Conceptualizing and measuring EI

The conceptualization and measurement of EI have evolved since its origins as a research field in 1990. Nowadays, two common approaches to conceptualizing EI are ability EI and trait EI. Ability EI refers to EI as a form of intelligence that processes emotional information and focuses on an individual's capacity to perceive, facilitate, understand, and manage emotions (^{Mayer & Salovey, 1997}). Ability EI is often measured through performance-based tests that simulate emotional scenarios, measuring an individual's competencies in real-time emotional navigation. An example of such an instrument is the Mayer-Salovey-Caruso Emotional Intelligence Test (MSCEIT; ^{Mayer et al., n.d.}). In contrast, trait EI centers around inherent emotional traits and dispositions as stable dispositions and personality facets (^{Petrides, 2011}). More specifically, trait EI has been defined as a constellation of emotional self-perceptions at the lower levels of personality hierarchies (^{Petrides et al., 2007}). Trait EI is typically assessed through self-report questionnaires that offer a subjective assessment of one's emotional tendencies, such as the Trait Emotional Intelligence Questionnaire (TEIQue; ^{Petrides, 2009}).

^{Joseph and Newman (2010}) have expanded on this distinction, differentiating between three EI models: performance-based ability, self-report ability, and mixed EI models. Performance-based ability EI models define EI as a set of emotion-related abilities that should be assessed via performance-based tests (e.g., MSCEIT). Self-report ability models also view EI as an ability but suggest measuring it through self-reported instruments that reflect participants' subjective perceptions about their emotional aptitudes. The most frequently used measurement tools within this alternative conceptualization are the Trait Meta-Mood Scale (TMMS; ^{Salovey et al., 1995}), the ^{Schutte Self-Report Emotional Intelligence Test (SSEIT; Schutte et al., 1998}), and the ^{Wong and Law Emotional Intelligence Scale (WLEIS; Wong & Law, 2002}). Mixed EI models understand EI as a broad construct combining emotion-related competencies, social skills, and personality features measured through self-report questionnaires. The ^{Bar-On Emotional Quotient-Inventory (EQ-i; Bar-On, 2006}) and the Emotional and Social Competence Inventory (ESCI; ^{Boyatzis & Goleman, 2007}) are commonly used by researchers to measure EI within this approach.

This study examines the dimensionality of EI in adolescents from Kazakhstan based on a self-reported ability model for three main reasons. First, although the trait EI model has gained popularity in recent years (^{Siegling et al. 2015}), the self-reported ability model is the most widely used procedure to measure EI in adolescents (^{Soriano-Sánchez & Jiménez-Vázquez, 2023}). Second, this seems to be the case in Kazakhstan, with limited research in this field, but some recent studies adopting self-reported ability EI conceptualizations in their designs (e.g., ^{Adilzhanova et al., 2024}; ^{Taibolatov et al., 2024}). Third, unlike trait EI models that conceptualize emotional intelligence as a set of stable personality traits, the self-reported ability model views EI as a set of competencies that can be developed and enhanced over time (^{Castillo-Gualda et al., 2018}; ^{Puertas-Molero et al., 2020}).

Various studies have examined the validity and reliability of self-reported EI scales across different populations and cultural contexts. To note, ^{Karabuschenko et al. (2016}) explored the capacity of individuals to recognize emotions in others of different cultures and identified key differences in some population groups, emphasizing the need to consider the development of EI across various social and ethnic contexts. Therefore, validating the various instruments in different contexts is an ongoing research process. For instance, the TMMS-24 has been validated in adolescent samples in Spain (^{Pedrosa et al., 2014}), Chile (Gong et al., 2020), Brazil (^{Câmara et al., 2023}), and France (^{Maria et al., 2016}), among many other countries. The factorial validity of the WLEIS has also been extensively explored in Spanish (^{Pacheco et al., 2019}), Italian (^{Iliceto & Fino, 2017}), American (^{LaPalme et al., 2016}), and Chinese (^{Li et al., 2012}) children and adolescents. Similarly, the psychometric properties of the SSEIT have been tested in adolescents and young adults across diverse contexts and cultures (e.g., ^{Gong & Paulson, 2018}; ^{Nassar et al., 2023}).

These studies contribute valuable insights into measuring and evaluating EI for diverse populations, emphasizing the significance of considering cultural and contextual factors when interpreting EI measures. By applying rigorous psychometric approaches to new sample groups, researchers can ensure the validity and reliability of the measurement of EI and enable an expanded understanding of this important topic.

The protective role of self-reported ability EI in adolescence

During the critical stages of adolescence, teenagers experience major emotional and social changes for which the development of EI is imperative (^{Megías-Robles et al., 2024}). Research has consistently pointed to the positive contributions of EI in adolescence, as measured through the self-reported ability model. For example, studies have shown that EI can positively influence adolescents' mental health and well-being states, including happiness (^{Abdollahi et al., 2019}; ^{Tejada-Gallardo et al., 2022}), quality of life (^{Soriano-Sánchez & Jiménez-Vázquez, 2023}), subjective well-being (^{Llamas-Díaz et al., 2022}), and eudaimonic well-being (^{Salavera & Usán, 2022}). Furthermore, there is evidence that high EI levels may predict teenagers' social adaptation (^{Mestre et al., 2006}) and contribute to reducing suicidal thoughts (^{Extremera et al., 2023}). Research has also identified the important role that EI has played in young people's educational outcomes. For example, ^{Somaa et al. (2021}) found consistent meta-analytic evidence that self-reported ability EI is a statistically significant predictor of academic achievement in young populations. Another recent meta-analysis supported this relationship after controlling for intelligence and Big Five personality factors (^{MacCann et al., 2020}). In addition, self-reported ability EI has been demonstrated to predict important outcomes later in life, such as job performance (^{Joseph et al., 2015}), health (^{Martins et al., 2010}), and civic engagement (^{Miao et al., 2017}).

These findings, collectively, emphasize the importance of understanding and evaluating EI to effectively support teenagers' development across multiple crucial aspects of their development. With much evidence to suggest that EI is associated with improved psychological, social, academic, and future outcomes, it is important to ensure that the measurement of EI is reliable and valid in different contexts, as well as differences and measurement invariant across important personal and socio-cultural characteristics, such as gender and language.

Gender, language, and self-reported ability EI

Gender differences in self-reported ability EI among adolescents have been a subject of interest in various studies. There is considerable consensus that females tend to report higher levels of performance-based ability EI (e.g., ^{D'Amico & Geraci, 2022}; ^{Gutiérrez-Cobo et al., 2016}). However, the evidence for gender differences in self-reported ability EI is less definitive, with findings showing that females sometimes score higher than males on only certain subscales. For example, ^{Ciarrochi et al. (2001}) found that females scored higher in total EI and perceived themselves as better at perceiving emotions, regulating others' emotions, and utilizing emotions, but not in regulating their emotions. Similarly, ^{Law et al. (2004}) reported that women tend to score higher than men in others' emotion appraisal and use of emotion, with no significant differences observed in self-emotion appraisal and emotion regulation. More recently, ^{Salavera and Usán (2022}) found that girls report higher levels of attention to their own and others' emotions, emotion regulation, and using emotions in problem-solving, but not in others. Furthermore, other studies have found no sex-related differences in self-reported ability EI across dimensions (e.g., ^{Joseph & Newman, 2010}). Nevertheless, there seems to be more agreement on the gender measurement invariance of self-reported ability EI measures, which has been consistently observed in different contexts and through several scales (e.g., ^{Kong, 2017}; ^{Pedrosa et al., 2014}).

Researchers have also recognized the significance of linguistic characteristics when measuring EI. For example, ^{Atamanova and Bogomaz (2018}) investigated differences in self-perceived EI competencies among Russian university students majoring in foreign languages studying Chinese, English, and Romance languages. The findings revealed significant variations in EI scores across linguistic groups. Students studying Chinese scored higher in recognizing, understanding, and managing others' emotions compared to students studying English and Romance languages. Additionally, students in the Romance language group outperformed their English-speaking peers in emotional expression, intrapersonal EI, and emotion management. The authors suggest that these between-group differences may be linked to the distinctive characteristics of the languages studied, as stated by ^{Smirnova (2017}). For instance, the tonal nature of the Chinese language necessitates a strong ability to discern the emotional nuances in speech, while Romance languages are inherently more emotionally expressive. Similarly, ^{El Ghoudani et al. (2018}) adapted the existing WLEIS scale for the Moroccan-Arabic-speaking population group and emphasized the importance of considering the cultural and linguistic backgrounds of research participants to ensure the validity and reliability of EI assessments.

Collectively, these results suggest that examining the dimensionality of self-perceived emotional aptitudes and understanding gender-and language-specific differences is important and can help develop targeted interventions to foster EI development and promote important health, social, and academic outcomes among adolescents with diverse demographic and cultural backgrounds.

The Present Study

In recent years, the concept of EI in adolescence has also attracted some attention among scholars in post-Soviet contexts (e.g., ^{Butvilas & Kovaitė, 2022}; ^{Ihnatovych & Liashch, 2020}). The recognition of and interest in EI has grown steadily as individuals, organizations, and educational institutions have begun to realize its potential relevance to social and educational outcomes. However, research on this topic in Central Asian countries, including Kazakhstan, is scarce. An exception is the study by ^{Adilzhanova et al. (2024}), which explored the link between self-perceived ability EI and subjective well-being among adolescents, revealing a positive correlation between the two. Another recent study examined the influence of EI on academic motivation, finding a significant positive association and gender differences in EI, with males scoring higher in intrapersonal EI, emotion management, and overall EI, and females excelling in interpersonal EI (^{Taibolatov et al., 2024}).

Interestingly, both studies used ^{Lyusin's Emotional Intelligence scale (EmIn), developed by Russian psychologist D. Lyusin in 2006}, to measure self-perceived ability EI. The EmIn consists of 46 Likert-style statements designed to elicit participants' level of agreement with the EI-focused statements (Lyusin, 2006). The questionnaire includes positively and negatively worded items, categorized into subscales to measure five self-reported EI skills and competencies: Understanding Own Emotions, Managing Own Emotions, Controlling Own Expressions, Understanding Others's Emotions, and Managing Others's Emotions.

The recent publication of ^{Adilzhanova et al.'s (2024}) and ^{Taibolatov et al.'s (2024}) studies suggests that the EmIn is being progressively adopted in Kazakhstani research and will be used in future studies in this context. Therefore, an analysis of the psychometric characteristics of the Kazakhstani version of the instrument is warranted. However, there has been a lack of sound measurement practice among academics and scholars regarding the use of this instrument. For example, ^{Lyusin's (2006}) original study lacks sufficient evidence for construct validity-the factor pattern matrix, based on a moderate sample size of N = 218, conflates the management of one's own emotions with the management of other's emotions in a single factor (the third factor, p. 12-13) calling to question the dimensionality of the proposed scale. Additionally, the reliance on alpha reliability, the lack of discriminant validity assessment, and the poor model performance in Taibolatov et al.'s study demonstrate challenges in measuring EI. This limitation has hindered the reliable assessment of emotional self-perceived EI competencies among local students and has problematized the development of specialized programs and interventions that aim to cultivate EI that are in alignment with the cultural and linguistic context of the country.

The current study uses the EmIn to explore the dimensionality of Kazakhstani adolescents' EI to bridge the existing gap in empirical research on assessing EI. More specifically, this study aims to examine the validity of Lyusin's model of EI among a large sample of Kazakhstani adolescents and to examine the potential role of gender and language in the factors in the model. By examining the alignment between Lyusin's model and self-perceived EI competencies of Kazakhstani adolescents, the research intends to provide insights into the applicability and relevance of the model in the specific cultural context of Kazakhstan.

Participants

The sample comprised 658 adolescents with a mean age of 13.09 (SD = 0.97, skew = -0.15), with 242 males (37.4%) and 416 (63.2%) females. In terms of grade, 270 (41.0%) were in Grade 8, 226 (34.3%) in Grade 7, and 162 (24.6%) in Grade 6. The sampled students were based on a convenience sample. Most students (n = 624) were from urban locations, while 34 were from rural locations. Overall, participants in the study were sampled from 11 different districts in Kazakhstan including Aksu (n = 44), Aktobe (n = 67), Baykonys (n = 11), Druzhba (n = 1), Kostanay (n = 93), Kyzylzhar (n = 10), Leninskiyi (n = 3), Pavlodar (n = 387), Semey (n = 33), Terenol (n = 3), and Zhelezenskyi (n = 6). In terms of schools, student participants were drawn from 12 unique public schools with the following representation from anonymized school IDs 1 to 12, respectively: n = 67, 129, 33, 11, 44, 14, 94, 10, 3, 108, 139, and 6.

Instruments

All EI items were Likert-style with the following response options: 0 = Strongly disagree, 1 = Slightly disagree, 2 = Slightly agree, 3 = Strongly agree. After the deletion of several items due to poor functioning, the final instrument was comprised the following four factors and 19 items: Understanding Own Emotions (3 items, combined score averages for students were: M = 1.99, SD = 0.61), Managing Own Emotions (4 items, M = 1.77, SD = 0.66), Understanding Other's Emotions (8 items, M = 1.86, SD = 0.55), Managing Other's Emotions (4 items, M = 1.91, SD = 0.65) (inclusion of items for Controlling Own Expressions, did not result in sufficient model fit). Table 1 provides details as to all descriptive statistics.

Table 1. Descriptive Statistics for All EmIn Items. 

^Notes.R = reversed-scored item (as presented as semantically-reversed); Med. = median; Skew = skewness; ICC = intraclass correlation coefficient by 12 school clusters, as calculated with the R misty package’s (^{Yanagida, 2023}) multilevel.icc function; 0 = Strongly disagree, 1 = Rather disagree, 2 = Rather agree, 3 = Strongly agree; items retained in the final four-factor model in bold; alpha, omega[total], and omega[total, polychoric] reliability estimates are based on the bold items retained in the final model.

Descriptive statistics also revealed that there was very little variance in EI variables due to between-school effects. Specifically, ICCs for each item ranged between 0.00 and 0.04, with only one instance of 0.15 for the managing other's emotions item, "I can comfort my parents and friends when they are sad or worried" (Table 1). Overall, this meant that very little of the variance in EI attributes, as measured by the 46 different items, could be attributable to between-school effects. It should be noted that the questionnaire was available to students in either Russian or Kazakh languages. As no examples of the survey could be found in Kazakh, the survey was translated by the first author (a bi-lingual Russian and Kazakh speaker) from Russian (source) to Kazakh (target). Thereafter, the Kazakh translation was checked for translation flaws, awkward sentence structure, and usage by a native Kazakh-speaking colleague. All data preparatory steps and cleaning are detailed in the Supplementary Materials, R Code.

Data Analysis

RQ1: Measurement Model for EI of Kazakhstani Adolescents

Adequate model fit, discriminant validity, and minimum item-factor loadings were ultimately reached with a four-factor solution that maintained most of the pre-conceived items. Figure 1 provides a visual illustration of the Four-Factor model of EI.

Figure 1. Four-Factor elf-Other Understand-Manage Model of Emotional Intelligence. 

The measurement model exhibited adequate model fit across all indices using both maximum likelihood (ML) and mean and variance adjusted weighted least squares (WLMSV) estimation methods (Table 2). All inter-factor correlations and item-factor loadings for the full WLSMV-estimated measurement model are presented in Figured B1 (Appendix B).

Table 2. Model Fit Indices for the Four-Factor Self-Other Understand-Manage Model of EI. 

^Note.Chi-Sq. = the chi-square statistic; df = degrees of freedom for the model; Ratio = the chi-square to degrees-of-freedom ratio for the model; the p value associated with the ratio; aall fit statistics are “scaled” versions.

Table 3 presents the results of the HT-MT_.85 test for discriminant validity. As all estimates were under the standard limit of .85, all four factors in the model were deemed to be sufficiently discriminant from each other. In addition, all upper 95% confidence intervals for the inter-factor correlation coefficients were statistically significantly different from 1.00 (Figure 1, coefficients in brackets).

It should be noted that a large number of items were dropped from the originally conceived 46 items (from 46 to 19) due to reversed item-factor loadings in the CFA models. Additionally, fit was not achievable with the inclusion of Notably, none of the reverse-coded items (those presented semantically negative) were retained. Additionally, due poor model fit, the inclusion of the proposed Controlling Own Expressions, was also dropped from the final model.

Table 3. Results for the HT-MT.85 Tests for Discriminant Validity. 

^Note.Underst. = Understanding.

Given that the Self-Other Understand-Manage model exhibited adequate fit, sufficient discriminant validity, and theoretical relevance, it was selected as the final measurement model for the current study.

RQ2: Measurement Invariance for Gender and Language

The tests for configural invariance for gender and language revealed RMSEA values of .05 and .05, respectively. Therefore, this suggested that the general factor structure was equivalent for both male and female gender groups, and Kazakh and Russian language groups (note that gamma also acceptable at = .958 and .961). The subsequent tests for metric invariance were also met with the change in CFI (from the model with loadings not held equivalent and loadings held equivalent) estimated at .0001, and .0005, respectively (note delta gamma = .000, and .000, respectively; results available to three decimal places). Therefore, the tests for scalar equivalence were also met with the change in CFI (from the model with loadings held equivalent; and loadings and intercepts held equivalent) estimated at .0050, and .0013, respectively (note delta gamma = .002, and .001, respectively). As the delta gamma statistic was above .001, the criteria for scalar invariance across the Kazakh and Russian survey languages was not met. Therefore, only subsequent analysis of differences in latent factor means for gender was undertaken.

RQ3: Differences in latent factor means by gender

Table 4 presents the results of the latent factor means difference tests.

Table 4. Differences in Latent Factor Means by Gender. 

^Notes.Standardized estimates reflect effect of being female on the latent factor mean; d = Cohen’s d; statistically significant estimates (p < .05) in bold and underlined; se = standard error, z = z-score, and p = probability.

Analysis revealed that females exhibited slightly higher levels of Understanding Other's Emotions than males (d = .20, small, p = .003), while males exhibited slightly higher levels of Managing Own Emotions (d = -.21, small, p = .006). No statistically significant differences between males and females were found for the Understanding Own Emotions and Managing Other's Emotions latent variables.

The proposed model

This study found low levels of between-school variation in EI among Kazakhstani adolescents. The finding suggests that different school environments have a limited influence on the overall development of EI in the Kazakhstani adolescent school population. The low between-school variation in EI suggests that individual differences play a more significant role than school-level factors in shaping adolescents' EI. Therefore, factors beyond the school environment, such as family dynamics, cultural influences, and personal experiences may have a stronger impact on the development of EI for adolescents in the country (^{Azpiazu et al., 2023}).

The initial purpose for including items with both positive and negative wording in the current study was to avoid acquiescence, affirmation, or agreement bias. However, ^{DeVellis (2012}) argues that the disadvantages of including negatively worded items outweigh any benefits (also see, ^{Johanson & Osborn, 2000}; ^{Motl et al., 2000}). Based on the findings of the current study, it appears that Kazakhstani adolescents may have experienced confusion or difficulty in responding to negatively-worded items related to EI. Therefore, in the future, researchers should consider excluding such items, especially when attempting to gauge EI among adolescents.

Results demonstrate especially strong positive relationships between Understanding Own Emotions and three other factors. Firstly, the coefficient of r = .81 (ML estimated coefficient) indicates a very strong positive correlation between Understanding Own Emotions and Managing Own Emotions, which suggests that individuals who possess a deep understanding of their own emotions are more likely to be effective in regulating their emotional responses, supportive of previous findings in the literature (e.g., ^{Harris et al., 2000}). In addition, the strong positive correlation (r = .70) between Understanding Own Emotions and Understanding Other's Emotions implies that individuals with a high level of emotional self-awareness also demonstrate a greater understanding of the emotions experienced by others. Individuals who possess a strong understanding of their own emotions also appear more capable of managing the emotions of others, as demonstrated by the strong correlation between Understanding Own Emotions and Managing Other's Emotions (r = .55).

The Understanding Own Emotions factor encompassed students' capacity to understand their sentiments toward others, the reasons behind their anger, and insight into their own emotional state. These results suggest that there may exist a cluster of emotional understanding pertaining to recognizing one's anger, sentiment toward others, and reasons for attraction/repulsion to others. It may be that these particular personal emotional capacities play an important part in determining how adolescents manage their own emotions, understand others' emotions, and ultimately manage other's emotions. The general pattern of inter-factor correlations in the current study are reflected in a similar first-order measurement model study by ^{Barchard and Christensen (2007}). In that study, among the equivalent four (self-other understand-manage) trait EI factors, the strongest correlation was between Recognition of Emotion in the Self and Regulation of Emotion in the Self (r = .48, p < .01). The second strongest correlation was between Recognition of Emotion in the Self and Recognition of Emotion in Others (r = .41, p < .01), while the correlation between Recognition of Emotion in the Self and Regulation of Emotion in Others was slightly smaller at r = .35 (p < .01). While these patterns were not discussed explicitly in that paper, they point to the centrality of the Understanding Own Emotions subscale of EI and provide some evidence of construct validity. The pattern of the three correlations in Barchard and Christensen (2007) and the current study are somewhat expected-one would assume more convergence between the understanding and regulation of self and more divergence between the understanding of self and regulation of others.

The factor Managing Own Emotions has moderate correlations with factors related to understanding and managing others' emotions (r =.49 and .52, respectively). This suggests that individuals who are adept at managing their emotions also tend to have a deeper understanding of others' emotions and the capacity to manage the emotions of others. In general, this suggests that individuals who are proficient in understanding the emotions of others also demonstrate a deep understanding of their own emotional experiences, and generally excel in assisting and supporting others in managing their emotions. It may also be that the ability to comprehend and relate to others' emotions may enhance self-awareness, provide insights into one's own emotional state, enabling more empathy and more appropriate emotional responses to the emotional needs of others (see ^{Juckel et al., 2018}).

Effects of gender on EI

The results indicated that the proposed model's general factor structure was comparable for both male and female groups. Therefore, furthermore, subsequent analyses were conducted to examine differences in latent factors between male and female participants. The results, presented in Table 4, indicate that there were slight variations in two out of the four factors based on gender. The female cohort demonstrated a slightly higher level of Understanding Other's Emotions, while male participants exhibited slightly higher levels on Managing Own Emotions factor. No significant differences were found in Managing Other's Emotions or in Understanding Own Emotions. These findings suggest that there are subtle gender differences in certain aspects of own's understanding and managing emotions. Females may have a slight advantage in understanding the emotions of others. This finding is supported by research that suggested that girls exhibit slightly more emotional empathy than boys (^{Watanabe et al., 2020}), while males may exhibit slightly higher levels of self-emotion management (see ^{Cabello et al., 2016}, ^{Ngondi et al., 2020}, for similar findings). However, it is important to note that the effect sizes for these differences were small, indicating that the practical significance of these findings may be limited.

Future research directions

Currently there are over 30 distinct and widely used measurement approaches for trait EI (^{O'Connor et al., 2019}). However, one fundamental similarity among the EI models is the recognition of emotional awareness and regulation as essential components of EI ability. They all emphasize the capacity to perceive, understand, and manage one's own emotions as well as recognize and empathize with the emotions of others. This focus on emotional awareness and regulation reflects the core notion that EI involves effectively navigating and utilizing emotions in social interactions and personal well-being. Therefore, our model might represent a first step for validly and reliably measuring EI among adolescents in Central Asia. However, some models rely on self-report questionnaires where individuals rate their emotional experiences and responses. Such models significantly diverge from Self-Other Understand-Manage related models. For example, ^{Schutte Self-Report Emotional Intelligence Test (SREIT; Schutte et al., 1998}) assesses EI by including factors related to optimism, emotion assessment, and social skills, whereas the Emotional and Social Competence Inventory (ESCI) model includes dimensions related to self and social awareness. Therefore, new lines of research into EI in Central could also include optimism, emotional assessment, and social skills. Researchers have sought to develop culturally sensitive EI measures to ensure the universality of the construct across diverse populations (^{Crnički, 2023}; ^{Mesquita & Frijda, 1992}), and we encourage more research in this area. Furthermore, cross-cultural research that incorporates an examination of multiple post-Soviet and Central Asian countries, such as Ukraine and Kyrgyzstan, would augur well.