Please use this identifier to cite or link to this item:
Title: Two studies on user-generated content in online platforms : review valence, self-presentation, and sales performance
Authors: Liu, Fuzhen
Degree: M.Phil.
Issue Date: 2021
Abstract: When the sharing economy encounters explosive growth of user-generated content (UGC), information asymmetry between sellers and buyers is a salient concern in carrying out online transactions. Trust-building signals gain prominence for customers to reduce their uncertainties in online activities. On the value of trust-building signals, it is desirable to understand how and why they matter for sales performance in the sharing economy context. So far, the literature remains unclear on two questions: 1) what signals from customer- and provider-generated content determine customer purchases? 2) what factors moderate the relationship between these trust-building signals and sales performance? In this thesis, we conduct two studies on trust-building signals derived from UGC in online platforms to identify and examine the factors influencing the signals-performance link from the contingency theory perspective. In the first study, using data from covering seventeen cities in China, we aim to examine the effects of review sentiment and average rating as well as their interactions with seller popularity and property quantity on sales performance. In terms of text mining, we adopt Naive Bayes (NB) to extract review sentiment. Through a time-lagged regression model and city and genre fixed effects controlled for this model, we find no difference between review sentiment and average rating in positively influencing sales performance. Notably, seller popularity strengthens while property quantity weakens the positive impact of review valence on sales performance. More importantly, the positive effect of review valence on sales performance is more prominent for hosts characterized by popularity and personalization, and such hosts own fewer listings to achieve more historical sales. Our findings provide new insights by comparing the performance effects of review sentiment and average rating and uncovering the moderating role of seller popularity and property quantity. The implications are helpful for service providers to enhance their service operations management and for policy makers to better regulate the sharing economy. In the second study, using data from Airbnb covering four cities in the United States, we investigate the effects of host self-presentation and social orientation embedded in self-presentation as well as their interactions with customer- (i.e., review rating) and marketer-generated (i.e., superhost) reputations on sales performance. Latent Dirichlet allocation (LDA) is employed to identify the topics included in self-presentation, and we find two formats, including social-oriented and official-oriented topics. Furthermore, we use the support vector machine (SVM) algorithm to predict the topic of social-oriented self-presentation. Using econometric analysis, we find that self-presentation, especially for social-oriented self-presentation, positively influences sales performance. Moreover, we provide evidence supporting that customer-generated reputation (i.e., review rating) strengthens, but marketer-generated reputation (i.e., superhost) weakens the positive effects of self-presentation on sales performance. This research provides new insights by uncovering the complementary effect of customer-generated reputation and the suppression effect of marketer-generated reputation in the link between self-presentation and sales performance. Our findings offer managerial implications for service providers to formulate marketing strategies through building self-presentation and an online reputation.
Subjects: User-generated content
Business networks
Online social networks -- Economic aspects
Consumer behavior
Hong Kong Polytechnic University -- Dissertations
Pages: ix, 87 pages : color illustrations
Appears in Collections:Thesis

Show full item record

Page views

Last Week
Last month
Citations as of Oct 1, 2023

Google ScholarTM


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.