Among the various theoretical paradigms used to approach emotional intelligence, multiple measures have been proposed. Questionnaires like the Trait Emotional Intelligence Questionnaire (TEIQue, Petrides, 2001), the Emotional Quotient Inventory (EQ-i, Bar-On, 1997), or the Mayer-Salovey-Caruso Emotional Intelligence Test (MSCEIT, Mayer et al., 2002), have also been adapted for younger populations for younger populations, including the TEIQue-Child Form (Mavroveli et al., 2008), EQ-i Youth Version (Bar-On & Parker, 2000), and the MSCEIT-Young Version (Rivers et al., 2012).

Some of these adaptations, despite being translated into Spanish, such as the EQ-i Youth Version (Ferrandiz et al., 2012), are directly derived from the adult questionnaire, without considering the differences between the emotional worlds of children and adults. Others, like the MSCEIT or the TEIQue, do not encompass earlier ages corresponding to the initial years of primary school or earlier. Additional questionnaires, such as the Assessment of Children’s Emotional Skills (ACES, Schultz et al., 2004), though assessed in younger children, concentrate on specific aspects of emotional competence, such as emotional recognition, and do not offer a comprehensive measure of emotional competence.

Recent works, like those presented by Pérez-González et al. (2022), and Bisquerra and López-Cassà (2021), which offer a synthesis of instruments for assessing emotional competencies, also highlight the need for validated instruments suitable for accurately evaluating emotional competencies within the early years of primary education. Furthermore, it is crucial that these instruments be tailored to the Hispanic context, offering valuable measures validated or developed based on samples of Spanish-speaking children.

To propose novel approaches to evaluate emotional competencies in Spanish children, the Psychopedagogical Guidance Research Group of the University of Barcelona (GROP, by its acronym in Catalan) has developed the Emotional Development Questionnaire for Primary Education (CDE-9-13, Pérez-Escoda et al., 2021), a questionnaire framed within Bisquerra and Pérez-Escoda’s model of emotional competence. This model defines emotional competencies as “a set of abilities, knowledge, skills, aptitudes, attitudes, and values required to understand, express, and appropriately regulate emotional experiences” (Bisquerra & Pérez-Escoda, 2007, p.69). Although framed within the theoretical perspective of emotional intelligence, this model also possesses an open character and considers other perspectives such as Gardner’s multiple intelligences theory (2000), neuroscience, positive psychology, and humanistic pedagogy. The model considers five basic emotional competencies: emotional awareness, emotional regulation, emotional autonomy, social competence, and competencies for life and well-being.

Emotional awareness involves the ability to recognize and understand one’s own emotions as well as the emotions of others, including the capacity to perceive the emotional atmosphere of a particular context. Emotional regulation encompasses the skill to effectively manage and control emotions. This includes recognizing the connection between emotions, thoughts, and behaviors, utilizing effective coping strategies, and being able to generate positive emotions within oneself, among other factors. Emotional autonomy encompasses various traits and aspects related to the self-management of emotions, including self-esteem, a positive outlook on life, responsibility, critical analysis of social norms, and personal self-efficacy. Social competence is about the ability to establish and maintain positive relationships with others, entailing mastering social skills, effective communication, respect, and pro-social attitudes, among other qualities. Lastly, competencies for life and well-being pertain to the ability to adopt appropriate and responsible behaviors to effectively navigate daily challenges as well as extraordinary life circumstances.

Another recent questionnaire based on this same theoretical framework is the Emotional Competence Assessment Questionnaire (ECAQ, Bartroli et al., 2022), which provided psychometric evidence to measure emotional competencies in small children between 3 and 5 years old. However, neither of these recent instruments covers the first cycle of primary education.

With the aim of addressing the research gap in evaluating emotional competencies in children during the first cycle of primary education, considering a relevant theoretical framework within the Hispanic context, and providing an easy-to-use and concise measure for teachers, the present study introduces a new instrument—the Emotional Development on Early Primary Education Scale (EDEPES). This scale is based on the theoretical model of Bisquerra and Pérez-Escoda (2007) and the GROP.

The specific objectives of the present study are as follows: (1) Investigate the EDEPES initial factorial structure through an exploratory factor analysis (EFA). (2) Conduct a confirmatory factor analysis (CFA) to scrutinize the factorial structure, comparing the results obtained through EFA with the model of emotional competence proposed by Bisquerra and Pérez-Escoda (2007). (3) Develop and propose the final form for the EDEPES. (4) Test the reliability of the final form of the EDEPES.

(5) Examine measurement invariance and potential gender differences across all emotional competencies within the different courses of the first cycle of primary school education. (6) Assess criterion validity by exploring the correlations between emotional competencies, children’s anxiety levels, and academic performance, as these variables have been observed to have relevant links with children’s emotional intelligence.

The sample considered for the factor analysis comprised 1113 primary school students aged 5 to 8 (M = 6.37; SD = .591), from three different Spanish autonomous communities: 90.3% from Catalonia, 1.7% from Navarre, and 8% from the Basque Country. Among them, 547 students were in their first year of primary school (48.8% girls and 51.2% boys), while 566 students were in their second year (50.2% girls and 49.8% boys). Approximately 19.9% of the schools where the students were enrolled were located in rural areas, while 80.1% were situated in urban areas. The minimum sample size was determined based on the recommended item-observation ratios of MacCallum et al. (1999), which suggest ratios between 1:5 and 1:10 to ensure sufficient sample power for factor analysis. Additionally, the suggestions proposed by Hu and Bentler (1999) recommended a minimum sample size of 250 or more. To assess test-retest reliability and criterion validity, a subsample of 430 children (M = 6.34; SD = .571) answered all the questionnaires a second time within a 6-month interval. This subsample comprised 233 students in their first year of primary school (54.5% girls and 45.5% boys) and 197 students in their second year (43.7% girls and 56.3% boys).

Emotional development on early primary education scale (EDEPES)

The initial version of the EDEPES, tested in the present article, consisted of 41 items, assessed using a 5-point Likert scale ranging from 0 (never) to 4 (almost always). Taking inspiration from the López-Cassà (2007) observation scale for emotional competence, a non-validated scale designed to propose a preliminary measure of emotional competencies within the model of Bisquerra and Pérez-Escoda (2007), a pool of 83 items was a priori considered. Through careful evaluation by experts, this item pool was subsequently reduced to 41 items that were deemed the most representative of the emotional competencies intended to be assessed. The group of experts consisted of ten professionals from different Spanish universities who had a deep understanding of the theoretical model within the questionnaire.

The preliminary item distribution of the EDEPES aimed to evaluate the five emotional competencies included in the model of Bisquerra and Pérez-Escoda (2007): emotional awareness (17 items), composed by the subscales of emotional expression, emotional recognition, and emotional ambivalence, emotional regulation (6 items), emotional autonomy (7 items), social competence (7 items), and life and well-being competencies (4 items). All the items were formulated in Spanish and underwent preliminary assessment in a pilot test to evaluate language-related issues.

The questionnaire was designed for teachers to fill out by observing and inquiring about individual children’s responses and behaviors. The assessment targeted children aged 5 to 8, corresponding to the first cycle of primary education in Spain. The preliminary scale aimed to generate scores for each dimension, as well as for the total measure of emotional competence. Example items include “The student is able to look for solutions to problems with the help of the teacher” and “The student can acknowledge his or her personal qualities”. The internal consistency of each dimension and the overall scale will be reported in the results section, as well as the full report of psychometric properties.

Child anxiety scale (CAS)

The CAS Spanish adaptation (Gillis, 2011) was utilized to assess anxiety levels and examine criterion validity. Children completed the scale, which consists of 20 dichotomous items (yes or no questions) yielding a single total score. Example items include, “Can you do things better than most children, or do other children do them better than you?” and “Do you think a lot of bad things happen to you, or do a few bad things happen to you?” Regarding internal consistency, the scale demonstrated a Kuder-Richardson formula 20 (KR20) of 0.58 at Time 1 and 0.66 at Time 2.

Academic performance

The school supplied data on students’ academic performance, indicated by their overall average grades on a scale from 0 to 10. The scores from all subjects included in the academic curriculum were utilized to compute the overall average. This quantitative measure was evaluated at two specific time points: the first trimester (Time 1) and the third and final trimester (Time 2), with an interval of approximately 6 months between the assessments.

First, after selecting items for the preliminary version of the EDEPES, the project obtained approval from the Ethics Committees of the University of Barcelona, the University of Lleida, and the Catalonian Health Service to ensure the protection of children’s data. Second, contact was established with several universities in other autonomous communities, disseminating information about the questionnaire and the project via email. Third, with the assistance of these institutions, researchers established contact with various educational centers. Detailed information about the questionnaire was sent to these centers by email, and a total of 23 schools from different autonomous communities expressed interest in the project. Six of these schools were private, while the others were public. Fourth, teachers within these centers who voluntarily agreed to participate were provided with prior instructions on accurately assessing students using the questionnaire. Seminars and online meetings were conducted to ensure that teachers from all centers had a clear understanding of how to administer the questionnaire. Fifth, parents were asked to provide and sign informed consent before the distribution of the questionnaire. Sixth, after obtaining parental consent, data collection took place online at the end of the first trimester and the end of the third trimester through a private platform from the University of Barcelona. Each student was assigned a code to guarantee anonymity throughout the entire process and facilitate identification for researchers at the two measurement points. In addition to the EDEPES, demographic questions, in conjunction with the CAS, were distributed to students. Teachers also provided academic performance reports at each assessment point. Seventh, at the conclusion of the study, a general report was provided to teachers and centers as a way of expressing gratitude for their collaboration.

Univariate and multivariate normality were assessed, also exploring the presence of outliers. Subsequently, the full sample was divided into two equal groups using the SOLOMON approach (Lorenzo-Seva, 2022), a splitting method specifically designed for factor analysis to create two equivalent subsamples.

Next, using one of the subsamples, an exploratory factor analysis (EFA) was performed to assess the factor validity of the EDEPES. After determining the recommended number of factors to retain and assigning the selected items to their respective dimensions, a confirmatory factor analysis (CFA) was conducted using the second subsample. Several models were evaluated, drawing inspiration from the initial theoretical model by Bisquerra and Pérez-Escoda (2007), the results of the EFA, and an approach suggested by the authors to strike a balance between the theoretical model and the EFA findings. Model fit was evaluated using the Chi-square test (χ2), Comparative Fit Index (CFI), Tucker-Lewis Index (TLI), Root Mean Square Error of Approximation (RMSEA), and Standard Root Mean Square Residuals (SRMR). CFI and TLI values above 0.90, and RMSEA and SRMR values below 0.08, indicated good model fit (Hu & Bentler, 1999). Differences between models were tested using the Chi-square difference test (Δχ2). Robust maximum likelihood estimation (MLM) was employed for all the models examined. The minimum sample size recommended for CFA using this estimator is 250 or more (Hu & Bentler, 1999).

First, univariate normality was examined, with item skewness and kurtosis values falling within the range of ±2, indicating univariate normality. Mardia’s test reported significant skewness (59351.98, p < .001) and kurtosis (249.69, p < .001) values, evidencing the absence of multivariate normality. No missing values were found in the full sample as the questionnaire was distributed online to control this aspect. Mahalanobis distance scores identified the presence of some outliers. However, none of these outliers were extreme values, so all of them were included in the analysis.

Subsequently, the matrix determinant, Kaiser-Meyer-Olkin measure of sampling adequacy (KMO), and Bartlett’s test were utilized to evaluate the appropriateness of conducting a factor analysis. The matrix determinant was 4.27e-15, Bartlett’s test was significant (1084.8, df = 40, p < .001), and KMO was .96. These values indicated that the data can be factorized. Before conducting the EFA, the sample was divided into two identical groups using the SOLOMON procedure (Lorenzo-Seva, 2022), resulting in a sample of 557 participants for the EFA and 556 participants for the CFA.

The next step involved performing the EFA. Parallel analysis (PA) was used to determine the number of recommended factors to retain, which was five. Oblimin rotation was also applied since the correlation between the factors was assumed, given the relationship established between the emotional competencies. The initial 5-factor solution yielded a model fit of χ2(820) = 18768.16, p < .001, TLI = .782, RMSEA = .09, RMSR = .04. Information pertaining to the preliminary 41-item EDEPES can be found in Supplemental Material 1, which includes data on item skewness and kurtosis values, factor loadings, explained variance, uniqueness, communalities, complexity values, and factor correlations.

Before proceeding with the CFA, items with factor loadings below .40 were decided to be eliminated. Specifically, item 5, which related to the expression of love, item 17 referring to emotional ambivalence, and item 39 from the life and well-being competencies dimension were removed.

The next step involved performing a CFA on the second subsample to test three different models: Model 1, which was a 5-factor model exclusively based on the theoretical model of emotional competence (Bisquerra & Pérez-Escoda, 2007); Model 2, a 4-factor model inspired from the EFA results; and Model 3, a 4-factor model aimed at finding a balance between the theoretical model and the model suggested by the EFA. In Model 3, it was decided to exclude the subdimension of emotional expression because the only items that showed adequate loadings were those related to negative emotions. Since this subdimension exhibited a negative polarization and did not capture the entire emotional experience, it was agreed to retain only the subdimension related to emotional recognition, which exhibited good factor loadings for both positive and negative emotions. Additionally, two items from the emotional autonomy subscale (“The student can appreciate the differences between people as positive aspects” and “The student can recognize their limitations”) were removed due to ambiguity. The wording of these questions led to confusion among experts regarding their appropriate subscale placement, and the results from the EFA also indicated similar factor loadings in two different subscales.

Results for Model 1 indicated robust model fit values of χ2(774) = 3901.12, p < .001, CFI = .73, TLI = .72, RMSEA = .10, SRMR = .09. Model 2 demonstrated improved model fit values of χ2(590) = 2671.10, p < .001, CFI = .79, TLI = .77, RMSEA = .09, SRMR = .08, with significant differences between the models (∆χ2 = 1264.80, ∆df = 184, p < .001). Model 3 exhibited the best model fit values, with χ2(346) = 1584.13, p < .001, CFI = .84, TLI = .82, RMSEA = .10, SRMR = .06, and significant differences compared to both Model 1 (Δχ2 = 2326.60, Δdf = 428, p < .001) and Model 2 (∆χ2 = 1086.70, ∆df = 244, p < .001). Consequently, Model 3 was selected to support the data and validate the structure of the questionnaire. After considering the modification index suggestions (Roseel, 2012), our final model demonstrated fair robust model fit values of χ2(336) = 1011.28, p < .001, CFI = .91, TLI = .90, RMSEA = .07, SRMR = .06.

Table 1 presents the items included in each emotional competence dimension of the EDEPES-28, along with their corresponding factor loadings and errors (SE). In our final model, we considered four emotional competencies: emotional awareness, emotional regulation, emotional autonomy, and social competence, along with a higher-order factor that measures global emotional competence.

Table 1
CFA factor loadings for EDEPES-28

Values for the four dimensions of the EDEPES-28 evidenced optimal reliability results for all the emotional competencies: emotional awareness (α = .90, ω = .89, CR = .89, AVE = .56), emotional regulation (α = .91, ω = .88, CR = .88, AVE = .61), emotional autonomy (α = .90, ω = .83, CR = .82, AVE = .49) and social competence ( = .92, ω = .90, CR = .90, AVE = .66). Only for emotional autonomy, AVE showed a little bit lower value than .50, reporting good values for the other indicators. Hierarchical omega for the global emotional competence also evidenced optimal internal consistency with a value of .87.

Intra-class correlations were tested to explore test-retest reliability for each emotional competence dimension: emotional awareness (.47 ICC [95% CI (.31, .58), F = 2.02, p < .001]), emotional regulation (.73 ICC [95% CI (.65, .80), F = 4.13, p < .001]), emotional autonomy (.66 ICC [95% CI (.54, .76), F = 3.50, p < .001]), social competence (.66 ICC [95% CI (.56, .74), F = 3.25, p < .001]) and the global emotional competence (.60 ICC [95% CI (.45, .71), F = 2.84, p < .001]). Cicchetti’s (1994) guidelines contemplate values between .40 and .59 as fair, so all emotional dimensions evidenced an adequate test-retest reliability.

Invariance related to sex was tested on our final 4-factor model of emotional competence, following the next steps and utilizing the CFA subsample (N = 556). First, separated models for girls and boys were evaluated, with both reporting adequate model fit (girls: χ2(336) = 687.55, p < .001, CFI = .92, TLI = .91, RMSEA = .07, SRMR = .08; boys: χ2(336) = 629.04, p < .001, CFI = .92, TLI = .91, RMSEA = .07, SRMR = .06). Figure 1 shows factor loadings for boys and girls, as well as items’ variances and covariances.

Figure 1
Model of the EDEPES-28 with factor loadings and covariances for both sexes

Second, configural equivalence was measured by performing a multigroup analysis, exhibiting again fair model fit values (χ2(670) = 1934.02, p < .001, CFI = .92, TLI = .91, RMSEA = .07, SRMR = .07). Third, equal factor loadings were assumed for both groups to test metric invariance, and forth, scalar invariance was evaluated fixing both loadings and intercepts. When comparing the models, using the Satorra-Bentler’s method (2001), no significant differences were found. Table 2 shows model fit values for the multigroup model and all the constrained models.

Table 2
Sex measurement invariance comparing differences between the multigroup model and the constrained models

Table 3
T-test results for sex differences per course

As the findings support measurement invariance, a comparison of mean differences between boys and girls was conducted to explore sex differences in each course for the latent variables of our four emotional competencies as well as the global emotional competence. In the first-year course, differences were found for the subscales of emotional regulation and social competence. Among second-year students, the differences were more prominent, indicating variations in all the subscales and the global emotional competence, with girls exhibiting higher scores. Table 3 presents means and standard deviations for boys and girls in both courses, along with t-test and effect size results.

Correlations between all the emotional competencies, academic achievement and children’s anxiety were explored to assess criterion validity, using a subsample of 430 students. Related with children’s anxiety, all the competencies at Time 1 showed significant negative associations with anxiety at Time 1 and Time 2 except for emotional awareness and emotional regulation, which only reported significant links at Time 1. Considering academic performance, all the emotional competencies evidenced positive correlations at both times. Discussed relationships are exposed in Table 4.

Table 4
Correlations for EDEPES-28 subscales and global emotional competence (Time 1) with anxiety and academic performance (Time 1 and Time 2)

Before presenting our conclusions, several limitations need to be considered. Firstly, we did not employ a random sampling method, and the children who participated were from pre-selected schools interested in establishing a collaboration agreement. Future research should aim to explore more diverse populations, including the evaluation of samples with special conditions, such as children with disabilities or specific diagnoses. This would allow us to observe whether the assessment of emotional competencies yields similar results in populations with special needs.

Secondly, children’s emotional competencies were assessed by teachers. Although the questionnaire is designed as a tool for teachers to evaluate emotional competencies, as self-report assessments are especially challenging in children of these ages, it is important to note that despite prior guidance to ensure proper use and reliable assessments, there remains a potential source of bias in this process. While we included a student self-report questionnaire for anxiety and assessed academic performance through evaluations by various teachers, future research should consider developing and incorporating separate questionnaires for students and parents, in addition to the teacher’s report. This approach would facilitate the comparison and contrast of results reported by different sources, providing a more comprehensive insight into the actual emotional competence of children.

Thirdly, criterion validity was only explored concerning the relationship with anxiety and academic performance. Although these variables were selected due to their relevant links with emotional competencies, future research should investigate associations between emotional competencies and other variables of interest, such as the use of specific emotion regulation strategies or children’s perception of the classroom climate.