主题：Estimating Search Models with Panel Data: Identification and Re-Examination of Preference Heterogeneity
主讲人：Prof. Xiaojing Dong, Santa Clara University
Professor Xiaojing Dong is a tenured Associate Professor of Marketing and Business Analytics at Santa Clara University. She got her B.E. from Tsinghua University in China, M.S. from MIT and PhD from Northwestern University.
Her research area applies Data Analytics technics to analyze and assist to improve Marketing and Business decisions. Her studies focus on customer level analysis, studying customer decision process, and how business actions can influence those decisions. Her papers have appeared in top academic journals, and some have attracted media attentions and been widely cited. She teaches cross-disciplinary classes, where not only their skills of computer programming, data statistics are sharpened, knowledge of the market and business insights are also emphasized.
With participation of ten top companies (谷歌, LinkedIn, Tesla, etc.) in the Silicon Valley, Prof. Dong founded a new Master of Science program in Business Analytics at the Leavey School of Business at Santa Clara University. The program has been attracted attentions from students from top schools globally, and major companies in seeking talents who can leverage big data technologies.
We study the estimation of preference heterogeneity in a setting where consumers are imperfectly informed and have to engage in costly search. In the presence of search costs, slight preference for one product over another can lead to substantially different purchase probabilities. As a consequence, search costs amplify how preference heterogeneity translates into differences in purchase probabilities across consumers for a given product, and generate a “seemingly” larger degree of preference heterogeneity. We estimate a model of search using a unique data-set from an online retailer that contains panel information on consumers' search and purchase behavior and show that when ignoring search costs, we overestimate standard deviations of product intercepts by 40%. Using the example of personalized pricing, we show that this bias has important consequences for targeted marketing. Finally, we show that panel dimension of the data is crucial for identifying preference heterogeneity separately from search costs.