Does the History of Investment Matter for an IPO? A Machine Learning Approach

Abstract

In this article, we use eXtreme Gradient Boosting (XGBoost) to examine the factors that determine the success of an IPO and how these factors interact. We find that, among various variables related to the investment history of the investor and investee firms, the total number of fund investors (TNFund), the capital under management of the firm investor (FCUM), the days between the first investment date and the founding date of the firm investor (FIFFD), and the days between the last investment received date and the founding date of the investee company (INLFD) are the four main determinants of a company’s IPO. Using SHapley Additive exPlanations (SHAP), we study the interaction effects between the main factors and discover that FCUM tends to interact with INLFD. We have tested our conclusions with different sub-samples, balanced samples using various balancing methods, and a variety of other machine learning methods such as Gaussian process regression, random forest, neural networks, and support vector machines, and we have found that our main conclusions are robust. JEL Classification: C45, G24, G34.