How and for whom does customer incivility affect employee outcomes? A three-wave survey

Frontline employees (FLEs) are considered the cornerstone of excellent service and customer satisfaction in many service organizations (Chaudhuri et al., 2023). However, their potential is not always fully realized because they are often placed in a vulnerable position relative to customers, which makes them likely to become victims of customer incivility (CI) (Chaudhuri et al., 2023). CI is defined as “low-intensity deviant behavior, perpetrated by someone in a customer or client role, with ambiguous intent to harm an employee, in violation of social norms of mutual respect and courtesy” (Sliter et al., 2010, p. 468). It often manifests as directing frustration and anger toward FLEs, treating them with condescension and making unreasonable requests (Sliter et al., 2012). In addition to being stressful for FLEs, CI can have severe implications for service organizations that prioritize customer satisfaction (Chaudhuri et al., 2023). Indeed, it leads to detrimental consequences, including decreased job satisfaction (Loh et al., 2022), the capacity to satisfy customers (Al-Hawari et al., 2020; Baker and Kim, 2024), customer problem-solving behavior (Bani-Melhem et al., 2022), customer-oriented citizenship behavior (Agnihotri et al., 2023; Mostafa, 2022), job performance (Choi et al., 2023; Tam and Trang, 2024), increased retaliation (Agnihotri and Bhattacharya, 2022) and service sabotage (Gaan and Shin, 2024; Laeeque and Ali, 2025). Despite the attention that CI has garnered, several important research gaps remain.

First, a more comprehensive understanding of CI mechanisms is necessary for service organizations to implement effective CI practices. While several mechanisms underlying CI have been proposed (e.g. emotional, motivational and problem-solving mechanisms) (Auh et al., 2024; Fujii, 2025), researchers have yet to determine which of these mechanisms is dominant in the relationship between CI and employee outcomes. In this respect, while most research on CI has focused on job performance, with limited attention to the process by which CI undermines service recovery performance (SRP) and service employee creativity (SC) has received limited attention (Table 1).

SRP refers to “the effectiveness of employees dealing with customer complaints to the satisfaction of employees” (Boshoff and Allen, 2000, p. 70) reflecting related yet distinct aspects of job performance. SC refers to “the amount of new ideas generated and novel behaviors exhibited while performing specific job activities” (Agnihotri et al., 2014, p. 165). FLEs are uniquely positioned to identify and solve customer problems creatively and can respond effectively to service failures through customer interactions, turning customers into satisfied and loyal ones (Coelho et al., 2021; Sommovigo et al., 2019). Despite its managerial relevance, scant research has examined the effect of CI on SRP (with exceptions of Sommovigo et al., 2019; Zahoor and Siddiqi, 2023). They identified the mediating effects of emotional exhaustion from an emotional mechanism perspective. However, these studies did not consider the possibility of alternative mechanisms and were derived from either a student or retail bank sample, which limits generalizability. Meanwhile, Hur et al. (2016) examined the indirect effects of CI on SC through intrinsic motivation from a motivational mechanism perspective but found no significant results. Given the importance of gaining a more nuanced understanding of the mechanisms underlying CI, additional motivational mediators warrant further investigation.

Second, prior research has identified the deleterious effects of CI on employee outcomes; however, the degree of its impact varies depending on individual characteristics (Bani-Melhem et al., 2022; Yoon, 2022). Research has focused on individual characteristics that serve as coping resources for CI, such as emotional intelligence (Laeeque and Ali, 2025), employee resilience (Al-Hawari et al., 2020; Bani-Melhem et al., 2022), core self-evaluations (Yoon, 2022), achievement orientation (Lee and Gong, 2024) and mindfulness (Tam and Trang, 2024). However, existing studies have primarily examined individual characteristics as moderators within specific companies or industries (with the exceptions of Al-Hawari et al., 2020; Lee and Gong, 2024). Therefore, further investigation is needed to determine what individual characteristics are more susceptible to CI across service industries. Because many service organizations struggle to manage CI proactively (Shin et al., 2022), identifying individual characteristics that are more susceptible to the effects of CI across service industries offers broadly applicable insights for practitioners. These research gaps thus raise the following research questions (RQ):

To address these questions, this study applies conservation of resources (COR) theory to disentangle the process through which CI affects SRP and SC via customer orientation (CO) and examines the moderating effects of regulatory focus on the relationship between CI and CO. COR theory provides an appropriate theoretical foundation for understanding CI because it is a motivational theory that explains how stress occurs from an evolutionary perspective (Hobfoll, 2001; Hobfoll et al., 2018). CI is considered a severe stressor that FLEs face daily (Chaudhuri et al., 2023). CO refers to “a work value that captures the extent to which employees’ job perceptions, attitudes, and behaviors are guided by an enduring belief in the importance of customer satisfaction” (Zablah et al., 2012, p. 24). While it has been conceptualized as either an attitude or an individual difference (Bani-Melhem, 2020; Zablah et al., 2012), this study views it as a malleable attitude variable and a crucial mediator. As a critical resource for FLEs (Zablah et al., 2012), CO prompts FLEs to identify and solve customer problems (Jeng, 2018; Sommovigo et al., 2019). However, experiencing CI can diminish FLEs’ CO (Hwang et al., 2025), leading to a decline in SRP and SC. Therefore, developing a nuanced understanding of the mechanisms underlying CI is beneficial for both theory and practice.

In addition, this study examines the moderating role of regulatory focus (i.e. promotion and prevention focus) as a personal characteristic. Promotion-focused individuals emphasize the presence of positive outcomes and attempt to approach their “ideal self,” hopes and aspirations. In contrast, prevention-focused individuals seek to avoid negative outcomes and are concerned with their “ought self,” duties and obligations (Higgins, 1997, 1998). Promotion and prevention focus involves a distinct approach to goal pursuit (Higgins, 1997, 1998). Consequently, when FLEs experience CI, each regulatory focus shapes how they manage their resources, potentially leading to different responses (Agnihotri and Bhattacharya, 2022; Auh et al., 2024). Identifying regulatory focus as a personal characteristic that is more susceptible to CI can also update recruitment strategies in service organizations.

This study contributes to the literature in two ways. First, it bolsters the research on CI by clarifying the primary mechanism underlying CI with robust evidence. Existing studies have highlighted the mediating roles of emotional exhaustion (Sommovigo et al., 2019; Zahoor and Siddiqi, 2023) and intrinsic motivation (Hur et al., 2016) as mechanisms of CI. In contrast, this study identifies CO as the primary mechanism underlying CI, while ruling out the possibility that these alternative variables explain the effects. Second, existing research on CI has found that the deleterious effects of CI on employee outcomes vary by individual characteristics. Unlike existing studies that rely on data from specific companies or industries, this study enriches the body of knowledge by examining the moderating role of regulatory focus using data derived from full-time FLEs working in high-contact service industries in Japan. In summary, this study offers service organizations widely applicable insights into how to effectively handle CI in terms of human resource (HR) practices.

The fundamental premise of COR theory is that humans seek to protect, retain and obtain resources that typically have value (Hobfoll, 2001; Hobfoll et al., 2018). In this theory, resources are generally classified as personal (e.g. optimism and hope), object (e.g. home and car), energy (e.g. knowledge and time) and condition (e.g. tenure and marriage) (Hobfoll, 2001). COR theory proposes two seemingly opposing principles: resource protection and resource investment (Hobfoll et al., 2018).

The principle of resource protection posits that when individuals suffer resource loss events they attempt to protect their existing resources and prevent further depletion (Hobfoll, 2001). For example, when FLEs deal with uncivil customers, their resources become depleted. Consequently, they tend to engage in service sabotage to prevent further resource loss (Gaan and Shin, 2024). In contrast, the principle of resource investment posits that when individuals experience resource depletion, they attempt to invest their existing resources or acquire additional resources to replenish them (Hobfoll, 2001). For example, when FLEs deplete their resources because of CI, they occasionally retaliate to replenish their lost resources (e.g. self-esteem and fairness) (Agnihotri and Bhattacharya, 2022).

In the following sections, this study develops hypotheses based on COR theory. The conceptual model is illustrated in Figure 1.

Contrary to most studies that focus on emotional mechanisms (Hur et al., 2016; Zahoor and Siddiqi, 2023), this study highlights CO as a mediator in the relationships among CI, SRP and SC. CO is “a work value that captures the extent to which employees’ job perceptions, attitudes and behaviors are guided by an enduring belief in the importance of customer satisfaction” (Zablah et al., 2012, p. 24). According to COR theory (Hobfoll et al., 2018), when FLEs deal with uncivil customers during service encounters, their valuable resources (e.g. self-esteem and energy) are drained. This is because dealing with CI requires additional resource investment, especially when customers are irritating and make unreasonable requests (Sliter et al., 2012). These customer behaviors can result in FLEs losing their personal resources. In fact, empirical evidence shows that CI diminishes self-efficacy (Hur et al., 2022). Consequently, FLEs who encounter incivility from customers experience personal resource depletion, leading to a decline in their CO (Hwang et al., 2025).

Nevertheless, FLEs with diminished CO will attempt to achieve their goals by allocating their limited CO to certain behaviors (e.g. SRP and SC), although they will be unable to provide excellent service to all customers due to limited resources. This can be explained by the resource investment principle, which posits that individuals who experience resource loss invest their limited resources to restore their resources (Hobfoll, 2001; Hobfoll et al., 2018). Specifically, FLEs who recognize the importance of customer satisfaction will attempt to allocate their limited CO to satisfy selected customers, thereby elevating their work engagement (Zablah et al., 2012). Consequently, engaged FLEs are more likely to dedicate their resources to certain behaviors (e.g. SRP and SC) to achieve their goals (Olugbade and Karatepe, 2019; Zablah et al., 2012). These arguments are supported by empirical evidence showing that CO mediates the relationship between CI and emotional labor (i.e. deep and surface acting) (Hwang et al., 2025) as well as those among CI, service performance and SC (Fujii, 2025). Accordingly, this study proposes the following hypotheses:

Extending the hedonic principle, regulatory focus theory presumes that individuals pursue their desired goals through two different approaches: promotion and prevention focus (Higgins, 1997, 1998). Promotion-focused individuals emphasize the presence of positive outcomes and attempt to approach their “ideal self,” hopes and aspirations. In contrast, prevention-focused individuals seek to avoid negative outcomes and are concerned with their “ought self,” duties and obligations (Higgins, 1997, 1998). These constructs are distinct, and individuals tend to lean toward either promotion or prevention focus, depending on cultural variations (Agnihotri and Bhattacharya, 2022; Kurman et al., 2015). Based on COR theory, this study posits that regulatory focus moderates the negative relationship between CI and CO.

FLEs with a high promotion focus pursue work success and their “ideal self” (Higgins, 1998). Hence, even when they suffer through customer’s uncivil behavior, their prioritization of goals gives them more confidence, reducing the impact of CI on their CO (Higgins, 1998; Hobfoll et al., 2018). Conversely, FLEs with a low promotion focus are more susceptible to the adverse effects of CI than those with a high promotion focus because they are relatively less concerned with achieving goals (e.g. customer satisfaction). As such, when they experience CI, they feel dejection and a loss of self-esteem (Higgins, 1997; Hobfoll et al., 2018), undermining their CO.

Meanwhile, FLEs with a high prevention focus tend to be more sensitive to the presence of negative outcomes than those with a low prevention focus (Higgins, 1997, 1998). CI is perceived as a severe negative event by FLEs with a high prevention focus. As such, FLEs experiencing CI feel that they have failed to fulfill their duties at work, triggering agitation and reducing their self-esteem (Hobfoll et al., 2018). Consequently, their CO will deteriorate further. Conversely, FLEs with a low prevention focus are less susceptible to the effects of CI because they pay less attention to the presence of negative outcomes (Higgins, 1997, 1998). When they experience CI, they are less likely to experience agitation or a loss of self-esteem (Hobfoll et al., 2018). Consequently, their CO is relatively unaffected by CI. In line with this reasoning, empirical evidence shows that regulatory focus moderates the relationship between CI and customer conflict-solving behavior (Auh et al., 2024). Accordingly, this study proposes the following hypotheses:

Data were obtained using Freeasy (https://freeasy24.research-plus.net/), an online research platform in Japan. Despite the limitations of nonprobability sampling (Zickar and Keith, 2023), this method efficiently reaches suitable respondents. Recent research on CI has relied on online research platforms to access various samples (e.g. Fujii, 2025; Shin et al., 2022), supporting this data collection procedure. The population comprises full-time FLEs who frequently interact with customers face-to-face and have been working in high-contact service industries for at least one year. Owing to the interpersonal nature of CI, FLEs in high-contact service industries are more exposed to it (Chaudhuri et al., 2023). Furthermore, engaging in SRP and SC requires experience.

Considering causality and common method bias, a three-wave survey with a one-month lag was administered. These time lags almost correspond to those reported by Auh et al. (2024) and Fujii (2025). A screening test was conducted to exclude those in managerial positions with less than one year of work experience, not full-time FLEs in high-contact service industries, or those serving customers online. The screening test identified 823 qualified respondents, who were also asked about their demographics. These individuals were invited to participate in the first survey, and 677 completed it. The first survey included questions on CI and regulatory focus. After one month, these 677 respondents were invited to participate in the second survey, of which 555 were completed (retention rate = 82.0%). The second survey included questions on CO, intrinsic motivation, emotional exhaustion and positive/negative affectivity. One month later, these 555 respondents were invited to participate in the third survey, of which 471 were completed (retention rate = 84.9%). The third survey included questions on SRP and SC. The relatively high response rate can be partly explained by the monetary incentives offered to respondents. Following Zickar and Keith (2023), careless respondents who answered each attention check question incorrectly (please select this item “strongly disagree” for waves 1 and 3; please select this item “strongly agree” for wave 2) were excluded (n wave 1 = 71; n wave 2 = 34; n wave 3 = 31). Furthermore, to avoid negative influences on the data analysis, respondents exhibiting straight-lining were excluded (n wave 1 = 1; n wave 2 = 1; n wave 3 = 10). In total, 323 usable responses were available for data analysis (usable response rate = 39.2%).

Regarding the sample characteristics (Table 2), the proportions of female and married respondents were 53.9% and 51.1%, respectively, aligning with the frame population. Their average age and tenure were 43.68 years (SD = 10.82) and 10.62 years (SD = 8.57), respectively. Most respondents held a bachelor’s degree or higher (48.0%), primarily served in business to customer contexts (75.9%) and worked in medical services (e.g. hospital, clinic and elderly care services) (44.9%). Following Shin et al. (2022), this study assessed nonresponse bias by comparing the respondents who answered the second survey (n = 555) with those who dropped out of the second survey (n = 122), and the respondents who answered the third survey (n = 471) with those who dropped out of the third survey (n = 84). The t-test detected no significant differences between the groups. Combined with the comparison of the frame population, nonresponse bias was not a significant problem.

This study relied on validated scales from the existing literature, which were originally developed in English. The author prepared both the original and the Japanese versions. Then, a bilingual translator at a translation company back-translated it into English. Another bilingual translator assessed the equivalence between the back-translated version and the original. The author asked three independent researchers to judge whether the item wording was semantically identical. Before conducting the main surveys, three pretests were conducted (n pretest 1 = 109; n pretest 2 = 107; n pretest 3 = 95). Researchers’ comments and pretests led to slight modifications to the wording of several items. Unless otherwise noted, all scales were measured using a five-point Likert scale (1 = “strongly disagree” to 5 = “strongly agree”) (Table 3).

CI was measured using an 11-item scale derived from Sliter et al. (2012). Following the existing literature (Fujii, 2025; Hur et al., 2016), respondents rated the frequency of CI in the past month using a five-point scale (1 = “never” to 5 = “very often”). A sample item includes “Customers treat employees as if they are inferior or stupid.” CO was measured using a five-item scale adapted from Brown et al. (2002). A sample item includes “I try to help customers achieve their goals.” SRP was measured using a five-item scale developed by Boshoff and Allen (2000). A sample item includes “I do not mind dealing with complaining customers.” SC was captured by a seven-item scale derived from Coelho et al. (2021), who studied FLEs’ creativity. A sample item includes “I deal with customers in a creative way.” Promotion and prevention focus were each measured by a five-item scale taken from Geng et al. (2018) who studied creativity among FLEs in East Asia. Each sample item includes “I often think how I will achieve success in my work” and “I focus my attention on avoiding failure at work.”

Building on extant research (Fujii, 2025; Olugbade and Karatepe, 2019), this study controlled for age, gender, education, tenure and positive/negative affectivity to minimize confounding effects on the dependent variables. Positive/negative affectivity was measured using a three-item scale derived from Choi et al. (2023). Moreover, based on extant research (Yoon, 2022; Zahoor and Siddiqi, 2023), emotional exhaustion and intrinsic motivation were included to rule out alternative explanations. When FLEs deal with uncivil customers during service encounters, their valuable resources are depleted and they become emotionally exhausted (Hobfoll et al., 2018; Zahoor and Siddiqi, 2023). Consequently, FLEs facing emotional exhaustion are less likely to allocate additional resources to SRP and SC (Fujii, 2025; Zahoor and Siddiqi, 2023). Emotional exhaustion was measured using the three-item scale developed by Maslach and Jackson (1981). Similarly, COR theory assumes that CI leads to a decrease in intrinsic motivation. Consequently, FLEs with lower intrinsic motivation are less likely to engage in SRP and SC to avoid further resource depletion (Sommovigo et al., 2019; Yoon, 2022). Intrinsic motivation was measured using Grant’s (2008) four-item scale, which was applied in Yoon (2022).

Self-administered surveys are prone to common method bias (Podsakoff et al., 2024). To alleviate this, several procedural and statistical approaches were used. Specifically, this study ensured respondent anonymity, guaranteed that each item had no right or wrong answers and used multiple scale types (i.e. frequency and agreement), which helps mitigate social desirability bias. Furthermore, temporal separation was introduced as a preferred method and interaction terms were included to obscure the research purpose.

The marker variable technique was used as the statistical approach (Podsakoff et al., 2024). Specifically, Machiavellianism, which is measured using a three-item scale and theoretically unrelated to the other constructs, was selected as the marker variable. Indeed, the correlations with the other constructs were insignificant or weak, and the results remained consistent regardless of the marker variable. In addition, the results of the full collinearity technique indicated that all variance inflation factors (VIFs) were below the cutoff point of 3.3 (Kock, 2015). Given that positive/negative affectivity was also controlled for, concerns about common method bias were minimal.

This study applied partial least squares structural equation modeling (PLS-SEM) as an analytical technique (Ringle et al., 2024). PLS-SEM is a causal-predictive approach, which is different from covariance-based SEM, thereby allowing researchers to derive practical implications (Hair et al., 2022). Moreover, PLS-SEM has the advantage of handling complex models (e.g. interaction terms) smoothly without wasting statistical power. Following Hair et al. (2022), this study validated the reflective measurement model and tested the hypotheses.

First, all but three items were above the 0.70 cutoff. Considering content validity, one prevention focus item was excluded. Second, all Cronbach’s alpha coefficients were above the 0.70 cutoff, except for positive/negative affectivity (range: 0.67–0.96). Nevertheless, all composite reliability scores exceeded the 0.70 cutoff (range: 0.82–0.96), and all average variance extracted (AVE) scores outperformed the threshold point of 0.5 (minimum value = 0.54). These results indicate that internal consistency reliability was satisfied. Third, all indicator outer loadings were statistically significant (p < 0.001) (Anderson and Gerbing, 1988), suggesting that convergent validity was satisfied. Finally, the Fornell–Larcker criterion indicated that all squared AVEs outperformed any pair of correlations. In addition, the heterotrait–monotrait (HTMT) method indicated that the highest HTMT score was 0.67, which was below the conservative cutoff point of 0.85 (Table 4). These results indicate that discriminant validity was satisfied. Descriptive statistics and correlations are delineated in Table 4.

Given that collinearity may lead to biased results due to the inflated standard error of the path coefficients, collinearity was examined using the VIF (Hair et al., 2022). All VIF values were below the conservative cutoff of 3, indicating that collinearity was not a critical issue. The hypotheses were tested using bootstrapping with 10,000 samples to derive t-values and confidence intervals (CIs) (Hair et al., 2022). The results are summarized in Table 5 and Figure 2.

As stated in H1, the results indicate that CI has a significant indirect effect on SRP through CO (β = −0.04, p < 0.05, 95% CI [−0.08, −0.01]). Therefore, H1 was supported. Similarly, as H2 posited, CI has a significant indirect effect on SC through CO (β = −0.05, p < 0.05, 95% CI [−0.10, −0.01]). Therefore, H2 was also supported. Although not hypothesized, combined with insignificant direct effects, CO showed full mediation.

Next, this study applied the two-stage approach, the most recommended method, to develop interaction terms (Hair et al., 2022). H3a posited that promotion focus moderates the negative relationship between CI and CO. However, the 95% CI included zero (β = 0.12, p < 0.05, 95% CI [−0.01, 0.22]); thus, H3a was not supported. By contrast, as stated in H3b, the moderating effect of prevention focus on the relationship between CI and CO was significant, and the 95% CI did not include zero (β = −0.17, p < 0.01, 95% CI [−0.29, −0.04]); thus, H3b was supported. While H3a was not supported, the interactive effects of promotion and prevention focus are presented graphically (Figures 3 and 4).

As shown in Figure 3, the negative effect of CI on CO was exacerbated for FLEs with a low promotion focus, whereas it was less pronounced for FLEs with a high promotion focus. Similarly, as Figure 4 shows, the negative effect of CI on CO was exacerbated for FLEs with a high prevention focus, whereas it was less pronounced for FLEs with a low prevention focus. These results remained stable even when emotional exhaustion and intrinsic motivation were included as alternative mediators. Indeed, emotional exhaustion showed no mediating effect on either the relationships between CI and SRP (β = 0.01, p > 0.10, 95% CI [−0.03, 0.05]) or SC (β = 0.01, p > 0.10, 95% CI [−0.02, 0.06]). Similarly, intrinsic motivation showed no mediating effect on either the relationships between CI and SRP (β = −0.01, p > 0.10, 95% CI [−0.05, 0.01]) or SC (β = −0.01, p > 0.10, 95% CI [−0.10, 0.01]).

Finally, the results explained 23.5% of the variance in SRP and 39.5% of the variance in SC. Although R² values for SRP were not necessarily reported (Sommovigo et al., 2019; Zahoor and Siddiqi, 2023), R² values for SC outperformed those of existing research (R² = 32.1 in Hur et al., 2016; R² = 32.1 in Fujii, 2025). Furthermore, the results of PLSpredict indicate that all Q2predict values were above zero, and most of the root mean square error values were lower than those of the linear regression model, except for one indicator (Table 6) (Hair et al., 2022). In addition, this study conducted cross-validated predictive ability test (CVPAT) to assess whether the predictive power of the model statistically exceeds that of the alternative models (Sharma et al., 2023). The PLS loss score of SC was lower than the naïve indicator averages (IA) and linear model (LM), whereas the PLS loss score of SRP in the model did not outperform those of the IA. Nevertheless, it showed a lower score than those of the LM, indicating that the proposed model has moderate predictive validity over the benchmark models (Table 7). Model comparison revealed that the average loss differences between the proposed model and the alternative models (Model 1 and Model 2) were statistically significant in terms of SC, but not SRP (Table 8). In summary, the model had sufficient explanatory and predictive power.

To eliminate other possible explanations, this study examined the moderating role of CO on the relationships between CI and SRP and SC. The results indicated that CO had no moderating effect on the relationships between CI and SRP (β = −0.15, p > 0.10, 95% CI [−0.40, 0.04]) or SC (β = 0.03, p > 0.10, 95% CI [−0.09, 0.12]).

In addition, following Vaithilingam et al. (2024), this study evaluated robustness in terms of normality, endogeneity, non-normality and unobserved heterogeneity. First, none of the indicators deviated from the acceptable range of −2 to 2 for skewness and kurtosis (Vaithilingam et al., 2024), indicating that normality was satisfied. Second, although the assumption of the Gaussian copula technique was not necessarily satisfied because of normality, Gaussian copula terms were included. The results indicate that none of the copula terms were significant. Combined with the fact that alternative mechanisms were controlled for, concerns regarding endogeneity were minimal. Third, by applying a regression equation specification error test, none of the squared constructs statistically explained the additional variance in each dependent variable (Vaithilingam et al., 2024). Finally, the application of the finite-mixture PLS technique (Sarstedt et al., 2022), Bayesian information criterion and consistent Akaike information criterion demonstrated that a single solution was better (Table 9). Combined with the limited sample size of each segment (Sarstedt et al., 2017), the data were analyzed at the aggregate level.

Drawing on COR theory, this study disentangled the underlying mechanisms between CI and SRP and SC and examined the moderating effects of regulatory focus on the relationship between CI and CO. The findings derived from data concerning FLEs working in high-contact service industries in Japan revealed that CO was a critical mediator, even when controlling for emotional exhaustion and intrinsic motivation. Moreover, prevention focus moderated the relationship between CI and CO, whereas promotion focus did not.

This study makes two significant contributions to the existing literature. First, using data from FLEs working in high-contact service industries, this study identified CO as the primary mechanism underlying the relationships between CI and SRP and SC. Existing studies have primarily relied on the emotional mechanism (Auh et al., 2024; Yoon, 2022) and examined job performance as a consequence of CI (Choi et al., 2023; Tam and Trang, 2024). However, the impact of CI on SRP, a related yet distinct construct of job performance (Boshoff and Allen, 2000), has garnered less attention. A few notable studies relying on the emotional mechanism have shown an indirect effect of CI on SRP through emotional exhaustion (Sommovigo et al., 2019; Zahoor and Siddiqi, 2023). Nevertheless, these studies did not consider alternative mechanisms and were based on a sample of either students or retail banks, which are difficult to generalize. From a motivational mechanism perspective, Yoon (2022) showed the indirect effects of CI on job performance through intrinsic motivation using while ruling out emotional exhaustion and prosocial motivation. In contrast, Hur et al. (2016) found that CI had insignificant indirect effects on SC through intrinsic motivation. These contrasting findings stimulate the need to clarify what mechanisms underlie the relationship between CI and employee outcomes. Therefore, this study contributes to the literature on CI by identifying CO as the primary mechanism underlying the relationships between CI and SRP and SC, while ruling out emotional exhaustion and intrinsic motivation as alternative explanations. Data collected from various FLEs working in high-contact service industries may help address sample limitations of existing studies.

Second, this study demonstrated that prevention focus moderates the negative relationship between CI and CO, whereas promotion focus does not. Taking also Figure 3 into consideration, a plausible explanation is that FLEs with a low promotion focus are less affected by CI because they are less interested in achieving goals than FLEs with a high promotion focus. Following this logic, even when they experience CI, their personal resources were less depleted (Higgins, 1997; Hobfoll et al., 2018). As a result, contrary to prediction based on COR theory, FLEs’ CO might be less diminished. Interestingly, the results of this study differ from those of Auh et al. (2024). Auh et al. (2024) showed a moderating effect of promotion focus on the relationship between CI and customer conflict-solving behavior across the three studies in Turkey, Germany and the USA. However, prevention focus was evident only in the German sample. These contrasting results may be partially explained by cultural differences. Trait-regulatory focus varies depending on the culture, and either promotion or prevention focus tends to be dominant in certain cultures (Kurman et al., 2015). Indeed, Kurman et al. (2015) demonstrated that Japanese participants were more prevention-focused and less promotion-focused compared to Israeli participants. Therefore, this study complements Auh et al.’s (2024) findings by revealing the moderating effect of prevention focus within the context of Japanese FLEs.

Existing research on CI has identified the moderating role of individual characteristics (e.g. emotional intelligence, resilience and achievement orientation). However, most studies have examined its role within specific companies or industries, limiting our understanding of its applicability to other service industries. This study expands the CI literature by identifying individual characteristics (i.e. prevention focus) that are more susceptible to the effects of CI across various service industries in Japan.

The current findings allow service and HR managers to reconsider their HR practices in handling CI. Although the deleterious impact of CI varies depending on industry, the following implications have potentially broad applicability to other service organizations because of the sample characteristics and the high predictive power of the model.

First, given that CO is a critical mediator, service organizations would benefit more from investing in CO training when FLEs suffer from CI (Sommovigo et al., 2019). Specifically, when service managers consider providing CO training to FLEs, it should include content such as role-playing and videos that take the position of customers (Auh et al., 2024). This type of training would effectively improve FLEs’ CO by restoring self-efficacy. Furthermore, considering the cumulative nature of CI and the delayed effect of training, service managers should measure FLEs’ CO scores two or three months later to assess its effectiveness. In addition to mental health practices aimed at alleviating emotional exhaustion among FLEs, such as a mentor system and organizational and manager interventions (Auh et al., 2024; Chaudhuri et al., 2023), the systematic implementation of CO training can help restore FLEs’ CO.

Second, if service organizations are struggling to proactively manage CI (Shin et al., 2022), then as a realistic approach, they can adapt by updating their firms’ recruitment and rotation strategies. Specifically, based on the findings of this study, HR managers should avoid employing and rotating FLEs with a high prevention focus rather than those with a low prevention focus. They can integrate this into practice by measuring candidates’ prevention focus scores as part of the recruitment process. Once HR managers have measured each employee’s regulatory focus, which is considered a relatively stable individual trait, they can use the scores when rotating employees into FLE roles.

In addition to the HR perspective, measuring the regulatory focus of FLEs would also enables service managers to identify which FLEs should be prioritized for support in advance. For instance, when newcomers are assigned to their departments, service managers can check whether they should be prioritized in the provision of support by accessing the database containing FLEs’ regulatory focus scores made by the HR function. HR managers can support service managers by providing a list of FLEs who potentially need support. Consequently, service managers can promptly support FLEs with a high prevention focus score when they experience CI. Furthermore, in cases where FLEs are confronted with very difficult customers, service managers should inform their employees that they can handle the situation on their behalf as a higher authority. These practices can help prevent FLEs with a high prevention focus from experiencing a decline in CO.

This study has several limitations. First, this study conducted data collection using an online research panel to identify appropriate respondents. However, data collection through nonprobability sampling makes it difficult to derive generalizable findings. Despite the high predictive power of the model, collecting data through probability sampling is recommended for future research. Moreover, this study’s sample was confined the sample to full-time FLEs who frequently interact with customers face to face and have been working in high-contact service industries for at least one year. Although the screening procedure was valid and consistent with previous research, it is important to note that this process may have resulted in discrepancies between the real world and the sample.

Second, the three-wave survey design can contribute to causal claims and mitigates the risk of common method bias. However, to more fully claim causality, future research should consider other research designs (e.g. longitudinal and experimental designs). Moreover, although this study implemented several procedural approaches to address social desirability bias, this bias may have been introduced by measuring SRP and SC using self-report data. Indeed, a meta-analytic study found that the relationship between personal and contextual factors and employee creativity was stronger with self-ratings than with non-self-ratings (Ng and Feldman, 2012). Future studies measuring SRP and SC should statistically control for social desirability bias, and obtain data from multiple sources, such as supervisors (Martinaityte et al., 2019) and customers (Dong et al., 2015). This will also help minimize common method bias (Podsakoff et al., 2024). In addition, this study considered CO as a malleable attitude construct (Zablah et al., 2012) and measured CI and CO at different time intervals. To corroborate the findings of this study, future studies should measure CO levels at the first and second waves to confirm changes in CO values.

Third, while this study included a wide range of respondents working as FLEs in high-contact service industries in Japan, the sample was primarily derived from medical services (44.9%). Among FLEs in this sector, nurses need to provide continuous care to patients with varying degrees of complexity and acuity (Thomas et al., 2022). Furthermore, most nurses are women and are less likely to be bestowed with authority compared to doctors (Thomas et al., 2022), making them more susceptible to incivility from patients. While the composition of this study’s sample reflects that of the frame population, the characteristics of the sample in this study may also restrict the generalizability of the results. In addition, while this study has moderate predictive ability, caution should be exercised in applying findings to other contexts because the predictive power of SRP did not outperform over alternative models. These results can be partially explained by the composition of sample and cultural factors. Therefore, researchers should replicate these findings in different cultural contexts to confirm their predictive validity.

Finally, while this study focuses on regulatory focus as a personal characteristic that moderates the impact of CI, future research should consider other moderators that can be exerted managerially (e.g. leadership). To my knowledge, the moderating effects of leadership on CI remain limited (notable exceptions are passive leadership: Bani-Melhem, 2020; and servant leadership: Mostafa, 2022). Therefore, researchers are strongly encouraged to investigate which types of leadership buffer the effects of CI.

The author would like to thank Taylor & Francis Editing Services (Link to Taylor & Francis Editing ServicesLink to the cited article) for English language editing and two anonymous reviewers for their helpful comments.