Assessing Emotional Self-Efficacy: Evaluating Validity and Dimensionality with Cross-Cultural Samples

Authors


  • An earlier version of this manuscript was presented at the 2008 annual meeting of the Academy of Management at Anaheim, CA. We warmly thank Bangseob Yoon and his colleagues for their tremendous cooperation in the Korean study administration. We also appreciate Yoonhyuk Jung's help with back-translation.

Sungwon Choi, Department of Business Administration, College of Government and Business, Yonsei University at Wonju, Wonju, Kangwondo, South Korea, 220-710. Email: swchoi33@yonsei.ac.kr

Abstract

The concept of emotional intelligence (EI) has frequently been conceptualised as a stable construct (i.e. an intelligence or disposition). As a theoretically distinct form of EI, we advance a measure of emotional self-efficacy (ESE), which is more dynamic and malleable than traditional EI measures. Two studies across two cultures yield a six-dimensional model of ESE. These six dimensions are shown to differ in their relationships with various outcomes. ESE is shown to predict low levels of stress and high levels of life satisfaction and coping beyond self-efficacy.

INTRODUCTION

Since its inception in the early 1990s (Mayer, DiPaolo, & Salovey, 1990), research on emotional intelligence (EI) has made significant contributions to an understanding of how individuals differ in their affective processes and how those differences impact their social functioning (Ashkanasy & Daus, 2005; Cote & Miners, 2006; Law, Wong, & Song, 2004; Mayer, Salovey, Caruso, & Sitarenios, 2003; Petrides & Furnham, 2000). Two major streams of literature have been developing within the broad EI framework: ability- and trait-based EI (Daus & Ashkanasy, 2005; Petrides, Furnham, & Mavroveli, 2007). The ability EI approach (Mayer & Salovey, 1997) utilises maximum-performance tests to quantify one's emotional capability. These tests are similar to IQ tests (Daus & Ashkanasy, 2005; MacCann & Roberts, 2008; Mayer, Caruso, & Salovey, 1999). On the other hand, trait EI is viewed as a person's predisposition (Furnham, 2006; Petrides, Frederickson, & Furnham, 2004) and is typically measured via self-reports. Within this broad EI framework, various theorists have tested divergent models and self-report measures such as WLEIS (Wong and Law's EI Survey; Wong & Law, 2002), SREIS (Self-Report EI Scale; Brackett, Rivers, Lerner, Salovey, & Shiffman, 2006), and EIS (Emotional Intelligence Survey; Schutte, Malouff, Hall, Haggerty, Cooper, Golden, & Dornheim, 1998).

An aim of this research is to clarify and redefine the meaning of self-report EI in the framework of self-efficacy. We contend that a person's self-perceptions of his/her emotional abilities should be viewed as a specific form of self-efficacy (Bandura, 1982), called emotional self-efficacy (ESE). Conceptualising EI in the light of self-efficacy has importance because it theoretically explains why people behave sub-optimally even though they know what is emotionally correct in social situations (Bandura, 1977). Studies in self-efficacy have shown that what a person does is an integrative orchestration of his/her cognitive, social, and affective skills, rather than what the person is simply able to do. A person is capable as much as he/she perceives him/herself to be, and these judgmental capabilities play a mediating role between ability and action (Bandura, 1977; Bandura, Adams, & Beyer, 1977). Two recent meta-analyses regarding EI (Joseph & Newman, 2010; O'Boyle, Humphrey, Polack, Hawver, & Story, 2011) demonstrate that ability EI has a consistent but only weak direct effect on individual performance. We argue that ESE facilitates this bridging role between individuals' ability and performance in the realm of emotions. We also contend that the conception of ESE can address the issue of empirically scant findings of ability EI.

As ESE is conceived of as efficacy in relation to EI, we use the ability-based model of EI (Mayer & Salovey, 1997) as a theoretical foundation to build the construct. The Mayer and Salovey model consists of the following four emotional abilities: to perceive emotions in self and others, use emotions to facilitate thought, understand emotional complexity, and manage emotions in self and others. We scanned the literature to draw upon three well-established measures of self-reported emotional abilities that collectively tap our proposed conceptualisation of ESE. Next, in line with Albert Bandura's (2006) guide for constructing self-efficacy scales, we theoretically draw upon items with wording congruent with the assessment of self-efficacy, which culminated in a comprehensive 24-item measure of the six proposed dimensions of ESE. We then assessed the factor structure, reliability, and validity of ESE, using multiple samples across two cultures.

EMOTIONAL SELF-EFFICACY

Linking self-efficacy and emotions is not new in the literature. Postulating the bridging role of self-efficacy between knowledge and action, Albert Bandura noted that perceptions of one's self-efficacy (i.e. judgments on how well one can execute courses of action required to deal with prospective situations; Bandura, 1982, p. 122) closely affect individuals' thought patterns, behavioral choices, and emotional reactions. Feelings of being inefficacious (regardless of being accurate or faulty) entail emotional arousal such as anxiety, fear, and apprehensiveness, and these emotions influence how much an individual thinks in a facilitating or debilitating manner, how much he/she expends effort on and perseveres with a task, and how much he/she is resilient against stressful and perturbing situations (Bandura, 1997). Perceived efficacy is a judgment of capability to execute given types of performance. Efficacy beliefs influence the courses of action people choose to pursue, the outcomes they expect their efforts to produce, how much stress and depression they experience in coping with taxing environmental demands, the life choices they make, and the accomplishments they realise (Bandura, 2006).

Along this line of reasoning, Bandura and colleagues (2003) introduced a concept of regulatory emotional self-efficacy (RESE), which refers to beliefs in one's capability to ameliorate negative emotional states and to experience positive emotions. The RESE and ESE both aim to conceptualise a person's belief in his/her emotional abilities. When comparing ESE to RESE, however, the ESE perspective is theoretically grounded in the ability-based model of EI, and it includes not only emotional regulation but also other facets of emotional abilities such as perceiving and understanding emotions. The ESE perspective differs from ability-based EI because it defines the construct as a person's belief in his/her emotional ability, not the ability per se (Kirk, Schutte, & Hine, 2008). We also argue that the ESE perspective differs from that of trait EI, although some researchers have implied that trait EI can be considered a self concept (Kluemper, 2008) or emotional self-efficacy (e.g. Petrides & Furnham, 2003; Petrides, Sangareasu, Furnham, & Fredrickson, 2006). However, Kirk and colleagues (2008) noted that equating trait EI with emotional self-efficacy may be an overgeneralisation. As trait EI researchers admit (e.g. Petrides, Pita, & Kokkinaki, 2007), trait EI includes various dispositional traits as well as emotional self-concept. This nomological distinction is important because the ESE perspective implies that a person's belief in his/her emotional ability should be dynamic and enhanced through experiences (Caprara, Giunta, Eisenberg, Gerbino, Pastorelli, & Tramontano, 2008; Kirk et al., 2008), whereas trait EI is typically conceptualised as a relatively stable disposition. We contend that ESE should develop (through training and education), as self-efficacy beliefs can be enhanced through mastery experiences as a result of individuals' capacities to reflect and learn from experience (Bandura, 1997).

Dimensionality of ESE

According to Bandura (2006), the efficacy belief system is not one global construct, but a differentiated set of self-beliefs linked to distinct realms of functioning. Individuals differ in the areas and levels in which their efficacy beliefs develop (Bandura, 2006). There is no all-purpose measure of perceived self-efficacy. The “one measure fits all” approach usually has limited predictive value because most of the items in an all-purpose test may have little or no relevance to the particular domain of functioning. Thus, measures of self-efficacy must be tailored to the specific domain of functioning that are the object of interest (Bandura, 2006).

In this tradition, we extend the domain of self-efficacy to six factors of ESE. The use of narrower sub-dimensions of a broader construct, such as ESE, is found in the literature regarding the bandwidth-fidelity debate (i.e. whether narrow or broad factors are preferable; see John, Hampson, & Goldberg, 1991, for review). Research has generally suggested that narrow personality traits are useful in the prediction of relevant criteria (e.g. Ashton, 1998; Hampson, John, & Goldberg, 1986; Judge, Erez, Bono, & Thoresen, 2003).

To establish a multidimensional measure of ESE, we first reviewed the Mayer and Salovey EI framework. The model is generally accepted as a four-dimensional model. The first dimension consists of an ability to perceive one's own emotions and to appraise emotional cues in others' verbal and nonverbal expressions. The second dimension relates to an ability to use emotions to facilitate thought processes such as problem solving and decision-making. The third dimension relates to an ability to understand how basic emotional experiences progress and combine into more complex emotions. Finally, the fourth dimension consists of one's ability to regulate one's own emotions and to manage others' emotions in social relations (Mayer & Salovey, 1997). Although this model seemingly contains four dimensions, a close reading of the description reveals that the first (perceiving emotions) and fourth (managing emotions) dimensions include two potentially independent reference points—self and other. Mayer and Salovey (1997) implied that an ability to perceive one's own emotions involves the same ability required to appraise others' emotions. This logic also applies to the ability to manage emotions in self versus others. However, when it comes to one's self-efficacy beliefs, one's ESE regarding perceiving emotions in self is likely to differ from ESE related to perceiving emotions in others. People should distinguish their capability of recognising their own feelings from their capability of appraising others' emotional state. Likewise, people distinguish their ability to control their own emotions from their efficacy beliefs in their ability to manage other people's emotions. Thus, we investigate whether the perceiving and managing dimensions are independent (perceiving emotions in self, perceiving emotions in others, managing emotions in self, and managing emotions in others) when assessing ESE.

The six-dimensional nature of ESE has been found (albeit implicitly) in prior research. For example, Wong and Law's (2002) Emotional Intelligence Survey (WLEIS) has 16 items, with four items for each of the four dimensions. These four dimensions were said to correspond with the four-factor model of EI (e.g. Law et al., 2004; Wong & Law, 2002). However, upon closer analysis, the four dimensions assessed by the measure, in fact, fail to tap the understanding emotions dimension, and instead, separate the perceiving emotions dimension into self and other sub-dimensions. Hence, the measure indeed consists of (1) perceiving emotions in self, (2) perceiving emotions in others, (3) use of emotions, and (4) managing emotions in self. A similar deficiency is found in another self-report measure. The Self-Report Emotional Intelligence Scale (SREIS; Brackett et al., 2006), which claims to correspond with the four-factor model, in fact includes items measuring managing emotions in self and others separately, and lacks items for perceiving emotion in self. Moreover, in assessing the psychometric soundness of this measure, Choi, Kluemper, and Sauley (2007) found that a five-factor model (i.e. separate managing emotions in self and others) fit significantly better than a four-factor solution.

Taken together, the theoretical and empirical evidence leads us to a test for a six-dimensional model of ESE with separate self- and other-referenced dimensions. We define emotional self-efficacy as a person's judgment on how well he/she is able to (1) perceive emotions in self, (2) perceive emotions in others, (3) use emotions to facilitate cognitive processes, (4) understand emotional complexity, (5) manage emotions in self, and (6) manage emotions in others.

Developing Hypotheses

Albert Bandura (2006) outlined procedures for the accurate measurement of self-efficacy constructs. He states that the construction of sound efficacy scales starts from a good theoretical and conceptual analysis of the relevant domain of functioning. We believe that our use of the Mayer-Salovey framework represents the domain of functioning for ESE. Self-efficacy relates to people's beliefs in their capabilities to produce given attainments, and therefore the efficacy measurement should be phrased in terms of “can do” rather than “will do” wording (Bandura, 1997).

As such, we used a priori theory to determine the six dimensions of ESE, followed by an assessment of existing self-reported EI measures to isolate theoretically relevant “can do” items that best reflect the six-dimensional model we propose. Thus, we propose that ESE extends beyond the four-dimensional model of ability-based EI, consisting of six distinct dimensions. In addition, as forms of self-efficacy, these six dimensions will be related to yet distinct from existing measures of self-efficacy. Therefore, Hypothesis 1 of the present study is about the six-dimensional nature of ESE, and Hypothesis 2 is about its discriminant validity from self-efficacy.

Hypothesis 1a: ESE consists of six distinct dimensions.

Hypothesis 1b: The six-dimensional model of ESE will fit significantly better than a four-dimensional model combining perceiving emotions in self/others and managing emotions in self/others.

Hypothesis 2: The six dimensions of ESE are distinct from self-efficacy.

To help establish the value of a new construct, researchers should demonstrate that the construct has enough ability to predict relevant outcomes beyond what is predicted by established constructs (Brackett & Mayer, 2003). Therefore, we hypothesise that ESE will predict nontrivial variance in outcomes after controlling for self-efficacy. In particular, we propose that some ESE dimensions (i.e. perceiving and managing emotions in self) will better predict outcomes than other dimensions. Though the remaining dimensions are discussed later when presenting future research, we focus more on these two dimensions to demonstrate the predictive power of ESE.

We propose that self-referenced ESE dimensions (perceiving and managing) are more theoretically related to individuals' psychological outcomes such as well-being and coping than the other dimensions of ESE. These relationships can be explained through transactional stress theory (Lazarus & Folkman, 1984, 1987). Transactional stress theory identifies primary and secondary appraisals as involved in resulting well-being and coping. Primary appraisals determine whether or not an individual will perceive an event as a challenge or a threat. The extent to which this event is seen as harmful depends on the person's cognitive characteristics that make them perceive vulnerability to that particular condition. Secondary appraisals involve evaluative judgments about whether any action can be taken to improve situations of perceived threat and depends upon the person's perceived control over the potential outcomes. For example, if there is a risk of harmful outcomes but the person perceives that these outcomes can be prevented, threat is likely to be minimal or absent. Beliefs about one's capability to manage these situations is an antecedent of secondary appraisal. Lazarus and Folkman (1987) conclude that people who are confident in their ability to adequately handle such interactions are less likely to see encounters as threatening and are more likely to use effective coping strategies. Further, the authors go on to argue for the investigation of more narrow forms of efficacy beliefs and that these beliefs about control likely transcend situations. Thus, we argue that a person's belief in perceiving and handling emotions in him/herself represents a specific form of efficacy that will transcend a wide range of situations, thereby predicting well-being and coping outcomes. If individuals have confidence that they can perceive and manage their own emotions effectively, this will result in positive appraisals in a variety of situations. Further, there is evidence that positive appraisals lead to benefits in adjusting to the event, and hence impact one's perception of well-being such as life satisfaction (Lengua & Long, 2002), coping efforts (Gamble, 1994), and anxiety or stress (Sandler, Kim-Bae, & MacKinnon, 2000).

Moreover, within the frameworks of both EI and self-efficacy, many empirical studies have demonstrated solid findings for these individual-level outcomes. For instance, self-efficacy in regulating negative emotions has been reported to decrease depression (Bandura, Caprara, Barbaranelli, Gerbino, & Pastorelli, 2003; Caprara, Steca, Cervone, & Artistico, 2003), while self-efficacy in expressing positive emotions has been associated with well-being (Caprara, Steca, Gerbino, Paciello, & Vecchio, 2006). Further, individuals with higher levels of EI are more likely to use rational coping and less likely to use avoidance coping strategies under challenging situations (Goldenberg, Matheson, & Mantler, 2006; Saklofske, Austin, Galloway, & Davidson, 2007; Schutte, Malouff, Thorsteinsson, Bhullar, & Rooke, 2007), are less stressed out (Ciarrochi, Dean, & Anderson, 2002; Nikolaou & Tsaousis, 2002; Salovey, Stroud, Woolery, & Epel, 2002), and are more satisfied with their lives (Ciarrochi, Chan, & Caputi, 2000; Law et al., 2004; Palmer, Donaldson, & Stough, 2002).

Taken together, we propose that the perceiving and managing emotions in self dimensions of ESE will predict low levels of stress, high levels of life satisfaction, low levels of avoidance coping, and high levels of rational coping. Further, we propose that these relationships hold even after controlling for self-efficacy. That is, the perceiving and managing emotions in self dimensions of ESE will explain significant nontrivial variance of the four outcomes when the influence of self-efficacy is accounted for.

Hypothesis 3: The perceiving emotions in self dimension of ESE will predict unique variance in stress (H3a), life satisfaction (H3b), avoidance coping (H3c), and rational coping (H3d) even when controlling for self-efficacy.

Hypothesis 4: The managing emotions in self dimension of ESE will predict unique variance in stress (H4a), life satisfaction (H4b), avoidance coping (H4c), and rational coping (H4d) even when controlling for self-efficacy.

We conducted two independent studies in two countries. Study 1, conducted in the US, consists of three measurement occasions, and Study 2, conducted in South Korea, consists of two measurement occasions. We established the six-dimensional ESE construct using exploratory and confirmatory factor analyses in the US study and cross-culturally verified the findings in the Korean study. Finally, we tested the predictive validity of ESE upon various outcomes in both studies.

STUDY 1

Participants and Procedures

Study 1 was conducted with a sample of undergraduate students enrolled in a large management course in a public university in the southern US. Three waves of surveys were administered, each two weeks apart. Of 900 students, 704 students participated in all three surveys (787 for time 1, 772 for time 2, and 812 for time 3). Of those respondents, 51 per cent were male, 87 per cent were Caucasian, and 67 per cent worked full or part-time. The mean age was 21 years old (ranging from 18 to 49). Confidentiality was assured, and respondents received extra credit by voluntarily participating in the on-line surveys. Three self-report EI measures (the WLEIS, SREIS, and EIS) were used in the first wave. In the second wave, emotional self-efficacy (ESE) and Bandura's self-efficacy were assessed. The outcome variables (stress, life satisfaction, avoidance, and rational coping) were measured in the final wave.

Measures

WLEIS.  The Wong and Law EI Survey (2002) is based on Mayer and Salovey's (1997) model and comprises 16 items (four items for each dimension): (1) perceiving emotions in self, (2) perceiving emotions in others, (3) use of emotions, and (4) managing emotions in self. A sample item reads, “I have good understanding of my own emotions.” A 7-point Likert scale which was originally recommended (1 =strongly disagree to 7 =strongly agree) was used (the result was multiplied by 5/7 for easier comparison), and Cronbach's alpha was .86.

SREIS.  Brackett et al.'s (2006) Self-Report EI Scale is based on Mayer and Salovey's (1997) model and is composed of 19 items covering five dimensions of EI: (1) perceiving emotions in others, (2) use of emotions, (3) understanding emotions, (4) managing emotions in self, and (5) managing emotions in others. A sample item is, “By looking at people's facial expressions, I recognise the emotions they are experiencing.” A 5-point Likert scale originally recommended was used (1 =very inaccurate to 5 =very accurate). Cronbach's alpha was .77.

EIS.  The Emotional Intelligence Survey developed by Schutte and colleagues (Schutte et al., 1998) is based on Salovey and Mayer's (1990) general EI model. The measure comprises 33 items, and no particular dimensional structure is assumed. A sample item includes, “I know when to speak about my personal problems to others.” Respondents were asked to report perceptions about their emotional capability using a 5-point Likert scale originally recommended (1 =strongly disagree to 5 =strongly agree). Cronbach's alpha was .88.

ESE-24.  Based on Mayer and Salovey's (1997) model, we built an item pool from the three widely used self-report EI scales. In this 68-item pool, we selected the items that fit the “can do” language associated with self-efficacy terminology. We then excluded items that do not tap the six dimensions of ESE. For example, some “will do” style items (e.g. I expect that I will do well on most things I try), motivation-related items (e.g. I am a self-motivating person), and personality-like items (e.g. I arrange events others enjoy) were filtered out in the content analysis. Exploratory and confirmatory factor analyses were conducted to identify the items most reflective of the six factors of the ESE construct (for details, see Table 1). We finalised the selection process by retaining 24 items representing the six dimensions of emotional self-efficacy (four items for each dimension). A 5-point Likert scale (1 =strongly disagree to 5 =strongly agree) was employed, and overall Cronbach's alpha was .87.

Table 1. Exploratory Factor Analysis with Six Dimensions
  1 2 3 4 5 6
  1. Note: Maximum likelihood estimation with promax rotation.

  2. Loadings below .20 were suppressed.

Perceiving emotions in self (Wong & Law, 2002)
1. I have a good sense of why I have certain feelings most of the time.83     
2. I have good understanding of my own emotions.86     
3. I really understand what I feel.68     
4. I always know whether or not I am happy.38     
Perceiving emotions in others (Wong & Law, 2002)     
1. I always know my friends' emotions from their behavior .61    
2. I am a good observer of others' emotions .73    
3. I am sensitive to the feelings and emotions of others .45    
4. I have good understanding of the emotions of people around me .71    
Use of emotions to facilitate thought (Schutte et al., 1998)
1. When my mood changes, I see new possibilities  .44   
2. When I am in a positive mood, solving problems is easy for me  .62   
3. When I am in a positive mood, I am able to come up with new ideas  .72   
4. When I feel a change in emotions, I tend to come up with new ideas  .41   
Understanding of emotional complexity (Brackett et al., 2006)
1. I have a rich vocabulary to describe my emotions   .75  
2. I could easily write a lot of synonyms for emotion words like happiness or sadness   .66  
3. I have the vocabulary to describe how most emotions progress from simple to complex feelings   .81  
4. My “feelings” vocabulary is probably better than most other persons' “feelings” vocabularies   .70  
Managing emotions in self (Wong & Law, 2002)
1. I am able to control my temper so that I can handle difficulties rationally    .73 
2. I am quite capable of controlling my own emotions    .85 
3. I can always calm down quickly when I am very angry    .72 
4. I have good control of my own emotions    .85 
Managing emotions in others (Brackett et al., 2006)
1. When someone I know is in a bad mood, I can help the person calm down and feel better quickly     .80
2. I know the strategies to make or improve other people's moods     .57
3. I am not very good at helping others to feel better when they are feeling down or angry (reversed)     .56
4. I am the type of person to whom others go when they need help with a difficult situation     .51

Self-Efficacy.  Self-efficacy was measured with ten items (Goldenberg et al., 2006), using a 5-point Likert scale (1 =strongly disagree to 5 =strongly agree). Participants responded to various statements regarding their belief in their abilities such as: “I can handle complex problems”; “I am quick to understand things.” The Cronbach's alpha was .79.

Stress.  Stress was assessed using a 14-item measure of Cohen, Kamarck, and Mermelstein (1983). A sample item is, “In the last month, how often have you been upset because of something that happened unexpectedly?” Respondents reported the frequency of stressful experiences on a 5-point Likert scale (1 =never to 5 =very often). Cronbach's alpha was .81.

Life Satisfaction.  Life satisfaction was assessed using a five-item measure of Diener, Emmons, Larsen, and Griffin (1985). Respondents indicated the extent to which they agree with statements such as “In most ways my life is close to my ideal”, using a 5-point Likert response format (1 =strongly disagree to 5 =strongly agree). Cronbach's alpha was .84.

Avoidance Coping.  Avoidance coping was assessed with four items from Carver's (1997) Brief Cope (i.e. substance and self-blame). A sample item reads, “I've been using alcohol or other drugs to make myself feel better.” Respondents reported the frequency of using such coping behaviors on a 4-point Likert scale (1 =I haven't been doing this at all to 4 =I've been doing this a lot). Cronbach's alpha was .78.

Rational Coping.  Rational coping was also assessed with four items from Carver's (1997) Brief Cope (i.e. planning and active coping). A sample item reads, “I've been concentrating my efforts on doing something about the situation I'm in.” Respondents reported the frequency of using such coping strategies on a 4-point Likert scale (1 =I haven't been doing this at all to 4 =I've been doing this a lot). Cronbach's alpha was .84.

Results

As stated earlier, we scanned the literature to draw upon three well-established measures of self-reported emotional abilities (the WLEIS, SREIS, and EIS) that collectively tap the six proposed dimensions of ESE. Using the first wave data, we began by conducting a content analysis of the 68 items to identify those items in line with Bandura's (2006) guide for constructing self-efficacy scales. This culminated in various candidate sets of ESE. With these candidate sets, we assess the best fitting six-dimensional model of ESE, using exploratory factor analysis (EFA) with maximum likelihood estimation and promax ortho-oblique rotation (a varimax solution shows a very similar result). The final set includes 24 items, four items for each dimension.

Table 1 shows detailed factor loadings of the final six-dimentional ESE scale extracted from the first wave data. All items are properly loaded on their proposed dimensions, and no significant cross-loadings were observed. It should be noted that there exist some remaining items (not selected in the 24-item version) that are interchangeable with the chosen ones. In such cases, we favored the items from the same measure, rather than mixing items from different measures. The six-dimensional nature of ESE is further supported by comparing six-factor and four-factor solutions. When constrained to four factors (contact the first author for details), perceiving emotion in self loaded on the same factor as managing emotion in self, rather than with perceiving emotion in others. Similarly, perceiving emotion in others loaded with managing emotion in others, rather than with perceiving emotion in self when a four-factor constraint was applied. This indicates that merging self and others dimensions may not be appropriate, and they should be treated as independent dimensions, lending support to Hypotheses 1a and 1b.

Using the second wave data, a confirmatory factor analysis (CFA) was conducted to test the validity of the six-dimensional ESE model. With LISREL 8.5, we assessed whether the proposed six-dimensional model is significantly better fitting than the four-dimensional solution (merging self and others dimensions). The CFA result revealed that the six-dimensional model fit well (X2= 642, df= 237, CFI = .97, RMSEA = .05), which is a significantly better fit than the four-dimensional model with self and others dimensions merged (ΔX2= 2154, Δdf= 9, p < .001, CFI = .87, RMSEA = .12), rendering additional support for Hypotheses 1a and 1b.

Table 2 summarises the basic descriptive statistics, reliability coefficients, and correlations of all variables across the three time periods. The three-time test–retest reliability of ESE-24 was .79. The dimensional level test–retest reliability coefficients were .78 (perceiving emotions in self), .80 (perceiving emotions in others), .64 (use of emotion), .87 (understanding of emotions), .85 (managing emotions in self), and .82 (managing emotions in others). Briefly, the six dimensions of ESE were closely correlated with the three EI measures (average correlation of .39; minimum of .11; maximum of .56), showing good convergent validity. Correlations between the six dimensions of ESE and self-efficacy ranged from .21 to .32, rendering support for discriminant validity of ESE from self-efficacy (Hypothesis 2).

Table 2. Descriptive Statistics and Correlations in the US Study
 MSD 1 2 3 4 5 6 7 8 9 10 11 12 13 14
  1. Note: N= 704; Cronbach's alphas appear on the diagonal in italics.

  2. Correlations of .08 and above are statistically significant at p < .05.

  3. Correlations of .11 and above are statistically significant at p < .01.

  4. Wong and Law EI Scale (WLEIS), Self-report EI Scale (SREIS), and EI Survey (EIS) were measured at Time 1; Stress, life satisfaction, avoidance coping, and rational coping were measured at Time 3; all other variables were measured at Time 2.

 1. Perceiving emotions (self)3.90.59 .81              
 2. Perceiving emotions (other)3.73.60.41 .81             
 3. Use of emotions3.74.49.19.16 .67            
 4. Understanding emotions3.37.72.23.26.19 .86           
 5. Managing emotions (self)3.57.75.48.26.12.11 .85          
 6. Managing emotions (other)3.62.57.35.48.20.30.27 .68         
 7. Self-efficacy3.76.48.32.21.30.31.21.32 .79        
 8. Stress2.84.49−.36−.05−.05−.06−.34−.14−.28 .81       
 9. Life Satisfaction3.47.74.30.04.12.04.25.15.25−.55 .84      
10. Avoidance Coping1.88.73−.27−.07−.03−.04−.24−.11−.17.48−.37 .78     
11. Rational Coping3.12.66.19.19.12.15.13.20.25−.09.18−.07 .84    
12. WLEIS3.79.46.51.38.11.16.56.38.33−.37.33−.28.27 .86   
13. SREIS3.48.36.38.49.21.56.29.55.35−.12.13−.08.27.48 .77  
14. EIS3.71.34.48.50.34.32.33.54.40−.22.25−.15.32.62.68 .88

To evaluate the exact amount of unique contribution the six-dimensional ESE makes to the outcome variables in the presence of self-efficacy (Hypotheses 3 and 4), we conducted hierarchical linear regressions and relative weight analyses. The former is a conventional tool to test incremental validity, and the latter is a recent technique developed to estimate the relative contribution each predictor uniquely makes to the total explained variance of a criterion variable. This technique has benefit over the traditional incremental variance approach because it addresses the variance attribution problem often occurring in hierarchical regression with collinear independent variables—the criterion variance explained by both old and new variables in hierarchical regressions tends to be blindly credited to the old variable, and therefore assessing incremental variance alone could lead to making misinformed decisions about the true importance of the new variable (LeBreton, Hargis, Griepentrog, Oswald, & Ployhart, 2007). The technique allows researchers to partition total criterion variance into its corresponding predictors simultaneously although the predictors are correlated with each other (Johnson, 2000; Johnson & LeBreton, 2004). Statistically, relative weight (i.e. the contribution each predictor makes to the total predicted criterion variance) can be calculated by (1) creating a new set of mutually orthogonal predictors through a singular value decomposition, (2) regressing the criterion variable on the new orthogonal predictors, and (3) scaling the contribution of the orthogonal predictors back to their original collinear counterparts (LeBreton & Tonidandel, 2008). Note that, in this study, the predictors (i.e. six dimensions of ESE and self-efficacy) and outcome variables were from different time periods. This temporal separation is to alleviate some concerns of common method variance (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003).

Table 3 summarises the results of hierarchical regressions and relative weight analyses. We expected that perceiving emotions in self dimension would significantly predict stress (H3a), life satisfaction (H3b), avoidance coping (H3c), and rational coping (H3d) in the presence of self-efficacy. All regression coefficients of perceiving emotion in self dimension were significant, except rational coping, which supports Hypotheses 3a, 3b, and 3c. Similarly, we hypothesised that managing emotions in self predicts the four outcomes in the presence of self-efficacy. The results show that the managing emotions in self dimension significantly predicted stress (H4a) and life satisfaction (H4b), but not avoidance coping (H4c) and rational coping (H4d). In summation, five out of eight hypotheses regarding the incremental criterion-related validity of ESE received strong support. The relative weight analyses further reveal that perceiving and managing one's own emotions are the strongest predictors of three of the four outcome measures. These two self-referenced factors jointly contributed 69 per cent of explained variance for stress, 56 per cent for life satisfaction, and 74 per cent for avoidance coping. For rational coping, it turned out that other-referenced dimensions (perceiving and managing) were more important (38 per cent of explained variance).

Table 3. Hierarchical Regression Results and Relative Weights of the US Study
  Stress Life Satisfaction Avoid Coping Rational Coping
Step 1
 R2.08**.06**.03**.06**
Step 2
 ΔR2.14**.06**.08**.06**
  β RW β RW β RW β RW
  1. Note: Step 1 includes self-efficacy; RW = Relative weight.

  2. ** p < .01; * p < .05.

 Self-efficacy−.23**27.32**38−.20**19.23**32
 Perceiving Emotion (self)−.15**39.17**41−.21**60.0610
 Perceiving Emotion (other).041.012−.023.07*14
 Use of Emotion.08*1−.031.051.099
 Understanding of Emotion.041−.061.062.0610
 Managing Emotion (self)−.09**30.05*15−.0414−.021
 Managing Emotion (other).031−.012.051.13**24

Discussion

The results of Study 1 provide evidence of the six-dimensional nature of emotional self-efficacy. Both EFA and CFA results support a six-factor solution as fitting significantly better than the other models. The ESE construct significantly and incrementally predicts important individual outcomes such as stress, life satisfaction, and different coping strategies over self-efficacy. The explained variances of the four outcome measures were significant when the six ESE dimensions were entered (stress: ΔR2= .14, life satisfaction: ΔR2= .06, avoidance coping: ΔR2= .08, and rational coping: ΔR2= .06). According to the relative weight analyses, the six dimensions of ESE collectively accounted for a considerable proportion of explained variance in the presence of self-efficacy: stress (73%), life satisfaction (62%), avoidance coping (81%), and rational coping (68%). These findings suggest that the criterion-related validity of ESE holds over self-efficacy, affirming the overall utility of ESE.

STUDY 2

Participants and Procedures

We validated our findings from Study 1 with an independent sample. Approximately 700 undergraduate students enrolled in multiple management courses in a South Korean university were asked to participate in the study. Two waves of the survey were administered, two weeks apart. A total of 321 students finished both on-line surveys (404 for the first wave and 329 for the second wave). Of those participants, 72 per cent were male, and the mean age was 22 years old (ranging from 17 to 39). Confidentiality was assured, and participants were entered into a $5 gift card draw as enticement. Similar to the US study, predictor variables were assessed in the first wave (i.e. the ESE-24, the three self-report EI measures, and self-efficacy), and the outcome variables were measured in the second wave.

Measures

All measures were translated into the Korean language by one of the authors, and then back-translated into English by a Korean doctoral student who was unfamiliar with the study. The back-translated measures were compared with the original English version. Discrepancies between the two versions were discussed among the authors, and this procedure was repeated until consensus was reached. The resulting Cronbach's alphas of the three EI measures were similar to the ones obtained in the US sample: .86 (the WLEIS), .76 (the SREIS), and .88 (the EIS). The same 24 ESE items used in the US study were employed. Cronbach's alpha of overall ESE was .86. Self-efficacy also showed similar reliability results to the US sample's (.79). Translated versions of stress, life satisfaction, avoidance coping, and rational coping were employed with their original response schemes. Cronbach's alphas were .87 (stress), .86 (life satisfaction), .74 (avoidance coping), and .85 (rational coping).

Results

In terms of the six-dimensional nature of ESE, we obtained very similar results to those found in the US study. A six-factor model fit well (X2= 519, df= 237, CFI = .95, RMSEA = .06), which was significantly better than a four-factor solution (ΔX2= 1304, Δdf= 9, p < .001, CFI = .83, RMSEA = .13). These results provide additional support for Hypotheses 1a and 1b.

Table 4 summarises the basic descriptive statistics, alphas, and correlations of the Korean study across the two time periods. Test–retest reliability of the ESE-24 in the Korean sample was .75. At the dimensional level, the test–retest reliability coefficients were .76 (perceiving emotions in self), .79 (perceiving emotions in others), .69 (use of emotions), .83 (understanding of emotions), .84 (managing emotions in self), and .83 (managing emotions in others). The average correlation between the six dimensions of ESE and the three EI measures was .50 (ranging from .08 to .78).

Table 4. Descriptive Statistics and Correlations in the Korean Study
 MSD 1 2 3 4 5 6 7 8 9 10 11 12 13 14
  1. Note: N= 321; Cronbach's alphas appear in the diagonal in italics.

  2. Correlations of .11 and above are statistically significant at p < .05.

  3. Correlations of .15 and above are statistically significant at p < .01.

  4. Stress, life satisfaction, avoidance coping, and rational coping were measured at Time 2; all other variables were measured at Time 1.

  5. Chronbach's alphas appear in the diagonal in italics.

 1. Perceiving emotions (self)3.21.57 .82              
 2. Perceiving emotions (other)3.12.58.44 .82             
 3. Use of emotions3.73.71.05.15 .83            
 4. Understanding emotions3.23.70.27.28.23 .75           
 5. Managing emotions (self)2.80.75.29.19−.09.31 .90          
 6. Managing emotions (other)3.37.69.26.36.19.50.23 .74         
 7. Self-efficacy3.48.54.32.23.12.39.34.35 .79        
 8. Stress3.06.89−.23−.01.11−.04−.21−.08−.28 .87       
 9. Life Satisfaction3.13.80.27.17.14.21.23.21.41−.35 .86      
10. Avoidance Coping2.37.76−.14−.04.05.11−.10.04−.15.40−.30 .74     
11. Rational Coping3.14.55.31.19.08.12.22.10.40−.17.30−.07 .85    
12. WLEIS3.63.52.72.64.08.39.71.42.49−.25.40−.16.40 .86   
13. SREIS3.30.40.38.50.25.76.44.78.46−.14.32.00.16.60 .76  
14. EIS3.61.40.44.56.55.47.33.58.41−.11.33−.05.28.64.72 .88

The discriminant validity of ESE from self-efficacy was confirmed with the Korean sample. Hypothesis 2 stated that ESE and self-efficacy should be modestly correlated. The correlations between ESE sub-dimensions and self-efficacy were similar to those of US study's (ranging from .12 to .39). This provides continuing support for Hypothesis 2.

We conducted hierarchical regressions and relative weight analyses to evaluate the unique contribution of the six dimensions of ESE to outcome variables in the presence of self-efficacy. The results appear in Table 5. We predicted that perceiving emotions in self would explain significant variance for stress (H3a), life satisfaction (H3b), avoidance coping (H3c), and rational coping (H3d) when self-efficacy was partialled out. The regression coefficients of perceiving emotions in self for all four outcomes were significant, giving full support to Hypothesis 3. On the other hand, no part of Hypothesis 4 received support. The managing emotions in self dimension did not predict significant variance when self-efficacy was present, rejecting Hypothesis 4 in the Korean study. Relative weight analyses further revealed that perceiving emotions in self explained 26 per cent for stress, 17 per cent for life satisfaction, 21 per cent for avoidance coping, and 25 per cent for rational coping, strengthening the support for Hypothesis 3.

Table 5. Hierarchical Regression Results and Relative Weights of the Korean Study
  Stress Life Satisfaction Avoid Coping Rational Coping
Step 1
 R2.08**.17**.02**.16**
Step 2
 ΔR2.07**.04*.06**.06**
  β RW β RW β RW β RW
  1. Note: Step 1 includes self-efficacy; RW = Relative weight.

  2. ** p < .01; * p < .05.

 Self-efficacy−.42**42.49**54−.27**33.36**54
 Perceiving Emotion (self)−.30**26.19*17−.17*21.18**25
 Perceiving Emotion (other).18*4.003−.012.056
 Use of Emotion.14*9.116.022.052
 Understanding of Emotion.123−.025.22**27−.072
 Managing Emotion (self)−.1215.0910−.079.0610
 Managing Emotion (other)−.043.035.055−.062

Discussion

Overall, the findings of the Korean study are similar to those of the US study. The six dimensions of ESE are observed, and the six-factor model is substantially better than the other models. After controlling for self-efficacy, the six dimensions of ESE significantly and incrementally predicted stress (ΔR2= .07), life satisfaction (ΔR2= .04), avoidance coping (ΔR2= .06), and rational coping (ΔR2= .06). The six dimensions of ESE in combination contributed 58 per cent of the explained variance in stress, 46 per cent in life satisfaction, 67 per cent in avoidance coping, and 46 per cent in rational coping. Similar to the US study, the perceiving one's own emotions dimension (but not others dimension) was determining in explaining outcome variables. One noteworthy variation from the US study was that the understanding dimension of ESE contributed a large amount of variance (27%) in avoidance coping. The predictive power of managing one's own emotions dimension also showed a substantial difference from the result of the US study. In addition, a few regression coefficients were in a somewhat counter-intuitive direction. For instance, the use of the emotions dimension significantly increases stress (.11). We have little theory to explain this divergence, but we deem that there might be a cultural difference in processing emotions between Western and Eastern cultures. Perhaps, Eastern culture puts more emphasis on regulating one's own emotions, and Western culture places more emphasis on emotional self-recognition. It would make interesting future research to find how people in various cultures give different weight to diverse ESE dimensions.

GENERAL DISCUSSION AND FUTURE DIRECTIONS

Bandura and his colleagues convincingly show that an individual's social behaviors are closely related to self-evaluative efficacy expectation, rather than controlled by immediate stimuli (Bandura et al., 1977). A person does not initiate and persist in certain behaviors if they have doubts whether they can successfully perform them. An activity believed to exceed one's capability is avoided even though one might in fact have the ability to perform it (Bandura, 1977). In this vein, a person's behaviors are jointly determined by his/her capability and evaluative expectation regarding such capabilities (Bandura, 1982). As Roberts, Zeidner, and Matthews (2001) noted, a person's belief in his/her emotional ability may influence outcomes, apart from whether the belief is accurate or not. Our conceptualisation of emotional self-efficacy reflects such a viewpoint; that an individual's belief in his/her emotional capability predicts important individual outcomes such as stress, life satisfaction, and coping behaviors.

This is one of the first attempts to clarify and validate the concept of ESE. We conducted a thorough evaluation of the dimensionality, reliability, and validity of ESE based on ability EI theory. The widely established four-branch model of EI, however, appears not to fit well in relation to ESE. Both exploratory and confirmatory results from multiple sources across different cultures indicate that a six-factor model fits significantly better than a four-factor model. In particular, the perceiving emotions dimension forms two distinct dimensions: perceiving emotions in self and perceiving emotions in others. Similarly, the managing emotions dimension forms two separate dimensions: managing emotions in self and managing emotions in others.

Our framework for emotional self-efficacy differs from an existing emotional self-efficacy measure (see Kirk et al., 2008) in several ways. Theoretically, the focus of the current study was to identify the multidimensional nature of ESE, and to make more fine grained predictions based on these dimensions is not possible with the existing uni-dimensional measure. Empirically, the current study utilises a multiple measurement design using cross-cultural samples, providing a more robust measurement validation than was conducted for the Kirk et al. scale. Practically, the length of the measure of the present study is substantially shorter than the existing measure, which increases the utility of our ESE-24 measure.

We want to emphasise that the ESE perspective theoretically differs from trait EI. The latter is defined as a general, relatively stable, long-term predisposition (Petrides et al., 2007). The efficacy perspective, on the other hand, defines the construct as a malleable construct that can be acquired and enhanced through learning (Bandura, 1982). A person's self-concept is increased or decreased by a response to emotional events, or facilitated by mastery experiences (Markus & Kunda, 1986). We propose that individuals consistently calibrate and modify their ESE in response to the surrounding affective events. Therefore, leadership training can increase a boss's self-efficacy in managing subordinates' emotions, and a recently blown-up deal can hamper a realtor's self-efficacy in reading customers' emotions. The possibility of recalibration makes ESE differ from the dispositional perspective of trait EI. An implication of this theoretical distinction is that ESE should be able to be developed through training and education, as self-efficacy beliefs can be enhanced (Bandura, 1997). Our test–retest reliability results show differences in the malleability of ESE traits, with the using emotions to facilitate thought dimension of ESE demonstrating more malleability (.64 in the US and .69 in Korea) than other ESE dimensions (ranging from .76 to .87) across cultures. Future research should investigate the effectiveness of training interventions on the six dimensions of ESE.

We also want to emphasise the cross-cultural nature of the current research. Study 1 was conducted in the United States, a Western culture where individualism and diversity are appreciated. Study 2 was conducted in South Korea, an Eastern culture where people are more collectivist and homogeneous (Kitayama, Mesquita, & Karasawa, 2006). It has been suggested that self-efficacy may differ across cultures (Bandura, 2002), as conceptions of emotional control may differ across cultures (Matsumoto, 2006). In these very different cultures, the multidimensional structure of ESE was shown to be invariant. Results also indicate that the relationships between the six ESE dimensions and criterion constructs were largely consistent across cultures. Due to its exploratory nature, however, this study did not explicitly hypothesise any particular patterns of cultural similarities (or differences). Future researchers may want to test more specific hypotheses about measurement equivalence across cultures using various forms of invariance testing (Vandenberg & Lance, 2000).

One major purpose of establishing six distinct ESE dimensions is to assess their independent nature. If these dimensions are found to differ from one another in their relationships with other constructs of interest, then there may be merit in the evaluation of an independent dimension in future studies. This approach has begun in ability-based EI research (see Kluemper, DeGroot, & Choi, 2012) and should be applied in the context of ESE. The findings suggest that the ESE dimensions differ in their relationships with outcome variables. Perceiving and managing emotions in self are shown to be stronger predictors of stress, life satisfaction, and coping than the other dimensions. Based on these results, future studies using ESE would be well served by developing theoretical rationales for specific dimensions. For example, Dudley, Orvis, Lebiecki, and Cortina (2006) found that facets of conscientiousness explained variance in job performance beyond the broad conscientiousness trait. In their meta-analysis, O'Boyle and others (2011) found that self-reported EI was a stronger predictor of job performance than was ability-based EI. Our findings indicate that it is likely that some dimensions of ESE might be more predictive than others. Potential outcomes for specific dimensions of ESE are numerous. For example, we suggest that one's efficacy about “use of emotions to facilitate thought” may be particularly relevant with creativity and decision-making. “Understanding emotional complexity” may emerge in emotionally ambiguous, ambivalent, and complex situations. “Perceiving and managing emotions in others” might relate when the individual interacts with others, such as with interpersonal conflict and performance related to customers.

We acknowledge that our six-dimensional model is somewhat deviant from the “sacred four-branch model” of Mayer and Salovey (1997). Many of those who currently accept the four-branch model could be doubtful about the independent nature of the self and others aspects of perceiving and managing emotions. We argue that one's belief in ability can be qualitatively different from the ability itself, and the dimensional structures of ESE versus EI could be so. By the same token, we have been careful to differentiate the terms “emotional self-efficacy” and “emotional intelligence” throughout the manuscript. The latter is an individual's emotional ability, and the former is a person's belief in his/her emotional ability.

This line of reasoning is also applicable to peer ratings of emotional ability. How others perceive someone's emotional capability may alter how they interact with the individual, regardless of the individual's actual emotional competence. It could be an interesting future approach to test whether the six-factor model extends to peer ratings of emotional ability. Perhaps, when individuals assess the level of emotional capability of other people, it might be difficult for raters to tell how well a ratee perceives and manages his/her own internal emotions. This opens the possibility that measurement methods change the structural nature of the ESE construct. In summary, we deem that a self-efficacy perspective of emotional capability brings a renewed and refreshing interest to extant research.

LIMITATIONS AND CONCLUSIONS

We want to mention a few limitations of the current study. First, the samples we collected in the two countries were mainly composed of college students. Future research may include a wider variety of age groups, occupations, and sub-cultures. Second, although we conducted CFAs across US and Korean samples, it is a limitation of the data to analyze exploratory and confirmatory factor analyses using the same study participants in the US study. Third, no relationship reported in the present study warrants inferences of causality. Although we gathered data at multiple time points, all relationships among variables are correlational by their nature. Fourth, we are aware that there are a number of EI studies not endorsed by Mayer and Salovey's model (e.g. Goleman, 1995). The present study does not encompass the findings and implications of that literature. Fifth, we did not test the incremental validity of ESE over ability EI measures (e.g. MSCEIT), therefore it is premature to estimate the incremental effect size of ESE beyond ability EI. Finally, the outcomes we selected were mainly self-reported variables. Future research might employ a larger selection of outcome variables such as objective performance, peer-rated behaviors, and relational outcomes.

To conclude, the present study demonstrates that (1) emotional self-efficacy (i.e. a person's belief in his/her emotional ability) consists of six dimensions (i.e. perceiving emotions in self, perceiving emotions in others, use of emotions to facilitate thought, understanding emotional complexity, managing emotions in self, and managing emotions in others), (2) these dimensions differ in relation to other constructs, and (3) the construct predicts important individual outcomes such as low levels of stress, high levels of life satisfaction, and constructive coping. We believe that the present study makes good conceptual and methodological progress in understanding emotional self-efficacy and its impact on human behavior.

Ancillary