Academic engagement has been assessed according to different theoretical standpoints (Alrashidi et al., 2016). On the one hand, one may refer to Fredicks’ model Fredricks et al. (2004), also known as the North American model, composed of three dimensions that measure academic engagement from a positive approach (i.e., behavioral, emotional and cognitive). Similarly, the model proposed by Schaufeli et al. (2002), known as the European model, measures academic engagement positively but using three other dimensions (i.e., absorption, vigor and dedication). However, Skinner et al. (2008) proposed that academic engagement should be measured from both a positive (i.e., engagement) and a negative (i.e., disaffection) perspective, each composed of two dimensions: cognitive and emotional. Thus, academic engagement would encompass four factors in total: behavioral engagement (i.e., persistence, attention and effort during the onset and execution of academic activities; Sinval et al., 2021), emotional engagement (i.e., the students’ positive and negative emotional responses to the learning process and class activities; Manwaring et al., 2017), behavioral disaffection (i.e., passivity and low student participation; Skinner et al., 2008) and emotional disaffection (i.e., boredom, anxiety and frustration experienced in the classroom; Skinner et al., 2008). We consider this theory to be of interest for our research, as it analyzes engagement from two standpoints, i.e., the positive and the negative.

Different studies (Liu et al., 2018; Zhen et al., 2020; Kuo et al., 2021) have highlighted the importance of academic self-efficacy on academic engagement after having analyzed the North American model by Fredricks et al. (2004). Other works (Azila-Gbettor and Abiemo, 2020; Zhao et al., 2021; Cai et al., 2022), following the European model by Schaufeli et al. (2002), have found that academic self-efficacy is a predictor of academic engagement. As can be seen, the scientific literature has already attested that academic self-efficacy predicts academic engagement, however, we are not aware of any studies that have delved into this relationship by having analyzed academic engagement from both positive and negative approaches. Moreover, research works that analyze the relationship between academic engagement and academic self-efficacy are virtually non-existent in the Mexican context. Therefore, and considering that academic self-efficacy positively predicts academic engagement, it is convenient to study how it relates to academic disaffection.

The theoretical construct of academic self-efficacy stems from the Social Cognitive Theory (Bandura, 1997) and is defined as the assumptions that students have about their own capabilities to organize and carry out the activities that are necessary to attain previously envisioned educational expectations (Gutiérrez and Tomás, 2019). Academic self-efficacy has a direct effect on satisfaction and the continuity of studies, thus helping prevent university students from dropping out (Lent et al., 2017). In this sense, students with high academic self-efficacy work harder, use more effective methods to deal with academic difficulties, are more willing to participate in learning activities, and have better performance compared to students with low academic self-efficacy (Fernandez-Rio et al., 2017; Gebauer et al., 2020). Hence, this variable (i.e., academic self-efficacy) relates to the students’ capabilities to identify opportunities and drawbacks in the environment, without prejudice to their engagement and motivation (Oriol-Granado et al., 2017). To our knowledge, although several studies have related academic self-efficacy to emotional intelligence (Bidhendi et al., 2018; Saeed and Ahmad, 2020; Pérez-González et al., 2022), only one study has analyzed the relationship between academic self-efficacy and emotional intelligence using structural equation modeling (SEM; Bonilla-Yucailla et al., 2022). These authors found that emotional intelligence positively and significantly predicts academic self-efficacy, with academic engagement playing a mediating role. However, said study only measured academic engagement from a positive perspective, so we consider it would be helpful to analyze this relationship taking the two possible academic engagement perspectives into consideration, i.e., positive and negative. In addition, given that the scientific literature fosters the role of academic self-efficacy as a predictor of academic engagement (Casas and Blanco-Blanco, 2017; Oriol-Granado et al., 2017; She et al., 2021; Akanni, 2022; Azila-Gbettor et al., 2022), we consider it interesting to study the relationships between these three constructs while analyzing the mediating role of academic self-efficacy between emotional intelligence and academic engagement.

A substantial body of literature has underscored the importance of emotional intelligence in the occurrence of positive emotional responses during the learning process (Thomas and Heath, 2022). In the university context, emotional intelligence has stood out as an adequate tool for coping with stressful situations, and for achieving a successful academic performance and emotional well-being (Parhiala et al., 2018; Guil et al., 2021). Salovey et al. (1995) suggest that emotional intelligence is composed of three dimensions and that it can be defined as an individual’s capacity to address (emotional attention), understand (emotional clarity), and alter (emotional repair) their own emotional states. Different studies have emphasized the important role of Salovey’s theory in terms of engagement with learning, satisfaction, and academic performance (Baños et al., 2019; Fernández-Lasarte and Axpe Sáez, 2019; Supervía et al., 2019).

Emotional intelligence (i.e., emotional attention, emotional clarity and emotional repair) has been recently related to the academic engagement of secondary (junior high) school students (Supervía et al., 2019; Supervía and Bordás, 2019; Usán-Supervía et al., 2019) using the European model by Schaufeli (2013). Similarly, emotional intelligence (albeit unidimensionally measured) has also been related to the academic engagement of university students (Thomas and Allen, 2020; Thomas and Heath, 2022) according to Skinner’s model Skinner et al. (2008), in which academic engagement is addressed from a positive and negative perspective. In fact, to our knowledge, these (Thomas and Allen, 2020; Thomas and Heath, 2022) are the only studies that have related emotional intelligence to Skinner’s model of academic engagement; said studies, however, did not measure emotional intelligence from the dimensions of attention, clarity and repair, but rather unidimensionally. These works pointed out the importance of considering the positive effect of emotional intelligence on behavioral engagement and emotional engagement. Conversely, students with high emotional intelligence relate negatively to behavioral disaffection and emotional disaffection (Thomas and Heath, 2022). As can be seen, scientific works that relate the dimensions of emotional intelligence (emotional attention, emotional clarity and emotional repair) to the academic engagement model by Skinner et al. (2008) are non-existent. Therefore, we consider it interesting to thoroughly analyze how these two constructs relate within the Mexican context.

Having analyzed the scientific literature and observed the importance of academic engagement in the education of university students, we consider the predictive analysis of emotional intelligence and academic self-efficacy on academic engagement to be relevant, and that studying it should lead to learning improvements that benefit Mexican university students. In summary: on the one hand, even though studies have related academic self-efficacy to either the North American model (Liu et al., 2018; Zhen et al., 2020; Kuo et al., 2021) or the European model of academic engagement (Azila-Gbettor and Abiemo, 2020; Zhao et al., 2021; Cai et al., 2022), these have only analyzed academic engagement from a positive perspective, without delving into how academic self-efficacy relates to academic disaffection. On the other hand, while few studies are known to have analyzed emotional intelligence unidimensionally with academic engagement and disaffection (Thomas and Allen, 2020; Thomas and Heath, 2022), others have analyzed how the dimensions of emotional intelligence correlate according to the European model of academic engagement, i.e., without measuring academic disaffection (Supervía et al., 2019; Supervía and Bordás, 2019; Usán-Supervía et al., 2019. Furthermore, as far as we are aware, the analysis of the role of academic self-efficacy as a mediator between emotional intelligence and academic engagement and disaffection has not been addressed. Therefore, this study intends to be a relevant contribution to the understanding of the relationship between the dimensions of emotional intelligence, academic self-efficacy and academic engagement, especially by incorporating the variables of academic disaffection, given that studies that have taken academic disaffection into consideration are scarce both in the Mexican university context and worldwide. Figure 1 shows the hypothesized model of this research to examine the aforementioned relationships. Thus, the purpose of this research is to analyze the mediation of academic self-efficacy between emotional intelligence and academic engagement. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) initiative (Von Elm et al., 2008) was used for the description of the study.

The design of this research was descriptive, observational, cross-sectional and non-randomized. Participants study at either of the three campuses of the Faculty of Sport of the Autonomous University of Baja California, Mexico (i.e., Campus Tijuana, Campus Mexicali and Campus Ensenada). Inclusion criteria for participation in the study were the following: (i) to be enrolled in the Bachelor’s degree in Physical Activity and Sport at either of the three aforementioned campuses and attend classes regularly; (ii) to give their informed consent for data collection; (iii) to duly fill out the data collection form with the different scales. An a priori analysis of the necessary sampling size was conducted to provide an answer to the proposed objective, considering a structural equation model (SEM) composed of six latent variables and 49 observable variables. The analysis was conducted using the Free Statistics Calculator v.4.0 software (Soper, 2022) and a minimum of 1,164 participants was calculated to detect effect sizes (f²) = 0.166), with a statistical power of 0.99% and a significance level of α = 0.05. The total number of participants was 1,164 university students of the Bachelor’s degree in Physical Education (349 women, 810 men and 5 other) from the Tijuana (n = 577), Mexicali (n = 357) and Ensenada (n = 230) University Campuses. Students were aged between 17 and 50 years old (M = 21.21; SD = 3.26). It must be mentioned that 14 individuals did not give their consent to participate in the research and that 29 questionnaires were discarded because they were filled incorrectly. Lastly, there were no lost values in the responses included in the study.

The general director of the Faculty of Sport and the deputy directors of the three campuses (Tijuana, Mexicali and Ensenada) of the Autonomous University of Baja California were contacted to inform them of the purpose of the research and to request permission to apply the questionnaires. Upon granted authorization, an online questionnaire was administered in person in the institution’s computer room in March 2022. Participants were taught how to use the scales and informed about the importance of the research, that their answers were anonymous and would therefore not affect their scores, and that they could abandon the study at any time if they so desired. All participants included in the study gave their prior consent for their responses to be used. The research protocol was approved by the Bioethics Committee of the University of Almeria (Ref:UALBIO2020/019).

Descriptive statistics, the correlations among variables and McDonald’s omega (ω) coefficient (McDonald, 1970) were initially calculated using SPSS v.28 for each dimension, assuming that values >0.70 indicate adequate reliability (Viladrich et al., 2017). Main analyses were performed using AMOS v.26, and a two-step SEM with latent variables following Kline (2016) was calculated to evaluate the predictive relationships of the dimensions of emotional intelligence on the dimensions of academic engagement, analyzing the mediating role of academic self-efficacy. In the first step of the SEM, known as the measurement model, the robustness of the bidirectional relationships between the model variables was assessed. In the second step, the predictive effects between the variables were examined, with the SEM effects being controlled according to the gender and campus of origin of the students. Due to the violation of the multivariate normality assumption (Mardia’s coefficient = 138.61; p < 0.001), the analysis was conducted using the maximum likelihood estimation method and the 5,000-iteration bootstrapping procedure (Kline, 2016). The SEM were assessed with the following goodness-of-fit indices: values of the chi-square/degrees of freedom ratio (χ²/gL), Comparative Fit Index (CFI), Tucker–Lewis Index (TLI), Root Mean Square Error of Approximation (RMSEA) with a confidence interval of 90% (CI), and Standardized Root Mean Square (SRMR). For the χ²/gL ratio, values <2.0 or < 5.0 are, respectively, considered excellent (Tabachnick and Fidell, 2019) or acceptable (Hu and Bentler, 1999); for the CFI and the TLI, values >0.95 are deemed excellent, whereas the range between 0.90 and 0.95 is considered acceptable; for RMSEA and SRMR, values <0.06 are considered excellent (Hu and Bentler, 1999; Marsh et al., 2004). The internal consistency of each instrument was assessed using McDonald’s ω, considering that values >0.70 are deemed acceptable. In this study, three factors of academic engagement (i.e., behavioral engagement, emotional engagement, and behavioral disaffection) showed reliability values <0.70, however, according to Taylor et al. (2008), these can be considered marginally acceptable due to the small number of items (three) in each dimension.

Descriptive statistics and correlations between the different variables are shown in Table 1.

The SEM showed acceptable goodness-of-fit indices during step 1: χ²/gL = 2.71, p < 0.001; CFI = 0.96; TLI = 0.95; RMSEA = 0.038 (90%CI = 0.035;0.041; p_close = 0.999), SRMR = 0.039. During step 2, the SEM showed a similar and acceptable fit: χ²/gL = 2.71, p < 0.001; CFI = 0.96; TLI = 0.95; RMSEA = 0.038 (90%CI = 0.035;0.041; p_close = 0.999), SRMR = 0.039. The model was controlled by the sex and campus of origin variables and reached an explained variance of 37% for behavioral engagement, 23% for emotional engagement, 12% for behavioral disaffection, 8% for emotional disaffection and 18% for academic self-efficacy (Figure 2). The relationships among the dimensions of emotional intelligence (i.e., emotional attention, emotional clarity and emotional repair), academic self-efficacy and the four dimensions of academic engagement (i.e., behavioral engagement, emotional engagement, behavioral disaffection and emotional disaffection) can be attested in Figure 2 and Table 2.

Figure 2 outlines the SEM and demonstrates that emotional clarity has a direct, positive, and significant relationship with academic self-efficacy. On the other hand, emotional repair correlates directly, positively, and significantly with emotional engagement and behavioral engagement, and negatively with emotional disaffection. Emotional attention shows a direct, positive and significant relationship only with behavioral engagement. Likewise, academic self-efficacy has a positive and significant direct effect on behavioral engagement and emotional engagement, and a negative and significant effect on behavioral disaffection and emotional disaffection. The mediating role of academic self-efficacy must be highlighted, as it indirectly, significantly and positively relates emotional clarity with behavioral engagement and emotional engagement, and significantly and negatively relates emotional clarity with behavioral disaffection and emotional disaffection. In addition, academic self-efficacy acts as a positive and significant mediating variable between emotional repair and behavioral engagement and between emotional repair and emotional engagement, and as a negative mediator between emotional repair and behavioral disaffection and between emotional repair and emotional disaffection. Lastly, the CI (95%) of R² can be attested in Figure 2, thereby confirming that these can be considered ES measurements (Dominguez-Lara, 2017).

The results of this research underscore the importance of emotional clarity, emotional repair and especially academic self-efficacy in the development of academic engagement in Mexican students. Therefore, those responsible for education matters should devise and establish psychoeducation groups that strengthen the academic self-efficacy of university students (Özhan, 2021). Accordingly, teachers should use different learning strategies to provide students with numerous learning tools and techniques, such as the ones proposed by different researchers (Baños et al., 2021, 2022; Gómez López et al., 2022; González-Fernández et al., 2022; Martín-Moya et al., 2022; Pérez-Suasnavas and Cela, 2022; Sánchez-Cabrero and Pericacho-Gómez, 2022;). In this way, students will be able to try and experience different educational resources, analyzing which of them will be more effective in their learning. To this end, the psycho-pedagogical areas can organize workshops for teachers to support the teaching and learning processes by creating positive classroom environments, acknowledging the motivations of each student and fostering academic engagement in them, understanding that, if they learn how to be efficient, they will obtain good results (Oriol-Granado et al., 2017; Pan, 2022). In other words, when university students develop academic self-efficacy, they will not only see their academic engagement improve (Álvarez-Pérez et al., 2021), but also feel efficient during their internships and in their future jobs (Martínez-Martínez and Ventura, 2020; Hong et al., 2021).

As a final consideration, students should be first taught how to be efficient not at the university stage, but in earlier years. Significant time must be allocated to inform middle and high school students about career guidance and job opportunities, and to teach them how to be efficient. However, students perceive that some schools do not prepare them adequately for higher education or do not provide them with the skills and information needed to thrive in higher education (Thomas and Maree, 2022).