Survey of ML/AI methods in finance
Published:
Minor R&R at Review of Corporate Finance
Abstract: [Title and abstract modified for anonymity.] We review and the ML/AI literature in finance and suggest avenues for future research.
Working Papers
DEI in Entrepreneurial Finance
Published:
R&R at Journal of Business Venturing (with J. Hackmann)
Abstract: [Title and abstract amended for anonymity.] We study how DEI impacts investement outcomes in private capital markets.
Social learning in decentralized autonomous organizations (DAOs)
Published:
Rej&R at Journal of Business Venturing (with E. Buddensiek)
Abstract: We document social learning in the form of information cascades in proposals put up for vote in decentralized autonomous organizations (DAOs). Exploring hand-collected data covering 19,450,710 votes from 924,095 unique voters in 3,317 DAO governance proposals, we estimate economically strong information cascades that run from influential through early to late voters. Several contextual factors impede social learning in DAOs, including the presence of cybercriminals in the DAO community and market sentiment, while the decentralization of token ownership is an important prerequisite for information cascades to unfold. Finally, we document a link between the potency of information cascades and the financial performance of DAOs. Stronger information cascades are associated with higher market capitalization, trading volume, and abnormal cryptocurrency token returns. Overall, our study shows how decentralized digital communities coordinate in ways that can translate into economic value in entrepreneurial settings.
Artificial Intelligence and Crowdfunding
Published:
(with J. Wiklund, Y. Yuan)
Abstract: Information asymmetries shape the (in)efficiency of entrepreneurial finance markets. We introduce a distinction between information-endowment asymmetry (unequal access to signals) and information-utilization asymmetry (unequal processing of signals). Our study explores whether the onset of generative artificial intelligence (AI) improves "signal conversion efficiency" in entrepreneurial finance; i.e., the effective utilization of the information endowment. Empirically, we engineer an AI agent that utilizes the information endowment in blockchainbased crowdfunding campaigns to generate investment recommendations, and compare those to recommendations by human experts and a human-built recommendation algorithm. Among all major signals of capital formation in blockchain-based crowdfunding, our AI agent’s advice outperforms statistically and economically: Including our AI agent’s advice as a signal for capital formation raises R-squared by up to 35%, and a one-standard-deviation increase in AI recommendation strength is associated with 48% higher funding amounts. Explainable AI in the form of Shapley values corroborates AI’s superior predictive power. Consistent with our signal conversion efficiency conjecture, we also find that AI-human advice congruence is positively associated with capital formation, while AI-human advice divergence predicts lower funding amounts if investment recommendations by humans are more optimistic than those by AI, highlighting the practical potential of AI as a corrective device for human biases in investment decisions.
Venture Tokenization and Growth
Published:
Rej&R at Strategic Entrepreneurship Journal (with J. Fuchs)
Abstract: After initial coin offerings (ICOs), decentralized digital platforms (DDPs) decide whether to go public or remain private. We explore the implications of the public-versus-private decision for the growth and decentralization of DDPs. Employing a difference-indifferences framework, we find that public DDPs scale faster post-listing relative to matched private DDPs. An important driver behind public DDPs’ superior growth is a spillover effect of financial speculation on fundamental platform activity, especially when DDPs are undervalued, hastening network effects. The going-public decision also facilitates DDP decentralization, although this stems mostly from the left tail of the token ownership distribution, while blockholders largely remain in control. Exploring the trade-off between going public through token exchange listings and remaining private with the help of institutional investors, we find that crypto fund-endorsed token listings yield more platform growth than unendorsed listings, while crypto fund backings without listings create the least value. Overall, our study suggests that early-stage startups may economically benefit from tokenization and creating liquid markets for venture tokens.
How Sustainability Segments Entrepreneurial Finance Markets
Published:
R&R at Journal of Business Venturing (with F. Xia, J. Thewissen, and S. Yan)
Abstract: We show that sustainability segments entrepreneurial finance markets by investor types in token-based crowdfunding, a market that is populated by individual and institutional investors alike. While ventures backed by institutional investor do not (need to) emphasize environmental, social, and governance (ESG) ambitions, non-backed ventures raising funds from masses of individual contributors display salient ESG orientations, with a disproportiantely pronounced governance dimension. ESG helps ventures unsuccessful in securing institutional investments partially compensate for the funding disadvantage, as individuals exhibit higher willingness-toinvest in sustainable ventures. Using artificial intelligence to separate ESG orientations into two latent sources, substantive and symbolic, we find that symbolic ESG is more prevalent in non-backed ventures, potentially suggesting "greenwashing." The startup valuations of individual investors are affected by substantive and symbolic ESG orientations, although symbolic claims do not compensate for the lack of institutional certification. Overall, sustainability-driven market segmentation is in line with the conjecture that individual investors perceive sustainability as a legitimization substitute for ventures that lack certification from institutional investors.
Generative Artificial Intelligence and Cryptocurrency Returns
Published:
Minor R&R at Journal of Banking and Finance (with H. Urban)
Abstract: [Title and abstract modified for anonymity.] We propose a novel ML/AI approach to improve return prediction for cryptocurrencies. Our approach is much simpler than existing methods and yet achieves twice the alpha of comparable models. We also theoretically justify the statistical outperformance of our model.
Emotions in New Venture Teams: Affects as Signals, Emotional Diversity, and Valuation Effects in Initial Coin Offerings (ICOs)
Published:
Rej&R at Strategic Entrepreneurship Journal
Abstract: New Venture Teams’ (NVTs’) collective emotions impact startup valuations through their intensity and diversity. I identify NVTs’ affective traits with artificial emotional intelligence by tracking 2,520 individuals across 165 NVTs during their Initial Coin Offerings (ICOs). The level of NVTs’ negative affects correlates with lower valuations, while within-NVT emotional diversity has a value-increasing effect. Intuitively, negative affects are associated with traits that may be prejudicial in dynamic entrepreneurial markets, but could be valuable if balanced by opposite traits in emotionally diverse NVTs. Moderated mediation analyses suggest that NVT affects have pronounced direct valuation effects. Overall, I extend the focus of the affective entrepreneurship literature from the entrepreneur to the team level, introduce the concept of emotional diversity, and explore the role of emotions in entrepreneurial finance.
The Discount for Lack of Marketability in Private Investments in Public Equity
Published:
Work in progress (with A. Bernardo, I. Welch)
Abstract: Our paper estimates that shares in Private Investments in Public Equity (PIPEs) offered a discount of 3-4% for each year during which these shares could not be resold. Our estimates make use of the duration of the resale restriction and information about the effects of a regulatory change. In 2008, the SEC amended Rule~144 to shorten the default statutory holding period. Our estimates are smaller than previous estimates and robust to various controls and endogeneity concerns. The discount can be twice as large in offerings in which marketability is a greater concern.