Search

We study the incentives of a digital business to collect and protect users’ data. The users’ data the business collects improve the service it provides to consumers, but they may also be accessed, at a cost, by strategic third parties in a way that harms users, imposing endogenous users’ privacy costs. We characterize how the revenue model of the business shapes its optimal data strategy: collection and protection of users’ data. A business with a more data-driven revenue model will collect more users’ data and provide more data protection than a similar business that is more usage driven. Consequently, if users have small direct benefit from data collection, then more usage-driven businesses generate larger consumer surplus than their more data-driven counterparts (the reverse holds if users have large direct benefit from data collection). Relative to the socially desired data strategy, the business may over- or undercollect users’ data and may over- or underprotect it. Restoring efficiency requires a two-pronged regulatory policy, covering both data collection and data protection; one such policy combines a minimal data protection requirement with a tax proportional to the amount of collected data. We finally show that existing regulation in the United States, which focuses only on data protection, may even harm consumer surplus and overall welfare. This paper was accepted by Itai Ashlagi, revenue management and market analytics.

International audience; For an efficient integration of autonomous vehicles on roads, human-like reasoning and decision making in complex traffic situations are needed. One of the key factors to achieve this goal is the estimation of the future behavior of the vehicles present in the scene. In this work, we propose a new approach to predict the motion of vehicles surrounding a target vehicle in a highway environment. Our approach is based on an LSTM encoder-decoder that uses a social pooling mechanism to model the interactions between all the neighboring vehicles. The originality of our social pooling module is that it combines both local and non-local operations. The non-local multi-head attention mechanism captures the relative importance of each vehicle despite the inter-vehicle distances to the target vehicle, while the local blocks represent nearby interactions between vehicles. This paper compares the proposed approach with the state-of-the-art using two naturalistic driving datasets: Next Generation Simulation (NGSIM) and the new highD Dataset. The proposed method outperforms existing ones in terms of RMS values of prediction error, which shows the effectiveness of combining local and non-local operations in such a context.

Modeling consumer heterogeneity helps practitioners understand market structures and devise effective marketing strategies. In this research we study finite mixture specifications for modeling consumer heterogeneity where each regression coefficient has its own finite mixture, that is, an attribute finite mixture model. An important challenge of such an approach to modeling heterogeneity lies in its estimation. A proposed Bayesian estimation approach, based on recent advances in reversible jump Markov Chain Monte Carlo (MCMC) methods, can estimate parameters for the attribute-based finite mixture model, assuming that the number of components for each finite mixture is a discrete random variable. An attribute specification has several advantages over traditional, vector-based, finite mixture specifications; specifically, the attribute mixture model offers a more appropriate aggregation of information than the vector specification facilitating estimation. In an extensive simulation study and an empirical application, we show that the attribute model can recover complex heterogeneity structures, making it dominant over traditional (vector) finite mixture regression models and a strong contender compared with mixture-of-normals models for modeling heterogeneity.The appendices for this paper are available at the following URL: http://ssrn.com/abstract=2370423