Reciprocal influences between self-assessed emotional intelligence and self-esteem

Abstract

Self-assessed emotional intelligence has appeared to maintain a correlation with self-esteem. Such a correlation is likely to signal that the two traits reflect a latent, superordinate trait of aptitude and/or maintain reciprocal effects on each other. Estimating the reciprocal effects given the superordinate trait is the task of this study. In this study, structural equation modelling is used to analyse survey data gathered from 405 undergraduates. Results verify that self-assessed emotional intelligence could form a second-order factor, which could identify a third-order factor of aptitude, together with self-esteem and grade point average. Given such a hierarchical factor model, self-esteem indicated a significant positive effect on self-assessed emotional intelligence. However, the positive effect of self-assessed emotional intelligence on self-esteem was not significant. The results favour the view about the top-down influence of self-esteem on self-assessed emotional intelligence.

Keywords::

• emotional intelligence
• self-esteem
• aptitude

Self-assessed emotional intelligence has appeared to be associated with self-esteem (Schutte, Malouff, Simunek, McKenley, & Hollander, 2002). Such association is likely to comprise reciprocal influences between self-assessed emotional intelligence and self-esteem. By identifying self-assessed emotional intelligence as a second-order factor composed of four first-order factors and a third-order factor of aptitude, the study attempts to disentangle the reciprocal influences. Such disentanglement vitally elucidates the nature of self-assessed emotional intelligence, particularly in relation to self-esteem. The elucidation is required in view of many unresolved issues about self-assessed emotional intelligence, notably involving its relationship with self-esteem (Zeidner, Matthews, & Roberts, 2009). As such, the aim of this study was to estimate the correlation between the trait factors of self-assessed emotional intelligence and self-esteem and to decompose the correlation into reciprocal influences and loadings on the superordinate factor of aptitude. The estimation essentially hinges on a causal or structural equation model for unraveling contemporaneously reciprocal effects (Greenland & Brumback, 2002; Holland, 1988). In fitting the model to data optimally, the modelling obtains effects proposed in the model.

Self-assessed emotional intelligence is an operationalisation of emotional intelligence that relies on the individual's own assessment, judgement or rating of indicators of emotional intelligence. Emotional intelligence means reasoning about one's and others' emotions (Mayer, Salovey, & Caruso, 2008). Such reasoning commonly operates in four ways: appraisal of one's emotion, understanding of others' emotion, use of emotion and regulation of emotion. Whereas self-assessment of emotional intelligence is one way, objective assessment of the intelligence is another, as scored according to the consensus, expert, or target of emotionally intelligent performance (Zeidner et al., 2009). The self-assessment upholds a trait or dispositional approach, which is distinct from the ability or performance approach championed by the objective assessment. That is, self-assessed emotional intelligence is not equal to objectively assessed emotional intelligence (Austin, Saklofsko, & Egan, 2005). Nevertheless, self-assessed emotional intelligence has its own merit in gauging a personality trait (Gignac, Palmer, Manocha, & Stough, 2005; Tett & Wang, 2005). Self-assessed emotional intelligence is also justified by the view that emotion and its assessment are private (Joseph & Newman, 2010). Meanwhile, self-assessed emotional intelligence is susceptible to self-aggrandisement and self-rating bias, such as acquiescence, which means rating oneself indiscriminately highly (Mayer et al., 2008). The former susceptibility suggests the entanglement between self-assessed emotional intelligence and self-esteem, both of which are self-concepts (Zeidner et al., 2009). Disentangling the two self-concepts and clarifying their relationship are therefore requisite for the understanding of self-assessed emotional intelligence.

Both self-assessed emotional intelligence and self-esteem are crucial because they tend to predict favourable performance, similar to the study by Viswesvaran (2004). The prediction is likely to reflect the good quality, such as aptitude, which underlies emotional intelligence, self-esteem, academic achievement and other performance and traits. Accordingly, many forms of performance, intelligence and even motivation, including emotional ones, maintain interrelationships strong enough to reflect the latent predisposition of aptitude (Ackerman & Wolman, 2007). That is, one who is high in aptitude is likely to hold high self-assessed emotional intelligence, self-esteem and academic achievement. This is plausible, given the premise that all emotional intelligence, self-esteem and academic achievement share a common root in aptitude (Mayer et al., 2008). Hence, one explanation for the relation between self-assessed emotional intelligence and self-esteem is their common root in the latent predisposition of aptitude.

Self-assessed emotional intelligence is also likely to be predictable by self-esteem, regardless of their common root in aptitude. The first possibility is the simple top-down influence from general self-concept to specific self-concept (Lyubomirsky, 2001). In this connection, self-esteem represents the general self-concept and self-assessed emotional intelligence registers the specific self-concept, such as emotional self-awareness (Zeidner et al., 2009). The second possibility is the logic of broaden-and-build theory, which posits that positive emotion fosters emotional intelligence (Tugade, Fredrickson, & Barrett, 2004). In this case, self-esteem is representative of positive emotion (Judge, Erez, Bono, & Thoresen, 2002). With self-esteem or positive emotion, one is able to broaden and build on one's strengths to behave intelligently, thus maintaining emotional intelligence. The third possibility is that proposed by cognitive-adaptive theory, which states that one selects environments and adapts to them according to one's personality and the adaptation would constitute one's emotional intelligence (Zeidner et al., 2009). Herewith, self-esteem may function as a personality trait that disposes one to socially facilitating environments, just as extraversion does (Tracy, Cheng, Robins, & Trzesniewski, 2009). Such environments would foster one's learning of emotional intelligence through cognitive-adaptive processing. The learning is possible because of the mutability or trainability of emotional intelligence (Tett & Wang, 2005). In all, the contribution of self-esteem to self-assessed emotional intelligence is analogous to the influence of one's identity on the intelligence (Seaton & Beaumont, 2008).

Self-assessed emotional intelligence may also affect self-esteem, in addition to their common root in aptitude. The simplest explanation is the bottom-up logic, which suggests the building of general self-concept from specific self-concepts (Schimmack, 2008). As such, self-assessed emotional intelligence would serve as a specific self-concept to compose the general self-concept of self-esteem. Another simple explanation is that of self-perception theory, which holds that self-esteem stems from perception about one's performance and ability (Redersdorff & Guimond, 2006). In this case, self-assessed emotional intelligence can be the self-perceived ability (Zeidner et al., 2009). A third explanation envisions mediation through the indirect paths of performance and social relationship, which are conducive to self-esteem (Locke, 2003). Such paths are in turn contingent on emotional intelligence, which would contribute to performance and social relationship (Mayer et al., 2008). More specifically, self-assessed emotional intelligence is likely to function to enhance the self (Zeidner et al., 2009). The contribution of self-enhancement to self-esteem is obvious (Spencer-Rodgers, Boucher, Mori, Wang, & Peng, 2009).

As a recapitulation, key hypotheses concerning relationships between the self-assessed emotional intelligence and self-esteem of a person are twofold:

• Self-esteem contributes to self-assessed emotional intelligence.

• Self-assessed emotional intelligence contributes to self-esteem.

Methods

A sample of 405 undergraduates of a university in Hong Kong, China provided data for the study. Each of the undergraduates completed a self-administered questionnaire in the classroom on a voluntary basis, as solicited by one of the researchers or their research assistant. The undergraduates were a mix of male (55.1%) and female (44.9%) students and students of different majors and years of study (see Table 1). There were more social science students (42.7%) than business, science or law students. The students had an average age of 20.0 years. Owing to the three-year undergraduate programme in Hong Kong, the student's year of study ranged from Year 1 (30.1%) to Year 3 (20.0%).

Table 1 Univariate statistics.

Variable	Scoring	M	SD	Skew	Kurtosis
Age	Years	20.0	0.9	0.038	− 0.134
Year of study	1–3	1.9	0.7	0.143	− 0.961
Latest GPA	0–4	3.0	0.5	− 0.109	− 0.480
Self-emotion appraisal	0–100	54.4	25.5	− 0.105	− 0.863
Other-emotion appraisal	0–100	54.5	24.5	− 0.227	− 0.644
Emotion use	0–100	54.0	22.8	− 0.096	− 0.560
Emotion regulation	0–100	55.2	24.0	− 0.172	− 0.709
Self-esteem	0–100	52.8	18.0	− 0.220	− 0.886
Acquiescence	0–100	52.2	7.7	− 0.083	− 0.005

Measurement

The questionnaire listed items for measuring self-assessed emotional intelligence, self-esteem and others in English. The former items accepted responses in four steps, with the lowest step (‘totally disagree’) generating a score of 0, the second step (‘disagree’) a score of 33.33, the third step (‘agree’) a score of 66.67 and the highest step (‘totally agree’) a score of 100. Such scoring just magnified the scale, without distorting the meaning of the rating.

Self-assessed emotional intelligence involved 16 items measuring four components: appraisal of self-emotion appraisal or understanding of others' emotion, use of emotion, and regulation of emotion (Law, Wong, & Song, 2004). Each of the components consisted of four of the items. A sample item was ‘I really understand what I feel’ for measuring self-emotion appraisal (see Table 2). Identification of emotional intelligence evolved from second-order confirmatory factor analysis incorporated in structural equation modelling.

Table 2 Standardised factor loadings.

Variable	Trait	Method
Self-emotion appraisal
I have a good sense of why I have certain feelings most of the time	0.669	0.311
I have good understanding of my own emotions	0.715	0.274
I really understand what I feel	0.758	0.319
I always know whether or not I am happy	0.731	0.314
Other-emotion appraisal
I always know my friends' emotions from their behaviour	0.773	0.238
I am a good observer of others' emotions	0.729	0.264
I am sensitive to the feelings and emotions of others	0.727	0.144
I have good understanding of the emotions of people around me	0.569	0.105
Emotion use
I always set goals for myself and then try my best to achieve them	0.622	0.189
I always tell myself I am a competent person	0.718	0.232
I am a self-motivating person	0.637	0.285
I would always encourage myself to try my best	0.587	0.214
Emotion regulation
I am able to control my temper so that I can handle difficulties rationally	0.633	0.250
I am quite capable of controlling my own emotions	0.615	0.360
I can always calm down quickly when I am very angry	0.638	0.347
I have good control of my own emotions	0.669	0.301
Self-esteem
I feel that I am a person of worth, at least on an equal plane with others	0.491	0.157
I feel that I have a number of good qualities	0.502	0.153
I am able to do things as well as most other people	0.602	0.185
I take a positive attitude of myself	0.540	0.171
On the whole, I am satisfied with myself	0.453	0.243
All in all, I am inclined to feel that I am a failure	− 0.325	0.342
I feel I do not have much to be proud of	− 0.463	0.321
I wish I could have more respect for myself	− 0.537	0.259
I certainly feel useless at times	− 0.554	0.223
At times I think I am no good at all	− 0.559	0.192
Emotional intelligence
Self-emotion appraisal	0.837	–
Other-emotion appraisal	0.602	–
Emotion use	0.751	–
Emotion regulation	0.793	–
Aptitude
Self-esteem	0.478	–
Emotional intelligence	0.479	–
GPA	0.480

Self-esteem involved 10 items, comprising five positively and five negatively phrased items (Rosenberg, 1965). A sample item was ‘I take a positive attitude of myself.’ Identification of self-esteem evolved from confirmatory factor analysis incorporated in structural equation modelling.

Academic achievement was in terms of the self-reported grade point average (GPA) of the latest academic result.

Aptitude was a latent factor identified by self-assessed emotional intelligence, self-esteem and GPA, predicated on third-order confirmatory factor analysis incorporated in structural equation modelling.

Acquiescence was the composite of all rating items, used as a control or method factor in structural equation modelling to filter out the rating tendency or method artefact (Gignac et al., 2005).

Modelling

Structural equation modelling incorporated a part of confirmatory factor analysis and a part of structural relation analysis (Kline, 2005; Muthen & Muthen, 2006). The modelling was particularly useful for decomposing the correlation between the trait factors of self-assessed emotional intelligence and self-esteem into their reciprocal effects and loadings on a higher order factor. This decomposition was the focal model (Model 2), as compared with a basal, comparison model (Model 1), which only estimated the correlation between the factors. In the focal model, the confirmatory factor analysis specified a third-order factor of aptitude, a second-order factor of emotional intelligence and five first-order trait factors of self-emotion appraisal, other-emotion appraisal, emotion use, emotion regulation and self-esteem (see Figure 1). This analysis also specified influences from acquiescence to the 26 items measuring self-assessed emotional intelligence and self-esteem in order to separate the trait factors from the method factor (Gignac et al., 2005). The specification was a kind of bi-factor analysis used to minimise the bias due to the method artefact of acquiescent rating (Jennrich & Bentler, 2011). Four of the first-order trait factors served to identify the second-order factor of emotional intelligence. This second-order factor, the first-order factor of self-esteem and GPA in turn identified the third-order factor of aptitude, with their loadings constrained to be equal. The corresponding premise posited that aptitude predisposed self-assessed emotional intelligence, self-esteem and self-reported GPA equally. In addition, the focal model (Model 2) estimated reciprocal influences between the self-assessed emotional intelligence factor and self-esteem factor. Instead of identifying the third-order factor and reciprocal influences, the comparison model (Model 1) allowed the factors of emotional intelligence, self-esteem and GPA to correlate with each other. Furthermore, both the focal and comparison models estimated the influences of background characteristics on emotional intelligence, self-esteem and GPA and residual correlations between consecutive items appearing in the questionnaire. Identification of the former served to control for the background influences of age, gender, major field of study and year of study. Meanwhile, estimation of the residual correlations between consecutive items served to accommodate the effect due to questionnaire design in order to maximise the fit of the model.

Figure 1 The model with a latent predisposition of aptitude and reciprocal effects between emotional intelligence and self-esteem. *Effect significant at the p < 0.05 level.

Results

Based on the simple averaging of item scores, all the self-emotional appraisal, other-emotion appraisal, emotion use, emotion regulation and self-esteem were within the middle range (M = 52–56). Their skewnesses and kurtoses were all very low, indicating that their distributions did not deviate from the normal distribution (Kline, 2005). This warranted the adequacy of the usual maximum likelihood estimation method in structural equation modelling.

The structural equation modelling manifested a good fit for both the focal (Model 2) and comparison models (Model 1) (see Table 3). Both models had comparative goodness-of-fit indices (CFI) above 0.95, standardised root-mean-square residuals (SRMR) below 0.05 and root-mean-square errors of approximation (RMSEA) below 0.07 (Marsh, Hau, & Wen, 2004). Nevertheless, the focal model outperformed the comparison model, in that the former model attained a higher CFI and a lower SRMR, RMSEA and Bayesian Information Criterion (BIC). The likelihood ratio chi-square (χ²) test also favoured the focal model, such that it fitted significantly better than did the comparison model (Δχ²(1) = 15, p < 0.001). Such comparison endorsed the adequacy of identifying the third-order factor of latent aptitude and reciprocal influences between self-assessed emotional intelligence and self-esteem.

Table 3 Goodness-of-fit.

Model	Specification	χ^2	df	BIC	RMSEA	SRMR	CFI
1.	Correlational	601	420	105,769	0.033	0.041	0.950
2.	Predispositional and reciprocal effect	586	419	105,759	0.031	0.038	0.954

Notes: Model 1 specified correlations among self-assessed emotional intelligence, self-esteem and GPA. Model 2 specified the loadings of self-assessed emotional intelligence, self-esteem and GPA on aptitude and reciprocal effects between self-assessed emotional intelligence and self-esteem.

Based on the focal model and the comparison model, the identification of all the first-order, second-order, and third-order factors was successful, in view of the substantial factor loadings (see Table 2 and Figure 1). The success revealed the structural validity of the factors, as organised hierarchically with items converging on their respective trait factors and meanwhile distinguished from the method factor (Onwuegbuzie, Daniel, & Collins, 2009). Discriminant validity emerged at the item level when the loading of an item on the trait factor was stronger than that of the same item on the method factor. Moreover, the four first-order trait factors of emotional intelligence converged on the second-order factor of emotional intelligence (λ>0.75).

In the comparison model (Model 1), self-assessed emotional intelligence showed a strong correlation of 0.698 (p < 0.001) with self-esteem, controlling for all the background factors. GPA also had significant correlations with self-assessed emotional intelligence (r = 0.305, p < 0.001) and self-esteem (r = 0.265, p < 0.001). A better understanding of these correlations was in terms of loadings on the third-order factor of aptitude and reciprocal influences in the focal model (Model 1, see Figure 1). The loadings showed that latent aptitude influenced all self-assessed emotional intelligence, self-esteem and GPA moderately (λ = 0.48). Meanwhile, self-esteem showed a significant positive effect (0.363) on emotional intelligence, which displayed a positive but not statistically significant effect (0.194) on self-esteem in return. The former supported Hypothesis 1, whereas the latter failed to support Hypothesis 2.

The above findings emerged, after controlling for background characteristics. Among the effects of background characteristics, only those of age, the business major and the year of study on the self-esteem factor were significant (see Table 4). The effects of age and the business major were positive, whereas the effect of the year of study was negative.

Table 4 Standardised effects of background characteristics.

Predictor	Emotional intelligence	Self-esteem	GPA
Age	0.003	0.235*	0.047
Female	− 0.041	0.003	− 0.012
Social science major	− 0.106	0.149	0.158
Business major	− 0.022	0.195**	0.135
Science major	− 0.006	0.086	0.089
Year of study	− 0.032	− 0.283***	− 0.022
R^2	0.647	0.514	0.238

Note: The law major is the reference category. *p < 0.05, ** p < 0.01, and *** p < 0.001.

Discussion

With the identification of first-order trait factors, the second-order factor of self-assessed emotional intelligence and the third-order factor of latent aptitude, self-esteem emitted a significant positive effect on self-assessed emotional intelligence. This finding supports Hypothesis 1. However, self-assessed emotional intelligence did not engender a significant effect on self-esteem in turn, even though the effect was a positive one. This finding fails to support Hypothesis 2. Overall, the findings reveal the top-down approach to the formation of the self-concept of emotional intelligence, given the valid identification of such a self-concept. The findings are also compatible with the broaden-and-build theory and cognitive-adaptive theory of emotional intelligence. Accordingly, self-esteem would sustain one's broadening and building on one's strengths to develop emotional intelligence, and self-esteem would facilitate one's engagement in social environments to enable one to learn about how to nurture emotional intelligence. The findings also suggest that the bottom-up approach to formation of self-esteem by the self-concept of emotional intelligence is not adequate. A possible reason for the failure of the bottom-up approach is that self-assessed emotional intelligence is only one of the many constituents of self-esteem (Lyubomirsky, King, & Diener, 2005). Alternatively, self-esteem is relatively stable as a personality trait (Trzesniewski, Donnellan, & Robins, 2003). As such, perceived and real performance based on emotional intelligence is not sufficient to enhance self-esteem. This implies that the self-enhancement function of self-assessed emotional intelligence is not as definite as expected, given the stronger reciprocal influence of self-esteem on self-assessed emotional intelligence.

Hence, self-estimated emotional intelligence neither exerted a significant influence on self-esteem nor significantly varied due to age, gender, the major field of study and year of study, given the predisposition of latent aptitude. The correlation between self-assessed emotional intelligence and self-esteem was strong, and this finding replicates previous research (Schutte et al., 2002). Furthermore, this correlation appears to result largely from the effect of self-esteem on self-assessed emotional intelligence and their common loadings on latent aptitude. The influence of self-esteem echoes the view that self-esteem is a force for conserving ideas about the self and developing skills for adaptation (Silbereisen & Wiesner, 2002). Meanwhile, self-assessed emotional intelligence is unlikely to develop during young adulthood, as consistent with previous research (Harrod & Sheer, 2005). A possible reason is due to the limitation of the sample of university students, whose study environments and other exposures do not seem to vary with age (Adams & Ryan, 2000). This invariance would constrain the development of emotional intelligence, which depends on socially challenging opportunity, according to cognitive-adaptive theory (Zeidner et al., 2009). Moreover, gender did not make a difference in self-assessed emotional intelligence. This finding, however, deviates from some expectation and research findings (Tett & Wang, 2005). A possibility, again, is the limitation of the sample of university students, who tend to show smaller gender differences than do older people (Lippa, 2005). This may be because of the similar student experiences in university, which already admits students of similar qualifications and provides them with uniform treatment. Male and female university students are also likely to socialise and mix among themselves (Stevens, Armstrong, & Arum, 2008). This would dampen the difference in social experience and eventually emotional intelligence between male and female students, with respect to cognitive-adaptive theory (Zeidner et al., 2009).

The positive effect of age on self-esteem is consistent with that found in research and theory (Oliver, 2003). One theory is the accumulation of various resources with age to raise self-esteem. Notably, the older person tends to encounter fewer negative experiences to erode self-esteem (Carr & Friedman, 2005). In contrast, the year of study showed a negative effect on self-esteem. This finding is compatible with the view that a higher level of university study is more competitive (Owens, 1992). The increased competition would pose a challenge to the student's self-esteem (Faunce, 2003). A business student was higher in self-esteem than were others, probably owing to business education that promotes the student's leadership and superiority (Posner, 2009). In this connection, the role of a leader helps foster the student's self-esteem.

Further research

The limitations of the study in sampling and design definitely necessitate further research for substantiating the findings of the study. Obviously, the design is a cross-sectional one, which can only rely on structural equation modelling to recover causal relationships. Such recovery is not definite, in the absence of clear evidence on temporal order. A panel design is indispensable to further research in order to ascertain temporal precedence in causal relationship in real life. Better sampling of people of diverse backgrounds and sociocultural contexts is certainly necessary to guarantee the generality of findings. Notably, the possibility of variation in the relationship between emotional intelligence and self-esteem due to the Chinese, Confucian tradition deserves further investigation. The possibility can emerge from the centrality of social relationship for the formation of self-esteem and emotion in Chinese, Confucian culture (Phillips, 2006). Hence, the nexus between self-esteem and self-assessed emotional intelligence may be especially strong in a Chinese, and this requires a test based on cross-cultural data.

Theories about the reciprocal influences between emotional intelligence and self-esteem also require further investigation. Such investigation needs to illuminate mechanisms underlying the influence of self-esteem on self-assessed emotional intelligence, predicated on the top-down approach, broaden-and-build theory and cognitive-adaptive theory. To do this, further research can illustrate such mediating processes of self-evaluation maintenance, relaxation, self-regulation, cognitive restructuring, social challenge and learning proposed by the theories (Zeidner et al., 2009). Meanwhile, the bottom-up formation of self-esteem by self-assessed emotional intelligence also deserves further investigation, particularly concerning the possible mechanisms of self-perception and performance (Redersdorff & Guimond, 2006). Performance, achievement and appraisal facilitated by emotional intelligence are mediators for such investigation. Identifying such mediating paths may demonstrate the indirect effects of self-assessed emotional intelligence on self-esteem.

Implications

Use and promotion of self-assessed emotional intelligence need to take note of its influential sources, including aptitude, self-esteem and even acquiescence or rating method, as unfolded by structural equation modelling. Hence, self-assessed emotional intelligence partly reflects aptitude, self-esteem and the rating tendency. Promotion of self-assessed emotional intelligence thereby would largely hinge on the fostering of aptitude and self-esteem. This fostering would benefit from the investment model, which emphasises early child parenting, coaching, discussion, reinforcement and rewarding socialisation (Zeidner et al., 2009). Apart from empowering aptitude and self-esteem in early life, sustaining self-esteem through the strengthening of social identity throughout one's lifespan is viable (Hogg, 2000). Such identity can be integral to the development of self-assessed emotional intelligence (Seaton & Beaumont, 2008).

Additional information

Notes on contributors

Chau-Kiu Cheung

Chau-kiu Cheung, PhD, is an associate professor at the City University of Hong Kong, China. He has recently published articles concerning resilience, social inclusion, character education, moral development, peer influence and class mobility. His current research addresses issues of grandparenting, drug abuse, risk society and Internet use.Hoi Yan Cheung is an assistant professor at the City University of Hong Kong, Department of Applied Social Studies. She is interested in doing research related to education. Currently, she is doing research on the risk-taking perceptions and behaviours of university students. Moreover, she is interested in doing comparative education studies.Ming-Tak Hue is associate professor at the Department of Special Education and Counselling, Hong Kong Institute of Education. He teaches graduate courses in self and personal growth, school guidance and counselling, classroom management, behaviour management and inclusive education. His research interests are in ethnic minority education, emotional and spiritual growth, and the development of guidance and discipline programmes.