Listening to the Market: Text Analysis Approaches to Consumer Research

Abstract

Language is central to human interaction, thinking, and sense-making. Witty marketing communicators and loquacious consumer research scholars can bend and shape language to great effect and be admired for it. Marketing has traditionally been a diverse discipline, borrowing from many parents and employing a variety of methods or approaches to understanding consumers and the market. This dissertation presents three essays exploring consumer behavior from a perspective influenced by theory and methods from language- and discourse-centric disciplines. The first essay employs ethnographic methods informed by discourse analysis of marketing communications and in-person service encounters to better understand services. We integrate the disconnected research streams of role theory in service encounters and cocreation. Our findings challenge common definitions of a service script, defining instead as the product of imagined service encounters that serve as a template for cocreation in consumers’ minds. The second essay examines U.S. news media on the topic of debt in order to reveal how public discourse frames debt to distribute responsibility and guide action. The data analysis begins at the qualitative level of discourse analysis and hermeneutics, followed by a corpus research approach, complemented with a neural network-based word embedding technique used in Natural Language Processing (NLP). Our findings reveal two dominant metaphors in public debt conversations: debt as weight, and debt as captivity. These metaphors frame the discourse, creating narratives with contrasting assignments of responsibility in the market and proposed marketing actions. The third essay utilizes the same U.S. news articles database of the second paper to answer different research questions. We join a growing and incredibly impactful new stream of consumer research studies harnessing the power of big textual data for marketing insight. We make a methodological contribution by developing and training a topic-detection Bi-directional long short term memory (Bi-LSTM) neural network to classify a large, unstructured corpus. We sequentially ran a dynamic Latent Dietrich Allocation (LDA) to identify the narratives predominant to each type of debt and how they change over the ten-year period (2010-2019).