MacCann, C., Jiang, Y., Brown, L. E. R., Double, K. S., Bucich, M., & Minbashian, A. (2020). Emotional intelligence predicts academic performance: A meta-analysis. Psychological Bulletin, 146(2), 150-186, Jan 2021. http://dx.doi.org/10.1037/bul0000219
Abstract: Schools and universities devote considerable time and resources to developing students’ social and emotional skills, such as emotional intelligence (EI). The goals of such programs are partly for personal development but partly to increase academic performance. The current meta-analysis examines the degree to which student EI is associated with academic performance. We found an overall effect of ρ = .20 using robust variance estimation (N = 42,529, k = 1,246 from 158 citations). The association is significantly stronger for ability EI (ρ = .24, k = 50) compared with self-rated (ρ = .12, k = 33) or mixed EI (ρ = .19, k = 90). Ability, self-rated, and mixed EI explained an additional 1.7%, 0.7%, and 2.3% of the variance, respectively, after controlling for intelligence and big five personality. Understanding and management branches of ability EI explained an additional 3.9% and 3.6%, respectively. Relative importance analysis suggests that EI is the third most important predictor for all three streams, after intelligence and conscientiousness. Moderators of the effect differed across the three EI streams. Ability EI was a stronger predictor of performance in humanities than science. Self-rated EI was a stronger predictor of grades than standardized test scores. We propose that three mechanisms underlie the EI/academic performance link: (a) regulating academic emotions, (b) building social relationships at school, and (c) academic content overlap with EI. Different streams of EI may affect performance through different mechanisms. We note some limitations, including the lack of evidence for a causal direction.
Public Significance Statement: This meta-analysis shows that emotional intelligence has a small to moderate association with academic performance, such that students with higher emotional intelligence tend to gain higher grades and achievement test scores. The association is stronger for skill-based emotional intelligence tasks than rating scales of emotional intelligence. It is strongest for skill-based tasks measuring understanding emotions and managing emotions.
Keywords: academic performance, emotional intelligence, intelligence, meta-analysis, personality
Results from these meta-analyses demonstrate that EI shows a small to moderate relationship with academic performance, of similar effect size to well-known noncognitive predictors (e.g., ρ = .20 for EI vs. ρ = .22 for conscientiousness, based on the current meta-analysis and Poropat [2009]). Ability EI was a significantly stronger predictor than self-report or mixed EI, as hypothesized. Within ability EI, understanding and management branches had a stronger effect than perception or facilitation branches. There is no evidence for selective publication of larger effects, for stronger effects in younger students, nor that effects differ depending on the proportion of ethnic minority students in the sample. For the other moderators, effects were mixed or limited. There is limited evidence that the effect is stronger: (a) for less female-dominated samples (this effect was significant for total EI, but not for any of the three streams), (b) for grades than standardized test scores (this was significant for total EI and Stream 2 only), and (c) for humanities versus mathematics/science performance (this was significant for ability EI only).
There was evidence of incremental validity of EI over intelligence and personality, but this was largely restricted to mixed EI (which explained an additional 2.3% of the variance) and the understanding and management branches of ability EI (which explained an additional 3.9% and 3.6% of the variance respectively). That is, self-rated EI, total ability EI, and the lower two branches of ability EI (emotion perception and facilitation) provide little to no explanatory power for academic performance over intelligence and personality. These differences across the three streams suggest that the underlying mechanisms accounting for the EI/performance relationship may differ for ability EI, self-rated EI, and mixed EI.
Why Does EI Predict Academic Performance? Insights Based on Moderators
In our introduction, we suggested there were three reasons why EI may predict academic performance. First, students with higher EI may be more able to regulate the negative emotions such as anxiety, boredom, and disappointment involved in academic performance. If this is true, emotion management would be responsible for the effects. Second, students with higher EI may be better able to manage the social world around them, forming better relationships with teachers, peers, and family. If this was the case, emotion management would again be responsible for the effect, and the effect would be stronger for grades than for standardized tests. Third, EI competencies may overlap with the academic competencies required for humanities subjects like history and language arts (e.g., understanding human motivations and emotions). In this case (a) understanding—the knowledge base of EI—would show the strongest effect and (b) the effect would be bigger for humanities than sciences. Based on the significant moderations, there is some support for each of these effects, with slightly different results for different streams of EI. We discuss the significance moderations below, with respect to these three proposed mechanisms.
Evidence for Mechanism 1: Is Emotion Management the Key Ingredient in EI?
Joseph and Newman (2010) proposed a “key conceptual role” of emotion management for predicting job performance (p. 69), proposing emotion management as the proximal predictor of performance. The facet-level moderation for ability EI provides partial support for this assumption, finding that management and understanding are jointly the strongest predictors of academic performance. Both these branches (management and understanding) showed significantly stronger effects than the two other branches. The effect was larger for understanding than management, but not significantly so (ρ = .35 vs. ρ = .26) and was equal after accounting for the effects of intelligence and personality (partial ρ = .22 in both cases). That is, both emotion understanding and emotion management are active ingredients in the prediction of academic performance. We believe this is consistent with an interpretation that EI affects academic performance through the regulation of academic emotions, but also due to the relevance of emotion content knowledge for academic performance in the humanities.
The critical role of emotion understanding for academic performance has implications for comparing ability EI with self-rated EI. For self-rated EI, many of the effects in our
meta-analysis used instruments that did not include emotion understanding content (because they were based on an older definition of EI that did not have emotion understanding in the definition). Specifically, 50% of the Stream 2 citations used the Schutte
Self-Report Scale, Trait Meta-Mood Scale, or the Wong-Law
Emotional Intelligence Scale, which do not include a subscale assessing emotion understanding. Given that ability EI shows the strongest relationship for emotion understanding, the difference in
effect size between ability EI and self-rated EI measures may in fact represent a difference in content (i.e., prediction is greater for tests that include emotion understanding content) rather than a difference in method (ability scales vs.
rating-scales). Many of the more recent self-rated EI tests
do include an emotion understanding component (e.g.,
Anguiano-Carrasco, MacCann, Geiger, Seybert, & Roberts, 2015;
Brackett et al., 2006).
Evidence for Mechanism 2: Are EI Competencies Required for Academic Content?
Moderation analyses largely support the idea that performance on academic tasks require some EI competencies. First, academic performance related significantly more strongly to ability EI than to the other two streams. This finding differs from meta-analyses predicting job performance, where ability EI is consistently the weakest predictor of the three streams (
Joseph & Newman, 2010;
Miao et al., 2017;
O’Boyle et al., 2011). This difference may relate to assessment methods. Academic performance is mainly assessed with objective tasks (i.e., evaluations of a product, such as an essay, lab report, speech, worksheet or test), whereas job performance is most often assessed via supervisor ratings. Similarly, ability EI is assessed with objective tasks and self-rated and mixed EI are assessed with
rating scales. We would expect stronger predictor-criterion relationships when predictor and criterion have the same method. As such, a higher relationship for ability EI (as compared to the other streams) may represent method bias rather than content overlap of academic and emotional knowledge.
However, ability EI (but not self-rated or mixed EI) relates more strongly to performance in humanities than sciences. This is one of the larger differences we found, where the effect was nearly twice as large for humanities as sciences (ρ = .38 vs. .21). Objective measurement of performance is similar across humanities and sciences. The academic processes (
social context and the student’s emotions and emotion regulation in the classroom) are also similar for different subdisciplines. Although subdisciplines differ in the degree of
social interaction involved, the degree of social interaction does not align with the humanities versus science categorization (e.g., science frequently involves lab partners or group work, whereas this is rare for mathematics). As such, we interpret this difference in subject areas to be largely attributable to a difference in academic
content, and specifically the relevance of emotion knowledge to subjects requiring an understanding of people and their interactions, motivations, and emotions (i.e., literature, history, geography, drama and other humanities subjects). The first standard of
The Standards for the English Language Arts (1996), as put forward by the
National Council of Teachers of English (1996) states that the purpose of reading texts is to “build an understanding of . . . themselves and the cultures of the United States and the world” and the second standard states that the purpose is “to build an understanding of the many dimensions (e.g., philosophical, ethical, aesthetic) of human experience” (p. 19). That is, broad statements of content for achievement in language arts inherently involve an
understanding of oneself and of others in terms of the intangible nature of being human—which we would argue is essentially emotions and social interactions. That is, understanding human emotions and the social and situational causes appear to be an underlying component of achievement in language arts.
In addition, the fact that the emotion understanding showed the strongest relationship to academic performance (as compared to the other four branches) supports the interpretation mentioned above, where understanding emotional content is a key part of the content of language arts education. It is possible to view emotion understanding as a kind of domain-specific knowledge, where the content domain is emotions. Content knowledge of emotion words, as well as the causes and consequences of emotions, appear highly relevant for understanding character motivations in literature as well as other academic subject matter relating to people and how they shape societies, countries and history (i.e., history, geography, psychology, sociology).
One possible interpretation is that the ability EI/academic performance association may be due to a third variable—reading comprehension. Because ability EI tests involve interpreting written text, reading comprehension ability may constitute construct-irrelevant variance on such tests (
AERA, APA, & NCME, 2014) that may partially explain the relationship between EI and academic performance. This particularly affects understanding and management tests, which involve more and more
complex text (e.g., most management tests involve a paragraph of text in each item stem). However, the fact that emotion understanding and management predicted academic performance over-and-above the effect of intelligence suggests that this confound does not account for the entirety of the relationship between ability EI and academic performance. Nevertheless, the relationship was greatly reduced, particularly for emotion understanding. Because the
partial correlations remained of small to moderate size after accounting for intelligence, our interpretation is that the bulk of the content overlap represents more than a reading comprehension method effect, particularly for emotion management. Taken together, results support the suggested mechanism whereby EI predicts academic performance because of the
emotional content required in academic subjects.
Evidence for Mechanism 3: Does EI Affect Academic Performance Through Interpersonal Processes?
If EI exerts an influence on academic performance via the ability to develop social relationships in the educational context, then EI should have a stronger effect on grades than standardized tests (as the social networking and relationship building with other students and teachers should have a stronger effect on grades than on standardized tests). This difference was significant only for self-rated EI and the three streams combined (not for ability or mixed EI). Self-rated EI did not relate to standardized test scores at all (ρ = −.03). In contrast, ability EI and mixed EI related to both grades and standardized tests. This suggests that academic performance relates to self-rated EI through relationship building only. In contrast, academic performance relates to ability EI and mixed EI through both relationship building and mechanisms related to regulating academic emotions.
For all three streams of EI, there is evidence that higher EI relates to building social relationships in a school environment. Ability EI relates to peer-nominations of reciprocal friendship in college students and to higher-quality of
social interactions with others (
Lopes et al., 2004;
Lopes, Salovey, Coté, Beers, & Petty, 2005). Self-rated EI predicts greater
social support in both high school and university students (
Ciarrochi, Chan, & Bajgar, 2001;
Kong, Zhao, & You, 2012). Mixed EI is associated with peer reports of cooperative behavior (
Mavroveli, Petrides, Rieffe, & Bakker, 2007;
Petrides, Sangareau, Furnham, & Frederickson, 2006). There is also evidence that both ability EI and mixed EI relate to using more effective strategies to regulate negative emotions (
Peña-Sarrionandia, Mikolajczak, & Gross, 2015).
Taken together with these findings, we propose that differences between the three streams of EI relate to the number of mechanisms that underlie the EI/performance relationship. Specifically, (a) self-rated EI predicts academic performance only through a relationship building pathway (students with higher emotional self-efficacy can build better relationships with teachers and peers), (b) mixed EI predicts academic performance through both relationship building and the regulation of academic emotions, and (c) ability EI predicts academic performance through relationship building, regulation of academic emotions, and also through emotion content knowledge requirements of some academic areas. This explanation accounts for the relatively greater prediction of academic performance by ability EI than mixed EI than self-rated EI and is consistent with the pattern of moderators we found.
The Relative Importance of EI to Academic Achievement
One of the critical drivers of EI’s early popularity was the idea that emotional skills are more important than intelligence in predicting life success. Indeed, the title of Daniel Goleman’s first book, the catalyst for EI’s snowballing popularity, was
Emotional intelligence: Why It Can Matter More Than IQ. The 1995 cover story of TIME magazine made similar claims, stating that “emotions, not IQ, may be the true measure of human intelligence” (
Gibbs, 1995, p. 60). These early claims were generally not borne out by research on job performance. Although EI predicts better job performance (
Joseph & Newman, 2010;
O’Boyle et al., 2011), a critical mass of research indicates that intelligence is a much stronger predictor and is in fact the single best predictor of job performance (
Ree & Earles, 1992;
Salgado et al., 2003;
Schmidt & Hunter, 1998). We found largely similar results for academic performance. Although EI predicts academic performance, intelligence was a much stronger predictor, with relative importance analysis indicating that
cognitive ability was the single most important predictor of academic performance.
Although the popular press hype about EI was not substantiated, we nevertheless believe that demonstrating a small to moderate effect size is informative for research and practice. Moreover, some of the recent changes occurring in education and assessment practices may increase the importance of noncognitive qualities, including EI.
The first such change to modern assessment and learning practice is the increasing use of group activities, including collaborative group assessments (
Ahles & Bosworth, 2004). Managing the social relationships and interpersonal conflicts of the group may thus become more and more reflected in students’ end of semester grades. A second change to education practices is the extent to which graduate attributes (also referred to as
21st century skills or
noncognitive constructs) are emphasized by schools and universities (e.g.,
Clarke, Double, & MacCann, 2017). Graduate attributes often include social-emotional skills such as leadership, communication, teamwork, and intercultural competencies, with some institutions explicitly including EI as a graduate attribute. For example, Australia uses Goleman’s model of EI as the basis for its national K–10
curriculum of personal and social competencies that students should be developing as they progress through primary and secondary education (
ACARA, 2017). Schools and universities are increasingly attempting to embed these graduate qualities within the content that is taught and assessed. As such, high grades might increasingly reflect skill development in these areas.
A third change to the classroom is the extent to which computers and technology are now an integral part of education. In tertiary education, there are a large and increasing number of online only courses or courses that have at least some online-only content (most famously, the Massive Open Online Courses [MOOCs]). There are two main differences between traditional face-to-face learning and online learning. First, in a traditional face-to-face university course, the schedule of learning is set by the schedule of the face-to-face lectures. In contrast, an online only course requires the learner to manage their own schedule of accessing online content, such that students with poor time management will not succeed (
MacCann, Fogarty, & Roberts, 2012). Second, in a traditional face-to-face course, communication with teachers and other students occurs through in-person conversation, with access to multiple channels of information (e.g., facial and vocal expression,
body language, and real-time clarification of misunderstandings). Online communication is more often based on text (e.g., discussion boards, emails, or computer chat). Most neurotypical people find it more difficult to detect another person’s emotions and
social needs from text rather than face-to-face contact. As such, greater emotional skills are required to build relationships with the instructor or other students in an online environment. Thus, social and emotional skills (both self-regulation and interpersonal skills) may become increasingly important as tertiary education involves a greater amount of online content.
Practical Implications
One of the major findings of this meta-analysis is that different parts of EI are differentially important for academic performance. Any applied uses of EI in education seem limited to the three parts of EI with nontrivial incremental validity: mixed EI, emotion management ability, and emotion understanding ability. There are three broad applications that might be considered: (a) identifying students at risk for failure, attrition, or underperformance; (b) selection decisions for high-stakes educational opportunities; and (c) policy decisions about the relative cost versus benefit of implementing SEL or EI training programs in schools.
The first two applications (identifying at risk students, and high-stakes selection) require careful consideration of response distortion issues. Particularly in a high-stakes selection context, test-takers are motivated to gain high scores and will distort their responses on
rating scales to ‘fake high’ (
Birkeland, Manson, Kisamore, Brannick, & Smith, 2006;
Viswesvaran & Ones, 1999). Faking is a consequential issue with personality scales, which use
self-report or observer-report ratings that test takers can fake. Observer-reports do not necessarily solve this problem, as the observers are often not impartial, but may be school staff with a vested interest in their students gaining entrance to prestigious colleges or programs. Faking is a problem for rating-scale measures of EI, but not for ability scales (
Day & Carroll, 2008;
Grubb & McDaniel, 2007;
Tett, Freund, Christiansen, Fox, & Coaster, 2012). The current
meta-analysis is the first to demonstrate that the relationship between EI and academic performance holds for ability-based tests as well as rating scales (in fact, the relationship is actually
higher for ability-based EI tests as compared to rating scales). As such, we demonstrate a pathway that might provide modest increments in high-stakes education selection decisions—using ability-based EI assessments of understanding and managing emotions (based on current results, other parts of ability EI are not important). Ability-based EI tests are already used for selection into medical schools in several countries, and evidence supports their use for selecting better candidates (
Libbrecht, Lievens, Carette, & Côté, 2014;
Lievens & Sackett, 2006). However, such tests are rarely used in other broader education selection contexts. If EI is considered as a selection procedure (perhaps as an add-on to intelligence and
personality assessment), we suggest that
ability tests of understanding and managing emotions be preferred over rating scales (due to response distortion) or tests of facilitation or management (due to low incremental prediction over intelligence and personality).
A national and international focus on
standardized tests to measure academic performance and milestones has lead schools, districts, states, and countries to focus on achievement in the narrow range of academic content that such tests focus on. Alongside this, classroom teachers face increasing challenges to their workload, including adapting the
curriculum to individual students’ needs, the
mainstreaming of students with special educational requirements, and adapting to rapidly changing curriculum and policy (
Skaalvik & Skaalvik, 2007). Against this background, devoting resources to teaching children EI skills can be seen as taking teacher resources and classroom time away from more critical activities that will increase test scores and achievement. What our
meta-analysis shows is that EI skills are in fact associated with higher academic performance. This implies that time spent teaching EI skills may not necessarily detract from student achievement, given that higher EI students also show higher achievement. Again, we highlight the different importance of the four EI abilities as a guide for where to focus skills training—a focus on perceiving emotions is likely to be less useful than a focus on understanding and managing emotions.
Our
meta-analysis also has implications for the effects of the such training programs (or a focus on EI more generally) on the known achievement gaps between
ethnic groups and between males and females. Although there is evidence that the Black-White achievement gap is slowly closing, differences in the achievement for minority students compared with White students remain substantial, at around 0.75 standard deviations for Black students and around 0.60 standard deviations for Hispanic students (
Hansen, Mann Levesque, Quintero, & Valant, 2018). There is also increasing evidence that males are falling behind females in terms of the grades they receive and their participation in higher education (
Fortin, Oreopoulos, & Phipps, 2015). Against this background, it is important to note that the effect of EI on academic performance does not appear to differ for minority students versus White students, and that gender differences are negligible and when significant favor males (who currently show lower achievement). These results imply, at the very least, that efforts to improve EI are unlikely to widen the achievement gaps.
The key role of emotion understanding and management is also important to consider in terms of EI training programs. Three recent meta-analyses on the effectiveness of EI training have reported significant increases in EI, with
effect sizes of .45, .46, .51, and .61 (
Hodzic, Scharfen, Ripoll, Holling, & Zenasni, 2018;
Mattingly & Kraiger, 2019;
Schutte, Malouff, & Thorsteinsson, 2013). Hodzic et al. found that programs based on the ability model were significantly more effective than those based on
mixed models (
g = .60 vs. .31), and that emotion understanding showed the largest increase of all the ability EI branches—significantly more than emotion facilitation (
g = .69 vs. .42). That is, it seems that programs are effective for increasing ability EI, and particularly its emotion understanding facet. This is highly relevant for our own
meta-analysis, where ability EI (and specifically emotion understanding) showed the highest association with academic performance. That is, EI training seems to produce the strongest increases in exactly those competencies that are most relevant for academic performance.
Although Hodzic et al. did not distinguish between EI training programs for workplace applications and EI training programs for schools and universities, several studies conducted in schools and universities report similar findings regarding the largest increases for emotion understanding. For example,
Pool and Qualter (2012) conducted a
training study in university students and found the largest increase in ability EI was for emotion understanding (and the second-largest for emotion management). Moreover, evidence from the RULER Feeling Words
curriculum (an EI development program for secondary school students) shows that EI training programs increase grades as well as social and emotional competencies. Specifically, students completing the RULER showed improved
school grades as well as improved teacher ratings of social and emotional competencies compared to
control groups (
Brackett, Rivers, Reyes, & Salovey, 2012). In fact, the relative increase in school grades was a larger effect than the relative increase in social and emotional competencies. That is, EI training programs are likely to increase academic performance as well as social and emotional outcomes, such that education decision-makers and policymakers are not faced with a decision of whether to invest in social/emotional wellbeing at the expense of student achievement—evidence suggests that these programs likely do both. This is a critical piece of information for schools deciding where to best allocate their resources.
Limitations
Our results demonstrated only that EI and academic performance are significantly associated, but not that higher EI
causes higher achievement. Only three of the citations reported a
longitudinal design, such that the empirical evidence for EI causing later achievement is very weak (
Costa & Faria, 2015;
Qualter et al., 2012;
Stewart & Chisholm, 2012). This association could occur because (a) higher EI causes increased academic performance, (b) higher achievement causes increased EI, or (c) there are one or more variables that influence both EI and academic performance. In the introduction, we outlined the reasons we believe theoretically that EI could cause later achievement. However, there are also feasible pathways by which greater academic performance could cause higher EI. Greater academic performance could feasibly result in increased self-esteem, greater opportunities for social and
emotional development, and higher expectations for
social skills and emotion regulation. High academic performance may act as a gateway for gifted and talented programs, streaming into enrichment activities, and a culture of high expectations from teachers, parents, and communities that permeate social and emotional behaviors as well as academic ones via the
halo effect (
Nisbett & Wilson, 1977). Conversely, low academic performance may act as a barrier to opportunities to develop social and emotional skills through loss of privileges for
academic failures (e.g., losing recess playtime or evening socialization to complete work or denied extracurricular activity participation because of course failure), the development of strong negative emotions surrounding school and schoolwork, and the correspondingly low expectations for social and emotional behaviors. It seems likely that the reality is
complex, with bidirectional effects of academic and emotional development, particularly in the earlier years of school.
One further limitation of the current article concerns the use of the meta-analytic
correlation matrix (used to test Hypotheses 10 and 11). This was composed of estimates taken from different journal articles from different research teams, and therefore did not use the same methods for estimation nor the same samples. Although we used RVE in the current study, all other sources for
effect sizes were obtained by aggregating multiple effect sizes from the same study. All studies except for
Poropat (2009) corrected for unreliability as well as
range restriction. The personality/academic performance estimation was not corrected for range restriction of measurement in either the predictor or criterion (
Poropat, 2009). The possible effect of this would be to underestimate the prediction and relative importance of
personality traits such as conscientiousness.
Future Research and Recommendations
One obvious future direction for further research is to test our three proposed mechanisms of the EI/performance relationship: (a) social relationship building, (b) regulation of academic emotions, and (c) content overlap between EI and academic subject matter. For point a, an analysis of content overlap between the competencies of EI and the different processes required for success in different disciplines could be undertaken by a panel of educators. Longitudinal research involving all three streams could test whether all three mediate ability EI, (a) and (b) mediate mixed EI, and (a) along mediates self-rated EI, as we proposed. As we mention above, there is a paucity of long-term longitudinal research on EI and academic performance. As such, examining mediators of the link as well as conducing lagged panel models to tease apart the direction of causation is important.
Although there is ample evidence that training EI works (e.g.,
Hodzic et al., 2018;
Mattingly & Kraiger, 2019;
Schutte et al., 2013), we are not aware of experimental studies on EI training that examine the effects of training different branches of EI. Such designs would isolate which facets of EI are most relevant for the improvement of which types of outcomes and would also provide stronger evidence for the causal direction from EI to academic performance.
Posts
