Reliability , construct and criterion-related validity of the Serbian adaptation of the Trait Emotional Intelligence Questionnaire ( TEIQue )

This paper presents evidence on the reliability and validity of the Serbian adaptation of the Trait Emotional Intelligence Questionnaire (TEIQue), an instrument designed to comprehensively assess emotional intelligence conceived as a constellation of emotionrelated self-perceptions. Study participants were 254 adults, who completed the Serbian TEIQue, NEO-FFI, MSCEIT, EQ-short, and RSPWB. The results indicate that the adapted TEIQue is a psychometrically sound assessment tool: internal consistencies were mostly acceptable at facet, generally good at factor, and excellent at whole-scale level; the fourfactor structure was confirmed by means of CFA; convergent-discriminant validity was established through meaningful associations with related constructs, indicating that trait EI is closely aligned with affect and self-efficacy related constructs from the realm of personality (i.e., E, N, C, and Empathy), but shows only moderate overlap with ability EI; finally, incremental validity was demonstrated in the prediction of psychological wellbeing, over and above the Big Five.

The term trait emotional intelligence (trait EI) was introduced by Petrides and Furnham (2000;2001) in an attempt to systematize proliferating operationalizations of EI by drawing on the fundamental psychometric distinction between measures of typical and maximum performance.In the proposed differentiation, trait EI was Corresponding author: zjolic@f.bg.ac.rsAcknowledgments.The research reported here was funded by the Ministry of Education and Science of the Republic of Serbia (Project No. 179018) and is part of a larger study which we conducted jointly with Dr. Nataša Hanak and Dr. Aleksandar Dimitrijević.Data collection within participants' working hours and environment was made possible by the managers of "Salford Capital Partners Inc. Beograd", to whom we remain grateful.We would especially like to thank Dr. K.V. Petrides, who was available with encouraging feedback throughout the process of adapting the TEIQue for the Serbian-speaking population.
intended to designate emotion-related self-perceptions and dispositions measured via self-report, as opposed to actual emotion-processing abilities measured with task-based tests and referred to as ability EI.
Since that time, trait EI (alternatively labelled emotional self-efficacy) has been established as a meaningful and useful construct located within the realm of personality, presumably at the lower levels of personality hierarchies (Petrides, Pita, & Kokkinaki, 2007).Although sharing as much as 65% of variance with the Big Five, trait EI has the explanatory advantage of covering emotion-related individual differences which otherwise remain scattered across the Big Five domains of personality (Petrides, Furnham, & Mavroveli, 2007).
What precisely is encompassed by trait EI? Sifting core components which reappeared in different early models of EI from those which were unique to one particular conceptualization, Petrides and Furnham (2001) identified 15 facets as representing the sampling domain of trait EI: adaptability, assertiveness, emotion expression, emotion management, emotion perception, emotion regulation, low impulsiveness, maintaining relationship skills, self-esteem, selfmotivation, social awareness, stress management, trait empathy, trait happiness, and trait optimism.According to trait EI theory, these facets are organized into four interrelated factors: well-being (traits pertaining to dispositional mood), self-control (self-efficacy in regulating emotions/impulses), emotionality (selfefficacy in perceiving and expressing emotions), and sociability (self-efficacy in interpersonal utilization and management of emotions) (Petrides, 2009).

Measuring Trait EI
Although there is an abundance of self-report instruments purporting to measure EI (see Pérez, Petrides, & Furnham, 2005 for an overview), most of them may be regarded only as "flawed measures of trait EI" (Petrides et al., 2007, p. 159), devoid of a sustainable theoretical basis and providing only partial coverage of the construct's sampling domain.The Trait Emotional Intelligence Questionnaire (TEIQue; Petrides, 2009), on the other hand, has been purposefully developed to comprehensively operationalize the construct in question.This 153-item inventory thus assesses all of the above listed trait EI components, yielding an appropriate number of scores at the facet, factor and whole-scale level (i.e., 15-4-1).
The TEIQue (including its various forms and translations) has been subject to rigorous psychometric evaluations, generally yielding rather favourable results on the inventory's psychometric properties.Internal consistencies reported in the manual (Petrides, 2009) and in more recent publications (e.g., Martskvishvili, Arutinov, & Mestvirishvili, 2013) are adequate at the facet and factor level, and very high at the level of global scores.Further, global trait EI and most of its 15 facets have been found to exhibit high temporal stability (test-retest reliability) (Petrides, 2009).

TEIQue Criterion Validity: Why the Assessment of Trait EI Matters
The criterion validity of the TEIQue -particularly the British and French version -has also been carefully studied and evidence obtained that trait EI (incrementally) predicts an array of important outcomes.
A comprehensive study by Petrides, Pérez-Gonzáles, and Furnham (2007) has established that low trait EI is associated with various indicators of psychopathology, including maladaptive coping styles, dysfunctional attitudes, self-reported depression, aggression, and personality disorders, with most relationships remaining significant when the Big Five are controlled for.Trait EI was further found to predict somatic complaints -above alexithymia and optimism (Mikolajczak, Luminet, & Menil, 2006), and above positive and negative affect (Andrei & Petrides, 2013) -and to mediate the paths between the Big Five and general health (Greven, Chamorro-Premuzic, Arteche, & Furnham, 2008;Johnson, Batey, & Holdsworth, 2009).A series of studies by Mikolajczak and colleagues (e.g., Mikolajczak et al., 2006;Mikolajczak, Luminet, Fillée, & de Timary, 2007b;Mikolajczak & Luminet, 2008;Mikolajczak, Petrides, Coumans, & Luminet, 2009) has shown that trait EI moderates the psychological and somatic/biological response to performancerelated stress.Moreover, it was demonstrated that trait EI predicts resistance to stress above the Big Five and other personality constructs (Mikolajczak et al., 2007b(Mikolajczak et al., , 2009)), and that it may in fact serve to down-regulate various negative emotions and not only stress (Mikolajczak et al., 2008).In line with this, Laborde and colleagues (Laborde, Brüll, Weber, & Anders, 2010) found trait EI to predict cognitive functioning on a stressful learning and decisionmaking task.
Trait EI assessments via the TEIQue have proved efficient not only vis-à-vis concurrent predictors (e.g., the Big Five), but also with respect to competing measures of EI: in a study by Gardner and Qualter (2010) global TEIQue scores were found to be a superior predictor of hostility, alcohol abuse, loneliness, happiness, and life satisfaction, in comparison with two other selfreport measures of EI.Thus, the TEIQue might arguably be the measure of first choice when predicting health, well-being, and even performance-related outcomes.

The Present Study
The purpose of the present study was to provide an independent psychometric evaluation of the TEIQue in a Serbian sample.In particular, our aim was to: (1) establish the distributional properties and internal consistencies of the current Serbian adaptation of the TEIQue; (2) test prior findings on the robustness of the proposed four-factor structure; (3) look into the relationship of trait EI scores with age and gender; (4) confirm the pattern of associations between trait EI and the Big Five, and explore the relations of the former with ability EI and empathy; (4) examine the TEIQue's incremental validity over the Big Five in predicting psychological well-being.In sum, the current study was designed so as to obtain further and expanded evidence on the reliability, factorial, convergent-discriminant, and criterion validity of the TEIQue.

Participants and procedure
Participants were 254 adults (137 males), aged 21 to 61 (M= 40.21,SD= 8.17) and employed at various positions within a large dairy company.Prior to data collection, participants were briefed about the general purpose of the study, and gave their informed consent to participate.By courtesy of the Company's management, data were collected during working hours, in two separate testing sessions with a 15-30 days intermission.

Measures
Trait Emotional Intelligence Questionnaire (TEIQue v.1.5)-Serbian adaptation.The TEIQue (Petrides, 2009) consists of 153 items, which respondents are asked to rate on a 7-point Likert scale, ranging from "completely disagree" to "completely agree".The rationale and psychometric properties of the instrument are described in the Introduction.
Participants in this study were administered the current Serbian adaptation of the TEIQue.This version of the instrument was arrived at after back translation and pilot testing of the initial Serbian translation, all of which had been carried out in cooperation with Dr. Petrides.Although the initial translation was already judged as good in terms of wholescale reliability (α= .95) and factorial validity (EFA yielded four recognizable factors; see Altaras Dimitrijević, Jolić Marjanović, Petrović, & Petrides, 2011 for details), some items were identified as compromising facet-level internal consistencies and eventually revised for the current Serbian adaptation of the TEIQue (12 items were rephrased by changing the wording or word order, 3 were shortened, and 4 were replaced with alternative items proposed by Dr. Petrides).
Participants also filled out TEIQue's "Question Section 2 -About you", thus providing data on age, gender, and other demographic variables.
The Mayer-Salovey-Caruso Emotional Intelligence Test Version 2.0 (MSCEIT v2.0).The MSCEIT (Mayer, Salovey, & Caruso, 2002) is a 141-item measure of ability EI.Test items are organized into eight tasks, yielding scores on four branches: Perceiving, Using, Understanding, and Managing emotions.The test also provides an overall ability EI score.In this study, the approved Serbian translation of the instrument was used and responses were scored by the publisher (Multi-Health Systems), using general consensus scoring.The Serbian MSCEIT was previously found to exhibit good reliability and convergent-discriminant validity (Altaras Dimitrijević & Jolić Marjanović, 2010).

NEO Five Factor Inventory (NEO-FFI).
The NEO-FFI (Costa & McCrae, 1992;McCrae & Costa, 2004) is a 60-item measure of the basic personality dimensions according to the five-factor model: Neuroticism (N), Extraversion (E), Openness (O), Agreeableness (A), and Conscientiousness (C).Each of the five subscales comprises 12 items, which are scored on a 5-point Likert scale.The scales' internal consistencies in the present study ranged from .59 to .81.
The Empathy Quotient Short (EQ-Short).The EQ-Short is a 22-item measure of empathy, derived from the original 40-item version via principal component analysis (Wakabayashi, Baron-Cohen, Wheelwright, Goldenfeld, Delaney, et al., 2006).Responses are given on a 4-point Likert-scale.The Serbian translation of the instrument (S-EQ) was found to have good psychometric properties (Dimitrijević, Hanak, Vukosavljević-Gvozden, & Opačić, 2012), and Cronbach's alpha in the present sample was .71

Ryff's Scales of Psychological Well-Being Short (RSPWB-S).
The RSPWB-S (Ryff & Keyes, 1995) is an 18-item inventory designed to provide a brief but comprehensive measure of psychological well-being conceived as a multifaceted construct.With respect to independent studies questioning the multidimensionality of the RSPWB-S (Springer & Hauser, 2006), we used only whole-scale level scores, for which internal consistency was found to be .74.

Distributional properties
As with the UK sample, the mean values of all TEIQue scores were somewhat above the theoretical average of a 7-point rating scale (i.e., 3.5), with Trait happiness eliciting the highest, and Emotion management the lowest mean scores.The greatest variability of scores was observed for Emotion expression, and the smallest for Relationships.Excluding Trait happiness, all TEIQue scores (facet, factor, and global) approximated normal distribution and the Kolmogorov-Smirnov tests yielding nonsignificant statistics (Table 1).

Reliability
Most of the subscales displayed acceptable to good reliability, with alphas above the recommended .70 level (Peterson, 1994) for ten out of 15 facets.Internal consistencies were consistently good at the factor and excellent at whole-scale level (Table 1).

Factor structure
The proposed structure of the TEIQue was tested by means of confirmatory factor analysis (CFA) using the maximum likelihood parameter estimation method and AMOS for SPSS software, with all relevant data available for all participants.The obtained fit indices suggested that a model equivalent to the theoretical structure of the TEIQue (allowing only for factor intercorrelations) did not fully fit the observed data (χ 2 (59, N = 254) = 187.21,TLI = .90,CFI = .92,RMSEA = .09,and SRMR = .05).Upon inspection of the modification indices, correlated errors were modelled between three pairs of facets (Figure 1): Trait happiness -Trait optimism, Emotion expression -Social awareness, and Trait empathy -Assertiveness.These minor changes yielded substantial improvements in all goodness-of-fit indices, which now attested to the model's good-to-excellent fit to the analysed data (χ 2 (56, N = 254) = 112.70,TLI = .95,CFI = .97,RMSEA = .06,and SRMR = .05).Of note, the allowed facet error correlations are all theoretically sustainable: happiness and optimism are expected to be closely related, as are the capability to communicate one's feelings and good social skills; similarly, it is conceivable that highly assertive persons can sometimes be less sensitive to others' feelings.Standardized parameter estimates indicate high factor loadings for all facets (.65 -.89), and interrelations between factors were in a similar range (.65 -.87).

Figure 1. CFA of the four-factor model of the TEIQue with allowed correlated errors
As a further test of factorial validity, zero-order correlations were calculated between regression factor scores obtained through principal component EFA in the present data set and factor scores obtained by applying the a priori scoring key.All correlations were very high: .93 for Well-being, .99 for Self-control, .94 for Emotionality, and .97 for Sociability.

Relationship with demographic variables
Correlations with age.All TEIQue scores, except Emotion perception, displayed significant negative correlations with age (Table 1).Although generally low, these correlations remained significant after controlling for the Big Five or level of education.Gender differences.Men scored higher than women on five facets and two factors (Table 1), with effect sizes being relatively small (d=.30-.36) for Impulsiveness (low), Stress management, Adaptability, Assertiveness, and Sociability, and medium for Emotion regulation (d= .54)and Self-control (d= .49).

Convergent-discriminant Validity
Relationships with the Big Five.TEIQue factor and global scores correlated significantly with all five personality traits (Table 3).N stood out as the strongest negative correlate of trait EI; correlations with E and C were medium-to-large and positive, while those with O and C were small, but still significant for the global and 3 out of 4 TEIQue factor scores.Together the Big Five explained 73% of TEIQue variance (F (5, 248) = 138.36,p<.001).

Relationships with empathy.
All TEIQue factors and the global score were significantly and positively related to the Empathy Quotient, with correlations ranging from moderate to large (Table 2).Moreover, the correlation between Empathy and the TEIQue factor of Emotionality remains large and significant (r= .60,p<.001) even when Trait empathy is excluded from the latter.
Relationships with ability EI.Significant positive correlations were also found between the TEIQue and the MSCEIT (Table 2): practically all TEIQue factors and global trait EI had small correlations with the two experiential branches (Perceiving and Using emotions), and moderateto-large correlations with the two strategic branches (Understanding and Managing emotions), as with global ability EI.

Criterion Validity
Zero-order correlations between the TEIQue and RSPWB-S were large, all reaching or exceeding .60 (Table 2).Moreover, the correlation between psychological well-being and trait EI remained very large and significant (r= .73,p<.001) even when the Well-being factor was excluded from the overall TEIQue score.
When entered as a single predictor in the regression analysis, global trait EI explained 55% of the variance in psychological well-being (F (1, 252) = 313.54,p<.001).
To test the incremental validity of trait EI in predicting the same criterion, two hierarchical regression analyses were performed, with either the Big Five or ability EI and Empathy entered in Step 1, and the TEIQue global score added in Step 2. As a result, two significant prediction models emerged, explaining 59% and 56% of the variance in well-being, with trait EI accounting for an additional 7% and 25%, respectively (Table 3).

Discussion
The present study investigated the psychometric properties of the Serbian adaptation of the TEIQue, shedding light on the instrument's distributional properties, reliability, factor-structure, associations with age and gender, but also addressing issues of convergent-discriminant and predictive validity by looking at the relationship of TEIQue variables with the Big Five, empathy, ability EI, and psychological well-being.
The distributional properties of the current Serbian translation of the TEIQue are good, with means and standard deviations comparable to those obtained for the original (Petrides, 2009).The basic requirement of normality was fulfilled for all TEIQue variables except Trait happiness, for which a negatively skewed distribution was observed.This is in accordance with results from the French sample and, as Mikolajczak et al. (2007a) have noted, should be not be regarded as a weakness, but rather as evidence of validity, since research has demonstrated that people generally tend to feel happy.
Overall, the accuracy of measurement indices obtained for the Serbian adaptation are somewhat lower than those reported for other versions (Freudenthaler et al., 2008;Mikolajczak et al., 2007a;Petrides, 2009), but still range from acceptable to good for most TEIQue scales (alphas are at or above .70for 10 out of 15 facets, and well above this threshold for all factors and the full scale).Facets with lower internal consistencies include Relationships, Selfmotivation, Trait Empathy, and Emotion Management which occupied the low end of alphas in the UK (Petrides, 2009), French (Mikolajczak et al., 2007a), German (Freudenthaler et al., 2008), and Georgian (Martskvishvili et al., 2013) samples as well, which may indicate that a general revision of these facets is required in order to improve accuracy of measurement.Most importantly, however, the reliability of the Serbian TEIQue is good at factor and excellent at whole-scale level.
The results of CFA, after allowing for correlated errors for three pairs of facets, strongly support the proposed four-factor format of the TEIQue.It is particulary noteworthy that the modifications introduced in the model were all theoretically sustainable and in two out of three instances (i.e., Trait happiness-Optimism and Assertiveness-Trait empathy) coincided with those made in the German TEIQue study (see Freudenthaler et al., 2008 for details).As further evidence of factorial validity, there was high convergence between empirically derived factor scores (obtained via principal component analysis) and those calculated on the basis of the a priori scoring key.In sum, factor analyses of the Serbian adaptation corroborate previous observations on the robustness of the four TEIQue dimensions, and lend support to the more general contention that the structure of emotion-related self-perceptions is not susceptible to cultural differences (Freudenthaler et al., 2008;Mikolajczak et al., 2007a), at least not within the broader context of the "Western world".
Gender differences on the Serbian TEIQue were all in favour of men and pertained primarily to factors of Self-control and Sociability, which yielded men as the higher-scoring group in the French (Mikolajczak et al., 2007a) and Georgian (Martskvishvili et al., 2013) samples as well.Gender differences with respect to global trait EI were not recovered in the Serbian sample.Also, the effect of gender on TEIQue scores seems to be less pronounced than in the UK sample.
All Serbian TEIQue variables except Emotion perception were significantly negatively related to age, which is at odds with previous findings of trivial positive associations (Mikolajczak et al., 2007a), but might be explained as a result of a more balanced representation of different age groups in our sample.Given the cross-sectional design of this study, the observed correlations do not necessarily imply that trait EI decreases with age; the fact that our younger participants tended to report higher emotional self-efficacy might reflect cohortdifferences: it is conceivable that these generations were socialized into a culture that puts more value on possessing and displaying emotion-related competencies than did the typical environment of our elder participants.
Trait EI factors and global score are significantly related to basic personality traits, empathy, and ability EI, in a way that supports the convergentdiscriminant validity of the Serbian TEIQue.First, the portion of common variance between trait EI and the Big Five is large, which is consistent with Petrides's conceptualization of EI as a personality construct; more specifically, the observed pattern of correlations -with N, E, and C as the most salient correlates of trait EI -mirrors the one obtained in Freudenthaler et al.'s (2008) study and fits the theoretical expectation that trait EI should be most strongly associated with affect and self-efficacy related traits.Second, a strong relationship is also found between trait EI and the cognate construct of empathy; as expected, this association is strongest for the TEIQue factor of Emotionality, which pertains to the perception and expression of emotions and thus remains closely aligned to EQ even when the Emotionality score is calculated so as to exclude Trait empathy.Finally, trait EI displayed moderate associations with a performancebased measure of EI -the MSCEIT.The amount of overlap between trait and ability EI tends to increase as a function of the hierarchy of MSCEIT branches (with Perceiving emotions as the lowest, and Managing emotions as the highest branch), but generally supports the distinctness of the two EIs.
That despite an array of moderate-to-large associations trait EI is not redundant with any of the aforementioned constructs (i.e., the Big Five, empathy, ability EI) is ultimately evidenced by the results of regression analyses testing the predictive validity of the Serbian TEIQue vis-à-vis psychological well-being.As these analyses show, the Serbian TEIQue acts as a powerful single predictor of psychological well-being (even when the Well-being factor is excluded from the global trait EI score), and in fact explains unique criterion variance over the Big Five (7%), as well as over empathy and ability EI (25%).These findings stand as important evidence of the incremental validity and utility of the Serbian TEIQue, but also as further testimony to the meaningfulness of the trait EI construct.
Before concluding, we would like to acknowledge several limitations of the present study and suggest possible directions for future research.First, although the sample was heterogeneous with respect to several demographic variables, all participants were employees of the same company, living and working in or near the country's capital; future studies with the Serbian TEIQue should draw samples from different organizational settings and parts of the country.Second, even though we were able to observe that the Serbian TEIQue is a superior predictor of psychological well-being, what remains missing is a comparison of the adapted TEIQue with other self-report measures of EI, as well as evidence of the instrument's predictive validity with respect to additional criteria, preferably measured by means other than self-report.Finally, the use of a relatively short assessment of the Big Five (i.e., the 60-item NEO-FFI rather than the 240-item NEO-PI-R) could have resulted in an underestimation of the shared variance between the basic personality domains and trait EI, and a subsequent overestimation of the TEIQue's incremental validity.
These limitations notwithstanding, the current findings present the Serbian TEIQue as a psychometrically sound instrument yielding data which is consistent with extant trait EI research and potentially useful in applied settings.

Table 2
Zero-order correlations of TEIQue factors and global score with ability EI, the Big Five, Empathy, and Psychological well-being **p< .01;*p< .05.