How Content Features of Charity Crowdfunding Projects Attract Potential Donors?

Abstract

This study investigates how the content features (e.g., images and texts) of donation projects affect potential donors’ participation. We collect textual and visual contents from one of the largest online donation crowdfunding platforms in South Korea. To extract features from the content, we use Deep Learning models for images and Latent Dirichlet Allocation (LDA) topic modeling for text contexts. We then construct variables representing visual and textual features. Finally, we estimate the effects of our independent variables on donors’ participation by using the Ordinary Least Squares (OLS) model. Our empirical results show that (1) Observing a small number of recipients in images attract more donors than a large number of recipients does; (2) Negative and positive emotions decrease potential donors’ willingness to help compared to neutral emotion; (3) Positive emotion in the image moderates the number of recipients’ negative effect; and (4) Since complex project description requires more effort to understand the recipients, potential donors are less likely to be engaged. Through this study, we hope to make contributions to the extant literature. In addition, our framework for content analysis will contribute to the future studies as we shed light on novel methodologies to measure image and text dimensions.