Determining IPO underpricing: an ML-MCDM approach

Purpose The study examines 498 initial public offerings (IPOs) listed during 2007–2022 from an emerging IPO market. Design/methodology/approach The study proposes a novel framework that utilizes the integration of machine learning (ML) outputs with multi-criteria decision-making–based ranking procedures. Findings Novel predictors, including the geopolitical risk index, underwriter network centrality and economic policy uncertainty, emerge as significant determinants. The IPOs most preferred by data-driven ranking frameworks perform economically and statistically better than the IPOs least preferred by the ranking frameworks. Practical implications The proposed framework aids investment in IPOs. It can also be generalized to other related domains for the selection of decision alternatives where decision makers face conflicting alternatives. Originality/value This is the first study in the finance domain that presents and proposes a hybrid ML framework with multi-criteria decision-making methods (VIKOR/TOPSIS). The study contributes methodologically and theoretically toward a better understanding of investment in IPOs.