Salovey and Mayer (1990) first discussed EI by dividing it into the following three categories of adaptive abilities: the appraisal and expression of emotions (AEE), regulation of emotions (RE), and utilization of emotions (UE) in problem-solving. Although emotion is at the heart of this model, it also encompasses social and cognitive functions related to AEE, RE, and UE. Goleman’s (1995) influential book Emotional Intelligence introduced many important correlates to EI and extended its structure to some extent to include specific social and communication skills that are influenced by emotional understanding and expression. Cooper and Sawaf’s (1997) well-known book Executive EQ outlined an EI model that links specific skills and tendencies through four building blocks: emotional literacy, well-being, depth, and alchemy. Beginning with the origin of EI as a concept, it has been understood as a set of interrelated abilities (Salovey and Mayer, 1990; Mayer and Salovey, 1997; MacCann et al., 2014) and an eclectic combination of many personality traits (Petrides and Furnham, 2003; Boyatzis and Sala, 2004; Tett et al., 2005; Bar-On, 2006), but the term EI came to be used to cover many different characteristics and concepts (Zeidner et al., 2004; Landy, 2005; Murphy and Sideman, 2006), leading to considerable confusion and misunderstanding of what EI is or should be (Ashkanasy and Daus, 2005; Gohm et al., 2005; Mayer, 2006).

To clarify the concepts in this field, researchers have developed two measurement models: ability EI scales and self-rating EI scales (Mayer et al., 2000). Ability EI scales require test-takers to demonstrate their knowledge by providing responses based on emotion-related information, whereas self-rating EI scales require participants to rate their agreement with a series of statements about themselves, thus reflecting their EI level. In related research, self-reported measures of EI were found to be more predictive of social functioning performance than performance measures (Brackett et al., 2006). In addition to the distinction between the two measurement models, there is a similar distinction between two theoretical EI models: mixed models and ability models. Mixed models primarily point to a broad compositional structure that leads to emotionally intelligent behavior, including emotion-related abilities, personality traits, and motivational factors (Bar-On, 2006; Petrides et al., 2007). Conversely, ability models define EI as a cognitive ability similar to language or numeracy but with its content focused on emotions rather than words or numbers (MacCann et al., 2014). Ashkanasy and Daus (2005) further distinguished between two rating scales: ratings of EI abilities and ratings of mixed models. The former are sometimes referred to as “self-perceptions of EI and emotional self-efficacy” (Qualter et al., 2012). Moreover, a recent meta-analysis divided EI measurement into ability, self-rated, and mixed EI (MacCann et al., 2020). In this study, we argue that there are two distinct types of assessments based on theoretical and empirical evidence (Joseph and Newman, 2010; O’Boyle et al., 2011): ability EI and mixed EI. The following describes the main ability models of EI and major mixed models of EI.

Mayer and Salovey (1997) developed a modified EI model consisting of the following four branches of EI: (a) perception, appraisal, and expression of emotions; (b) emotional facilitation of thinking; (c) understanding, analyzing, and employing emotional knowledge; and (d) reflective RE to promote emotional and intellectual development. Based on these four branches, the Mayer–Salovey–Caruso Emotional Intelligence Test (MSCEIT) became the best-known assessment of ability EI (Mayer et al., 2003). Additionally, there are some alternative measurements for only one or two of the four branches (e.g., Freudenthaler and Neubauer, 2007; MacCann and Roberts, 2008) as well as the predecessor of the MSCEIT (i.e., the MEIS; Mayer et al., 2000). The youth versions of the MSCEIT and MEIS have frequently been used in research in the field of schooling (Rivers et al., 2012). The Schutte Self-Report Emotional Intelligence Test (SREIT) developed by Schutte et al. (1998) was one of the earliest scales to assess self-reported EI. Specifically, the SREIT is a 33-item scale based on an earlier definition of EI than the four-branch hierarchical model proposed by Salovey and Mayer (1990). Wong’s Emotional Intelligence Scale is another measure of EI that utilizes this earlier definition with four components: perceiving one’s own emotions; perceiving others’ emotions; using emotions; and managing emotions (Law et al., 2004). Subsequently, the Self-Rating Emotional Intelligence Scale was developed to accurately capture the four-branch model of EI. This 19-item scale consists of 5 subscales: perceiving emotions; using emotions; understanding emotions; managing one’s own emotions; and managing others’ emotions (Brackett et al., 2006).

Appraisal and Expression of Emotions theoretically includes the ability to recognize one’s own emotions (i.e., internal stimuli); the ability to express emotions in response to external stimuli; the ability to accurately express one’s emotions; the ability to distinguish between true and false emotions; and the knowledge of display rules for emotional expression in different cultural contexts (Salovey and Mayer, 1990; Mayer and Salovey, 1997; Mayer et al., 2016). However, in actual tests of EI, AEE is primarily used to measure an individual’s capacity to identify the type and intensity of an emotion present in external stimuli, such as facial expressions, micro expressions, tone-of-voice, body postures, landscapes, and evocative art (Nowicki and Duke, 1994; Matsumoto et al., 2000; Mayer et al., 2003; Schlegel et al., 2014). Once an emotion is perceived, it serves as an input for cognitive systems (Mayer et al., 2001). In this regard, individuals can understand the domain-specific knowledge of emotional content, which includes knowledge about emotions and emotion-related phenomena. According to Mayer et al. (2001), AEE is the central locus for abstract processing and reasoning about emotions and emotional information, including a vocabulary of emotional terms, the antecedents and consequences of emotions, the ways in which emotions may combine or change over time, and the possible effects that particular situations may have on an individual’s present or future emotions (Salovey and Mayer, 1990; Mayer and Salovey, 1997; Mayer et al., 2016).

Regulation of Emotion refers to regulating one’s own and others’ emotions to achieve desired outcomes, such as personal development (Mayer and Salovey, 1997; Mayer et al., 2001, 2016). Because it involves managing one’s own and others’ emotions (i.e., intrinsic and extrinsic RE; Gross and Thompson, 2007; Gross, 2008), this facet includes knowledge about RE and related cognitive strategies (Mayer et al., 2016). RE is also based on personal goals, such as upregulating and downregulating emotions to achieve personal growth. Hence, RE not only represents knowledge regarding how to regulate emotions but also serves as an important motivating factor in deciding when and why to regulate emotions (Mayer et al., 2001).

Utilization of Emotion refers to using emotion and affective information as an input or guidance for cognitive tasks or decision-making. It involves using existing emotions to guide task selection or approaches and generate new emotions to help accomplish specific tasks (Mayer et al., 2016). In this case, positive emotions can lead to innovative thinking because they give individuals a broad perspective, whereas anxiety is a form of hypervigilance to threats, i.e., when using existing emotions, individuals use their current emotional state as a key reference (Fredrickson, 2001; Bar-Haim et al., 2007). Additionally, tasks can be chosen to take advantage of emotional states, which may contribute to performance. For instance, one might choose to welcome new colleagues when in a good mood but wait to attend a funeral or other solemn occasion until the mood becomes more serious.

LS, an important aspect of individual subjective well-being (Kjell et al., 2016), reflects an individual’s overall evaluation of their life experience (Diener et al., 2002; Schimmack et al., 2004). Factors influencing LS have been discussed in previous studies, including financial status (Johnson and Krueger, 2006; Boyce et al., 2010; D’Ambrosio et al., 2020), educational level (Kousha and Mohseni, 1997; Meeks and Murrell, 2001; Park et al., 2020), and self-esteem (Utsey et al., 2000; Diener and Diener, 2009; Liang et al., 2020). The factors influencing LS have attracted the attention of researchers because it has a significant impact on physical and psychological health (Strine et al., 2008; Garcia and Moradi, 2013; D’Ambrosio et al., 2020) and is an important component of subjective well-being (Diener, 1984). Positive affect (PA) and subjective well-being have been found to predict greater job dedication and loyalty as well as lower rates of burnout and turnover (Rhoades and Eisenberger, 2002). Rural teachers are generally less satisfied with life than urban teachers (Qiao and Lina, 2019). In view of the work pressures that teachers face, exploring the factors influencing their LS and examining PA and negative affect (NA) will help them better cope with their daily life and work.

EI plays a nonnegligible role among the many factors that affect LS. Individuals with high trait EI report higher LS (Urquijo et al., 2016). This positive association remains significant after controlling for many sociodemographic variables (e.g., Extremera and Fernández-Berrocal, 2005; Gohm et al., 2005; Ruiz-Aranda et al., 2014). Another study found that trait EI is more strongly associated with LS than performance-based EI (Sánchez-Álvarez et al., 2016). The conceptualization and assessment of EI are controversial, but two main ideas are accepted. The first views EI as a set of skills, similar to cognitive intelligence. For instance, Salovey and Mayer (1990) defined ability EI as “the subset of social intelligence that involves the ability to monitor one’s own and others’ feelings and emotions to discriminate among them and to use this information to guide one’s thinking and actions” (p. 189). The second approach primarily considers EI as a set of traits, similar to personality, with Petrides et al. (2007) defining trait EI as “a distinct and compound construct that lies at the lower levels of personality hierarchies” (p. 283).

Previous research has examined the association between trait EI and LS in diverse population samples, namely, in university graduates (Amdurer et al., 2014), university employees (Hafiz and Chouhan, 2015), and employees of secondary schools. This result was also obtained in groups of teachers (Mehta and Mehta, 2015) and undergraduates (Afolabi and Balogun, 2017). However, little attention has been paid to rural teachers. One study found that the LS of rural teachers in China is generally not high (Qiao and Lina, 2019). Among rural teachers who lack resources and face greater pressure, studying the relationship between LS and trait EI will not only enable the adoption of steps to improve the physical and mental health of rural teachers but also help develop measures to stimulate their enthusiasm for education and teaching.

The positive effects of trait EI on LS are well established, but the exact mechanism of this effect remains unclear. Several studies have identified the mediating role of numerous variables, such as self-evaluation characteristics, perceived stress, and social support (e.g., Zeidner et al., 2012; Kong and Zhao, 2013; Sun et al., 2014; Urquijo et al., 2016). Here, we attempt to identify the mechanisms behind this association from the perspective of PA and NA. In particular, for rural teachers, we investigate how they use their emotional resources and abilities to improve their own LS and the mechanisms behind the improvement (Hypothesis 1).

A total of 577 participants were included in this study, all of whom were living in county areas in central China, that is, rural areas that are governed under the category of county-type administrative regions in China. Among these, 148 were male (26%) and 429 were female (74%). The average age of the participants was 28.40 years (SD = 6.75). The participants had an average of 15 years of education (SD = 0.627), and most had a university degree (97.57%). Nearly two-thirds were married (65.34%). More than three-quarters of the participants (79%) had no part-time administrative job. The sample was made up of 72.27% teachers at the primary level and 24.96% at the junior high school level. Nearly 60% of the participants were second-level teachers or below, which is representative of the average level of teachers of basic education in China. Thus, the data analysis of these 577 teachers can be regarded as reflecting the psychological characteristics of Chinese rural teachers of basic education. Although the study participants have certain structural differences, there still existed strong homogeneity due to homogeneity among the rural areas of central China. The survey data analysis also reveals some commonalities. In the subsequent analysis, structural differences in terms of subjects such as age, gender, and teaching level were controlled (Table 1).

All data used in this study were collected online via questionnaires. Three measures were used, one for trait EI, PANAS, and general LS. Additionally, basic demographic information was collected from all participants.

In this section, we first evaluate the theoretical context of each core measurement (EI\PA\NA\LS) and then test the reliability and validity of their scales using Bartlett’s test, the Kaiser–Meyer–Olkin (KMO) test, and Cronbach’s alpha coefficient.

The Chinese version of the questionnaire used in the study was developed via back-translation (Brislin, 1986). First, the original English version was translated into Chinese by two individuals who were fluent in English. Second, the Chinese version was translated back into English by two other individuals who were unfamiliar with the original English version. Third, the translated versions were compared, and adjustments were made to ensure that the Chinese version was comparable with the original English version. Fourth, a panel of five researchers reviewed the Chinese and original English versions to finalize the version used in this study.

The participants were recruited via social media in the early fall semester of 2022, during the COVID-19 pandemic. Because the lockdown in China was relatively eased during this period, the teaching life in the school was somewhat normal. Those who were willing to participate in the study were provided with an informed consent form and web link to respond to the questionnaires online. The participants were required to sign the informed consent form before completing the surveys. Participation was voluntary, and participants could withdraw from the process at any time. Because the participants were anonymous, no personal information was collected. No financial compensation was provided to the participants. The study was reviewed and approved by the East China Normal University institutional review board.

This study adopted a combination of data-driven and theory-driven methods for analyses. First, this study used a data-driven GLASSO model to establish an association network of all variables to present the original relationships among variables. This method has now attracted attention in the field of psychometrics, especially to screen core variables, verify the measurement structure, and explore the relationship among variables (Epskamp et al., 2018; Epskamp and Fried, 2018). In our analysis, we used R Studio to call the GLASSO degree package to implement this function and used the social network analysis tool Gephi to visualize the results of the model. Additionally, to prevent false correlations and multicollinearity among the indicators, the penalty term lambda was set to 0.1 in the model parameter setting.

Second, according to the theoretical framework of the current study and the analysis results of the GLASSO model, we used an SEM to test its core content and ensure the robustness of the core mechanism. In the software tool, we used the Structural Equation Module in STATA 17 to complete the second step analysis.

Figure 1 presents the analysis results of the GLASSO model. This model is also a visual representation of the correlation coefficient matrix and the measurements from the reliability and validity tests. Furthermore, it shows the potential structure of each measurement dimension. The network has four subgroups (communities), which correspond to the four types of measurement that this study focused on, and the structure of the four measurements was relatively clear. We found that the EI subgroup influenced the variables of the PA subgroup, which indirectly affected LS. Some EI variables also directly affected LS. However, the network connections between the EI and NA subgroups were relatively sparse, which indicates that the correlation between the two types of variables (EI and NA) was relatively low. Furthermore, there were relatively few connections between NA and LS variables, and the relationship between the two may not be obvious.

In addition, we found that the five variables—EI33, EI28, EI5, EI4, and EI1—had a marginal position in the network, with deviations in their subgroup classification (EI33, EI28, and EI5), thereby indicating that the structural directivity of these five indicators was relatively weak. We tested the five variables using CFA and found that their contributions to the measured latent variables were relatively low, with factor loadings less than 0.4. Therefore, to ensure the accuracy of the following analysis, we eliminate them from the SEM analysis.

In the Introduction, we suggested three mechanisms that explained why EI could be correlated with LS. First, those with higher EI may be better able to regulate negative feelings, such as anxiety, stress, and disappointment, which are all connected to LS (Roeser et al., 2013; Carroll et al., 2022). If this is true, RE could be responsible for these effects. Second, those with higher EI might be better at navigating their social world and building stronger relationships with students, colleagues, and their respective families (Jennings and Greenberg, 2009). Thus, RE would be responsible for the effects. Third, the appraisal and expression of emotional abilities may overlap with the need for educational proficiency and subject matter understanding. For teachers to be more successful, they must become better at comprehending human emotion and displaying knowledge; those that possess such skills are more likely to have positive experiences in school (MacCann et al., 2014). In this case, the knowledge reserve of appraisal, understanding, and expression (a facet of EI) will have the most significant impact. We elaborate on the importance of these three mechanisms in the following sections.

Teaching is an emotionally draining profession that can have negative effects on teachers’ well-being (Luque-Reca et al., 2022). In particular, rural teachers may be at risk due to the pressure of teaching and caring for students while learning new teaching skills. To obtain higher LS, such individuals must make the most of their positive emotions (Szczygieł and Mikolajczak, 2017). Negative emotions can have a similar effect, and those with higher RE can reduce the duration of such emotions, which can increase LS. Another possible explanation is that teachers who perceive themselves as being successful in managing their emotions may experience a sense of self-efficacy, which can lead to increased PA and higher LS (Vergara et al., 2015). Hence, the ability to regulate emotions is crucial for rural teachers. The model results from the EI facets suggest that PA and NA partially mediate the correlation between RE and LS.

There is evidence that higher EI is associated with stronger relationships within school settings (Lopes et al., 2004, 2005). Similarly, higher EI has been found to be an effective predictor of social support within the educational environment (Ciarrochi et al., 2001; Kong et al., 2012). Rural teachers with strong EI skills may be more capable of managing their lives and cultivating positive relationships with students, colleagues, and their respective families (Jennings and Greenberg, 2009). In the three facet-level subdimensional models, RE has the best model fit (AIC = 36407.088, BIC = 36707.779). As shown in the AEE and UE models, RE also has an indirect impact on LS through PA in addition to its direct impact, which is more likely to result from making it easier for rural teachers to access resources or improve achievement. Meanwhile, the indirect effect of RE on LS is more likely due to the help it provides to build a social support network for managing stress. According to the results, RE is more about improving PA, which has an indirect impact on LS. For example, teachers resonate with students in daily teaching to obtain more positive feedback, thereby improving their sense of accomplishment and LS.

In Mechanism 3, we expected that the AEE facet of EI would be the strongest factor in predicting LS. However, the empirical model results did not support this assumption. In contrast, we found that the AEE model had the least predictive accuracy (AIC = 40826.774, BIC = 40162.328). Previous studies have shown that the perception and understanding facets of ability EI are the most important factors in predicting students’ academic performance in the school environment (MacCann et al., 2020). Therefore, we question why the AEE model has the least explanatory power for LS. We believe that the SREIT scale used in this study does not include the understanding-branch subscale, making the use of the scale different from previous research. The potential overlap between EI and teaching skills appears to be less important for their LS possibly because larger vocabularies and better expressive skills are not characteristic of teaching and school life in rural areas. Moreover, rural teachers are usually required to teach using textbooks and relying on examinations instead of exploratory teaching. Thus, the characteristics of this teaching mode do not necessitate more vocabulary knowledge and the ability to express oneself. In comparison, rural teachers may benefit more from having the ability to regulate and appropriately use emotions than from having a large vocabulary and expressive abilities.