Predicting Financial Distress through Ranking Working Capital Management Components Using Random Forest Algorithm

The primary objective of this research was to analyze the relative importance of working capital management factors in predicting financial distress among companies. The study population consisted of 167 companies listed on the Tehran Stock Exchange (TSE) from 2019 to 2023. 7 key working capital management indicators were selected based on their potential impacts on financial distress. Using Zavgren’s (1985) financial distress prediction model, the sample companies were classified into distressed and healthy groups. In the first step, a random forest algorithm was employed to assess the predictive power of the seven working capital management indicators in classifying companies as distressed or healthy. The results indicated that these indicators could successfully identify and predict the financial distress status of the companies with up to 85% accuracy. In the second step, the unique feature of the random forest algorithm was leveraged to rank the importance of each working capital component in achieving this 85% classification accuracy. The findings showed that the Average Collection Period (ACP) was significantly more important than the other working capital components in predicting financial distress.Keywords: Financial Distress, Working Capital Management, Average Collection Period (ACP), Random Forest AlgorithmJEL Classification: G01, G30, C38 IntroductionIn recent years, financial distress and bankruptcy have become increasingly prevalent issues for business enterprises. The financial literature offers various definitions to describe the state of financial distress and bankruptcy. While some researchers equate financial distress with bankruptcy, financial distress is more accurately viewed as a precursor to bankruptcy – a stage of financial decline that may or may not ultimately lead to a company's bankruptcy. Simply put, financial distress reflects a business entity's inability or weakness in fulfilling its obligations to creditors (Gerged et al., 2022). Given the rapid growth of joint-stock companies and the emergence of severe financial crises at both micro- and macro-economic scales, it is crucial to identify the key factors that can predict a company's financial health before it reaches the stage of bankruptcy, i.e., during the financial distress phase (Pourheydari et al., 2010). Evidence suggests that working capital management is a significant factor influencing the financial distress of business enterprises (Geng et al., 2015). Companies experiencing financial distress and bankruptcy often exhibit weaknesses in working capital management, particularly in cash control. Therefore, the aim of this study was to evaluate the predictive power of working capital management components in forecasting financial distress and rank the importance of each component in this prediction process.Materials & MethodsThe raw financial statement data for this research were extracted from Rahavard Novin Database and the Codal website. These data were then systematically organized in Excel. After applying certain eligibility criteria, a sample of 167 companies was identified as the accessible statistical population. To classify the sample companies into distressed and healthy groups, which served as the target variable (label), Zavgren’s (1985) financial distress prediction model was utilized. Subsequently, the predictive power of 7 key working capital management components in forecasting financial distress was tested using Python software and the random forest algorithm.The random forest method is based on ensemble learning, wherein the data are split into training and testing sets. During the learning phase, the model attempts to identify the inherent pattern or the relationship between the dependent variable (financial distress) and each explanatory variable (working capital management components) with the validity of this learning measured by the testing data. The random forest method employs a bagging approach, creating subsets from the entire dataset and determining the final result based on the average outcomes of these subsets. This approach helps to significantly mitigate the overfitting problem.One notable feature of the random forest algorithm is its ability to rank the importance of the input features in determining the trend of the target variables. This capability was leveraged in this research to answer the second research question, which focused on the relative importance of each working capital component in predicting financial distress. Research FindingsThe model achieved an accuracy of 85%, indicating that it could correctly predict whether a company was in financial distress or not based on what it learned during the training phase. Additionally, the model's F1-Score metric was 0.89 for identifying healthy companies and 0.76 for predicting distressed companies. These scores, being close to 1, suggested that the model's estimations were performed with a high degree of accuracy.The analysis of the relative importance of each working capital management component in achieving this 85% accuracy rate revealed some key insights. The Average Collection Period (ACP) was identified as the most important factor in predicting financial distress. Following the ACP, the Current Ratio (CR) ranked second, the Average Payable Period (APP) ranked third, and the Inventory Turnover In Days (ITID) ranked fourth in importance.These findings suggested that the initial signs of financial trouble for a company often stemmed from its failure to collect receivables in a timely manner, leading to an increased collection period. If the company's management did not effectively address this issue, other problems could likely arise, ultimately pushing the business entity into a state of financial distress. Discussion of Results & ConclusionThe results of the data analysis using the random forest algorithm indicated that working capital management indicators had an 85% predictive power for identifying financial distress in companies. This finding is consistent with those of the previous studies by Habib and Kayani (2022), Morshed (2020), and Li et al. (2018). Regarding the second research objective, which aimed to rank the importance of each working capital management component in predicting financial distress, the analysis revealed that the Average Collection Period (ACP) was the most significant factor. This suggested that a company's inability to collect receivables in a timely manner was a crucial early indicator of impending financial distress.An increase in the ACP could lead to a serious risk of bad debts and liquidity problems for the company. As a result, the company's management might need to secure additional working capital to fund operations, which could potentially increase the Weighted Average Cost of Capital (WACC). However, if the company failed to generate adequate returns to cover these elevated financing costs, it might ultimately fall into a state of financial distress (Panigrahi, 2014). Given the notable importance of the ACP compared to other working capital management components, it appeared that many of the underlying issues leading to financial distress stemmed from poor performance in collecting receivables. Therefore, this research underscored the critical need for robust management practices of receivables to maintain liquidity and avoid the escalating costs and risks associated with financial distress.