Three Essays on Digital Platforms, Start-up Accelerators and Venture Capital Firms

In this thesis, I present three chapters related to my two areas of research: digital platforms (first chapter) and entrepreneurial finance (second and third chapters). In the first chapter, I explore advertising on social media platforms, which surged in recent years due to platforms’ ability to capture user attention and broker this attention to advertisers. I develop a two-sided model with heterogeneous users, who differ in product preferences and annoyance from ads, and heterogeneous advertisers, who vary in product type and quality, located on an infinite plane. Advertisers choose whether to join the platform and their targeting reach, while the platform sets prices and users decide whether to participate and to click on ads. I analyze two cases: one with full user participation driven by a high standalone benefit, and one with partial participation when there is no standalone benefit. In both cases, improvements in attention brokering ability raise advertising prices and yet lead to greater advertiser entry, which reduces average ad quality and relevance. Under partial participation, users who are most annoyed by ads exit, leading the platform to raise prices further to account for the negative externalities of ads. The welfare analysis shows that platform profits and advertiser surplus always increase with attention brokering; user welfare follows an inverted-U shape when the standalone benefit is high but increases monotonically otherwise. A regulator focused on users would restrict brokering only when the standalone benefit is high; one maximizing the welfare of all platform participants would choose intermediate brokering under high standalone benefit and full brokering otherwise. In the second chapter, I examine start-up accelerators and the role of similarity among start-ups within their cohorts. While similarity can improve post-acceleration performance through exposure to relevant knowledge, it can also intensify competition and reduce the benefits of acceleration. I hypothesize and show in the data an inverted U-shaped relationship between start-ups’ business similarity and their post-acceleration performance. Using data on 2,225 start-ups across 129 cohorts from 8 U.S. accelerators (2005-2018), I find that performance initially rises with similarity but falls beyond a certain point. Decomposing business similarity into technology and market similarity suggests that the interaction of these two dimensions contribute to the observed inverted-U: while higher similarity on either dimension has a positive impact on performance, high similarity along both dimensions vanishes these beneficial effects. These results offer guidance for accelerator managers designing cohorts and for start-ups evaluating accelerators. In the third chapter, I employ machine learning methods to isolate and quantify the persistent effect (if any) of general partners on performance across multiple funds. Analyzing a panel dataset of 29,021 quarterly observations covering 722 funds managed by 811 general partners (1997-2022), I document statistically significant albeit modest effects of general partners on performance persistence. These magnitudes are substantially smaller than those reported in the literature, highlighting the limited external validity of traditional methods for the task at hand. Moreover, the general partner effect consistently exceeds that of venture capital firms, suggesting that individual-level analyses provide greater insights than firm-level ones. These results indicate that most of the variation in venture capital performance is not attributable to the organizational characteristics of venture capital firms and can be explained only partially by the individuals managing them.