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

This study aims to investigate the effects of psychological safety on work engagement with moderating mediation roles of surface acting and deep acting. The paper examines the mediating role of deep acting in the relationship of psychological safety to work engagement during the COVID-19 pandemic. The empirical data of this study has been collected via simple random sampling through an online survey with a self-report questionnaire involving funeral service workers (N = 533) from cemeteries, morgues, crematoria, and funeral agencies in Taiwan area. Results reveal that inner feeling (deep acting) and psychological safety positively affect work engagement, while facial expression (surface acting) and psychological safety negatively affect work engagement. Also, deep acting strengthened the positive relationship between psychological safety and work engagement, resulting in a moderated mediation effect whereby surface acting moderates the indirect effect of psychological safety and deep acting on work engagement. The findings of this study provide theoretical and practical implications within the scope of psychological safety and work engagement in the frequency of interaction of emotional labor in the workplace.

• Funeral Service Worker
• Facial Expression
• Inner Feeling
• Psychological Safety
• Work Engagement
• COVID-19 Pandemic

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

Figure 1. Conceptual framework model adopted in this study.

3. Methods

3.1. Sampling Procedure and Participants

According to the statistical data of Taiwan region’s Labor Department 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.

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

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

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

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

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:

$\text{GOF}=\sqrt{\overline{AVE}\times {\bar{R}}^2}=\sqrt{0.634\times 0.351}=0.796$

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

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

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

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

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

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.

References

1. Aplin-Houtz, M. J., Sanders, M. G., Lane, E. K.,&Jefferies, T. C. (2022). Coughing Customer: The Interplay of the Behavioral Immune System and Emotional Labor of Frontline Workers. Employee Responsibilities and Rights Journal, 35, 395-415. >https://doi.org/10.1007/s10672-022-09416-3

2. Ashforth, B. E.,&Humphrey, R. H. (1993). Emotional Labor in Service Roles: The Influence of Identity. Academy of Management Review, 18, 88-115. >https://doi.org/10.2307/258824

3. Bagozzi, R. P.,&Yi, Y. (1988). On the Evaluation of Structural Equation Models. Journal of the Academy of Marketing Science, 16, 74-94. >https://doi.org/10.1007/BF02723327

4. Bakker, A. B. (2022). The Social Psychology of Work Engagement: State of the Field. Career Development International, 27, 36-53. >https://doi.org/10.1108/CDI-08-2021-0213

5. Brotheridge, C. M.,&Grandey, A. A. (2002). Emotional Labor and Burnout: Comparing Two Perspectives of “People Work”. Journal of Vocational Behavior, 60, 17-39. >https://doi.org/10.1006/jvbe.2001.1815

6. Brotheridge, C. M.,&Lee, R. T. (2002). Testing a Conservation of Resources Model of the Dynamics of Emotional Labor. Journal of Occupational Health Psychology, 7, 57-67. >https://doi.org/10.1037/1076-8998.7.1.57

7. Cho, C. J. (2020). The Work of the Funeral Director: Emotional Labor, Family, and Friends. University of Central Oklahoma.

8. Cropanzano, R., Weiss, H. M.,&Elias, S. M. (2003). The Impact of Display Rules and Emotional Labor on Psychological Well-Being at Work. In P. L. Perrewe,&D. C. Ganster (Eds.), Emotional and Physiological Processes and Positive Intervention Strategies (pp. 45-89). Emerald Group Publishing Limited. >https://doi.org/10.1016/S1479-3555(03)03002-6

9. Dickinson, E. (1960). Because I could not stop for Death. ProQuest LLC.

10. Dijkstra, T. K.,&Henseler, J. (2015). Consistent Partial Least Squares Path Modeling. MIS Quarterly, 39, 297-316. >https://doi.org/10.25300/MISQ/2015/39.2.02

11. Edmondson, A. (1999). Psychological Safety and Learning Behavior in Work Teams. Administrative Science Quarterly, 44, 350-383. >https://doi.org/10.2307/2666999

12. Fornell, C.,&Larcker, D. F. (1981). Evaluating Structural Equation Models with Unobservable Variables and Measurement Error. Journal of Marketing Research, 18, 39-50. >https://doi.org/10.1177/002224378101800104

13. Fu, C. S. (2015). The Effect of Emotional Labor on Job Involvement in Preschool Teachers: Verifying the Mediating Effect of Psychological Capital. Turkish Online Journal of Educational Technology-TOJET, 14, 145-156.

14. Glomb, T. M.,&Tews, M. J. (2004). Emotional Labor: A Conceptualization and Scale Development. Journal of Vocational Behavior, 64, 1-23. >https://doi.org/10.1016/S0001-8791(03)00038-1

15. Grandey, A. A. (2003). When “the Show Must Go on”: Surface Acting and Deep Acting as Determinants of Emotional Exhaustion and Peer-Rated Service Delivery. Academy of management Journal, 46, 86-96. >https://doi.org/10.2307/30040678

16. Hair, J. F., Risher, J. J., Sarstedt, M.,&Ringle, C. M. (2019). When to Use and How to Report the Results of PLS-SEM. European Business Review, 31, 2-24. >https://doi.org/10.1108/EBR-11-2018-0203

17. Hair, J. F., Sarstedt, M., Ringle, C. M.,&Mena, J. A. (2012). An Assessment of the Use of Partial Least Squares Structural Equation Modeling in Marketing Research. Journal of the Academy of Marketing Science, 40, 414-433. >https://doi.org/10.1007/s11747-011-0261-6

18. Hayes, A. F. (2017). Introduction to Mediation, Moderation, and Conditional Process Analysis: A Regression-Based Approach. Guilford Publications.

19. Henseler, J. (2012). PLS-MGA: A Non-Parametric Approach to Partial Least Squares-Based Multi-Group Analysis. In W. A. Gaul, et al. (Eds.), Challenges at the Interface of Data Analysis, Computer Science, and Optimization (pp. 495-501). Springer. >https://doi.org/10.1007/978-3-642-24466-7_50

20. Hicks, E., Mashinchi, G., Heppner, H., Bean, P.,&Maxson, C. (2022). PTSD Prevalence Rates of US Funeral Directors during the COVID-19 Pandemic: A Brief Report. Advanced Journal of Social Science, 10, 45-51. >https://doi.org/10.21467/ajss.10.1.45-51

21. Hill, N. S., Zhang, H., Zhang, X.,&Ziwei, Y. (2020). The Impact of Surface and Deep Acting on Employee Creativity. Creativity Research Journal, 32, 287-298. >https://doi.org/10.1080/10400419.2020.1821564

22. Hochschild Arlie, R. (1983). The Managed Heart: Commercialization of Human Feeling. University of California Press.

23. Hur, W. M., Rhee, S. Y.,&Ahn, K. H. (2016). Positive Psychological Capital and Emotional Labor in Korea: The Job Demands-Resources Approach. The International Journal of Human Resource Management, 27, 477-500. >https://doi.org/10.1080/09585192.2015.1020445

24. Kahn, W. A. (1990). Psychological Conditions of Personal Engagement and Disengagement at Work. Academy of Management Journal, 33, 692-724. >https://doi.org/10.2307/256287

25. Lin, S. P. (2000). The Development Organization Emotional Labor Scale. Sun Yat-Sen Management Review, 3, 427-447.

26. Ministry of Interior MOI (2020). Yearly Bulletin of Interior Statistics. >https://www.moi.gov.tw/english/cl.aspx?n=7768

27. Ministry of Labor MOL (2022). Labor Statistics Survey 2021. >https://statfy.mol.gov.tw/statistic_DB.aspx

28. Olivier, A. L.,&Rothmann, S. (2007). Antecedents of Work Engagement in a Multinational Company. SA Journal of Industrial Psychology, 33, 49-56. >https://hdl.handle.net/10520/EJC186810 >https://doi.org/10.4102/sajip.v33i3.396

29. Raykov, T. (2009). Evaluation of Scale Reliability for Unidimensional Measures Using Latent Variable Modeling. Measurement and Evaluation in Counseling and Development, 42, 223-232. >https://doi.org/10.1177/0748175609344096

30. Schaufeli, W. B., Bakker, A. B.,&Salanova, M. (2006). The Measurement of Work Engagement with a Brief Questionnaire: A Cross-National Study. Educational and Psychological Measurement, 66, 701-716. >https://doi.org/10.1177/0013164405282471

31. Schaufeli, W. B., Salanova, M., Gonzalez-Roma, V.,&Bakker, A. B. (2002). The Measurement of Engagement and Burnout: A Two Sample Confirmatory Factor Analytic Approach. Journal of Happiness Studies, 3, 71-92. >https://doi.org/10.1023/A:1015630930326

32. Seo, E. C., Sim, Y. K., Kim, I., Seo, J. P., Ha, M. S.,&Kim, S. E. (2023). The Mediating Effect of Presenteeism on the Relationship between Emotional Labor and Work Engagement of Coaches for Disability Sports. International Journal of Environmental Research and Public Health, 20, Article No. 919. >https://doi.org/10.3390/ijerph20020919

33. Steiger, M., Bharucha, T. J., Venkatagiri, S., Riedl, M. J.,&Lease, M. (2021). The Psychological Well-Being of Content Moderators: The Emotional Labor of Commercial Moderation and Avenues for Improving Support. In Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems (pp. 1-14). Association for Computing Machinery. >https://doi.org/10.1145/3411764.3445092

34. Subramony, M., Golubovskaya, M., Keating, B., Solnet, D., Field, J.,&Witheriff, M. (2022). The Influence of Pandemic-Related Workplace Safety Practices on Frontline Service Employee Wellbeing Outcomes. Journal of Business Research, 149, 363-374. >https://doi.org/10.1016/j.jbusres.2022.05.040

35. Tenenhaus, M., Vinzi, V. E., Chatelin, Y. M.,&Lauro, C. (2005). PLS Path Modeling. Computational Statistics&Data Analysis, 48, 159-205. >https://doi.org/10.1016/j.csda.2004.03.005

36. Van Dijk, P. A.,&Brown, A. K. (2006). Emotional Labour and Negative Job Outcomes: An Evaluation of the Mediating Role of Emotional Dissonance. Journal of Management & Organization, 12, 101-115. >https://doi.org/10.5172/jmo.2006.12.2.101

37. Wetzels, M., Odekerken-Schröder, G.,&Van Oppen, C. (2009). Using PLS Path Modeling for Assessing Hierarchical Construct Models: Guidelines and Empirical Illustration. MIS Quarterly, 33, 177-195. >https://doi.org/10.2307/20650284

38. Wharton, A. S. (1999). The Psychosocial Consequences of Emotional Labor. The Annals of the American Academy of Political and Social Science, 561, 158-176. >https://doi.org/10.1177/000271629956100111

39. Wharton, A. S. (2009). The Sociology of Emotional Labor. Annual Review of Sociology, 35, 147-165. >https://doi.org/10.1146/annurev-soc-070308-115944

40. Wong, K. K. K. (2019). Mastering Partial Least Squares Structural Equation Modeling (PLS-Sem) with Smartpls in 38 Hours. IUniverse.

41. Yoo, J. J.,&Arnold, T. J. (2014). Customer Orientation, Engagement, and Developing Positive Emotional Labor. The Service Industries Journal, 34, 1272-1288. >https://doi.org/10.1080/02642069.2014.942653