Crowdfunding and the sustainability dilemma: Do sustainable-oriented campaigns perform better? A machine learning approach

Sustainability has emerged as an important attribute in crowdfunding campaigns, yet the channels through which sustainability signals affect fundraising outcomes differ across crowdfunding models. Drawing on signaling theory, we argue that in reward-based crowdfunding (RBC), where backers are motivated by ethical values and prosocial preferences, sustainability disclosures act as an inherently credible signal that directly increases participation and funding. In equity crowdfunding (EC), where investors apply a more instrumental, risk–return logic, the same signal is effective only when communicated concisely, because excessively long messages impose cognitive costs and reduce perceived reliability. To test these mechanisms, we analyze 592 Italian campaigns (337 RBC from Kickstarter and 255 EC from CrowdFundMe, spanning 2015–2024) and we employ a machine learning-based text classification approach to objectively identify sustainability disclosures, overcoming the biases of traditional keyword methods. Our results confirm the proposed channels: in RBC, sustainability has a direct positive effect on the likelihood of success, the amount of capital raised and the number of backers; in EC, the positive effect emerges only when descriptions are concise, highlighting the moderating role of message length. This study contributes to the crowdfunding literature by combining a novel machine-learning text-analysis approach with an explicit test of the moderating effect of message length, explaining potential mechanisms underlying sustainability signals. Furthermore, we offer actionable guidance for entrepreneurs, platforms and policy makers on how to structure sustainability communication to maximize impact across different crowdfunding models.