1. Introduction
Forecasting corporate financial troubles has become an essential topic of interest over the past few decades due to its great impact on publicly listed companies, current and potential stakeholders, and even a country’s economy [
1]. Financial resource providers need to evaluate the financial risk of a corporate before they make a financing decision or grant credit judgments on firms in order to avoid or prevent any tremendous financial shock and/or loss. Corporate suppliers and partners that conduct credit transactions with corporates also require a more detailed illustration of their financial status. If a prediction model is useful, then top-level managers can initiate some initial prevention such as adjusting their capital structure or modifying their financial leverage to avoid any deterioration in corporate status before financial trouble erupts. Current and potential investors can also utilize such a model to change their investment strategy as well as allocate monetary resources to more profitable places [
2].
Multivariate discriminant analysis (MDA) was the most frequently utilized forecasting model before the 1980s. Altman [
3] introduced a very famous forecasting architecture, the “Z-score”, that incorporated MDA with five financial ratios (i.e., working capital to total assets, retained earnings to total assets, earnings before interest and taxes to total assets, market value of equity to book value of total liabilities, and sales to total assets) so as to discriminate between healthy corporates and non-healthy corporates. Although this model performs a satisfactory job in forecasting quality, it also comes with some statistical challenges, such as linear separability, independent predictors, and multivariate normality that usually do not hold in real applications. To overcome these obstacles, the literature has proposed the linear probability model (LPM) and logit or probit regression models. Meyer and Pifer [
4] employed LPM to handle the task of the corporate financial bankruptcy forecasting task. Martin [
5] assessed banks’ financial troubles by relying on a logit model. Dimitras et al. [
6] provided a detailed review of statistical-based approaches in financial crisis forecasting, indicating that the logit model achieves optimal forecasting performance.
In contrast with those studies that have broadly examined financial crisis prediction and credit risk forecasting, very few have looked into firm value forecasting. Poor firm management is widely recognized as being the main trigger for a financial crisis, and thus firm value can appropriately reflect the quality of corporate management. If managers can run their business with efficiency and target maximizing shareholders’ wealth, then investors will likely pay more than average to own their stock. The higher the firm value is, the stronger and more developed it is.
How to increase firm value as well as sound a corporate’s competitive edge turns out to be an essential task in this highly turbulent economic atmosphere. Although coming up with some generally accepted conclusions is quite difficult, it is widely acknowledged that corporates with good corporate social responsibility (CSR, which considers the voluntary integration of social and environmental concerns in a business operation and its interaction with stakeholders such as investors, shareholders, employees, suppliers, bankers, and regulators) have the prescribed means for addressing the challenge of globalization and increasing their competitive advantages (Organization for Economic Co-operation and Development, OECD). That is the reason why so many executives and researchers have devoted considerable amount of time and efforts to investigate the influence of CSR on firm value.
Although there are many different types of definitions and dimensions of CSR in the extant studies, Carroll [
7,
8,
9] defined four CSR dimensions: a corporate should (1) obey the laws and regulations announced by governments in its daily operations, (2) make products or provide services for customers to achieve suitable profitability in the process, (3) meet shareholders’ expectations and protect their wealth, and (4) strengthen and increase human welfare or firm reputation. Based on these perspectives, CSR consists of numerous factors, such as community involvement, labor security, environmental protection, human rights, and business standards. CSR may also function similarly to advertising, by enlarging a firm’s profit spread, increasing the demand for products and services, eliminating buyers’ price sensitivity, and solidifying consumer loyalty [
10,
11,
12].
Most research works attempt to identify the link between CSR and firm value in order to examine why firms engage in CSR [
13,
14]. Unfortunately, there is no conclusive theory that can explain the relation between CSR and firm value, although two dominant theories do exist. The agency theory [
15] indicates that corporates performing CSR activities see a decrease in firm value when managers use the firm’s limited resources to draw benefits of personal reputation at the expense of shareholders [
16]. On the other hand, the conflict resolution theory notes that corporates with high CSR activities can lead to higher firm value by mitigating conflicts of interest between managers and investors, raising firm reputation, and enhancing firm profitability [
17]. It also views CSR as a strategic investment to increase a firm’s competitive edge. The existing research on the relation between CSR and firm value is mixed and sometimes confusing [
18]. One of the possible reasons for not reaching a consensus conclusion comes from the effect of the quality–quantity trade-off among each one of the CSR dimensions [
2,
19,
20]. CSR encompasses economic, environmental, business, and social behaviors. Only using one synthesized indicator as a proxy to depict a corporate’s CSR performance is not reliable and trustworthy. Therefore, there is an urgent requirement to decompose CSR into some dimensions and further examine the impact of each dimension on firm value.
How to determine the most essential dimension on firm value is quite similar to handling the task of feature selection. The fundamental concept of feature selection is identifying a subset from the original set of features without impeding the model’s forecasting performance as well as improving the quality of the data and facilitating the calculation efficiency [
21]. However, most related works that considered feature selection are based on one pre-decided method. It is widely deemed that different method adoptions are likely to yield different outcomes (i.e., different selected features). If we can apply a number of dissimilar feature selection approaches and then combine the selection results, then we not only can realize the most essential feature that all the feature selection approaches “agree” on, but also enhance the model’s forecasting accuracy over utilizing one feature selection approach [
22].
This basic idea of combining multiple feature selection approaches is inspired by the ensemble learning theory—that is, the combination strategy is able to complement the error made by a singular method. By doing so, decision makers can realize which dimension of CSR has the greatest influence on firm value. Managers can then consider the potential implications to allocate valuable resources to an appropriate place so as to maximize stakeholders’ wealth and sustain the firm’s reputation. The selected outcome can then be entered into an emerging neural network-based model, namely support vector machine (SVM), to construct the firm value forecasting model. SVM [
23], grounded on statistical learning theory, produces an optimal separating hyperplane to discriminate two dissimilar class labels. There are some benefits in performing SVM [
24]: (1) there are only two parameters to be decided, (2) the solution of SVM is optimal and unique, and (3) the model has greater tolerance on extreme values. Due to these advantages, SVM was performed by this study. Investors can take the proposed model as a roadmap to adjust their investment portfolios so as to reach the goal of sustainable development.
The rest of this article is organized as follows.
Section 2 reviews the existing literature of CSR’s impact on firm value.
Section 3 proposes our research design.
Section 4 shows the experimental results.
Section 5 concludes.
5. Conclusions and Further Research
The many empirical research works up to date have identified no conclusive pattern in the relation between CSR and firm value. Ignoring CSR’s multi-dimensional characteristics is one of the possible reasons for this absence of a consensus conclusion. Given this concern, this study followed the “Research Report on Corporate Social Responsibility of China” to decompose CSR into four dimensions and further examine the impact of each CSR’s dimension on firm value. The focus of previous studies has been to identify “the single best” mechanism that is most precise for a pre-decided financial task, but this reliance on a single mechanism may be misguided and could contain some biases. To reach a more sound research outcome, a multiple combination strategy, grounded on the ensemble learning theory, was conducted herein. The basic idea of the ensemble learning theory is to complement the error made by a singular mechanism. Through different combinations of adopted strategies, users can realize the most representative features from an over-abundant database and find the most influential dimension on firm value.
The results herein indicate that X4: Environmental responsibility is the most essential element on firm value determination. The reason is because the Chinese government has placed much more emphasis on environmental protection and retains “vote power” over major decisions. In other words, if a corporate pollutes the environment, then the government has the right to delist the corporate regardless of major investors’ decisions. Managers of firms can consider the potential implications of these results and allocate valuable resources to an appropriate place in order to enhance their firm’s CSR performance, increase firm value, and reach the goal of sustainable development. Investors can look to invest in firms that have better resource utilization efficiency so as to maximize their wealth under anticipated risk exposure.
Certainly, this study has some limitations. First of all, this research was an exploratory study carried out with a high level of technology and a small sample. Larger samples with greater explanatory power will allow for more complex assessments in the future. Second, the effects of corporate social responsibility implementation include economic, social, and environmental impacts. All of these effects have short-term and long-term effects. Furthermore, some companies have implemented corporate social responsibility for some time, but some companies have begun to implement corporate social responsibility in accordance with government regulations. Although this study only attempts to explore the impact of stock prices (Tobin Q). Future research can continue to explore the long-term effectiveness of CSR through long-term concepts such as customer loyalty and/or sustainable value index. Finally, other studies combining and using different multiple attributes decision makings (MADMs) can provide insight into the unrecognized facets of CSR in this study. Future research can use different research methods such as time series analysis and prediction method to continue to study this issue.