The dissertation consists of three chapters examining U.S. public firms' innovation and patenting activities and their relationship with patent policy and economic growth. In the first chapter, I empirically study the effect of patent publications on firm-level innovation and patenting. Previous works have studied the effect of patent monopoly rights and knowledge disclosure on innovations. The proposed chapter supplements these studies by analyzing the disincentive effect of patent publications on firm innovations through costly knowledge disclosure. Exploiting the American Inventors’ Protection Act of 1999 as a natural experiment that shortened the time it took for patents to get published, I show the negative effect of earlier patent publications on manufacturing firms' patenting and innovation activities. The benchmark analysis shows that the average decline of 10 months in patent publication lag resulted in 13 percentage points lower firm-level patent growth rate during 2001-2005. In the second chapter, I build an endogenous growth model with a patent system. By modeling patenting decisions endogenously, I also introduce patent protection and information disclosure mechanisms through patents. Traditional innovation and growth models assume that innovators patent whenever they innovate and consider patenting and innovating as the same. However, this assumption is no longer innocuous if patenting has an implicit cost to the innovator e.g., the cost of disclosing valuable information. Therefore, to analyze the impact of the patent system’s disclosure mechanism on firm innovation, one must at a minimum work with a model distinguishing between the two concepts. Using my model, I show that a higher patent disclosure policy reduces firm patenting intensity as firms strategically opt out of patenting. In the absence of patents, there is less knowledge diffusion in the economy, which leads to less industry competition and growth. The third chapter studies the effect of firms' ability to build on their previous innovation on firm growth. While innovating, firms can either develop fully novel exploratory ideas or exploit their existing ideas. Using firm patent data, I document that U.S. manufacturing firms' innovation became more exploitative and that their patent growth rate simultaneously declined after 2000. To rationalize these changes in firm innovation, I build a firm-level endogenous growth model with both initial exploratory and subsequent exploitative innovations. Estimating my model using 1990-2000 microdata, I show that a decline in the usefulness of exploratory innovations as a foundation for future exploitation can match a shift in the composition of innovation we saw over this period, resulting in a 0.8 percentage point decline in firm average growth and a 9% decline in firm market value post-2000.

L'innovation engendre la croissance économique et permet d'accroître le niveau de vie des populations. Cette proposition de Joseph Alois Schumpeter (1912) a été plutôt confortée que contestée par les faits empiriques. De nombreux travaux ont en effet montré que le lien entre innovation, croissance et niveau de vie existe. Ainsi, il existe un assez large consensus sur l'idée qu'encourager l'innovation devrait constituer l'un des objectifs prioritaires de la politique économique. Cependant, de nombreuses innovations peuvent affecter la richesse, la santé ou même la liberté de l'ensemble de la société ou de certains de ses membres. Par exemple, les semences OGM peuvent permettre d'accroître la production agricole, mais peuvent également avoir un impact négatif sur la santé humaine et l'environnement. C'est également le cas de nombreuses innovations à caractère médical, les nanotechnologies, certains logiciels, certains produits financiers, etc. Partant de l’hypothèse que l’innovation peut être génératrice d’effets négatifs, cette thèse propose un cadre théorique pour guider le décideur public dans la mise en œuvre d’une politique avisée. Dans un premier temps, nous évaluons tour à tour la performance de plusieurs instruments permettant de réglementer une innovation potentiellement nuisible. Nous examinons deux règles de responsabilité distinctes et nous caractérisons les conditions sous lesquelles chaque réponse réglementaire devrait être mise en oeuvre par les pouvoirs publics. Contrairement à des travaux antérieurs, nous montrons notamment que même si l’innovateur a une responsabilité illimitée, les autorisations peuvent être optimales. D’autre part, en cas de responsabilité limitée, nous établissons que les pénalités ne devraient jamais être privilégiées. Dans un second temps, nous déterminons la politique de brevet optimale lorsque l’innovation est à l’origine d’une externalité négative. Dans ce cas, la largeur du brevet influence non seulement la perte sèche induite par la propriété industrielle mais également la taille de l’effet externe. Nous proposons plusieurs interprétations du concept de largeur et nous caractérisons la structure optimale du brevet. Par ailleurs, nous nous penchons sur la question de savoir quelle récompense attribuer aux innovateurs et nous montrons notamment qu’une politique de brevet modulée gratifiant celles et ceux ayant mis en œuvre une stratégie de Recherche & Développement (R&D) précautionneuse d’une protection industrielle plus forte pourrait être un instrument à même de contenir l’émergence d’innovations génératrices de nuisances. Enfin, nous considérons que plusieurs entreprises sont engagées dans une course à l’innovation et nous caractérisons les conditions sous lesquelles une augmentation de la rivalité conduit ces dernières à précipiter le processus de R&D au risque d’introduire une innovation nuisible. Nous proposons plusieurs méthodes de mesure du degré de rivalité et nous montrons notamment que, dès lors qu’il existe un certain équilibre entre les coûts et les bénéfices engendrés par la précaution, la rivalité rend les entreprises plus enclines à négliger les actions préventives/correctives et à briser la séquentialité du processus de R&D. / Innovation drives economic growth and it increases the standard of living of populations. This proposition of Joseph Alois Schumpeter (1912) has been rather confirmed than challenged by empirical evidence. Indeed, much research has shown that the link between innovation, economic growth and standard of living does exist. Hence, there is a fairly broad consensus on the idea that fostering innovation should constitute one of the main objectives of the economic policy. However, many innovations can affect the wealth, the health, or even the freedom of the society as a whole or of some of its members. For instance, genetically-modified seeds enable farmers to increase agricultural production but they may also have a negative effect on public health and the environment. This is also the case of nanotechnologies, certain software, some financial products, etc. On the assumption that innovations may induce social harm, this thesis provides a theoretical framework in order to guide the regulator in implementing a sound policy. First, we successively evaluate the performance of several instruments available for regulating potentially harmful innovations. We examine two distinct liability rules and we characterize the conditions under which each regulatory response should be implemented. Unlike previous work, we show that even though the innovator has unlimited liability, authorizations may be optimal. Furthermore, in case of limited liability, we establish that penalties should never be favored. Second, we determine the optimal patent policy when the innovation generates a negative externality. In that case, the patent breadth not only influences the deadweight loss induced by the industrial property, but also the size of the external effect. We provide several interpretations of the concept of breadth and we compute the optimal patent structure. Besides, we address the question of how much to reward innovators and we show that a modulated patent policy giving stronger protection to those who have implemented safe Research & Development (R&D) strategies might be an instrument with the potential to prevent the emergence of harmful innovations. Last, we consider several firms engaged in an innovation race and we determine the conditions under which an increase in rivalry leads them to hasten the R&D process at the risk of introducing a harmful innovation. We provide several methods of measurement for the degree of competition and we show that whenever there is some balance between the cost and the benefit induced by safety, then rivalry makes firms more inclined to overlook preventive/corrective actions and to break the sequentiality of R&D.