Read on Twitter
1/n Here a detailed thread on my paper published at @RevEconDyn (Vol. January 2021, https://authors.elsevier.com/a/1cK0~3uolWYEmc). In addition to the contributions by Carvalho, Atalay, and @DBaqaee, and Fahri, the paper shows why deviating from Cobb-Douglas production technologies is important...
2/n I show that a “new feature” of the production network: network density/diversification—whether sectors use a few of many intermediates in production—is relevant for macro volatility (I say new because Dupor 1999 and Acemoglu et al 2012 showed its irrelevance for volatility)
3/n contemporaneous to this paper, @_Herskovic shows how production network density is also important for asset prices. @DrDaronAcemoglu et al (2015) showed that banking network density matters for systemic risk and Acemoglu and @pabloazar show how density matters for GDP growth.
4/n I start by using cross-country data on production network structures and show that (after controlling for development), i) network density/diversification shapes sectoral sales shares (Domar weights) and macro volatility.
5/n The theory part of the paper shows that a version of existing production network models can account for the facts. What we need is i) to deviate from Cobb-Douglas and ii) a cost of complexity in the bundle of intermediates (naturally embedded in standard CES aggregators)
6/n when the elasticity of subs. between intermediates and labor is above one, all else equal, a denser network implies lower intermediate input shares, lower Domar weights, and, therefore, lower macro volatility. I then estimate production elasticities for the U.S and find that
7/n while aggregate estimates suggest unitary elasticity between intermediates and labor (so no role for network density), there is important sectoral heterogeneity: service sectors (largest fraction of economy) display an elasticity that is substantially larger than one.
8/n Then, assuming that sectoral technology is the same across countries, I calibrate the model to match countries' network structures. Assuming that sectoral productivity shocks are the main driver of fluctuations, the non-unitary elasticity model can deliver the facts.
9/n As an aside product, the model provides a rationale for why service sectors display lower intermediate shares than non-service sectors: services have a more diversified set of suppliers and higher substitutability between intermediates and labor. Thanks for reading!
10/n This was my JMP 4 years ago (thanks a ton to my advisors E. Young, @ToshiMukoyama, S. Osotimehin, and Pierre Sarte). I was also lucky to have the feedback and support from leading macro experts: thanks to @DBaqaee, E. Atalay, and V. Carvalho, @xgabaix @MacroInPieces, and +
11/n might be of interest to @Basile_G @ErnestoPasten2 @ErnestLiuEcon @SakiBigio (whose work inspired me to do research in this area), and @ArnaudDyevre (who wrote a nice thread on the importance of BF's work).
