Essays on the herding behavior in online reviews and its consequence
Online consumer review (OCR) is an important source of product information. But how do reviews from friends affect the contents of subsequent reviews? And will the subsequent review contents also be influenced by prior reviews from users from the same city (i.e., neighborhood effect)? Although resea...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis-Doctor of Philosophy |
| Language: | English |
| Published: |
Nanyang Technological University
2023
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/168490 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | Online consumer review (OCR) is an important source of product information. But how do reviews from friends affect the contents of subsequent reviews? And will the subsequent review contents also be influenced by prior reviews from users from the same city (i.e., neighborhood effect)? Although researchers have examined the social influence of prior reviews on the rating/valence of subsequent reviews, a clear understanding is missing of the social influence on review contents and how this influence is associated with review helpfulness.
In Chapter 1, leveraging Yelp review data from 118 cities in North America, we employ a Naïve Bayes algorithm and hybrid matching method that combines propensity score matching (PSM) and Mahalanobis distance matching to estimate friends’ social effect and neighborhood effect on content similarity, thus resolving the endogeneity problem that similarity among review contents from friends, or users from the same city, can be attributed to commonalities among friends or consumers from the same city. This study shows the superior performance of the hybrid matching method compared to the Mahalanobis distance matching and PSM. We find that subsequent review contents converge with the review contents from friends for a given restaurant, and the social influence from friends is stronger in restaurants with low review volume. By estimating a causal forest, we show that reviews from certain types of users are prone to the neighborhood effect.
In Chapter 2, we further examine the effect of review topic similarity on perceived review helpfulness through negative binomial regression. Our results suggest that consumers consider reviews with topics similar to the prior reviews less helpful. This research expands the social influence literature on the contents of online reviews and provides insight into online review recommendation systems. |
|---|