Quantitative Marketing 2: Modeling & Methods
(A2024-119767)

AUTHORS

Alina Ferecatu, Rotterdam School of Management, Erasmus University; Arash Laghaie, Nova School of Business and Economics; Jason Roos, Erasmus University; Matteo Fina, Goethe University Frankfurt am Main; Dominik Papies, University of Tübingen; Iris Steenkamp, Bocconi University; Rafael P. Greminger, University College London, School of Management; Daniela Schmitt, Nova School of Business and Economics

ABSTRACT

In the data-rich business landscape, effective decision-making hinges not just on data volume, but on the ability of firms and analysts to extract valid and actionable insights. Quantitative marketing, at the intersection of marketing, economics, statistics, and machine learning, therefore, remains pivotal in this era of data-driven decision-making. Against this backdrop, we aim to stimulate the application of advanced empirical methods and innovative quantitative marketing approaches within the EMAC community through a series of 5 special sessions. This special session includes four papers that cover advances in marketing modeling and methods. These advances in methods are crucial to allow researchers to better tackle substantive research questions. The four papers deal with how to leverage data from hypothetical discrete-choice experiments, how to disentangle perceptions of quantity from latent value in models of risky choice (e.g., Prospect Theory), how to accurately approximate the marginal treatment effect in the context of policy optimization (e.g., optimizing targeted promotions), and how to empirically identify simple and complex utility components from discrete choice data under the rational inattention framework. In the first paper, the authors develop a process model of decision to model choice as generated by a common set of invariant ('deep') preference parameters but explicitly account for differential decision effort between hypothetical and incentive-aligned experimental settings. The authors further develop a framework that helps decrease the data collection cost by leveraging a large amount of hypothetical data while keeping with the goal of predicting incentivized choices. The second paper updates the Prospect Theory model to disentangle perceptions of quantity from latent value, typically conflated in models of risky choice. The updated model assumes prospects (e.g., win $254 with probability .27) are first distorted by the perceptual system (s(254; .27)), due to the left-digit bias for instance, then imputed into the valuation system – thus V($254, . 27)=v(s(254, .27)). The updated model outperforms Prospect Theory in in-sample and out-of-sample fits of archival datasets. The third paper estimates the marginal treatment effect (MTE), a valuable tool for evaluating causal effects, when individuals determine their own treatment status based on unobserved factors, and an instrumental variable (IV) is available. Estimating the MTE is challenging when the data has a small number of IV levels, often the case in randomized experiments. The paper shows how to leverage identifying information from higher moments of the outcome than available levels of an IV to improve the accuracy of polynomial approximations of the MTE in the context of policy optimization (e.g., optimizing targeted promotions). The fourth paper uses simulated data to empirically investigate the conditions under which analysts can separate simple utility components, assumed to be processed effortlessly and at no cost, from complex utility components or attributes, using only discrete choice data. The author shows that under a discrete choice model grounded in the rational inattention framework, the empirical distinction between simple and complex task aspects is generally achievable if utility is additively separable in these aspects.