The impact of soft information extracted from descriptive text on crowdfunding performance

Crowdfunding provides an alternative way of financing, but its success is heavily challenged by information asymmetry. Based on the signal theory, this paper investigates the impact of hard information and soft information as project quality signals on crowdfunding performance. To this end, we extract soft information in two ways, including the topics from project descriptions by the Latent Dirichlet Allocation (LDA) model and the investor sentiments from project comments by the sentiment analysis. Based on a dataset of 957 projects from the Dreamore, a major crowdfunding platform in China, the empirical results show that the impacts of experience of a creator, comment quantity, and the number of backers on crowdfunding performance are significantly positive, whereas reward execution moderates the impact of goal. In addition to the hard information, we find that the soft information also contributes greatly to crowdfunding performance. Specifically, the topics of project descriptions play an important role in crowdfunding performance. The positive sentiment in project comments has a positive impact on crowdfunding performance, whereas the negative one has a negative impact. These empirical findings are significant for crowdfunding platforms, creators, and investors from both theoretical and practical perspectives.