Deliberate premarket underpricing: New evidence on IPO pricing using machine learning

Abstract

We propose a nonlinear approach based on stochastic frontier and Deep Neural Networks (DNN) to estimate the pricing efficiency and the level of premarket inefficiencies for IPOs, using information available before the IPO day and without any distributional assumptions. We apply the proposed approach in the US IPO market to estimate deliberate (premarket) underpricing and find that the IPO offer prices are about 12.43% less than the estimated maximum offer prices on average. We further show that only a few determinants of the value of firms impact the pricing and deliberate underpricing of IPOs. Negative net income and EBITDA play the most important roles in determining the IPO maximum offer price, among various pricing variables. Proceeds followed by underwriter reputation negatively impact the premarket underpricing, and the IPO market activity, measured by the number of new issues, is the most important market cycle proxy that influences the premarket underpricing. We show that aftermarket mispricing is attributed more to offer size and underwriter reputation. The proposed DNN-based method is an easy to implement approach and can be used by academics and practitioners to estimate maximum offer prices and disentangle initial returns into deliberate premarket underpricing and aftermarket mispricing.