Application of Multivariate Logistic Model in Prospective Identification of Initial Public Offering Risk from the Perspective of Investment Banking

This paper explores Initial Public Offering (IPO) risk identification from the perspective of investment banking, focusing on the misreporting of R&D expenses among companies listed on the Science and Technology Innovation Board between 2019 and 2024. Firms penalized or rejected due to false Research and Development (R&D) disclosures were selected as the experimental group, while companies successfully listed without disputes served as the control group. A logistic regression model incorporating six core variables—such as project anomaly index and R&D personnel salary dispersion—was constructed. Parameters were estimated using the maximum likelihood method, with Lasso regression applied to refine variable selection and improve model clarity. The analysis shows that project anomaly index and R&D salary dispersion significantly raise the likelihood of misreporting, whereas auditor industry expertise reduces this risk. The model achieves strong predictive performance, with an AUC of 0.89 and an overall accuracy of 85%. The findings support a forward-looking framework for investment banks to assess IPO risks and strengthen audit quality.