Search results

Structural models of demand founded on the classic work of Berry, Levinsohn, and Pakes (1995) link variation in aggregate market shares for a product to the influence of product attributes on heterogeneous consumer tastes. We consider implementing these models in settings with complicated products where consumer preferences for product attributes are sparse, that is, where a small proportion of a high-dimensional product characteristics influence consumer tastes. We propose a multistep estimator to efficiently perform uniform inference. Our estimator employs a penalized pre-estimation model specification stage to consistently estimate nonlinear features of the BLP model. We then perform selection via a Triple-LASSO for explanatory controls, treatment selection controls, and instrument selection. After selecting variables, we use an unpenalized GMM estimator for inference. Monte Carlo simulations verify the performance of these estimators.

We propose a generic algorithm for numerically accurate likelihood evaluation of a broad class of spatial models characterized by a high-dimensional latent Gaussian process and non-Gaussian response variables. The class of models under consideration includes specifications for discrete choices, event counts and limited-dependent variables (truncation, censoring, and sample selection) among others. Our algorithm relies upon a novel implementation of efficient importance sampling (EIS) specifically designed to exploit typical sparsity of high-dimensional spatial precision (or covariance) matrices. It is numerically very accurate and computationally feasible even for very high-dimensional latent processes. Thus, maximum likelihood (ML) estimation of high-dimensional non-Gaussian spatial models, hitherto considered to be computationally prohibitive, becomes feasible. We illustrate our approach with ML estimation of a spatial probit for US presidential voting decisions and spatial count data models (Poisson and Negbin) for firm location choices.

This paper investigates the capital structure of a large sample of U.S. private firms from 2004 to 2013. There is a considerable heterogeneity in private firm capital structure not only in terms of the level of leverage but also with regard to the issuance of specific debt instruments. Leverage, debt type usage, and debt specialization are dynamic and strongly related to observable firm characteristics largely in support of contract theory. Unobservable firm and industry characteristics are strong determinants of leverage levels and debt specialization. Macro credit conditions are not related to private firm leverage but are strong determinants of the degree to which firms diversify their debt capital structures.

We provide an overview of how artificial intelligence is transforming the identification, structuring, and prioritization of customer needs – known as the voice of the customer (VOC). First, we summarize how the VOC helps firms gain insights on using user-generated data. Second, we discuss the types of user-generated data and the challenges associated with analyzing each type of data. Third, we describe common methods, matched to the firms' goals and the structure of the data, that are used to analyze the VOC. Fourth, and most importantly, we map the methods to relevant applications, providing guidance to select the appropriate method to address the desired research questions.

This article estimates the loan spread equation taking into account the endogenous matching between banks and firms in the loan market. To overcome the endogeneity problem, I supplement the loan spread equation with a two-sided matching model and estimate them jointly. Bayesian inference is feasible using a Gibbs sampling algorithm that performs Markov chain Monte Carlo (MCMC) simulations. I find that medium-sized banks and firms tend to be the most attractive partners, and that liquidity is also a consideration in choosing partners. Furthermore, banks with higher monitoring ability charge higher spreads, and firms that are more leveraged or less liquid are charged higher spreads.

This chapter reviews the recent developments at the intersection of personalization and AI in marketing and related fields. We provide a formal definition of personalized policy and review the methodological approaches available for personalization. We discuss scalability, generalizability, and counterfactual validity issues and briefly touch upon advanced methods for online/interactive/dynamic settings. We then summarize the three evaluation approaches for static policies – the Direct method, the Inverse Propensity Score (IPS) estimator, and the Doubly Robust (DR) method. Next, we present a summary of the evaluation approaches for special cases such as continuous actions and dynamic settings. We then summarize the findings on the returns to personalization across various domains, including content recommendation, advertising, and promotions. Next, we discuss the work on the intersection between personalization and welfare. We focus on four of these welfare notions that have been studied in the literature: (1) search costs, (2) privacy, (3) fairness, and (4) polarization. We conclude with a discussion of the remaining challenges and some directions for future research.

The expansion of marketing data is encouraging the growing use of deep learning (DL) in marketing. I summarize the intuition behind deep learning and explain the mechanisms of six popular algorithms: three discriminative (convolutional neural network (CNN), recurrent neural network (RNN), and Transformer), two generative (variational autoencoder (VAE) and generative adversarial networks (GAN)), and one RL (DQN). I discuss what marketing problems DL is useful for and what fueled its growth in recent years. I emphasize the power and flexibility of DL for modeling unstructured data when formal theories and knowledge are absent. I also describe future research directions.

In this chapter, we propose a nonconventional methodology, the graph theory, which is especially relevant for the study of high-dimensional financial data. We illustrate the advantages of this method in the context of systemic risk in derivative markets, a main subject nowadays in finance. A key issue is that this methodology can be used in various areas. Numerous applications have now to face the challenge of analyzing gigantic financial data sets, which are more and more frequent. We offer a pedagogical introduction to the use of the graph theory in finance and to some tools provided by this method. As we focus on systemic risk, we first examine correlation-based graphs in order to investigate markets integration and inter/cross-market linkages. We then restrain the analysis to a subset of these graphs, the so-called “minimum spanning trees.” We study their topological and dynamic properties and discuss the relevance of these tools as well as the robustness of the empirical results relying on them.

This article reviews the quantitative marketing literature on artificial intelligence (AI) through an economics lens. We apply the framework in Prediction Machines: The Simple Economics of Artificial Intelligence to systematically categorize 96 research papers on AI in marketing academia into five levels of impact, which are prediction, decision, tool, strategy, and society. For each paper, we further identify each individual component of a task, the research question, the AI model used, and the broad decision type. Overall, we find there are fewer marketing papers focusing on strategy and society, and accordingly, we discuss future research opportunities in those areas.