A Moderated Mediation Model for Facial Expression, Inner Feeling, and Psychological Safety to Work Engagement: An Empirical Study for Funeral Service Workers in Taiwan Area ()
1. Introduction
“Because I could not stop for Death.”
—Emily Dickinson (Dickinson, 1960)
The COVID-19 pandemic has underscored the importance of workplace safety (Subramony et al., 2022), particularly for frontline service industries like funeral directors. Frontline employees continued their duties despite the risk of contracting the virus through direct customer interactions (Aplin-Houtz et al., 2022). Hochschild (1983: p. 236) estimated that 38.1% of all occupations entail “substantial emotional labor.” Moreover, Wharton (1999) noted that more than half of workers engage in jobs that demand at least some emotional labor (EL).
Hochschild (1983) defined emotional labor as an occupation involving frequent public interactions, where employees must manage their own and others’ emotions, which are monitored and enforced by management. She argued that in service transactions, common expectations about appropriate emotional reactions exist, termed “feeling rules,” which describe the institutional norms regarding the appropriate type and amount of emotion to be displayed in specific situations. This includes surface acting, which involves managing facial expressions, and deep acting, which aligns genuine inner feelings with external displays. We propose that facial expression corresponds to surface acting, and inner feeling corresponds to deep acting, in the context of employee-focused emotional labor (regulation of feelings and emotional expression; Brotheridge & Grandey, 2002).
Ashforth and Humphrey (1993) observed that funeral service workers are expected to be somber and reserved. Cho (2020) further explained that these workers must balance providing services to grieving families with displaying a range of emotions, including strength, compassion, sympathy, and vulnerability. According to the Ministry of Interior and the Ministry of Labor, Taiwan area had 173,156 deaths in 2020, with 3,724 funeral service companies operating (MOI, 2020; MOL, 2022). This means each company handled an average of 47 deaths annually, indicating that funeral service workers face high job demands and emotional burdens. The negative mental health impacts have been exacerbated by the COVID-19 pandemic, leading Hicks et al. (2022) to emphasize the need for increased awareness of the stress, physical, and mental health issues funeral directors are experiencing due to COVID-19.
Previous studies have focused on minimizing the cost of emotional labor, primarily examining psychological capital (Fu, 2015; Hur, Rhee, & Ahn, 2016) and psychological well-being (Cropanzano, Weiss, & Elias, 2003; Steiger et al., 2021), with few addressing psychological safety. We suggest that psychological safety can reduce the cost of emotional labor and enhance employee work engagement. Edmondson (1999) defined psychological safety as the perception of the consequences of taking interpersonal risks in a specific context, such as the workplace. Therefore, funeral service workers who experience greater psychological safety at their workplace are likely to incur lower emotional labor costs and show increased work engagement during the COVID-19 pandemic.
Work engagement is defined as a positive, significant, and work-related mental state (Schaufeli et al., 2002). Various studies (Seo et al., 2023; Yoo & Arnold, 2014) have consistently found that while surface acting negatively affects work engagement, deep acting has a positive impact. We suggest that to enhance genuine feelings about work through psychological safety, aligning required emotions with deep acting, rather than expressing insincere or surface feelings, would be beneficial.
Due to the limited literature on funeral service workers’ display rules regarding emotional labor, this study explores psychological safety and work engagement by understanding the mediating effects of deep acting and the moderating effects of both surface acting and deep acting on work engagement among funeral service workers. The aim of this paper is to expand the literature on how emotional labor, psychological safety, and work engagement affect funeral service workers.
2. Literature Review and Hypotheses Development
2.1. Psychological Safety, Deep Acting and Work Engagement (WE)
Safety is a fundamental human need and an essential precondition for employee work engagement (Kahn, 1990; Subramony et al., 2022). Psychological safety refers to the freedom of self-expression without fear of negative consequences to self-image, status, or career (Kahn, 1990). Funeral service workers who experience a team environment perceived as psychologically safe at the workplace improve work engagement during the COVID-19 pandemic.
Schaufeli et al. (2002) described work engagement as a positive and fulfilling attitude towards work, characterized by vigor, dedication, and absorption. Vigor involves high energy levels, mental resilience, effort, and persistence. Dedication encompasses the significance, enthusiasm, inspiration, pride, and challenge that work presents. Absorption means being fully focused and joyfully immersed in work, experiencing swift passage of time and difficulty detaching from work tasks (Schaufeli, Bakker, & Salanova, 2006).
Kahn (1990) argued that engaged employees exhibit their physical, emotional, and cognitive traits while performing tasks, benefiting both the organization and the individual. Building on Kahn’s work, work engagement is defined as a “psychological connection with the performance of work tasks, rather than an attitude towards the characteristics of the organization or the work itself.” This implies that optimizing the psychological safety of the workplace contributes to the engagement of funeral service workers. Therefore, we predict:
Hypothesis 1: Psychological safety is positively linked with work engagement.
Deep acting (DA) is defined as the induction of genuine feelings (Hochschild, 1983: p. 35). Previous research has shown how emotion work impacts employees’ mental and emotional resources, influencing their work outcomes (Brotheridge & Lee, 2002; Hill et al., 2020). Emotional labor that involves deep acting requires an effort to genuinely experience the emotions one is supposed to display (Wharton, 2009). Deep acting also involves modifying physiological arousal and cognitions through various techniques (Brotheridge & Grandey, 2002). Moreover, funeral service workers use deep acting to experience somberness, grief, and mourning, which enhances psychological strength. This practice helps them navigate the emotional complexity of “the separation between loved ones in life or death” with profound personal empathy.
Olivier and Rothmann (2007) investigated the antecedents of work engagement and found that psychological meaningfulness and psychological availability are significant predictors. Psychological meaningfulness is defined as the feeling of receiving a return on the investment of oneself in terms of physical, cognitive, or emotional energy (Kahn, 1990: pp. 703-704). Psychological availability is the sense of having the necessary physical, emotional, or psychological resources to engage at a given moment (Kahn, 1990: p. 714). The psychological safety and emotional labor of funeral service workers may also relate to psychological meaningfulness and availability. Due to their sympathy towards customers’ feelings of sorrow and grief, funeral service workers may enhance their inner feelings at work, thereby spontaneously evoking vigor, dedication, and absorption. The following hypotheses are formulated regarding the relationship between psychological safety, deep acting, and work engagement.
Hypothesis 2(H2): Psychological safety is significantly positively linked with deep acting.
Hypothesis 3(H3): Deep acting is significantly positively linked with work engagement.
2.2. Surface/Deep Acting as a Moderator
Surface acting (SA), defined by Hochschild (1983) as using the body, not the soul, as the main tool of trade, involves faking or suppressing emotions at work. This behavior is linked to increased emotional exhaustion among workers (Van Dijk & Brown, 2006; Glomb & Tews, 2004) and contributes to feelings of depersonalization and reduced personal accomplishment (Brotheridge & Grandey, 2002). Surface acting requires managing emotions by suppressing genuine feelings and displaying desired emotional expressions (Hill et al., 2020). Despite funeral service workers feeling psychologically safe, they are still mandated to display facial expressions to customers, which diminishes their genuine feelings (deep acting) at all times. If employees maintain authentic expressions (deep acting) while feeling psychologically safe in customer interactions, their work engagement is enhanced. In essence, deep acting strengthens the link between psychological safety and work engagement. Thus, we hypothesized the following:
Hypothesis 4: The relationship between psychological safety and deep acting is moderated by surface acting, weakening the relationship.
Hypothesis 5: The relationship between psychological safety and work engagement is moderated by deep acting, strengthening the relationship.
2.3. Mediating Effect of Deep Acting
Kahn (1990) claims that meaningfulness, safety, and availability are psychological conditions influencing employee engagement, as employees invest their whole selves (body and mind) in their jobs based on their psychological experiences of self-in-role. Meanwhile, Bakker (2022) argued that work engagement results from various social-psychological processes, is contagious, and can be significantly influenced by colleagues, leaders, and intimate partners. Grandey (2003) defined deep acting as the extent to which an employee modifies their feelings to match expressions. Therefore, despite funeral service workers fearing the risk of coronavirus infection, they still engage in deep acting at work, supported by their colleagues, leaders, and partners. Based on these findings, we hypothesized the following:
Hypothesis 6: Employees’ psychological safety positively influences work engagement via the mediating role of deep acting.
Hypothesis 7: Surface acting moderates the relationship between psychological safety and deep acting, strengthening the mediation effect of deep acting when surface acting is low rather than high. Figure 1 presents the proposed conceptual framework model for our hypotheses.
Figure 1. Conceptual framework model adopted in this study.
3. Methods
3.1. Sampling Procedure and Participants
According to the Taiwan area Ministry of Labor’s statistical data for 2022, presented in July 2021, the total number of funeral service companies is 3724. This study distributed the questionnaire through Google Forms and posted the links on social platforms (e.g., the Funeral Service Association social group) such as Line, Facebook, Twitter, and WeChat to increase exposure to our research context. From July 2022 to 30 December 2022, the study collected samples from employees in the Taiwan area funeral industry. Over 600 funeral service employees completed the questionnaire, with 533 valid responses, achieving an 88.83% response rate.
3.2. Measurement
The main instrument of this research was a questionnaire, which consisted of four parts: 1) demographic characteristics (age, gender, marital status, etc.); 2) work engagement (9 items); 3) emotional labor, measuring surface (3 items) and deep acting (4 items); 4) psychological safety (7 items). The details are as follows:
To measure work engagement, we administered the Utrecht Work Engagement Scale (UWES; Schaufeli et al., 2006), which comprises three subscales: vigor (3 items, e.g., “I am always full of energy at work”), dedication (3 items, e.g., “My work inspires me a lot”), and absorption (3 items, e.g., “I am so engaged in my work that I lose track of time”). Items are rated on a 5-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). The scale’s alpha reliability was 0.87.
Emotional labor was measured by revising items developed by Lin (2000). We assessed both surface acting (SA; also known as facial expression; 3 items, e.g., “You often have to pay attention to body language when interacting with customers”) and deep acting (DA; also known as inner feeling; 4 items, e.g., “You must do your best to motivate consumer behavior, even if you are in a bad mood and must overcome this condition”). Subjects responded to these questions on a 5-point scale, ranging from 1 for “strongly disagree” to 5 for “strongly agree.” The alpha reliability of this scale was 0.90.
The psychological safety scale by Edmondson (1999) was used to assess the extent to which an individual feels comfortable being themselves and expressing their viewpoints at work, as opposed to feeling threatened in the workplace. This study employed a 7-item scale with a 5-point Likert response format, ranging from 1 (strongly disagree) to 5 (strongly agree). Sample items include “I can assert myself without any restraint and express my ideas in the company”. The scale’s alpha reliability was 0.82.
3.3. Data Analysis
SmartPLS 3.0 and SPSS 26 were used for statistical analysis. Descriptive statistics, including demographic variables, means, standard deviations, and correlations, were utilized for data analysis. Confirmatory Factor Analysis (CFA) tested the factorial validity of the measuring instruments. Composite reliability was assessed using Raykov’s rho coefficients with a cut-off value of 0.70 (Raykov, 2009). The Partial Least Squares Path Modeling Method (Hair et al., 2019) was employed to examine the hypothesized relationships, and Hayes PROCESS macro-Model 7 (ver. 4.1) was used to test the moderated mediation effects.
3.4. Some Common Mistakes
The word “data” is plural, not singular.
The subscript for the permeability of vacuum 0, and other common scientific constants, is zero with subscript formatting, not a lowercase letter “o”.
In American English, commas, semi-/colons, periods, question and exclamation marks are located within quotation marks only when a complete thought or name is cited, such as a title or full quotation. When quotation marks are used, instead of a bold or italic typeface, to highlight a word or phrase, punctuation should appear outside of the quotation marks. A parenthetical phrase or statement at the end of a sentence is punctuated outside of the closing parenthesis (like this). (A parenthetical sentence is punctuated within the parentheses.)
A graph within a graph is an “inset”, not an “insert”. The word alternatively is preferred to the word “alternately” (unless you really mean something that alternates).
Do not use the word “essentially” to mean “approximately” or “effectively”.
In your paper title, if the words “that uses” can accurately replace the word “using”, capitalize the “u”; if not, keep using lower-cased.
Be aware of the different meanings of the homophones “affect” and “effect”, “complement” and “compliment”, “discreet” and “discrete”, “principal” and “principle”.
Do not confuse “imply” and “infer”.
The prefix “non” is not a word; it should be joined to the word it modifies, usually without a hyphen.
There is no period after the “et” but a period after the “al” in the Latin abbreviation “et al.”.
The abbreviation “i.e.” means “that is”, and the abbreviation “e.g.” means “for example”.
4. Results
4.1. Description of Sample
The demographic analysis of research respondents shows that most were above 50 years old (22.9%). Of these 533 respondents, 320 (60%) were men, and 213 (40%) were women. Regarding tenure, 231 (43.3%) of the respondents had over 15 years of experience, 67 (12.6%) had 4 - 5 years, and 1 (0.2%) had one year or less. The majority of participants’ education level was a Bachelor’s degree (254; 47.7%), followed by Senior High School (94; 17.6%), College (92; 17.3%), Master’s (72; 13.5%), Junior High School (12; 2.3%), and Ph.D. (9; 1.7%).
4.2. Common Method Bias
There is a possibility of common method bias when measuring constructs in the same survey or self-report questionnaire. To mitigate this, we employed procedural and statistical methods. The questionnaire items were arranged in a random order, distributing the four constructs throughout, and we ensured complete protection of the respondents’ anonymity (Podsakoff et al., 2003). Additionally, we conducted Harman’s single factor test using Exploratory Factor Analysis (EFA). The EFA results revealed four distinct factors with eigenvalues above 1, explaining 65.72% of the total variance. The first extracted factor accounted for 41.83% of the total variance, which is less than 50%, indicating it did not represent the majority of the variance (see Table 1).
Table 1. Common method bias (CMV).
| Component |
Initial eigenvalues |
Extraction sums of squared loadings |
| Total |
% of variance |
Cumulative % |
Total |
% of variance |
Cumulative % |
| 1 |
9.62 |
41.83 |
41.83 |
9.62 |
41.83 |
41.83 |
| 2 |
2.64 |
11.47 |
53.31 |
2.64 |
11.47 |
53.31 |
| 3 |
1.76 |
7.65 |
60.95 |
1.76 |
7.65 |
60.95 |
| 4 |
1.10 |
4.77 |
65.72 |
1.10 |
4.77 |
65.72 |
4.3. Descriptive Statistics
Pearson correlations, means, and standard deviations for all study variables are presented in Table 2 (N = 533), showing that all research variables are significantly correlated, with psychological safety being strongly correlated with work engagement.
Table 2. Mean, SD and inter-correlations between survey variables (N = 533).
| Var. |
M |
SD |
DA |
SA |
PS |
VG |
DC |
AP |
WE |
| DA |
3.838 |
0.642 |
- |
|
|
|
|
|
|
| SA |
4.267 |
0.612 |
0.518** |
- |
|
|
|
|
|
| PS |
3.559 |
0.802 |
0.287** |
0.322** |
- |
|
|
|
|
| VG |
3.560 |
0.893 |
0.339** |
0.428** |
0.567** |
- |
|
|
|
| DC |
3.784 |
0.832 |
0.349** |
0.420** |
0.617** |
0.746** |
- |
|
|
| AP |
3.894 |
0.739 |
0.402** |
0.451** |
0.526** |
0.679** |
0.765** |
- |
|
| WE |
3.746 |
0.744 |
0.399** |
0.477** |
0.631** |
0.903** |
0.925** |
0.888** |
- |
Note: DA = deep acting; SA = surface acting; PS = psychological safety; VG = vigor; DC = dedication; AP = absorption; WE = work engagement; *p < 0.05; **p < 0.01.
4.4. Assessment of Measure Model
In this study, item loading, Cronbach’s alpha, average variance extracted (AVE), and composite reliability (CR) were critically analyzed, with results presented in Table 3. According to Hair et al. (2012), item loading should exceed 0.6, a criterion met by all items in our study. The inner VIF values for the research variables were all below 5, indicating that multicollinearity was not an issue. Regarding AVE, Hair et al. (2019) suggested it should surpass 0.5; in our research, AVE values ranged from 0.617 to 0.648, which is acceptable and indicates good construct reliability. Additionally, CR values varied from 0.84 to 0.943, aligning with the values recommended by Hair et al. (2019). In addition, Table 4 presents Dijkstra and Henseler (2015) rho_A coefficient, which Wong (2019) highlighted as the preferred method for assessing the reliability of PLS construct scores. According to the modern view of PLS, the rho_A coefficient is recommended over Cronbach’s alpha and composite reliability.
Table 3. Measure model of parameter estimation (N = 533).
| Construct |
Item |
Loading |
VIF |
α |
CR |
AVE |
| WE |
vg01 |
0.824 |
3.711 |
0.932 |
0.943 |
0.648 |
| vg02 |
0.819 |
3.517 |
|
|
|
| vg03 |
0.819 |
2.212 |
|
|
|
| dc04 |
0.758 |
3.476 |
|
|
|
| dc05 |
0.868 |
2.649 |
|
|
|
| dc06 |
0.811 |
3.465 |
|
|
|
| ap07 |
0.862 |
2.703 |
|
|
|
| ap08 |
0.817 |
2.384 |
|
|
|
| ap09 |
0.758 |
1.804 |
|
|
|
| DA |
da01 |
0.786 |
1.746 |
0.794 |
0.866 |
0.617 |
| da02 |
0.803 |
1.835 |
|
|
|
| da04 |
0.790 |
1.503 |
|
|
|
| da07 |
0.763 |
1.485 |
|
|
|
| SA |
sa01 |
0.732 |
1.375 |
0.719 |
0.840 |
0.638 |
| sa02 |
0.833 |
1.557 |
|
|
|
| sa03 |
0.827 |
1.376 |
|
|
|
| PS |
ps01 |
0.801 |
2.598 |
0.936 |
0.945 |
0.634 |
| ps02 |
0.825 |
3.573 |
|
|
|
| ps03 |
0.838 |
3.979 |
|
|
|
| ps04 |
0.779 |
2.596 |
|
|
|
| ps05 |
0.728 |
2.176 |
|
|
|
| ps06 |
0.792 |
2.617 |
|
|
|
| ps07 |
0.842 |
2.963 |
|
|
|
Table 4. Composite reliability and average extracted variance (N = 533).
| Variable |
α |
rho_A |
CR |
AVE |
| DA |
0.749 |
0.797 |
0.866 |
0.617 |
| SA |
0.719 |
0.739 |
0.840 |
0.638 |
| PS |
0.936 |
0.937 |
0.945 |
0.634 |
| WE |
0.932 |
0.934 |
0.943 |
0.648 |
We followed Tenenhaus et al. (2005) and computed the Goodness of Fit (GOF) to understand the overall quality of the proposed model. The GOF is calculated as:
The results show that the GOF is 0.796, surpassing the cut-off criterion of 0.36 for a large effect size (Wetzels et al., 2009).
Table 4 displays construct reliability and validity values, indicating that Cronbach’s alpha for surface acting (SA) is acceptable at 0.719, and for psychological safety (PS), it is very strong at 0.932. These figures confirm the model’s validity and reliability. Cronbach’s alpha measures internal consistency, but Bagozzi and Yi (1988), along with Hair et al. (2012), have suggested that Composite Reliability (CR) offers more precise consistency values. Fornell and Larcker (1981) recommend that CR values for all latent variables should exceed 0.6, with our results showing CR values ranging from 0.84 to 0.945, meeting this criterion. The Average Variance Extracted (AVE) assesses the extent to which variances in measured variables explain the latent variables, with a recommended standard value above 0.5 (Fornell & Larcker, 1981). Our findings also indicate that the AVE estimates for all constructs exceed the recommended threshold of 0.5 (DA = 0.617; SA = 0.638; PS = 0.634; WE = 0.648).
The Fornell-Larcker criterion (1981; Table 5) is a widely used and conservative method for assessing discriminant validity in PLS-SEM. To establish discriminant validity, the square root of the average variance extracted (AVE) of each latent variable should be greater than the latent variable correlations (LVC). Table 3 and Table 4 clearly show that discriminant validity is met for this research.
Table 5. Fornell-Larcker criterion to test discriminant validity.
| Variable |
DA |
SA |
PS |
WE |
| DA |
0.786* |
|
|
|
| SA |
0.535 |
0.799* |
|
|
| PS |
0.276 |
0.328 |
0.796* |
|
| WE |
0.412 |
0.477 |
0.637 |
0.805* |
Note: * = the square roots of AVE and within the diagonal are the correlations among variables.
4.5. Structural Model Analysis
PLS path modeling analysis (Henseler, 2012) was conducted using the non-parametric bootstrap procedure to provide a significant test based on the path coefficient for hypothesis testing in this study. Path coefficients represent the hypothesized relationships among the constructs. Table 6 and Figure 2 summarize the study’s findings, demonstrating the relationship between the path coefficients, standard deviation (STDEV), probability value (p-value), and each construct’s outcome. Both Table 6 and Figure 2 show that psychological safety positively and significantly affects work engagement, supporting hypothesis H1 (Beta = 0.544; t-value = 15.269; p = 0.000). Hypothesis 2 supported the proposed relationship between psychological safety and deep acting (Beta = 0.106; t value = 2.453; p = 0.014). Hypothesis 3 supported significant positive effects between deep acting and work engagement (Beta = 0.263; t value = 7.037; p < 0.001). The analysis showed that surface acting had negative moderating effects and insignificant direct effects on the relationship between psychological safety and deep acting, with a path coefficient (Beta = 0.031; t value = 0.3; p = 0.549), supporting hypothesis 4. Hypothesis 5, which reported moderating effects of deep acting on the relationship between psychological safety and work engagement, was supported by a path coefficient (Beta = 0.079; t value = 2.43; p = 0.015). Additionally, a strong correlation was found between psychological safety and work engagement (Beta = 0.544; t value = 15.269; p < 0.001).
Table 6. Path coefficient of structural equation model analysis.
| Procedure |
Beta |
Mean |
STDEV |
t value |
p |
| Deep acting → work engagement (H3) |
0.263 |
0.267 |
0.037 |
7.034 |
0.000 |
| Moderating Effect 1 → Deep acting (H4) |
0.031 |
0.038 |
0.052 |
0.600 |
0.549 |
| Moderating Effect 2 → work engagement (H5) |
0.079 |
0.075 |
0.032 |
2.430 |
0.015 |
| psychological safety → Deep acting (H2) |
0.106 |
0.105 |
0.043 |
2.453 |
0.014 |
| psychological safety → surface acting |
0.328 |
0.329 |
0.046 |
7.202 |
0.000 |
| psychological safety → work engagement (H1) |
0.544 |
0.547 |
0.036 |
15.269 |
0.000 |
| surface acting → Deep acting |
0.507 |
0.506 |
0.035 |
14.522 |
0.000 |
Note: STDEV = standard deviation.
Figure 2. Results of structural equation models analysis in this study.
The SMART-PLS algorithm was run, determining a maximum of 300 interactions with a stop criterion of 7, using a path weighting scheme to maximize the R-square value for the model’s endogenous latent variables, as shown in Figure 2.
R squared is a coefficient of determination that indicates the variance explained by an independent variable on a dependent variable. Deep acting increases by 7.9% due to psychological safety, while work engagement increases by 17.5% due to psychological safety. Moreover, the adjusted R squared for all constructs is less than the R squared, indicating that all the independent variables significantly reveal (Table 7).
Table 7. Moderation and mediation analysis in PLS-SEM (N = 533).
| Hypothesis |
Path |
Coefficient |
t value |
p-value |
Support |
| H4 |
SA* × PS → DA |
−0.019 |
0.178 |
0.859 |
No |
| H5 |
DA# × PS → WE |
0.095 |
2.597 |
0.009*** |
Yes |
| Mediation effect of DA on PS and WE in PLS-SEM |
| Specific In. effect |
Original sample |
Sample mean |
STDEV |
T statistic |
p |
| PS → DA → WE |
0.116 |
0.121 |
0.027 |
4.284 |
0.000*** |
| Path coefficients |
Original sample |
Sample mean |
STDEV |
T statistic |
p |
| DA → WE |
0.414 |
0.419 |
0.04 |
10.429 |
0.000*** |
| PS → DA |
0.28 |
0.287 |
0.046 |
6.101 |
0.000*** |
| Summary of R2 (H6) |
| Endogenous construct |
R square |
R square adjusted |
| Deep acting |
0.079 |
0.077 |
| Work engagement |
0.172 |
0.17 |
Note: * moderator; #mediator; *** = p < 0.001; Bootstrapping = 5000; STDEV = Standard deviation. Asterisks *, **, *** indicate two-tailed significance at the 0.10, 0.05, and 0.01 levels, respectively.
4.6. Moderated Mediation Effect Analysis
To test moderated mediation effects, we applied Model 7 of the Process Macro by Hayes (2017), as shown in Table 8. This model assesses both the moderating and mediating effects. Following Hayes’s (2017) guidelines, we probed conditional indirect effects for significance at ±1SD and examined the index of moderated mediation as an additional test of significance for the conditional indirect effect. The results, as depicted in Table 8, indicate that the indirect effect of surface acting on work engagement through psychological safety was significant, as evidenced by the index of moderated mediation. The bias-corrected 95% confidence interval for the index ranged from −0.022 to 0.059, not including zero. Therefore, we plotted the predictive relationship between psychological safety and work engagement, separately for low and high levels of surface acting (M ± 1SD). The results showed that as surface acting decreased among funeral service workers, the predictive effect of psychological safety on work engagement gradually increased.
Table 8. Results of moderated mediation model (N = 533).
| Outcome variable |
R |
R2 |
F |
p |
| Deep acting |
0.534 |
0.286 |
70.491 |
0.000*** |
|
B |
SE |
t |
p |
| Constant |
1.794 |
0.635 |
2.826 |
0.004*** |
| Psychological safety |
−0.323 |
0.186 |
−0.173 |
0.862 |
| Surface acting |
0.392 |
0.146 |
2.686 |
0.007*** |
| PS × SA |
0.032 |
0.042 |
0.761 |
0.447 |
| Outcome variable |
R |
R2 |
F |
p |
| Work engagement |
0.671 |
0.450 |
216.694 |
0.000*** |
|
B |
SE |
t |
p |
| Constant |
0.833 |
0.160 |
5.191 |
0.000*** |
| Psychological safety |
0.522 |
0.031 |
16.734 |
0.000*** |
| Deep acting |
0.275 |
0.039 |
7.054 |
0.000*** |
| Conditional indirect effect |
| Group |
SA |
Effect |
SE |
LLCI |
ULCI |
| M − 1 SD |
3.655 |
0.023 |
0.018 |
−0.019 |
0.053 |
| M |
4.267 |
0.028 |
0.011 |
0.008 |
0.049 |
| M + 1 SD |
4.879 |
0.034 |
0.015 |
0.009 |
0.068 |
| Conditional direct effect |
| Effect |
SE |
t |
p |
LLCI |
ULCI |
| 0.522 |
0.031 |
16.734 |
0.000 |
0.461 |
0.584 |
| Index of moderated mediation |
|
Index |
Boot SE |
Boot LLCI |
Boot ULCI |
| Surface acting |
0.009 |
0.021 |
−0.022 |
0.059 |
Note: W = moderator; Number of bootstrap samples = 5000; Level of confidence = 95%. Asterisks *, **, *** indicate two-tailed significance at the 0.10, 0.05, and 0.01 levels, respectively.
5. Discussion
Our study aimed to examine the relationship between psychological safety, surface acting (facial or fake expressions), deep acting (inner or genuine feelings), and work engagement (H1-H3). Initially, we hypothesized that deep acting would mediate (H5) and moderate (H6) the relationship between psychological safety and work engagement. We also explored surface acting as a moderator between psychological safety and deep acting, which was not supported (H4). Funeral service workers often appear generous and sublime, offering their services with sympathy, which may lead them to exhibit genuine rather than fake feelings. The results indicated that funeral service workers who experienced higher perceived psychological safety through their leaders, teams, and colleagues were more likely to reduce their reliance on facial expressions and enhance their psychological well-being, thus positively linking psychological safety to deep acting (e.g., Cropanzano et al., 2003; Hur et al., 2016). Our findings supported the hypothesized moderated mediation model (H7; see Table 8). The results demonstrate that surface acting moderated the relationship between psychological safety and deep acting, such that the mediation relationship is stronger when surface acting is low rather than high. This study builds on previous research (e.g., Fu, 2015; Steiger et al., 2021) by providing a detailed examination of moderated mediation effects.
5.1. Theoretical Implications
Our study contributes to the literature in several areas, including psychological safety, emotional labor, work engagement, and the funeral service industry. First, it demonstrates how psychological safety can minimize the cost of emotional labor by serving as a prerequisite for deep acting and work engagement (Fu, 2015; Hur, Rhee, & Ahn, 2016). Second, while previous studies have generally agreed that surface acting negatively affects work engagement and deep acting has a positive effect (Seo et al., 2023; Yoo & Arnold, 2014), our study builds on this evidence by examining the moderated mediation role of emotional display rules, thereby enriching the literature. Lastly, few studies have explored psychological safety, emotional display rules, and work engagement among funeral service workers, a gap our study addresses, especially relevant in the era of the COVID-19 pandemic risk.
5.2. Practical Implications
The main practical implication of this study is that funeral service workers prefer to express their inner feelings rather than their facial expressions at work. This aligns with Cho’s (2020) perspective that funeral service workers must navigate a difficult path in serving grief-stricken families and exhibit a wide range of emotions, including strength, compassion, sympathy, somberness, and vulnerability. Hence, we suggest that funeral service employers should create a psychologically safe environment to improve employee engagement. Furthermore, the recruitment process should prioritize candidates from related funeral studies departments, such as life-and-death studies, funeral service, and funeral science, as these professionals possess increasingly valuable skills. Moreover, since work engagement stems from the psychological experiences of self-in-role (Kahn, 1990) and is influenced by the collective mood of colleagues (Bakker, 2022), we recommend that funeral service managers foster a team spirit among employees and promote a harmonious workplace environment, rather than focusing excessively on doctrines, rules, and training.
5.3. Research Limitations and Future Studies
Possible limitations of the present study should be considered. One major limitation is the reliance on self-report questionnaires, which may not accurately capture all respondents’ true feelings. Including both customers (grief-stricken families) and funeral service workers in future studies could provide more comprehensive insights than self-reported data alone. Although our study demonstrated the relationship among psychological safety, emotional display rules, and work engagement, and also highlighted the negative effect of surface acting and the positive effect of deep acting, many potential predictors, mediators, and moderators were not explored. Additionally, the question of how to minimize the cost of emotional labor remains, despite the extensive literature on the topic. Therefore, we suggest incorporating more predictors related to emotional labor and work engagement in future research.
6. Conclusion
Understanding how psychological safety influences workers’ emotional labor display (facial expression/inner feeling) and work engagement is particularly important in today’s funeral service industry. This study has raised more concern about these issues, especially among frontline service employees in the funeral service sector amid the spread of coronavirus. Indeed, “I could not stop for death” (Dickinson, 1960), which has extended the range of fear and sympathy.