Research
Working Papers
Output Without Workers: Automation and Structural Transformation
Buera and Kaboski (2009) pose a puzzle in structural transformation: why do sectoral value-added and employment shares often diverge? In this paper, value-added shares are generated by the price and income effects, while employment shares require a third force: automation’s displacement effect. Inverting the growth accounting of Jones and Tonetti (2026), I quantify this displacement effect from national-accounts data alone. In the US, displacement accounts for 64% of the model-predicted agriculture–manufacturing and 35% of the services–manufacturing employment reallocation over 1951–1989, rising to 188% and 73% over 1990–2019. After 1990, manufacturing overtakes agriculture as the sector with the fastest automation displacement.
Work in Progress
Stable Networks, Falling Growth
Since the late 1990s, R&D intensity has risen sharply in small US firms but only modestly in large firms, where sales have grown. A firm’s sales are the sum of its links to customers, so I ask which suppliers keep their customers and whether they innovate. In Compustat, supplier-customer links have lasted longer since 2000 and last longer at large suppliers with stable sales; suppliers whose links last are R&D-light. I build a growth model: the buyer picks the supplier with the lower effective price, which penalizes volatility, and new products are more volatile than mature ones. Under limit pricing, a mature leader charges a higher markup, is harder to replace and, because innovating means giving up its mature product, innovates less. In illustrative simulations, a rising new-product discount lengthens leader tenure, raises markups and, beyond a point, slows growth.