Date of Award
Fall 2022
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Management
First Advisor
Kumar, Vineet
Abstract
Emotion impacts consumer behavior all throughout the customer journey. While there exists a large body of research on emotion in the behavioral marketing and psychology literatures, quantifying the impact of emotion in the field enriches our understanding of the generalizability of behavioral theories and informs managerial decision-making. In my dissertation, I quantify emotion in two ways: 1) by converting unstructured audio data into structured data using deep learning methods that incorporate domain knowledge from psychology and music theory and 2) by measuring the impact of emotion-related marketing actions on consumer behavior in field settings. In the first essay, which is joint work with K. Sudhir and Subroto Roy, I focus on the emotion of gratitude in the context of ``thank you'' letters routinely sent by non-profits. While gratitude expressions can generate reciprocity and increase giving, an accompanying ``ask for more'' call to action can decrease giving if potential donors perceive persuasive intent. Using a large-scale natural field experiment, we find that asking for more reduces giving for loyal donors but increases giving for non-loyal donors. The results replicate in a lab experiment, which further shows that loyals feel worse and view the charity more negatively when asked to give relative to non-loyals. Targeting ``ask for more'' messages in our field setting can increase donations by 12.8-17.5\%. In the second essay, which is joint work with Vineet Kumar and K. Sudhir, I do not focus on a particular emotion but instead characterize emotion more broadly using the two underlying dimensions of valence and arousal. I develop a theory-based, explainable deep learning convolutional neural network (CNN) classifier to predict the time-varying emotional response to music. The key advantage of a theory-based architecture is that we can connect how the predicted emotional response is related to explainable features of the music, providing transparency into the model. We then illustrate an application of our model in a digital advertising setting by using the model's predictions to identify optimal emotion-based ad insertion positions in videos. Finally, the third essay, which is joint work with Vineet Kumar and Ravi Dhar, explores the question of whether matching ad emotion with overall context emotion---the emotion elicited by the content where an ad is placed---is an effective ad targeting strategy, as has been suggested in past research. In contrast to the second essay, here we do not think about when to insert an ad based on emotion but instead which types of emotional content should be targeted, focusing on the emotions of happiness and sadness. In contrast to extant research, we consider the self-selection of emotional content in determining ad effectiveness. We quantify the effect of ad-context emotion matching and the effect of content self-selection using a lab experiment and find that the selection effect overwhelms the emotion matching effect. We observe the same pattern of results in a field study conducted on YouTube, where individuals self-select the content they watch. The results suggest that the benefit of ad targeting on emotion stems from finding individuals who are more receptive to ads based on content emotion. Overall, the essays contribute to our understanding of how to measure emotion and how emotion-related marketing actions impact consumer engagement using a diverse set of methods.
Recommended Citation
Fong, Hortense, "Three Essays On Quantifying Emotion" (2022). Yale Graduate School of Arts and Sciences Dissertations. 767.
https://elischolar.library.yale.edu/gsas_dissertations/767
