zhihu_data

Published: 4 June 2021| Version 1 | DOI: 10.17632/gf5jhrz683.1
Contributor:
wei zhuang

Description

Data acquisition: We crawled daily user data from 440 knowledge suppliers who provide knowledge services in the paid Q&A section of Zhihu from November 15, 2020, to April 6, 2021, obtaining a total of 14,303 records, including transaction data of knowledge services and personal information of knowledge suppliers. After data preprocessing, the information of knowledge suppliers who suspended the Q&A service and some data missing information were deleted. Finally, we obtained a dataset composed of 12,419 records for this study. Research hypothesis: H1a——The degree of recognition of the supplier’s content (upvotes) has a positive influence on knowledge payment behavior. H1b——The degree of usefulness of the supplier’s content (favorites) has a positive influence on knowledge payment behavior. H1c——The number of followers has a positive influence on knowledge payment behavior. H2a——The amount of published content has a positive influence on knowledge payment behavior. H2b——User ratings of knowledge services has a positive influence on knowledge payment behavior. H3——Whether the knowledge supplier is authenticated has a positive influence on knowledge payment behavior. Notable findings: Our empirical study finds that whether the knowledge supplier is authenticated has no obvious influence on user payment behavior, which is different from the previous studies on knowledge platforms in professional fields. In the context of grassroots knowledge sharing communities, consumers care more about the suppliers’ online presences such as content contributions and user interactions rather than their offline backgrounds and statuses, which also coincides with the saying "expert in civil".

Files

Steps to reproduce

Since the independent variables such as the number of upvotes, the number of favorites, the amount of published content, and the number of followers change over time, the model is estimated using panel data analysis with fixed effects, this can also reduce the heterogeneity caused by unobservable individual characteristics. In addition, we use the t-1 period of independent variables for regression, so as to reduce the endogeneity that may exist among variables. Meanwhile, the dependent variable is a discrete non-negative integer, so we apply panel negative binomial regression as the main method for estimation.

Institutions

Shanghai University of Finance and Economics

Categories

Knowledge

Licence