Academic stress refers to students’ attitudes and behaviour toward the pressure of accomplishing academic demands and achievements (Karaman et al., 2019). Students’ perceived expectations of academic performance cause academic stress (Ang and Huan, 2006; Deb et al., 2014). Usually, students feel academic pressure to meet the expectation of their parents (Yi et al., 2006; Bedewy and Gabriel, 2015; Deb et al., 2015), academicians (Bedewy and Gabriel, 2015; Jayanthi et al., 2015) and others. So, it is crucial to identify whether the demographic factor of the students (e.g., gender) and their parents’ educational qualifications contribute to academic stress. Deb et al. (2015) concluded that the parent’s educational level and occupation impact their expectations over their children’s academic performance, which causes stress. Although Huan et al. (2006) denied any relationship between gender and academic stress, Misra and McKean (2000) and Jayanthi et al. (2015) reported higher anxiety and academic stress among female students compared to male students and Tangade et al. (2011) reported the opposite, i.e., higher academic stress among male students compared to their counterparts. Besides, the tire of education contributes to academic stress, where students studying in the final year of undergraduate level encounter a higher level of stress (Tangade et al., 2011). Extant literature (Kohn and Frazer, 1986; Misra and Castillo, 2004; Harikiran et al., 2012; Pulido-Martos et al., 2012) also explained that students’ perceived course load and difficulty, test anxiety, final grades, excessive homework, and financial hardship of family cause stress among the students. Khan et al. (2013) explained that severe academic stress and delayed identification hamper students’ ability to study efficiently with proper time management and consequently downgrade students’ academic performance.

Stress has become a widespread phenomenon during the COVID-19 pandemic. COVID-19 resulted in fear, mental stress, anxiety, and depression among university students in Bangladesh (Islam et al., 2020; Sifat, 2021). The pandemic has caused a higher level of mental stress and depression among students (Cao et al., 2020; Liu et al., 2020), which may lead to academic stress as mental health and stress levels impact academic performance (Sano et al., 2015). World Health Organization (WHO) and health experts have warned about the manifestation of these diseases during this situation which may lead to self-harm, chaos in the family, and suicidal cases (Sifat, 2020; WHO, 2020).

The term emotional intelligence has been used occasionally and inconsistent basis before 1990 (Mayer et al., 2011) and in 1990, Salovey and Mayer (1990) defined EI as the subset of social intelligence. The best-selling book ‘Emotional Intelligence’ by Goleman (1995) drew the attention of the term to many researchers (Uzzaman and Karim, 2017), which is an emerging area of study, especially for the educational, medical, and organizational behavior researchers (Chandra, 2021). According to Goleman (1995), some abilities, e.g., recognizing, understanding, and regulating emotions, have a positive relationship with people. In other words, EI is the ability to control interpersonal and intrapersonal emotions to create positive outcomes. Goleman (1995) argued that EI is twice as important as technical skills and more vital than intelligence quotient (IQ) in forecasting positive outcomes; and people should be evaluated not only by their intelligence or professional competence, but also by their own behaviors toward themselves and others.

EI has been found as an important interpreter of students’ activities, such as managing health and academic achievements. Higher levels of emotional intelligence help students to deal with the transition challenges from primary to secondary school (Qualter et al., 2007). Por et al. (2011) highlighted the potential value of EI for nursing students to cope actively and effectively in dealing with stress. Watson and Watson (2016) also indicated an explanatory capacity of emotional intelligence over the academic stress of a sample of American Hispanic students.

Ghosh (2014) concluded a positive correlation between emotional intelligence and academic achievements that might reduce students’ academic stress as academic achievement is negatively correlated with academic stress. Besides, Mohzan et al. (2013) and Pool and Qualter (2012) concluded that emotional intelligence is one of the key determinants of academic success. Iqbal et al. (2021) studied the impact of emotional intelligence on the academic performance of undergraduate students in crisis situation during COVID-19 and found a positive correlation between these two. Valenti et al. (2021) reported that individuals with higher EI are more productive in the workplace and can cope with stress in difficult situations. This study will examine whether such ability to cope with stress through EI is applicable for students to cope with academic stress during a crisis. In light of the discussion, the current study proposes the following hypothesis:

Social support refers to the experience of how an individual is appreciated, loved, and taken care of by the persons present in someone’s life (Gurung, 2006). Deficiency in social support causes depression, loneliness, anxiety, and many psychological problems (Eskin, 2003). Based on the findings of Stroebe and Stroebe (1996) on the impact of social support to reduce stress and other relevant literature, Snyder (1999) argue that emotionally intelligent individuals might access to rich social support network and it would help to reduce their stress. Besides, from previous literature (Yarcheski and Mahon, 1999; Soman et al., 2016; Poots and Cassidy, 2020), the role of social support as a mediator in reducing the adverse impact of stress is well established. Since one’s primary sources of reference are family and friends, support from these sources substantially impacts academic success (Yasin and Dzulkifli, 2010). Through this support, academic stress could be minimized to help them achieve their academic goals. Coffman and Gilligan (2002) concluded that social support largely contributes to stress management and higher life satisfaction among students.

Social distancing and isolation created the need for social support during this pandemic, and EI may contribute positively (Zysberg and Zisberg, 2020). Valenti et al. (2021) found a close relationship between EI and occupational stress mediated by social support during COVID-19. Ursin et al. (2020) explained that effective social support helps students to deal with stress. Besides, Rayle and Chung (2007) confirmed that social support helps to deal with academic stress more effectively. Based on the previous studies, the current study offers the following hypothesis (Figure 1):

In this study, a deductive approach is used with a quantitative research design because quantitative research is best for analyzing the relationships between quantitative data and looking into the causes of empirical results (Creswell, 2011; Sekaran and Bougie, 2016). The target respondents of this study are tertiary-level students studying business, engineering, social science, and science in Bangladesh.

The respondent pool was limited to the four public universities in Bangladesh: University of Dhaka, Jahangirnagar University, Rajshahi University, and Chittagong University. These institutions were chosen because they represent the nation’s largest and most prestigious public universities. A public university is a higher education institution funded and regulated by the government. Besides, these institutions hold a great deal of sway in higher education because of their self-regulated status, which gives them more freedom in running their own academic and administrative operations. Furthermore, respondents were only included if they willingly agreed to engage in the study, provided responses in English, and were prepared to spend 10 to 15 min completing the survey.

We chose graduate and undergraduate students in business, engineering, social science, and science as the target respondents for this study because these fields are offered at our chosen university. Besides, these disciplines are prevalent and extensively represented in the country’s higher education institutions. In addition, we selected university-level students because of their degree of maturity and ability to take an active role in the survey process. Furthermore, the recent pandemic has significantly impacted university students, especially those attending public universities.

Through a convenience sampling technique, 500 questionnaire survey forms were given via Qualtrics to the targeted respondents, who were assured that their identities would remain confidential and that their responses would be used solely for research reasons. Besides, they were also notified that their participation is voluntary and that they may opt-out at any moment. While conducting the survey, we followed up with calls and emails after 1 week, 1 month, and 2 months to ensure a prompt response and full collaboration. Despite the attempt, 59 students out of 500 did not return the survey, and 14 responses were eliminated.

Out of the 427 respondents, 57% were male, and the rest were female. 33% of respondents were graduate students, but the majority, 67%, were undergraduate students from various academic levels. Since undergraduate students make up the bulk of the intended audience, most respondents fall within the age range of 21 to 25. In addition, the remaining 12.1 and 11.7% of the respondents fall within the age range of (less than or equal to) 20 years and (more than or equal to) 26 years, respectively. On the other hand, roughly half of the respondents who participated in the survey completed their schooling in the city area, while the remainder were from towns and villages. Most of the respondents (54.3%) live with their parents or other family members (8.2%), while the rest of the students live in dorms (26.6%) and with friends in shared private houses (11%; Table 1).

This study employed the G*Power to conduct power analyses specific to model setups. As the model could only have two predictors, this study required 138 participants to test the hypothesis. Besides a sample size of 200 responses or above is appropriate for a sophisticated path model (Kline, 2015). However, we set out to collect data larger than the required number to increase the robustness of the results.

In the current study, the conceptual model was evaluated using SmartPLS 4 (4.0.8.9). PLS-SEM was chosen because CB-SEM is often impractical for estimating complicated models with many latent variables (Hair et al., 2017). A complicated model is characterized by the presence of more than seven variables (Hair et al., 2014). The model used in this current study contains 11 lower-order and three higher-order constructs. Besides, CB-SEM is incapable of running the formative construct. Therefore, PLS-SEM is considered more appropriate for this study.

Typically, PLS models are studied and interpreted in two steps (Hair et al., 2017): the measurement model to assess the reliability and validity, and the structural model to test the stated research hypotheses. In addition, SPSS 26 software was used to perform frequency analysis to acquire descriptive information on the demographic profile.

All unobserved constructs used in this study were measured using questions adapted from existing scales (Zimet et al., 1988; Wong and Law, 2002; Bedewy and Gabriel, 2015) and all employed construct items were measured on a Five Point Likert Scale (1 = strongly disagree and 5 = strongly agree; Table 4).

Hair et al. (2017) advised demonstrating convergent and construct validity when evaluating a measuring model. To establish convergent validity, we utilized the loading of each item on its associated variable, average variance extracted (AVE), and composite reliability (CR; Hair et al., 2014).

Table 5 shows that all item loadings exceeded the recommended value of 0.7. However, the items self-emotion 4 (0.564), regulation of emotion 3 (0.645), pressure to perform 4 (0.524), workload 3 (0.673) and time restraints 3 (0.645) were retained due to the acceptable values for AVE and CR (Hair et al., 2014). We found CR values above 0.70 (Chin, 2010) to be indicative of construct validity (Table 1). Table 5 shows that the AVE values for each latent variable in the measurement model surpassed the required value of 0.5 (Hair et al., 2017).

Besides, to test the discriminant validity, we used the Heterotrait–Monotrait ratio of correlations (HTMT; Zailani et al., 2019). The HTMT ratio should be lower than 0.85 to verify discriminant validity (Kline, 2016). Table 3 demonstrates that all values passed the HTMT ≤0.85 threshold, validating the measurement model’s discriminant validity.

In addition, we modeled emotional intelligence as the reflective-formative, and social support and perceived academic stress as the reflective-reflective, second-order constructs. Emotional intelligence is established through self-emotion, others’ emotions, significant others’ emotions, and emotion regulation. Since emotional intelligence is the ability to control one’s emotions to express them effectively and responsibly, all the dimensions are equally important to measure one’s EI. Therefore, it is not possible to quantify EI by omitting any lower-order constructs. In addition, emerging research on emotional intelligence considers EI as a formative rather than a reflective construct (Gentina et al., 2018; Iqbal et al., 2021).

Likewise, we modeled social support and perceived academic stress as reflective-reflective second-order constructs. Family support, friends’ support, and significant others’ support are the dimensions of social support, while pressure to perform, workload, academic self-perception, and time restraints are the dimensions of perceived academic stress. We used second order because it reduces the number of relationships (and thus the number of hypotheses to be tested) in the structural model, making the PLS Path model more compact and easier to comprehend (Hair et al., 2014). We employed a two-stage procedure to generate these second-order constructs (Becker et al., 2012), which included computing the latent variable scores and using these values as the items for the second-order constructs. To create these second-order constructs, we used a two-stage approach (Becker et al., 2012), which involves calculating the latent variable scores and using these scores as the items for the second-order constructs. The items used to measure this study’s two second-order reflective-reflective constructs are valid and reliable, as shown in Table 6.

Emotional intelligence, on the contrary, was modeled as Type II second-order measurement (reflective-formative); the accuracy of this formative measurement was assessed by looking at the weights, t-values, and the variance inflation factor (VIF). Table 6 displays the assessment for formative measurement.

Sarstedt et al. (2021) indicated that the path coefficients and their accompanying t-values were derived using 5,000 bootstrap subsamples to test the proposed study hypotheses. Table 7 and Figures 2, 3 demonstrate that EI has a statistically significant negative association with PAS (β = −0.408, t = 7.327, p < 0.05). Therefore, H1 is supported. Likewise, it is evident from Table 7 and Figure 2 that EI has a significant indirect effect on PAS by social support (SS; β = −0.060, t = 2.661, p < 0.05). Besides, the direct effect of EI on SS (β = 0.389, t = 8.046, p < 0.05) and SS on PAS (β = −0.152, t = 2.951, p < 0.05) are also significant. Thus, H2 is supported, and it can be said that SS mediates the relationship between EI and PAS, and the mediation is partial mediation.

Moreover, the study delves deeper and examines the mediating effect of family, friends and significant other support in the linkage between EI-PAS under social support. Table 7 and Figure 3 highlight that the indirect effect of EI on PAS mediated by friend’s support (FRS; β = −0.012, t = 0.803, p > 0.05, CI has value 0 included) and significant others support (SOS; β = −0.009, t = 0.568, p > 0.05, CI has value 0 included) is not statistically significant. So, FRS and SOS do not mediate the EI-PAS linkage. Thus H2(a) and H2(c) are not supported. However, Table 7 and Figure 3 highlight that the indirect effect of EI on PAS mediated by family support (FMS) is statistically significant (β = −0.012, t = 0.803, p > 0.05). Besides, the direct effect of EI on FMS (β = 0.320, p < 0.05) and FMS on PAS (β = −0.131, p < 0.05) are also significant. Thus, H2 (b) is supported, and it can be said that FMS mediates the relationship between EI and PAS, and the mediation is also partial mediation.

In the current study, the R 2 values, a measure of the accuracy of the model’s predictions, were 0.250 and 0.151 for perceived academic stress and social support, respectively. In addition, the Stone-Geisser Q 2 (cross-validated redundancy) value was computed to quantify the predictive relevance based on a blindfolding procedure in PLS (Teo et al., 2017; Saghapour et al., 2018). According to Chin (2010), the model exhibits predictive relevance if the value of Q 2 is greater than 0. The results show that the Q 2 values for perceived academic stress (0.210) and social support (0.136) are more than 0, verifying the predictive relevance of the endogenous variables in this study.

In structural models, on the other hand, the effect size is regarded as one of the most crucial indicators of statistical power. Therefore, Cohen (1988) established the thresholds for measuring the influence of effect size f 2 at the structural level: 0.02 for a small, 0.15 for a medium, and 0.35 for a large effect. Table 7 revealed that for f 2 emotional intelligence has a moderate effect on both perceived academic stress and social support, while social support has a small effect on perceived academic stress. Moreover, we evaluated the structural model’s collinearity. Results indicate that the VIF values of all predictor variables are below the cautious threshold of 5 (1.111 to 1.591), indicating the absence of multicollinearity problems (Hair et al., 2017).

A one-way between-groups analysis of variance (ANOVA) was performed to investigate the impact of gender (male and female) and current tier of academic study (1st year, 2nd year, 3rd year, 4th year, and Master) on emotional intelligence and perceived academic stress, respectively. Investigation of the skewness, kurtosis, and Shapiro–Wilk statistics showed that the assumption of normality for the dependent variable (emotional intelligence and perceived academic stress) was not violated. Besides, Levene’s statistics were not significant for both EI [F (1,424) = 1.788, p = 0.0.182] and PAS [F (4,421) = 2.114, p = 0.078]; hence the assumption of homogeneity variance can be made.

There was a statistically significant difference in EI score at the p < 0.05 level for the two groups: F (1,424) = 7.893, p = 0.002. Therefore, it is comprehended that male students have higher EI than females, as evident from this current study. The effect size, calculated using eta squared, was 0.018. Based on Cohen’s criteria, the effect size appears to be small. Therefore, despite the result showing that males and females have different emotional intelligence, the difference is negligible and limited to the particular context.

Likewise, there was a statistically significant difference at the p < 0.05 level in perceived academic stress score for five groups: F (4,421) = 5.374, p = 0.000. The effect size, calculated eta squared, was 0.048. Post-hoc comparisons using the Tukey HSD test revealed that the degree of academic stress in each year of schooling is very different. Student stress levels vary widely from first year to final year, third year to final year, and final year to post-graduation. Among the five groups, master’s students (M = 2.970, SD = 0.570) have fewer level of stress, whereas final-year students (M = 3.970, SD = 0.689) have higher stress levels. In particular, the level of perceived academic stress is significantly different between the second-year (M = 3.329, SD = 0.559) and master’s students, as well as final-year students and masters’ students. However, inter-year levels of perceived academic stress among the first-year (M = 3.049, SD = 0.509), third-year (M = 3.052, SD = 0.618) and master’s students did not differ significantly.

Despite plenty of research in the field of emotional intelligence, academic stress, and social support, there is limited evidence on the association between EI and academic stress among university students during the COVID-19 pandemic. Besides, no prior studies examine the mediating influence of social support on the relationship between emotional intelligence and academic stress, especially due to online classes during the COVID-19 pandemic. Therefore, this study analyzes the relationship between emotional intelligence and perceived academic stress and the mediation effect of social support (family, friends, and significant others) in the relationship between emotional intelligence and academic stress of tertiary-level students when they continue academic activities during a crisis like the COVID-19 pandemic.

In the first hypothesis, emotional intelligence is assumed to have a negative association with academic stress. The result validates the hypothesis that students with greater EI experience less stress (Por et al., 2011; Valenti et al., 2021). EI is a personal resource that helps to combat negative situations positively (Gupta and Kumar, 2010). Valenti et al. (2021) also reported a similar result on occupational stress and EI during a challenging time, and MacCann et al. (2020) underscored that EI is a vital tool to cope with students’ stress and negative emotions. Emotional intelligence is thus a protective personal resource for students navigating academic stress, as it enables them to effectively manage and comprehend their emotions, nurturing positive relationships with others while coping with negative emotions.

The second hypothesis hypothesized that social support mediates the negative association between emotional intelligence and academic stress caused by online classes. The hypothesis was tested, and the results revealed that social support mediates the connection between EI and PAS due to online classes. Though academic stress in the context of pandemic-induced online classes has not yet been investigated, prior research has examined the mediating effect of social support in the EI and perceived stress (Ursin et al., 2020; Zysberg and Zisberg, 2020; Valenti et al., 2021) and found that social support mediates the EI and perceived stress linkage. Similarly, this study’s results suggest that students can rely on social support to help them cope with academic stress in the face of unforeseen circumstances.

Students’ communities are more susceptible to stress and trauma than working professionals because of the understanding of subjective well-being of life and EI increase with age (Chen et al., 2016). While pursuing tertiary education, they mostly migrate from rural to urban areas and stay away from family members, making their life more stressful. Under these circumstances, social support through frequent interactions, positive feedback, and counselling can drive down stress and improve mental health. They may also develop emotional intelligence through this process because they might better regulate disruptive emotions.

Under social support in the second hypothesis, the study also investigates the specific mediating effect of familial support, friend support, and significant other support on the relationship between EI and PAS. The results indicate that family support significantly mediates the relationship between emotional intelligence and perceived academic stress in the context of online classes. However, it was found that the mediating effects of friend support and significant other support were not statistically significant. These findings are relevant for students in public universities in Bangladesh, many of whom encounter economic challenges and lack access to necessary resources for online learning. The pandemic exacerbated these difficulties, as students struggled with limited access to smartphones, personal computers, and related equipment. The burden imposed on students was alleviated to a greater extent when they received support from their families, who played a crucial role in providing necessary resources and fostering an emotionally stable environment. In contrast, the limited social detachment caused by the pandemic led to decreased access to friends and others, resulting in insignificant mediating effects of friend and significant other support. These findings highlight the pivotal role of family support in mitigating academic stress and underscore the impact of economic constraints on students’ online learning experiences.

In addition to the hypotheses, the study provides some insightful findings. It is improbable for all students to possess an equivalent level of EI, leading them to react distinctively in response to stressful situations encountered throughout their academic journey and challenging periods. Therefore, the study examined levels of EI between male and female students. The result reveals that male students possess a significantly higher EI than female students, indicating the higher vulnerability of female students during a crisis that demands more empathetic interactions. However, the effect size of the difference of EI between the male and female is very small (η p 2 = 0.018), implying that the generalizability of the findings is limited within the Bangladeshi context.

Likewise, the study also examined the degree of academic stress in each year of schooling at the tertiary level. The result reveals that student stress levels vary widely from first year to final year, third year to final year, and final year to post-graduation. The effect size of the differences is moderate (η p 2 = 0.048), which implies substantial disparities between student years in school and their perspectives on academic stress.

In a world of growing stress arising out of many academic, environmental, socio-cultural, and psychological factors (Chandra, 2021), it is essential to promote the students’ EI. Promoting skills related to emotion management of students will enable them to achieve better academic results (MacCann et al., 2011). Since stress grows from uncontrolled factors, actions are needed to harness EI skills. EI training to be included in tertiary education curriculum that can help students combat growing stress levels in their academic life. Since EI is more of learning skills than inherited, training can boost EI. Higher education institutions can also practice assessing students’ EI and stress levels to design coping mechanisms for different groups of students. This will result in better stress management at the institutional level, protecting vulnerable students from distress and a tendency to commit suicide. Institutions can also convey the assessment results to the concerned family members to take corrective measures for handling those with higher levels of stress or lower levels of EI.

Apart from teaching EI skills, higher education institutions also need to promote an environment of empathy and emotional support for students by providing them with access to mental health services, encouraging open dialogue about their emotional experiences, and educating faculty and staff teams about the need for emotional intelligence and interactions with positive and encouraging language inside and outside the classroom. Since COVID-19 can negatively affect a person’s emotional and psychological wellbeing, institutions should prioritize providing students with counseling and other mental health services. Institutions may also consider incorporating mindfulness practices and other stress-reduction techniques into the curriculum to promote emotional health and resilience. Although EI instruction might help students deal with the pressures of institutions, it is vital to remember that stress is a complicated issue with many potential causes. Therefore, institutions must adopt a comprehensive approach to address emotional, environmental, social, and psychological factors contributing to stress. Besides, systemic concerns, such as high academic expectations, financial stress, and limited access to resources, can also be addressed by institutions. Therefore, adjustments in policy, such as reducing workload or increasing support for disadvantaged students, as well as community outreach and advocacy efforts, are necessary to raise awareness of the issue. Besides, to design better architecture for handling students’ academic stress, university administrators need to facilitate institutional-level integration with subjects like clinical psychology. Furthermore, family members, friends, and significant people from society, like teachers, relatives, and role models, can emotionally connect with students, assist students in developing their EI, and reduce academic stress.

This study’s findings are important and well-supported, but the cross-sectional survey design limits our ability to conclude the temporal relationships among participants’ emotional intelligence, social support, and perceived academic stress. Besides, participants in the study are restricted to Bangladesh, so the results may not be generalizable to other economic, cultural, and geographical contexts. Finally, this study used self-report measures, which could have introduced response or social desirability bias into the study.

Henceforth, research in the future could use longitudinal or experimental designs to investigate the links between EI, social support, and academic stress. Possible areas for future research also include how cultural and environmental factors, such as differences in social norms, educational systems, and job opportunities, affect the association between emotional intelligence, social support, and academic stress. Other than perceived academic stress, academic outcomes like performance and motivation may also be influenced by emotional intelligence, which might be explored in future studies. Furthermore, future research could investigate the potential mediators and moderators (e.g., coping strategies or personality traits) in the linkage between emotional intelligence and academic stress. As the current study reported that gender and academic seniority cause a significant difference in stress levels, future studies can address the potential sources of stress for each group and their coping strategies.