Putty-Clay Automation

Even as automation and AI reshape labor markets, their adoption differs significantly across firms. This paper studies the coexistence of technologies in an economy where automation capital is putty-clay so that firms cannot adjust their mix of human and automated inputs after capital is installed. It has two main findings. The first is that the effects of increased automation differ depending on which firms become more automated. Increased automation at frontier firms reduces the labor share, whereas automation among laggard firms raises wages and the labor share. I validate this theoretical finding by constructing cross-firm measures of routine employment shares as a proxy for automation and using moments of this distribution to study the equilibrium effects of greater automation among different groups of firms. Given the coexistence of diverse technologies in the economy, I then ask whether the competitive equilibrium allocation of capital across automation technologies is efficient. The second main finding is that, for a given level of frontier technology, there is overinvestment in the lowest and underinvestment in higher automation technologies. I estimate the model to match the empirical cross-firm distribution of routine shares and find that these inefficiencies lead to substantial productivity losses (1-9% in European region-industry pairs) and that wages in the decentralized social optimum are higher (1.5-14%) than in the competitive equilibrium.