Bankruptcy risk models and firm performance in Indian context an empirical approach

Abstract

The study aims to depict the determinants of bankruptcy in Indian context and how firm performance variables are associated with it. The study also intends to build a bankruptcy prediction model to forecast the bankruptcy condition for especially Indian companies as the business structure is quite complex. The study also proposes to give anecdote to industries about the threshold of each variables influencing bankruptcy of the firm. Hence, we propose the title bankruptcy prediction models and firm performance in Indian context an empirical approach. To justify our objectives, we performed exploratory data analysis by reviewing past studies and then we collected data of Indian listed companies in the construction, iron and steel, power and textile sectors from secondary database (i.e., CMIE Prowess IQ) to empirically test our model. Initially, we applied traditional techniques like logistic regression and principal component analysis (PCA) to extract features from the dataset. We introduced a novel machine learning technique, known as explainable artificial intelligence (xAI) algorithms for feature extraction and threshold analysis i.e., Locally Interpretable Model-Agnostic Explanations (LIME) and SHapley Additive exPlanations (SHAP). Then to eliminate the time invariances and make our prediction process dynamic we introduced the Malmquist Data Envelopment Analysis (MDEA) model and give the results. We finally, observed that the mostly affected sectors were iron and steel, followed by the textile sectors mainly due to the debt-ridden conditions and lack of working capital movements. We also suggest the minimum threshold for efficiency for each industry and segment wise below which a firm might get bankrupt and thus a distress signal can be sent to the management and the policymakers in advance to prevent the firm from going into liquidation. We also suggest the policymakers to make data disclosure mandatory for all companies in all sectors so that more important variables like sentiment of audit report.