THE NEXUS BETWEEN CASH CONVERSION CYCLE, WORKING CAPITAL FINANCE, AND FIRM PERFORMANCE: EVIDENCE FROM NOVEL MACHINE LEARNING APPROACHES

This study examines the moderating role of the cash conversion cycle (CCC) while investigating the effects of working capital finance (WCF) on firm performance. Using more than 18000 observations from Chinese manufacturing firms, we computed several proxies for each variable of the study and merged these proxies via Principal Component Analysis (PCA) to create one master proxy for each variable. These master proxies contain all the essential information of individual proxies. Hence, they are more useful in producing reliable results than individual proxies. We also compared the predicting power of 15 econometric and machine learning estimators to select the best estimator. Based on the highest R2 value, we used two machine learning estimators, K-Nearest Neighbors (KNN), and Artificial Neural Networks (ANN) for subsequent analysis. To strengthen the empirical analysis, we employed another machine learning technique, i.e., the Bagging method, which is an ensembling technique that uses multiple estimators simultaneously to improve the accuracy and generalization of results. We used the Bagging method with 50KNN estimators. The findings unfold that the sensitivity level of firm performance to short-term debts shifts when the CCC period of firms fluctuates. More precisely, the WCF-performance relationship in firms with extended CCC is more sensitive compared with this relationship in the full sample. On segregating the three elements of CCC, we observe that the WCF-performance relationship in firms carrying extended account receivable (AR) days or extended Inventory days is more sensitive than the full sample. These findings are useful for firms' management for revising the optimal level of short-term debts according to CCC fluctuation. Also, the lending agencies can use these results for the assessment of firms' risk levels and adjustment of the interest rate.