Cumulative Innovation and Dynamic R&D Spillovers (Job Market Paper)
Abstract: While much theoretical attention has focused on the important role of dynamic knowledge spillovers for economic growth, such spillovers have been difficult to empirically measure. Using a panel of US firms and a network of corporate patent citations, this paper estimates the dynamic spillovers of corporate R&D on firm productivity, value, and innovation activity. Causal effects are estimated with an instrumental variables strategy that exploits the persistence of the network as well as variation in tax incentives. The positive effect of dynamic spillovers on firm productivity is economically important, and at least as large as that of own R&D investments. Dynamic spillovers accrue mainly for so-called ”complex” technologies that build cumulatively on multiple components, they exhibit little depreciation over time, are larger for established firms than for VC-backed startups, and do not decrease in magnitude with geographic distance. Around 40% of dynamic spillovers are re-absorbed by the original innovator, and –accounting for other spillovers– these estimates suggest that the social returns to R&D are about three times as large as the private returns.
Patent Duration and Cumulative Innovation: Evidence from a Quasi-Natural Experiment (joint with Jean-Noel Barrot)
Abstract: Cumulative innovation is at the core of economic growth, but the impact of patent policy on it is not well understood. This paper investigates whether patent term duration affects the rate and direction of follow-on innovation. We use a quasi-natural experiment that lengthened the term on existing patents in the US, and leverage a kink in the extension formula to identify the effect of patent term increases. We find no statistically significant impact of patent extensions on subsequent innovation, neither locally around the kink using a sharp "Regression Kink Design" nor on average on the population of treated patents. We further analyze whether the null average effect could be masking important heterogeneous effects, and find no such evidence.
Durable Crises (joint with Nicolas Caramp and Pascual Restrepo)
Abstract: Durable goods consumption is highly cyclical: it falls substantially during recessions and rises sharply during booms. Using U.S. County Business Patterns data between 1988 and 2014, this paper studies how durable manufacturing industries amplify business cycle fluctuations. We show that employment in durable manufacturing industries is more cyclical than in other industries, and the cyclicality is amplified in general equilibrium at the commuting zone level. We provide evidence of three mechanisms that generate amplification. First, employment changes propagate through input-output linkages, which amplify effects on local aggregate employment because industries co-locate. Second, the reduction of employment in durables negatively affects employment in non-tradable sectors, consistent with the existence of demand externalities. Third, we find that workers do not completely reallocate to other less cyclical tradable industries. Our estimates suggest that consumer durables amplify the impact of aggregate shocks on employment volatility by up to 40%.
Research in Progress
Knowledge Diffusion and the Patent Citation Network
Abstract: Innovation activities form networks through which information flows and diffuses across researchers, inventors, and firms. This paper studies knowledge diffusion patterns across the patent citation network. Efficient algorithms are applied to compute centrality measures on a large, yet sparse, network. The betweenness and eigenvector centrality measures paint a more complete picture of knowledge diffusion than the more commonly used degree centrality, or simple forward-citation count. Using a subset of corporate patents held by publicly listed firms, this paper identifies key players in the innovation network.
Error Dependence Beyond Two-Way Cluster: A Panel Spatial Autoregressive model for Non-Contemporaneous Spatial Dependence (joint with Lucciano Villacorta)
Abstract: Considering error dependence when computing standard errors is crucial for conducting valid inference in applied work. One-way cluster and two-way cluster estimators are used for robust standard errors to time and (contemporaneous) spatial dependence in panel data. However, more complicated structures of dependence across time and space are not usually taken into account. This paper proposes a panel version of the spatial autoregressive model to consider non-contemporaneous spatial dependence. The covariance estimator consists of a non-parametric two-way cluster, and a parametric part with a known NTxNT weighting matrix for the non-contemporaneous spatial dependence. The model is suitable for settings where the researcher has a prior knowledge about the dependence structure. We apply this estimator to Colino (2016), in which non-contemporaneous spatial dependence arises because of dynamic spillovers of R&D accruing between firms and across time.