Abstract: Individual-based transmission models have found applications in fields such as epidemiology, information diffusion, cyber security, and marketing science. These models involve stochastic rules that specify how a number of individuals would infect one another, recover or be removed from the population. Contrary to compartmental models which are widely-used, individual-based models do not assume interchangeability of individuals and that all pairwise contacts are equally likely. Such modelling flexibility comes at the price of high computational cost when fitting these models to data. I will discuss on-going research to design computationally feasible methods that are scalable to large population sizes. 

Abstract: Economic nationalism refers to government actions that enhance national interest by protecting and aiding domestic firms at the expense of foreign ones. While prior research has highlighted discriminatory laws and government policies, we contend that economic nationalism may manifest as policies that are non-discriminatory but are nonetheless unequally and differently enforced among domestic and foreign firms by actors entrusted to implement them. We explore this subtler mechanism of economic nationalism in the context of judicial rulings in property rights disputes.  Through its interpretation and enforcement of policies and laws, we expect the legal system to create a disadvantage for foreign firms. In a sample of 58,754 intellectual property lawsuits in the United States between 1963 and 2016, we find domestic patent holders are 10% more likely to successfully defend their patents in lawsuits against foreign (compared to other domestic) firms. Similarly, foreign patent holders are 17% less likely to win against domestic (compared to other foreign) challengers. Our results are robust to controlling for firm and patent characteristics and appear unlikely to be driven by differences in legal capabilities. Thus, the legal system appears to be another source of economic nationalism, giving domestic firms the home-court advantage.

Abstract: We explore the impact of gender on employee reactions to and support of their CEO. Employing the lens of group competition theory to understand the female leadership penalty, we theorize that negative reactions toward women in senior positions are motivated in part by the perceived threat these women pose to the existing gender hierarchy. As women increasingly move into positions previously held by men, and as diversity programs lead to shifts in the balance of power, gender is likely to become more salient and trigger negative reactions. Exploiting data from close to a million employee reviews of U.S. listed firms from Glassdoor.com, we show that female CEOs receive lower employee approval ratings compared to male CEOs, and that this is driven primarily by male employees. Our results are robust to controls for firm performance, reviewer ratings of employment conditions, as well as indicators of managerial quality as measured by educational credentials, board memberships, and other achievements. We further find that the organization’s diversity ratings moderate the relationship between CEO gender and employee approval, such that female CEOs are especially penalized among firms that have made substantial progress in promoting women and minorities. Implications for female leadership are discussed. 

Abstract: Customers often have negative service experiences. But might a subtle way customers describe such experiences shed light on how likely they are to voice or escalate their complaints (e.g., share negative word of mouth or dispute an offered resolution)? A multimethod exploration, combining automated textual analysis of over 160,000 consumer complaints with experiments, demonstrates the important role of passive voice. Consumers who complain using passive voice are more likely to spread negative word of mouth or dispute the resolution offered by the company. Greater use of passive voice indicates consumers attribute more fault to the company (rather than themselves), which leads them to voice or escalate their complaint. These findings shed light on an understudied aspect of language (i.e., linguistic structure), deepen understanding of how language reflects attitudes and intentions, and provide insight into how managers should respond to dissatisfied customers. 

Abstract: We build a new global database of effective tax rates on factor incomes covering 1965 to 2018.  While labor taxation has steadily risen everywhere, we uncover an asymmetric evolution of capital taxation: effective tax rates on capital have fallen in OECD countries, but risen in developing countries since the mid-1990s. To explain this asymmetry, we show that the large increase in trade since the 1990s exerts two opposing forces: globalisation pushes countries into a "tax race to the bottom" to attract capital, but also improves countries' tax capacity, in particular via the share of output produced in large corporations where capital is easier to tax. This latter force was dominant in developing countries, but not in OECD countries .

Abstract: Online innovation communities are increasingly becoming a primary source of innovation for organizations. Past research on product development processes has mostly focused on individual product characteristics, with novelty — meaning how uncommon a product is compared to preexisting ones — having a prominent role in these endeavors. However, while some have recommended striving for the development of novel products, others have advocated against novelty as it not only creates uncertainty, but also increases search costs for consumers. 

In this work, we examined products developed in the largest open innovation community for 3D printed products to date, Thingiverse. In order to provide insights on the role of novelty in a product’s success, we collected more than 35,000 product designs that were created over 4.5 years by more than 8,000 creators.

Our work provides strong evidence that products that were novel either in terms of their form, or in terms of their function were used more than their counterparts. However, in contrast to past findings and this overall strong preference for novel products, consumption is lower when a product is novel both in terms of its form as well as in terms of its function. In addition, our work demonstrates how the structure of the design landscape — meaning products that preexist — affects the subsequent development of novel designs.

Understanding these relationships can help managers and contributors alike to predict which products are more likely to become successful. Our research also highlights the importance of studying design landscapes in general, rather than merely focusing on individual users or product designs, as well as provides practical advice on designing and managing online innovation communities.

Cerema (Centre d'études et d'expertise sur les risques, l'environnement, la mobilité et l'aménagement - Centre for Studies on Risks, the Environment, Mobility and Urban Planning) is a public institution dedicated to supporting public policies, under the supervision of the ministry for ecological transition and regional cohesion. 

Abstract: This paper investigates whether, and if so, for which topics, companies incorporate in their disclosures content discussed by individual and institutional investors to address their information needs. We use LDA, an increasingly popular machine learning technique, to identify firm-related topics discussed by individual (via tweets posted on Seeking Alpha) and institutional (via earnings conference call transcripts) investors and compare them with topics discussed in the subsequent MD&A section of the 10Ks. We find a positive and statistically significant association between the proportion of text dedicated to a subset of topics in individual investors’ discussions (i.e., financing activities, investment and future growth, R&D, and revenue forecast) and the proportion of text on the same topics in subsequent firm disclosures. A similar relation holds for all topics regarding the discussions of institutional investors. Consistent with the notion that firms weigh the costs and benefits of addressing investors’ information demands, we find that this association is stronger when the tone of investors’ discussions is more negative, for more powerful investors, and for firms subject to lower proprietary costs. Overall, our findings extend the research on corporate disclosures and their content.  

Abstract: Multiple-stage estimation is a widely used technique in econometrics to handle issues such as endogeneity, missing data, and latent variables. This approach involves estimating controls, which are then plugged into a model that can be estimated using conventional methods. The asymptotic properties of the resulting "plug-in" estimator are of interest for hypothesis tests and can be difficult to determine as they depend on the precision of the control estimator. We examine these properties for a general class of control estimators when the final-stage estimator is an M-estimator. Our approach allows for the use of Bayesian or nonparametric estimators for the controls. We introduce a simple simulation method to approximate the variance and distribution of the plug-in estimator, taking into account the uncertainty of the control estimates. Our method can be easily applied to complex models as we do not compute the control and plug-in estimators on several subsamples, as is required in the Bootstrap approach. We applied the method to an auction model to study bidder behavior in the weekly open market operations conducted by the Central Bank of West African States. 

Abstract: While there’s a lot of buzz around AI, I’d like to put this field in perspective by quickly summarizing how we got there and discussing what AI and robots are really capable of doing today. My critique will then focus on the use made in recent years of AI and robotics that tends to robotize humans to fill the gaps of current AI. Finally, I will talk about the work we are doing on the modeling of cognitive mechanisms and the interest of using robots as models to progress in domains as different as place recognition, the construction of autobiographical memories or resilience to stress through the importance of rhythm in interactions. 

Abstract: Most seats in private cars are empty when drivers hit the road. Carpooling could thus represent a low-cost strategy to reduce carbon emissions in the transportation sector. Using revealed preference data from actual long-distance carpooling trips in France, we estimate passengers' preferences for the different characteristics of a ride. We find that passengers are highly price-elastic and value significantly the convenience of pick-up and drop-off locations. In contrast, their value of time once in the car is significantly lower than typical reference values. Finally, we discuss the effectiveness of a number of counterfactual policies aimed at promoting carpooling.