This paper investigates the effect of export shocks on innovation. On the one hand a positive shock increases market size and therefore innovation incentives for all firms. On the other hand it increases competition as more firms enter the export market. This in turn reduces profits and therefore innovation incentives particularly for firms with low productivity. Overall the positive impact of the export shock on innovation is magnified for high productivity firms, whereas it may negatively affect innovation in low productivity firms. We test this prediction with patent, customs and production data covering all French manufacturing firms. To address potential endogeneity issues, we construct firm-level export proxies which respond to aggregate conditions in a firm's export destinations but are exogenous to firm-level decisions. We show that patenting robustly increases more with export demand for initially more productive firms. This effect is reversed for the least productive firms as the negative competition effect dominates. 5

We analyze how demand conditions faced by a firm impacts its innovation decisions. To disentangle the direction of causality between innovation and demand conditions, we construct a firm-level export demand shock which responds to aggregate conditions in a firm's export destinations but is exogenous to firm-level decisions. Using exhaustive data covering the French manufacturing sector, we show that French firms respond to exogenous growth shocks in their export destinations by patenting more; and that this response is entirely driven by the subset of initially more productive firms. The patent response arises 3 to 5 years after a demand shock, highlighting the time required to innovate. In contrast, the demand shock raises contemporaneous sales and employment for all firms, without any notable differences between high and low productivity firms. We show that this finding of a skewed innovation response to common demand shocks arises naturally from a model of endogenous innovation and competition with firm heterogeneity. The market size increase drives all firms to innovate more by increasing the innovation rents; yet by inducing more entry and thus more competition, it also discourages innovation by low productivity firms.

We study the effects of fiscal incentives for self-employment using new French tax data from 1994 to 2012. France serves as a good quasi-laboratory: It has three fiscal regimes -or modes of taxation -for the self-employed, which differ in their financial payoffs and in their administrative simplicity. These regimes have changed extensively over time -offering the opportunity to study how people learn about them and understand them. We find that the self-employed respond to the tax and administrative notches created by the eligibility thresholds: there is strong bunching right before the eligibility thresholds, which we use to estimate self-employed taxable income elasticities and the value of administrative simplicity. Even a small preference for administrative simplicity could explain the bunching observed. There is a sizable cost of tax complexity; agents are not immediately able to understand what the right regime choice is and there is evidence for costly learning over time. The cost of complexity is regressive because it affects mostly the uneducated, low income, and low skill agents. Agents who can be viewed as more informed and knowledgeable (e.g., the more educated or high-skilled) are more likely to make the correct regime choice and to learn faster.

We study the effects of fiscal incentives for self-employment using new French tax data from 1994 to 2012. France serves as a good quasi-laboratory: It has three fiscal regimes -or modes of taxation -for the self-employed, which differ in their financial payoffs and in their administrative simplicity. These regimes have changed extensively over time -offering the opportunity to study how people learn about them and understand them. We find that the self-employed respond to the tax and administrative notches created by the eligibility thresholds: there is strong bunching right before the eligibility thresholds, which we use to estimate self-employed taxable income elasticities and the value of administrative simplicity. Even a small preference for administrative simplicity could explain the bunching observed. There is a sizable cost of tax complexity; agents are not immediately able to understand what the right regime choice is and there is evidence for costly learning over time. The cost of complexity is regressive because it affects mostly the uneducated, low income, and low skill agents. Agents who can be viewed as more informed and knowledgeable (e.g., the more educated or high-skilled) are more likely to make the correct regime choice and to learn faster.