*2.2. Combination Weight-Cloud Evaluation Comprehensive Evaluation Model*

#### 2.2.1. Combination Weight Model

The analytic hierarchy process (AHP) is a method of subjective empowerment, and its basic idea is to use the systematic idea of decomposition followed by synthesis to organize and synthesize people's subjective judgments, realize the organic combination of qualitative and quantitative analysis, and complete quantitative decision-making [31,32]. The general steps of the research using this method are: (1) establishing the hierarchical structure model; (2) constructing the judgment matrix; (3) calculating the index weights; (4) testing the consistency of the judgment matrix. In the specific operation, due to the problems of large calculation workload and tedious testing process, this study uses Yahhp software for subjective weight measurement. The entropy weight (EW) method is a method of objective assignment of weights, the core of which is to use the amount of data information of each indicator to determine the weight; when the evaluation data value of an evaluation indicator differs greatly, its entropy value is smaller, indicating that the evaluator has a greater difference in the sensitivity degree of the indicator, that is, the indicator can provide more reference information for the evaluation of the merits, and has greater significance within the evaluation system [33,34]. The general steps when using this method for research are: (1) standardization of data; (2) calculation of the entropy value of each indicator; (3) calculation of the weight vector of each indicator. This study used AHP-EW for combined weighting to obtain more accurate and objective weights. The specific formula can be found in the related literature [35].

#### 2.2.2. Cloud Evaluation Model

The cloud model is a kind of evaluation method based on probability statistics and fuzzy set theory, and its evaluation results can be expressed by cloud digital features (*Ex*, *En*, *He*), which is schematically shown in Figure 1. When cloud model evaluation method is used, it can be realized by the cloud generator (CG), and four types of each cloud generator algorithm are shown in Figure 2. Specific algorithms can be found in the related literature.

#### 2.2.3. Comprehensive Evaluation Model

This study uses a combination of combined weights and cloud model to evaluate the energy saving and environmental protection management status of the company, and the specific steps are as follows. When using this method for evaluation, the general steps are: (1) establish the weight factor set *W* = {*ω*1, *ω*2,..., *ωn*} of indicators; (2) determine the indicator set and the evaluation language domain *V* = {*V*1, *V*2,..., *Vm*}, in this study, the evaluation language is divided into five levels: vigilance-level, improvement-level, transition-level, acceptable-level, and declarable-level; (3) determine the cloud parameter matrix (*Ex*, *En*, *He*) for each level of each indicator; (4) calculate the affiliation degree of each sample and each indicator; and (5) determine the evaluation level. The specific formula for each step is referred to in the related literature [36].

**Figure 1.** Normal cloud and digital features (*Ex* = 0, *En* = 1, *He* = 0.1).

**Figure 2.** Schematic diagram of cloud generator. (**a**) Forward CG. (**b**) Backward CG. (**c**) X-conditional CG. (**d**) Y-conditional CG.

#### *2.3. Data Collection and Samples*

According to the Guidelines on Industry Classification of Listed Companies issued by CSRC, listed companies in the energy industry from 2006–2017 were selected for this study (industry codes 06, 07, 25, 44, 45, 46). The sample was also carefully screened (e.g., shaving off ST and \*ST companies; shaving off companies listed after 2006, etc.), and after sample screening, 59 companies with 378 sample observations finally remained. It is worth noting that there are still 78 companies that did not release any ESEP-related reports during 2006–2018 and did not participate in this evaluation study.

The original data of this study can be divided into quantitative index data and qualitative index data. Quantitative indicators such as COD per ten thousand yuan output value, SO2, NOX, solid waste emissions, comprehensive energy consumption of ten thousand yuan output value (ton of standard coal/ten thousand yuan), etc. can be obtained or calculated through the social responsibility report, CSMAR database, enterprise official website and other channels. Quantitative indicators are difficult to be quantified by themselves, and they need to be quantified in combination with expert scoring and information disclosure measurement methods. Referring to relevant literature [37,38], this study uses 1–5 score points for quantification (see quantification standard of indicators in Table 2).


**Table 2.** Institutional indicators—Quantitative scoring standard.

#### **3. Results and Discussion**
