Exploring investors' expectancies and its impact on project funding success likelihood in crowdfunding by using text analytics and Bayesian networks

Abstract

Crowdfunding has become immensely popular today, allowing entrepreneurs to present innovative ideas to a broad audience of potential investors. However, due to the online nature of crowdfunding, investors have to use clues only available on the website to decide on whether to invest or not. Grounded in language expectancy theory (LET), we proposed and tested hypotheses suggesting that, when no knowledge of the entrepreneur is available, investors have to use language expectancy as a way to inform them on their investment decisions. Furthermore, we propose that communication content in entrepreneurial narratives such as vague communication and linguistics affect the quality of information leading to a violation of language expectancy. We also postulate that this will manifest in affect intensity and in two-sided persuasion. We separated the description and risks and challenges (R&C) sections from Kickstarter and performed discrete analyses with regressions to further test two-sidedness. Next, we sought to understand causal knowledge of the underlying target class by implementing a Bayesian Network to find the conditional probability of the variables before attempting to find a near-optimal probability of funding success using a genetic algorithm. We found robust support for our hypotheses and helped shed light on which information is received and interpreted by investors leading to a greater likelihood of funding success. Overall, this approach sheds new light on the role of language within crowdfunding literature.