Impact mechanism of ESG ratings on corporate financing costs: a hybrid machine learning analysis using marginal effect and ESG rating effect in developed and developing countries

The goal of this study is to conduct a comparative analysis of the impact mechanisms of environmental, social, and governance (ESG) ratings on the costs of corporate finance within the framework of hybrid machine learning models. This will be accomplished by providing a comprehensive analysis of the situation. Our training and testing of three hybrid models—RF+NN, SVM+GB, and LSTM+XGB—can be accomplished with the use of an online dataset. These models are RF+NN, SVM+GB, and LSTM+XGB. In comparison to the other two models, the LSTM+XGB model has superior performance, as shown by the data. The accuracy is 0.90, the precision is 0.85, the recall is 0.95, the F1-score is 0.90, the mean squared error is 0.01 and the mean absolute error is 0.06. All of these results are achieved by the system. The ESG rating is the most important of the three models, according to the findings of the research that was conducted on the significance of characteristic scores. With a score of 0.35 for feature significance, the LSTM+XGB model is the most important feature despite the fact that this value indicates the greatest possible score. Clarification of the impact mechanisms of environmental, social, and governance ratings on the costs of corporate financing is provided by the research. This is accomplished by demonstrating the significance of hybrid machine learning models in the analysis of complex financial events. ESG ratings, corporate finance costs, and GDP per capita were compared across rich and developing countries using RF+NN, SVM+GB, and LSTM+XGB models. Developed countries have greater GDP per capita, cheaper finance costs, and better ESG rankings. For developed countries, the LSTM+XGB model yielded the highest average ESG ratings (81.1 ± 1.7) and lowest financing costs (3.2% ± 0.4 ESG rating and financing cost were negatively correlated in developed and developing countries. Both nation groups had substantial negative coefficients for ESG rating models in regression analysis.