Relationship Between Emotional Contagion and Leisure Benefits in Sports and Leisure: The Mediating Role of Leisure Involvement and Flow

The purposes of the study were to test the mediating role of leisure involvement and flow on the relationship between emotional contagion and leisure benefits in sports and leisure. The participants were college student in Taiwan selected by the purposive sampling method. Sports and Leisure Emotional Contagion Scale, Sports and Leisure Involvement Scale, Sports and leisure Flow Scale, and Sports and Leisure Benefits Scale, are used as instuments after compiled by authors, and retested using PLS-CFA. The CFA results showed that four scales have a good reliability and validity. The data were analyzed using PLS-SEM. Results of the study indicate that leisure involvement and flow had full mediating effects between emotional contagion and leisure benefits. The results could be applied in both sports and leisure area of mathematics instruction, and emotional contagion could elevate leisure benefits by leisure involvement and flow.

Leisure involvement affects leisure benefits. Involvement is the degree to which one exerts oneself in a matter . High involvement in an activity can improve exercise benefits (Lee & Hung, 2015). Empirical studies have indicated that exercise or leisure involvement has positive influences on leisure benefits, for instance, Yong and Jinxia's (2013) study on cyclists, Chung's (2015) study on slow pitch softball competitors, Ho and Yang's (2016) study on university students, Huang and Sun's (2017) study on pétanque competitors, and Kung's (2017) study on tennis competitors. The results of these studies demonstrated that exercise or leisure involvement had positive influences on leisure benefits, which is to say that benefits accrue with the effort exerted (Yu & Tian, 2013). Therefore, leisure involvement is deduced to affect leisure benefits, and the following is posited: H 5 : Among university students, leisure involvement and leisure benefits are positively correlated.
Last, as mentioned, emotional contagion affects leisure benefits and also affects flow, and flow can affect leisure benefits; as such, flow may play a mediating role between emotional contagion and leisure benefits. Similarly, leisure involvement also plays a mediating role. Thus, the following are posited: H 6 : Among university students, flow and leisure involvement has a mediating effect on the relationship between emotional contagion and leisure benefits.
Thus, the goal of this study was, in addition to verifying the correlations among emotional contagion, leisure involvement, flow, and leisure benefits in sports and leisure, to verify the mediating effects of flow and leisure involvement. Furthermore, past studies have not verified the mediating effects of flow and leisure involvement on the relationship between emotional contagion and leisure benefits, and this research gap may be filled by the findings of this study.

Participants
Two batches of participants were recruited: survey pretest participants (N = 81) and formal participants (N = 100). The demographic variables of the participants are presented in Table 1.

Research Instruments
Four instruments measuring emotional contagion in sports and leisure were created for the study. Each instrument used 5-point Likert scales, with 1 to 5 points assigned to the responses (does not match at all, slightly matches, somewhat matches, mostly matches, and fully matches). The factors of the four instruments were very clear, and confirmatory factor analysis (CFA) was therefore used to revise the instruments. Furthermore, because the pretest sample comprised 81 people, variance-based SmartPLS 3.0, which is suitable for small sample sizes (30100), was used.

Sports and Leisure Emotional Contagion Scale
The five factors of Doherty's (1997) Emotional Contagion Scale were referenced in this study; combined with jel.ccsenet.org Journal of Education and Learning Vol. 11, No. 6;2022 sports and leisure scenarios, the scale items were revised to form an initial draft comprised of 27 items in total for Love (6 items, e.g., On the sports field, seeing the support and comfort between my peers made my heart full of warmth), Happiness (6 items, e.g. When my sports peers smile warmly at me, I feel warm inside and smile too.), Fear (6 items, e.g., During a game, sitting next to an angry crowd made me feel restless.), Anger (5 items, e.g., Seeing teammates accusing each other of negligence made me uncomfortable.), and Sadness (4 items, e.g., Watching fans cry as the team they support loses a game, I unconsciously shed tears.).
Four practitioners with backgrounds in sports and leisure were invited to revise the draft. Two practitioners were university students; one had expertise in cycling, and the other had expertise in tennis. The third practitioner was employed in an indigenous hunter school. The fourth practitioner was a high school teacher with expertise in badminton.
The PLS-CFA test results indicated that in composite reliability (CR), the five factors met the benchmark of .70. However, for average variance extracted (AVE), Love = .46 and Anger = .47, failed to meet the benchmark of .50; the reason for their not meeting the benchmark was the factor loading λ. In structural equation modeling (SEM), the benchmark for λ is .50.95, and Hair et al. (2014) argued that observation indicators with λ values between .40 and .70 may be eliminated if doing so will result in the CR or AVE values meeting the benchmark. Therefore, "Love 6: When teammates or companions hug me enthusiastically, I return the hug" (λ = .57) and "Anger 5: When I see other fans accuse the referee of unfairness, I join in" (λ = .46) were eliminated from the instrument.
After these two items were removed, the factor AVEs improved; Love = .51, and Anger = .56, meeting the benchmark of AVE being equal to or greater than 50%. The final CFA results indicated that among the 25 items: Love (5 items), Happiness (6 items), Fear (6 items), Anger (4 items), and Sadness (4 items), the λ values were .60.86; the factor CR values were between .83 and .90, and the AVE values were .51.59. Based on the cross-factor loadings, the observation indicators were more highly correlated with the factor to which they belonged than to other factors and, therefore, the discriminant validity (DV) of the results was also consistent.
These numbers indicate that the revised sports and leisure emotional contagion scale had acceptable reliability and validity.

Sports and Leisure Involvement Scale
McIntyre (1989) proposed that leisure involvement is a multidimensional concept with three factors: attraction, self-expression, and centrality. These factors can explain leisure involvement (Kyle & Mowen, 2005). In Taiwan, for instance, Lin et al. (2016) and Tsao et al. (2016) used them to measure leisure involvement. They have also been used to measure leisure involvement in western countries, such as in studies by Kyle and Mowen (2005) and by Sosianika et al. (2020). The factors were also used in the present study.
The first draft of the Leisure Involvement Scale was drafted in reference to Chang's (2007) Activity Involvement Scale for hikers and Yeh et al.'s (2007) Online Game Leisure Involvement Scale. Therefore, with reference to the content of the two scales, language about sports and leisure were added to the items to create five items on attraction (e.g., The sports and leisure activities I do is very interesting for me.), five items on centrality (e.g., The sports and leisure I am engaged in occupy a central position in my life), and four items on self-expression (e.g., The sports and leisure activities I am engaged in made me truly express myself.).
The CFA results of the Leisure Involvement Scale (see Table 3) indicated that the CR values of the three factors were between .92 and .96, exceeding the benchmark of .70. The AVE values were between .71 and .83, exceeding the benchmark of .50. The λ values of the items were between .71 and .93, thus meeting the benchmark of .50.95. Therefore, all the items were retained. In the cross-factor loadings, the observation indicators were more highly correlated with the factor to which they belonged than to other factors, and the DV results were therefore also consistent.

Sports and Leisure Flow Scale
The Flow Scale was drafted referencing the Flow Subscale (9 items) in Hou's (2012) Positive Emotion Scale of College Students. The subscale was based on flow theory (Nakamura & Csikszentmihalyi, 2009) and referenced the Flow Questionnaire proposed by Csikszentmihalyi and Csikszentmihalyi (1992) and the Game Flow Experience and Motivation Questionnaire developed by Wang and Chen (2010). Wallis (2020) argued that flow can be viewed as a single-factor concept. The items in Hou's (2012) scale had a theoretical basis and, like this study, were designed for university students. Therefore, the Flow Subscale was selected for this study. The items were revised based on a sports and leisure perspective. For instance, the original item "When I am interested in and committed to a problem that has not yet been completed, I think about it all the time" was revised to "When I am interested in a sports and leisure problem that hasn't been resolved yet, I think about it all the time." The first draft of the Flow Scale had nine items, and the CFA results (  (Hung & Crompton, 2006;Ho, 2008;Akgül et al., 2018) have proposed that leisure benefits can be distinguished into three types: 1) Physiological benefits (such as disease prevention and control); 2) Social benefits (such as social relationship and family bonding developments); 3) Psychological benefits (such as self-growth and mental stress relieving). In Taiwan, the most recent studies-e.g., Chin et al. (2020) and Lai et al. (2020)-also used these three elements. In this study, the implications ascribed by Huang et al. (2012) for the three types of leisure benefits and the tennis player scale developed by Huang et al. (2014) were referenced. The tennis player scale also measured three factors, namely psychological, physiological, and social benefits, and was used to measure the benefits gained by members of the public engaged in tennis.
The Sports and Leisure Benefits Scale comprised 21 items in total, with 9 items on psychological benefits, 7 on physiological benefits, and 5 on social benefits. The CFA results indicated all three factors had CRs greater than .90, exceeding the benchmark of .70, and the AVEs were between .65 and .75, exceeding the benchmark of .50. The factor loadings (λ) of the 12 observation indicators were between .62 and .91, meeting the benchmark. In the crossfactor loadings, the observation indicators were more highly correlated with the factor to which they belonged than to other factors; therefore, the DV results were also consistent. Therefore, all the items were retained.

Data Analysis
The formal sample size in this study was 100 people, and variance-based SmartPLS 3.0, which is suitable for small sample sizes (30100), was used to verify the hypotheses. Furthermore, whether the statistical effects were statistically meaningful was analyzed using the built-in bootstrapping function in SmartPLS to perform 5000 extractions.

Results
From Baron and Kenny's (1986) perspective, the analysis of mediating effects involves three steps: 1) confirm that a correlation exists between the exogenous variable (emotional contagion) and endogenous variable (leisure benefits); 2) confirm that correlations exist among the exogenous variable, the mediating variables (flow and leisure involvement), and the endogenous variables; and 3) after establishing steps 1 and 2, verify the mediating effects. The mediating effects were verified using partial least squares (PLS) SEM; this method was also recognized by Hair et al. (2014).

Verifying Correlations Among Emotional Contagion, Flow, Leisure Involvement, and Leisure Benefits
Steps 1 and 2 verified whether the correlations among emotional contagion, flow, leisure involvement, and leisure benefits were significant. As shown in Table 2, the path coefficients among the exogenous variable (emotional contagion), mediating variables (flow and leisure involvement), and endogenous variable (leisure benefits) all met significance. The path coefficients (β) among the four latent variables were .51 or greater, and every t value was 6.30 or greater; all the p values were less than .001. Within the 95% confidence interval, 0 did not fall between the lower and upper limits, indicating that significance was met. Therefore, steps 1 and 2 established that significant correlations were present among emotional contagion, flow, leisure involvement, and leisure benefits.

Verifying the Mediating Effects of Flow and Leisure Involvement Between Emotional Contagion and Leisure Benefits
After verifying the presence of correlations in steps 1 and 2, step 3 was carried out. This involved placing the four latent variables into one model, with flow and leisure involvement as the mediating variables, to analyze their mediating effects between emotional contagion and leisure benefits. And according to Hair et al. (2014), the fitness of the mediating effect model can be determined by the measurement model and the structural model.

Fitness of the Measurement Model
In PLS path modeling, the measurement model is known as the outer model. According to Table 3, the observation indicators of the four latent variables had factor loadings (λ) between .81 and 1.00 (Figure 1), which meets the benchmark of being between .50 and .95; the latent variables' CRs were .93 or greater, exceeding the CR ≥ .60 benchmark; and the AVEs were .71 or greater, exceeding the ≥ .50 benchmark. As for DV, as long as the square root of the AVE of a latent variable is greater than the value of the variable's correlation with other variables, then DV is present (Fornell & Larcker, 1981); the results also fit this benchmark. In summary, λ, CR, AVE, and DV were in line with the benchmarks, indicating model fitness. Furthermore, because flow had only one dimension and no variance, its CR, AVE, and DV were 1.00.

Fitness of the Structural Model
In PLS-SEM, the structural model is also known as the inner model. As for its fitness, first, collinearity was determined based on the variance inflation factors (VIFs). The VIFs among the four latent variables were between 1.00 and 4.52, that is, less than 5.00, indicating the problem of collinearity is small.
Second, the coefficient of determination (R 2 ) represents the accuracy of the prediction; it also represents the overall explanatory effect of all the exogenous variables on the endogenous variables in the model (Hair et al., 2014). In the mediating effects model, the endogenous variable was leisure benefits, and, in relation, emotional contagion, flow, and leisure involvement were the exogenous variables. The analysis found R 2 to be .74, which represented jel.ccsenet.org Journal of Education and Learning Vol. 11, No. 6;2022 the overall explanatory power of emotional contagion, leisure involvement, and flow for leisure benefits. The explanatory power was moderate, according to the R 2 benchmarks proposed by Hair et al. (.75 for high, .50 for moderate, and .25 for low).
Third, for effect size f 2 , which is the change in R 2 after eliminating a specific exogenous variable, was .17 for leisure benefits, .13 for leisure involvement, and .04 for emotional contagion. According to Cohen (1992), an f 2 of .02 indicates a small effect size, .15 a moderate effect size, and .35 a large effect size. Based on these benchmarks, emotional contagion, leisure involvement, and flow had small to moderate levels of effects on leisure benefits.
Last, for Q 2 , which is predictive correlation (Note 1), the benchmark is greater than zero. The results for Q 2 were leisure involvement = .20, flow = .249, and leisure benefits = .62, indicating that predictive correlations were present; emotional contagion had predictive correlations with leisure involvement, flow, and leisure benefits.

Analysis of the Mediating Effects
The results verifying the mediating effects models indicated that both the measurement and structural models were fit. Therefore, the mediating effects were analyzed. Mediating effects can be direct, indirect, or total (see Table 4).
Direct effects can be captured with the coefficient values (β) of the five paths involving emotional contagion, leisure involvement, flow, and leisure benefits in the mediating effects model. The coefficient values were as follows: leisure involvement to leisure benefits = .38 (p = .007), flow to leisure benefits = .44 (p < .001), emotional contagion to leisure benefits = .11 (p = .094), emotional contagion to leisure involvement = .51 (p < .001), and emotional contagion to flow = .50 (p < .001). In addition to the direct effect of emotional contagion on leisure benefits achieving significance, the other four paths of direct effect also achieved significance, as 0 did not appear between the lower and upper limits of the 95% confidence interval.
Indirect effects are described as the product of two path coefficients, which are the exogenous variable to mediating variable and the mediating variable to endogenous variable path coefficients. This mediating model had two indirect pathways. One was the emotional contagion to leisure involvement to leisure benefits pathway; in this pathway, leisure involvement was the mediating variable, and the value of the indirect effect was .20 (.51 times .38). The second pathway was emotional contagion to flow to leisure benefits, and the value of the indirect effect was .22 (.50 times .44).
The total effect is the sum of the direct effect and the indirect effect. The total effects of pathways without mediating variables are the direct effects. For the two pathways with mediating variables, the total effect of the emotional contagion-leisure involvement-leisure benefits pathway was .31, and the total effect of the emotional contagion-flow-leisure benefits pathway was .33.
In the mediating model, the direct effect of emotional contagion on leisure benefits was .11 (p > .05) and did not reach significance. Accordingly, leisure involvement and flow had full mediating effects between emotional contagion and leisure benefits. Whether these two mediating effects were statistically meaningful was verified using the built-in bootstrapping function in SmartPLS. For leisure involvement, t = 2.52, p = .012, and 95% CI [.05, .34]. The results for flow were t = 2.75, p = .006, and 95% CI [.08, .40]. The results indicated that the two mediating effects were statistically meaningful.

Discussion
Six hypotheses were posited in this study and, based on the analysis of the mediating model, only "H 1 : Among university students, emotional contagion and leisure benefits are positively correlated" was not supported. The jel.ccsenet.org Journal of Education and Learning Vol. 11, No. 6;2022 other hypotheses were supported. Discussions as below: No research on the relationships of emotional contagion (including positive and negative emotions) with other variables in sports and leisure has been conducted in Taiwan. According to doctrine, theory, and inference, "H 2 : Among university students, emotional contagion and leisure involvement are positively correlated" was proposed. Some studies have argued that positive emotions, by being shared, cause others on a team to feel them and influence the team's work or sports involvement (Chang et al., 2017;Savardelavar & Arvin, 2012). As H 2 was supported, positive or negative emotional contagion in sports and leisure affects leisure involvement.
Similarly, the correlation between emotional contagion and flow among university students has been substantiated by other studies. Based on doctrine and theoretical inference, a positive correlation was posited; "H 3 : Among university students, emotional contagion and flow are positively correlated," which was supported by the empirical results. During sports and leisure, members unconsciously and synchronously imitate the emotional expressions of others (Hatfield et al., 1994). Using sports spectating as an example, emotional contagion allows individuals to be intensely focused on watching the game, which is in line with the phenomenon of flow (as mentioned in the theory of flow), which is to be in a state of enthusiasm and total concentration, forgetting one's self (Nakamura & Csikszentmihalyi, 2009).
"H 4 : Among university students, flow and leisure benefits are positively correlated" was also supported. This is consistent with Chen's (2011) discovery that flow and leisure benefits are significantly correlated among participants in road running. Road running is a leisure sport and, thus, the results of the current study are supported. The correlation can be explained using broaden-and-build theory (Fredrickson, 2001(Fredrickson, , 2003, as flow is a positive emotion, and positive emotions can broaden individuals' abilities in behavioral thinking and building lasting new resources. These include intellectual resources, such as developing problem-solving techniques and absorbing new knowledge; social resources, such as solidifying existing and new social relationships (interpersonal relationships); physical resources, such as increasing physical coordination and improving health; and psychological resources, such as increasing resilience and positive attitudes and expanding one's sense of identity and goals. These four types of resources demonstrate the implications of leisure benefits-that physical benefits, psychological benefits, and social benefits are consistent.
"H 1 : Among university students, emotional contagion and leisure benefits are positively correlated" not being supported may have been due to the effects of leisure involvement and flow in the mediating model. In the first of the three steps verifying the mediating effects, the presence of a significant relationship between the exogenous and endogenous variables was confirmed; which is to say, a significant relationship was observed between emotional contagion and leisure benefits. However, once leisure involvement and flow were introduced, emotional contagion and leisure benefits were no longer significantly correlated. This also indicates that leisure involvement and flow have full mediating effects between emotional contagion and leisure benefits. Therefore, "H 5 : Among university students, flow has a mediating effect on the relationship between emotional contagion and leisure benefits" and "H 6 : Among university students, leisure involvement has a mediating effect on the relationship between emotional contagion and leisure benefits" were supported. This also indicates that emotional contagion in sports and leisure increases leisure involvement and flow, which then increase perceptions of leisure benefits. Therefore, the mediating roles of leisure involvement and flow cannot be overlooked in studies with university students on relationships between emotional contagion and leisure benefits in sports and leisure.
In terms of research limitations, this study is aimed at college students, so the results of the study are only applicable to college students. When inferred to other levels of education, it must be carefully considered. In addition, this study adopted purposive sampling, the sample size is 220 college students. For future research, the sample size can expand. Third, the different dimensions of emotional contagion and leisure benefits, flow and leisure involvement have been proposed by different scholars, so that the inference of results is limited to the same dimensions of four scales in the present study.
In In terms of further research, the present study focuses on the mediating effect of flow and leisure involvement, and also found that the two have a fully mediating role between the relationship of the emotional contagion and leisure benefit. Future studies can replace different mediating or regulating variables according to the literature.

Conclusion
Past studies have described a positive correlation between emotional contagion and leisure benefits, but in the mediating model in this study, the correlation was nonsignificant. This indicates that the correlation between emotional contagion and leisure benefits is produced by leisure involvement, while flow exerting their mediating effects between them and promoting the leisure benefits.

Notes
Note 1. Q 2 = 1 − (Σ D SSE D )/(Σ D SSO D ). D: missing distance. SSE: the sum of the square of the predicted error. SSO: the sum of the square of the observed value.

Copyrights
Copyright for this article is retained by the author, with first publication rights granted to the journal.
This is an open-access article distributed under the terms and conditions of the Creative Commons Attribution license (http://creativecommons.org/licenses/by/4.0/).