Klusowski, Jason MatthewHuang, Rosalind K.2025-08-062025-08-062025-04-10https://theses-dissertations.princeton.edu/handle/88435/dsp0108612r99tThis thesis investigates how early-stage venture capital (VC) investor characteristics, particularly their syndication network centrality and firm-specific attributes, relate to the long-term outcomes of technology startups. Motivated by the “living dead” phenomenon in the space and theories of information flow and social embeddedness, we build a large co-investment network using pre-2014 deal data sourced from Pitchbook and focus on startups in North America that began raising seed or angel funding in 2014. We propose a set of centrality measures and firm quality indicators, transforming and aggregating them to the portfolio company level to serve as model features. Outcomes are categorized into a three-class framework capturing positive, neutral, and negative exits by 2024. To address multicollinearity and enhance model interpretability, we implement a multi-stage feature selection pipeline involving distributional analysis, correlation pruning, and variance inflation factor (VIF) diagnostics. Several classification models – XGBoost, Random Forest, and Logistic Regression – are trained and evaluated. Despite modest overall accuracy (approx. 45%), SHAP analysis reveals that certain features, particularly Closeness centrality, investor age, and exit history, consistently contribute to predictive power. Interestingly, Betweenness centrality demonstrates a reverse relationship, suggesting that occupying a brokerage position may not always confer informational advantage. The analysis highlights the nuanced ways different centrality constructs interact to shape investor influence. While the dataset’s size and scope limit generalizability, the framework offers a scalable approach to understanding investor signal quality. Our findings suggest that startup success is not merely a function of capital raised, but also of the network structure and embeddedness of early investors. This contributes novel empirical insight into how network-informed screening could enhance investment decision-making at each stage.en-USPrinceton University Senior Theses