus, in the standard view, the oil industry is no more important for understanding economic fluctuations than say the retail sales industry, since they are about the same size.

The standard view isn’t arbitrary, Hulten proved that it was true under certain assumptions. Hulten’s theorem appeared to offer a very useful simplification and so it diverted the attention of economists away from trying to model the structure of production.

But I’ve always been skeptical. One reason is that rapid increases in the price of oil have preceded almost all U.S. recessions (see Hamilton’s papers) and such increases appear to be much more important than the size of the oil sector would allow. More generally, although the canonical real business cycle model emphasizes aggregate technology shocks, I’ve always looked favorably on variations with sectoral shocks–this is one reason why Modern Principles is one of the few principles of economics textbooks to devote significant attention to real shocks, including oil shocks, and the mechanisms that can transmit and amplify these shocks from one sector to the wider economy.

The sectoral approach gets new support in a recent paper, The Macroeconomic Impact of Microeconomic Shocks: Beyond Hulten’s Theorem (non-gated) by Baqaee and Farhi. The authors show that Hulten’s theorem offers a true simplification only under very restrictive conditions. Relax those conditions and what seem like second-order mechanisms can have first-order effects. The structure of production–how industries are linked to one another, elasticities of substitution, reallocation speeds and so forth–matters in theory.

In a calibration Baqaee and Farhi consider “the response of GDP to shocks to specific industries” and:

…It turns out that for a large negative shock, the “oil and gas” industry produces the largest negative response in GDP – this despite the fact that the oil and gas industry is not the largest industry in the economy.
Hulten’s simplification came at too high a price. Further progress in understanding business fluctuations will come from developing microeconomic foundations to macroeconomics–not simply by subjecting a utility-maximizing representative agent to an optimization problem but by investigating the real structure of production in all of its complexity.


Comments for this post are closed