5.1 Case selection
Based on the causative factor model derived from grounded theory, this study identified a total of six condition variables. Following the scholars' recommendations on the number of cases for fsQCA (Zhang et al., 2017), and under the premise of determining six condition variables, 45 representative cases were selected for fsQCA analysis. As a case-oriented research method, case selection is crucial, and the QCA method adheres to the principle of theoretical sampling rather than random sampling (Du and Jia, 2017). The selection of typical cases in this study adheres to the following criteria:
(1) Representativeness: The selected accident cases are widely reported in the news or have garnered significant attention on the internet, demonstrating a certain level of representativeness.
(2) Diversity: The chosen accident cases exhibit diverse characteristics in terms of accident outcomes, types, and other dimensions, enhancing the explanatory electric power of the configuration path of causative factors in electric power production safety incidents.
(3) Completeness: The selected accident cases have comprehensive records of the entire incident, including crucial information such as the time and location of the accident, accident type, causes, and outcomes.
5.2 Variable configuration and assignment
The setup and assignment of variables in fsQCA are crucial steps, determined by researchers based on their experience and relevant theoretical judgments. This process relies on external knowledge rather than the inherent attributes of the data itself (Du and Jia, 2017).
1 Variable Configuration
In this study, six factors derived from grounded theory were established as condition variables for the causative model of electric power production safety incidents. These variables include human unsafe behavior, equipment factors, enterprise safety management, on-site safety management, safety qualifications of production personnel, and environmental factors. The outcome variable in this study is the severity level of electric power production safety incidents. The classification of accident severity refers to the "Regulations on Reporting and Investigation of Production Safety Accidents" (Legislative Affairs Office of the State Council of China, 2007).
2 Variable assignment
In fsQCA, membership scores range from 0, indicating complete non-membership, to values between 0 and 0.5, signifying fuzzy non-membership. The midpoint, 0.5, represents the crossover point of the result set. Values between 0.5 and 1 indicate fuzzy membership, while a membership score of 1 denotes complete membership (Zhang and Du, 2019). Considering the actual circumstances of safety incidents and the assignment rules from relevant studies (Li and Feng, 2023), this study assigns membership values for condition variables based on their occurrence in the cases. If a condition variable appears in a case, it is considered to have fuzzy membership, and a fuzzy value between 0.5 and 1 is assigned. The specific assignment rules are outlined in Table 2.
Table 2 Assignment rules
Unsafe Human Behavior
|
Errors or Mistaken Operations
|
The absence of the appeal situation in the case is assigned a value of 0; satisfaction of one condition is assigned a value of 0.6; satisfaction of two conditions is assigned a value of 0.8; satisfaction of all conditions is assigned a value of 1.
|
Failure to Adhere to Safety Protocols
|
Violations of Operational Procedures
|
Enterprise Safety Management
|
Lack of Safety Training
|
The absence of the appeal situation in the case is assigned a value of 0; satisfaction of one condition is assigned a value of 0.6; satisfaction of two conditions is assigned a value of 0.8; satisfaction of all conditions is assigned a value of 1.
|
Absence or Non-Implementation of Safety Regulations
|
Negligence by Enterprise Management
|
On-site Safety Management
|
Insufficient On-site Safety Supervision
|
In the case, a score of 0 is assigned when there is no appeal situation; a score of 0.7 is assigned when one condition is met, and a score of 1 is assigned when all conditions are met.
|
Incomplete On-site Hazard Inspections
|
Equipment Factors
|
Equipment Malfunction
|
The absence of the appeal situation in the case is assigned a value of 0; satisfaction of one condition is assigned a value of 0.6; satisfaction of two conditions is assigned a value of 0.8; satisfaction of all conditions is assigned a value of 1.
|
Inherent Equipment Defects
|
Quality Issues with Equipment
|
Safety Competence of Production Personnel
|
Diminished Safety Awareness
|
The absence of the appeal situation in the case is assigned a value of 0; satisfaction of one condition is assigned a value of 0.6; satisfaction of two conditions is assigned a value of 0.8; satisfaction of all conditions is assigned a value of 1.
|
Limited Risk Identification Capabilities
|
Inadequate Job Skills and Competencies
|
Environmental Factors
|
Production Environment
|
In the case, a score of 0 is assigned when there is no appeal situation, and a score of 1 is assigned when the condition occurs.
|
Natural Environment
|
Severity of Incidents
|
Major Incidents
|
1
|
Significant Incidents
|
0.8
|
Ordinary Incidents
|
0.6
|
Potential Hazardous Incidents
|
0.4
|
5.3 Univariate necessary condition analysis
The univariate necessary condition analysis in this study was conducted using the R programming language in conjunction with the fsQCA 4.1 software. NCA serves the dual purpose of identifying specific conditions that are necessary for a particular outcome and analyzing the effect size of these conditions (Dul, 2016). The effect size ranges from 0 to 1, with larger values indicating a more substantial effect, and values below 0.1 suggesting a minimal effect. NCA methodology is versatile, capable of handling both continuous and discrete variables. In cases where both x and y are continuous or discrete, and there are five or more levels, the analysis employs Ceiling Regression (CR) to generate an upper limit function. However, if x and y are binary variables or discrete variables with fewer than 5 levels, the method uses Ceiling Envelopment (CE) to generate the function (Du et al., 2020). Given the characteristics of the data in this study, particularly the presence of discrete variables with fewer than 5 levels, CE was employed as the method for function generation.
In Table 3, this study presents the results of the R language NCA. In the NCA method, for a condition to be deemed necessary, two criteria must be met: the effect size (d) should not be less than 0.1, and the Monte Carlo simulations of permutation tests should indicate that the effect size is significant (Dul et al., 2020). From the results in Table 3, it is observed that the p-values for all conditional variables are greater than 0.01, suggesting that none of them can be considered necessary conditions for causing electric power production safety accidents. Furthermore, this study employs QCA to verify necessary conditions, as shown in Table 4. The consistency of individual condition necessity is below 0.9, aligning with the NCA results, indicating the absence of necessary conditions within the conditional variables for causing electric power production safety accidents.
Table 3 R Language necessity analysis results
Conditional Variables
|
accuracy
|
Ceiling zone
|
Scope
|
Effect size(d)
|
p-value
|
Unsafe Human Behavior
|
100%
|
0.160
|
0.5
|
0.333
|
0.701
|
Enterprise Safety Management
|
100%
|
0.320
|
0.6
|
0.533
|
0.049
|
On-site Safety Management
|
100%
|
0.000
|
0.6
|
0.000
|
1.000
|
Equipment Factors
|
100%
|
0.200
|
0.6
|
0.333
|
0.432
|
Safety Qualities of Production Personnel
|
100%
|
0.160
|
0.6
|
0.267
|
0.193
|
Environmental Factors
|
100%
|
0.000
|
0.6
|
0.000
|
1.000
|
Table 4 QCA software necessity analysis results
Variables
|
Consistency
|
Coverage
|
UB
|
0.875
|
0.856
|
∼UB
|
0.551
|
0.872
|
ES
|
0.801
|
0.826
|
∼ES
|
0.544
|
0.796
|
OS
|
0.287
|
0.907
|
∼OS
|
0.868
|
0.648
|
EF
|
0.790
|
0.771
|
∼EF
|
0.496
|
0.789
|
SQ
|
0.500
|
0.907
|
∼SQ
|
0.750
|
0.580
|
EVF
|
0.154
|
0.600
|
∼EVF
|
0.845
|
0.605
|
5.4 Configuration analysis and interpretation
1 Configuration analysis
This study utilized fsQCA4.1 software to analyze the configurations leading to electrical power production safety incidents, representing different causal configurations resulting in diverse occurrences of safety incidents. The original consistency threshold was set at 0.8, the PRI (Pattern-Response-Implication) consistency threshold at 0.75, and the case frequency threshold at 1 (Frambach et al., 2016). Given the aim of exploring the impact of conditional variables on electrical power production safety incidents and the absence of necessary conditions in the univariate necessary condition analysis, the "Present or Absent" option was chosen for counterfactual analysis. By comparing the nested relationship between intermediate and reduced solutions, the core conditions of each solution were identified: conditions appearing in both intermediate and reduced solutions were considered core conditions, while those appearing only in the intermediate solution were considered marginal conditions (Du and Jia, 2017).
QCA Analysis results are presented in Table 5. The consistency of all configuration sets exceeds 0.8, indicating that all cases meet the consistency requirement. In other words, the seven configurations in Table 3 are sufficient conditions leading to electric power production safety incidents. Each configuration involves at least two variables, reaffirming that the occurrence of electric power production safety incidents is the result of the combined action of multiple conditions. The overall consistency is 0.97, exceeding 0.9, indicating that, as a whole, the configurations are sufficient conditions for the occurrence of electric power production safety incidents. The overall coverage is 0.83, indicating that the seven configurations can explain 83% of the cases of electric power production safety incidents. Following the configurational theorizing process, equivalent configurations discovered in this study are named (Fainshmidt et al., 2020), meaning they share the same core conditions (Fiss, 2011). In Table 5, the seven path combinations are categorized into three types of incidents, namely, management deficiency, low safety competence, and unsafe behavior.
Table 5 Configuration of factors affecting electric power production safety accidents
2 Configuration explanation
(1) Management Deficiency Type: When there are deficiencies in enterprise safety management and on-site safety management, it is prone to causing electrical production safety accidents. The management deficiency type includes three paths, namely C1a, C1b, and C1c, all of which have core conditions of deficiencies in enterprise safety management and on-site safety management. In C1a, deficiencies in enterprise safety management and on-site safety management serve as core conditions, while unsafe behavior is a marginal condition. This indicates that during on-site operations, workers exhibited unsafe behavior or had deficiencies in unsafe behavior, but due to inadequate enterprise safety management and on-site safety management, timely intervention did not occur, ultimately resulting in a safety accident. In C1b and C1c, deficiencies in enterprise safety management and on-site safety management serve as core conditions, with equipment factors and environmental factors serving as marginal conditions. These two configurations indicate that due to deficiencies in enterprise safety management and on-site safety management, potential safety hazards in the working equipment or working environment on-site were not correctly identified before production operations, leading to the occurrence of safety accidents. An example of a management deficiency type accident is seen in the case of "Electrical Accident at a Wind Power Company," where due to insufficient safety education and training of employees by the company, an employee was allowed to go to work. On-site management failed to effectively supervise and did not detect Sun's erroneous behavior, leading to his accidental contact with the high-voltage side B-phase cable inside the box, causing the plug with a load to disconnect, generating an arc and resulting in fatal electrocution.
(2) Low Safety Competence Type: The low safety competence type encompasses two configuration paths, C2a and C2b, both with the core condition of inadequate safety competence among production personnel. Unsafe behavior concurrently serves as the edge condition for both paths, indicating an inherent connection between safety competence and unsafe behavior. From the causation model constructed based on grounded theory, inadequate safety competence among production personnel includes factors such as a lack of safety awareness, poor risk identification capabilities, and insufficient job skills and abilities. These factors have the potential to lead production personnel to engage in unsafe behavior, consequently resulting in safety accidents. An illustrative case of the low safety competence type is evident in the "Electrical Shock Fatality Incident at a Certain Electric power Generation Limited Company," where weak safety awareness and inadequate job skills of the operators, combined with an unclear understanding of the energized components of on-site equipment, led to a fatal electric shock incident when the operator, in the presence of energized static contacts on the upper part of the generator switchgear, opened the isolating baffle inside the cabinet and came into contact with the static contacts at the generator outlet switch, resulting in an electric shock fatality.
(3) Unsafe Behavior Type: The unsafe behavior type comprises two configuration paths, namely C3a and C3b, both with the core condition of unsafe behavior by production personnel. According to the causation model inferred from grounded theory for safety incidents in electric power generation, unsafe behavior in the electric power production process primarily includes errors or operational mistakes, failure to adhere to safety protection requirements, and engaging in unauthorized actions. For example, in the "Fall Accident during Maintenance Project at a certain Electric power Engineering Limited Company," on-site personnel, including Mr. Na, violated enterprise safety management regulations. High-altitude work was conducted without proper work permits, and during the operation, the safety harness was improperly removed, resulting in an accidental fall and subsequent incident.
3 Robustness test
This study conducted a robustness analysis of the fsQCA results, commonly employed methods include adjusting calibration thresholds, changing PRI consistency thresholds, adding or removing cases, altering frequency thresholds, and introducing other conditions (Zhang and Du, 2019). Initially, this paper increased the PRI consistency threshold from 0.75 to 0.8, following the referenced methods, and found that the three identified accident types remained inducible. The overall consistency remained largely unchanged, while the overall coverage decreased from 0.83 to 0.77. Subsequently, 4 cases were randomly selected and removed, constituting a 10% reduction in the dataset. The solutions remained similar, indicating the robustness of the research findings.