The COR theory holds that individuals always strive to acquire new resources and protect existing ones (Bickerton & Miner, 2023; Halbesleben et al., 2014; Hobfoll, 1989; Hobfoll et al., 2018), which consists of two competing tenets – ‘resource conservation’ and ‘resource acquisition’. The ‘resource conservation’ tenet states that individuals take actions to preserve existing resources to prevent further losses when they face the loss of resources (Halbesleben et al., 2014; Hobfoll, 1989). Conversely, the rarely mentioned ‘resource-acquisition’ tenet of COR theory posits that ‘individuals are driven to obtain resources to accomplish their objectives’ (Halbesleben et al., 2014; Hobfoll, 1989; Zhu et al., 2021). Employees need to resist the temptation of cyberloafing using resources because cyberloafing is an act of instant pleasure (Wagner et al., 2012; Zhu et al., 2021), and engaging in cyberloafing may be a strategy to conserve the available resources when employees’ resources are consumed a lot (Yui et al., 2021). Thus, cyberloafing is related with resources.

Leaders are an important source for employees to acquire resources in the workplace (Hobfoll et al., 2018; Peng et al., 2023). Responsible leaders care about the needs and interests of employees and treat employees fairly (Maak & Pless, 2006; Zhou et al., 2022), making employees feel respected (Osman et al., 2023). Thus, responsible leaders develop good relationship with employees (Osman et al., 2023). Good relationship is a valuable resource (Owens et al., 2016) that is likely to provide employees positive emotional and psychological resources to resist the temptation of cyberloafing. Osman et al., (2023) found that responsible leadership was positively related to relational energy. Relational energy is able to foster psychological resources (Osman et al., 2023).

‘Responsible leaders prioritise effective engagement with stakeholders, distribute responsibilities equitably, collaboratively address issues, and empower stakeholders by involving them in the decision-making process’ (Voegtlin, 2016). This indicates that responsible leaders possess a pronounced sense of responsibility. Thus, strong sense of responsibility is one of the key attributes of responsible leadership (Antunes & Franco, 2016). The sense of responsibility, an important type of psychological resource (Halbesleben et al., 2014; Zhu et al., 2021), enables individuals to stay focussed on job tasks (Abbaas & Ibrahim, 2012; Zhu et al., 2021). Responsible leaders create incentives to foster employees’ responsibility and pass on responsibility to employees (Lu et al., 2022). Employees’ psychological resources are enriched in this process (Zhu et al., 2021). Consequently, employees are motivated to work hard and, with their complete concentration on work, have the ability to consciously resist cyberloafing. Gökçearslan et al., (2016) found that self-control is an effective tool for reducing employees’ cyberloafing behaviour.

Conversely, ‘responsible leaders not only concentrate on meeting economic obligations and ensuring organisational functionality, but also shoulder societal and environmental responsibilities’ (Miska et al., 2014).

They need to balance the external pressure of conflicting interests and demands by stakeholders (Maak & Pless, 2006). Responsible leaders ask employees to ‘discern the needs of external stakeholders, engage in dialogue with these stakeholders, or mediate disputes among them’ (Zhu et al., 2021). These requests go beyond the scope of employees’ standard job responsibilities, consuming a large number of employees’ resources such as time and energy. Given that the amount of resources such as time and energy available to individuals is limited (Hobfoll, 1989, 2001). When employees’ resources such as time and energy become depleted to a certain level where employees experience significant loss of resources, the resource-conservation motivation is triggered to prevent further loss of resources. Consequently, employees become less engaged in their work and engage in more cyberloafing behaviour. Previous empirical studies have shown that inadequate resources (e.g. lack of sleep) can lead employees to engage in cyberloafing behaviour (Wagner et al., 2012). On the basis of the above analysis, we proposed the following hypothesis:

The COR theory highlights that individual traits affect individuals’ resource acquisition and conservation process (Halbesleben et al., 2014; Hobfoll, 1989). We introduce self-efficacy, a key individual trait, to comprehend how the impact of responsible leadership varies depending on individual characteristics. According to the COR theory and literature related to self-efficacy (Bandura, 2012; Gist & Mitchell, 1992; Halbesleben et al., 2014; Hobfoll, 1989), we argue that for employees with high self-efficacy, responsible leadership is more likely to trigger resource-conservation motivation. For employees with low self-efficacy, responsible leadership is more likely to trigger resource-acquisition motivation.

Employees with high self-efficacy are confident in their abilities (Bandura, 2012), and are less influenced by contextual factors like leadership because of their insensitivity to external environmental factors (Gist & Mitchell, 1992; Pan et al., 2011; Saks, 1994). Self-efficacy is an important individual resource. Employees with high self-efficacy naturally possess high individual resources, thus they are able to resist the allure of indulging in cyberloafing mainly relying on their individual resource, so it is not necessary for them to acquire resources from leaders to resist the allure of indulging in cyberloafing. However, employees with low self-efficacy lack confidence in their abilities (Gist & Mitchell, 1992; Pan et al., 2011; Saks, 1994), leading them to believe that they possess insufficient resources and have to acquire resources from leaders to resist the allure of indulging in cyberloafing. Once responsible leaders care about the needs and interests of employees (Maak & Pless, 2006), employees with low self-efficacy will be greatly encouraged, which will become individual important psychological resources. Hence, we argue that responsible leadership is more likely to trigger resource-acquisition process for employees with low self-efficacy.

Employees with high self-efficacy are confident in their abilities (Bandura, 2012). ‘When they experience expanded roles or tasks from responsible leadership, employees are confident to complete tasks and then tend to allocate more resources to meet leader demands’ (Zhu et al., 2021). The empirical study by Seo and Ilies (2009) shows that employees with high self-efficacy exhibit higher levels of effort and consistency. Thus, employees with high self-efficacy put in great effort to deal with expanded roles or tasks from responsible leadership (Zhu et al., 2021). ‘Given that the total pool of personal resources is limited, their strategy of resource distribution will gradually exhaust their individual resources’ such as time, energy, cognitive attention (Zhu et al., 2021). According to the COR theory, individuals have to replenish their resources by having a break in order to avoid stress (Kim et al., 2017). Cyberloafing, such as browsing sports or entertainment-related websites, is possibly a useful way to have a break to conserve the available resources (Yui et al., 2021). Yui et al., (2021) found that job burnout had a positive effect on cyberloafing. Following COR theory, resource depletion will drive individuals to preserve their resources (Halbesleben et al., 2014; Hobfoll, 1989). Hence, we argue that resource depletion results in seeking immediate pleasure from cyberloafing. However, employees with low self-efficacy lack the confidence in performing tasks (Gist & Mitchell, 1992) and believe that they possess insufficient resources. As a result, they may not allocate resources in expanded roles or tasks from responsible leadership, leading to suffer less from resource loss and being less likely to motivate resource conservation (Zhu et al., 2021). Therefore, we argue that responsible leadership is more likely to trigger resource-conservation process for employees with high self-efficacy. On the basis of the above analysis, we proposed the following hypothesis:

The questionnaires were distributed via the network platform ‘Sojump’ to full-time employees in the enterprises who work on computers for over 1 year. Five leaders provided assistance with the distribution, and they sent the questionnaire link to their colleagues. A total of 400 questionnaires were distributed, out of which 318 were returned. Thirteen sets of invalid questionnaires were excluded according to the following criteria (Xu et al., 2021): overly long or short answer time, a high frequency of identical responses, or observable response patterns. The final sample was 305, representing a valid response rate of 76.25%. Half of the participants (51.80%) were male; 77.40% were married; 53.80% held a bachelor’s degree, 20.30% held a postgraduate degree or above, and only 6.90% had a secondary school degree or below. The average age of the participants was 36.86 years (standard deviation [SD] = 8.80), and their average duration of employment in the current organisation was 6.39 years (SD = 5.70).

Confirmatory Factor Analysis (CFA) was used to test the discriminant validity among the variables, and the Harman single-factor method was used to test common method bias. Hierarchical regression was then employed to test all hypotheses. Finally, we used the Bootstrap and Bayesian analysis to conduct robustness tests. Following Dawson (2014), we set up the following mathematical equation to test H1 and H2.

Cyberloafing = β₀ + β₁gender + β₂age + β₃marriage + β₄job tenure + β₅education level + β₆responsible leadership + β₇squared responsible leadership

The regression slope β₇ determines the U-shaped relationship between responsible leadership and cyberloafing (H1).

Cyberloafing = β₀ + β₁gender + β₂age + β₃marriage + β₄job tenure + β₅education level + β₆responsible leadership + β₇squared responsible leadership+ β₈self-efficacy + β₉self-efficacy* responsible leadership + β₁₀self-efficacy* squared responsible leadership.

The regression slope β₁₀ determines whether the U-shaped relationship between responsible leadership and cyberloafing is moderated by self-efficacy (H2).

To test the distinctiveness of our variables, we performed CFA. Results (see Table 1) showed that the three-factor model (χ²/degrees of freedom [df] = 1.37, comparative fit index [CFI] = 0.992, Tucker–Lewis index [TLI] = 0.988, root mean square error of approximation [RMSEA] = 0.035) provided a significantly better fit than the other hypothesised models, indicating good discriminant validity.

To reduce the influence of common method bias, we followed the recommendations of Podsakoff et al., (2003) in our study design. Specifically, we employed different scale anchors for the dependent variable (cyberloafing behaviour) and the independent variable (responsible leadership) to mitigate the effect of participants answering identically to the same scale anchors. In the statistical analysis phase, we conducted both exploratory factor analysis (EFA) and CFA, using Harman’s single factor test to examine common method bias, following the advice of Podsakoff et al., (2003). The results of EFA showed that the unrotated principal component analysis (PCA) explained only 24.40% of the total variance, while the one-factor model in CFA provided a poorer fit compared to the three-factor model (Δχ2 (3) = 396.546, p < 0.001). These findings suggest that common method bias was not serious and did not systematically affect the relationship between the variables in our study.

The results (see Table 2) showed that there was no significant correlation between responsible leadership and cyberloafing (r = –0.047, p > 0.1). This indicated that the relationship between responsible leadership and cyberloafing was likely to be U-shaped relationship, moderated by other variables, or non-existent. Overall, these findings offer initial evidence that support our hypotheses.

However, the high correlation between responsible leadership and self-efficacy may induce multicollinearity. We test multicollinearity by variance inflation factor (VIF). The results showed the VIF values of responsible leadership and self-efficacy was 1.278 and 1.215 respectively, which is less than 10, indicating lowly collinearity. That is to say, the high correlation between responsible leadership and self-efficacy does not impact the relationships between key variables in the study.

The results of the hierarchical regression analysis using Mplus statistical software to test each of the hypotheses are presented in Table 3. The results revealed that the effect of the squared term of responsible leadership on cyberloafing behaviour was both positive and significant (β = 0.117, p < 0.05), which supported Hypothesis 1. Specifically, the result indicated that there was a U-shaped relationship between responsible leadership and cyberloafing behaviour. Furthermore, the product term of self-efficacy and responsible leadership-squared displayed a positive and significant coefficient (β = 0.195, p < 0.01), indicating that self-efficacy served as a moderator for the U-shaped relationship between responsible leadership and cyberloafing behaviour. This result supported Hypothesis 2.

In order to clearly illustrate the U-shaped relationship between responsible leadership and cyberloafing behaviour and the moderating effect of self-efficacy, we followed the recommendation of Toothaker (1994) to plot the U-shaped effect and interaction effect. Figure 2 and Figure 3 present the results of the curve relationship between responsible leadership and cyberloafing behaviour and the moderating effect of self-efficacy, respectively. As shown in Figure 2, responsible leadership was negatively related to cyberloafing before the threshold, then becoming positively related to cyberloafing beyond the threshold. Figure 3 demonstrates the moderating effect of self-efficacy on the relationship between responsible leadership and cyberloafing behaviour. Specifically, individuals with high levels of self-efficacy demonstrated a nearly positive relationship between responsible leadership and cyberloafing behaviour, while those with low levels of self-efficacy exhibit an almost negative relationship.

To establish the reliability of the study’s findings, we conducted two robustness tests: the bias-corrected Bootstrap and Bayesian analysis, which do not rely on normal distribution assumptions and are appropriate for small sample sizes (Miočević et al., 2017). The bias-corrected Bootstrap analysis provides reliable confidence intervals (CIs) for parameter estimates (Fang & Zhang, 2012), while Bayesian estimation has high statistical power (Miočević et al., 2017). The results of bias-corrected Bootstrap analysis showed that the effect of the squared term of responsible leadership on cyberloafing behaviour was significantly positive (β = 0.117, 95% CI = [0.006, 0.224], Bootstrap = 5000). The coefficient of the product term of self-efficacy and responsible leadership-squared was also significantly positive (β = 0.195, 95% CI = [0.013, 0.393], Bootstrap = 5000). These results confirm that both Hypothesis 1 and Hypothesis 2 were supported.

Recently, Bayesian analysis has gained popularity among scholars (Miočević et al., 2017; Yuan & MacKinnon, 2009). It differs from traditional frequentist analysis by considering the estimated parameter as an uncertain variable and expressing possible values as a distribution. Bayesian analysis aims to find this distribution and determines the most likely value of the estimated parameter, such as the mean (Wang et al., 2017). In this study, we performed Bayesian analysis using a non-informative prior and two Markov chains with 30 000 iterations, as displayed in Figure 4, Figure 5, Figure 6, Figure 7, Figure 8 and Figure 9.

Trace plots in Figure 5 and Figure 8 showed that the two Markov chains almost overlap, indicating a stable distribution. Additionally, the autocorrelation coefficient in Figure 6 and Figure 9 is under 0.1, while the potential scale reduction (PSR) is 1.0 after 15 300 iterations. These findings suggest that parameter estimates in our study are well-converged. The p-values of the posterior predictive checking (PPC) for Model 3 and Model 4 were 0.344 and 0.628 respectively, indicating a good model fit. The mean value of the coefficient for the effect of responsible leadership-squared on cyberloafing behaviour was 0.113, and the 90% credibility interval (0.016, 0.208) did not include 0. Therefore, Hypothesis 1 receives weak support. The mean value of the coefficient for the product of self-efficacy and responsible leadership-squared was 0.188, with a 95% credibility interval (0.021, 0.351) did not including 0, supporting Hypothesis 2. We summarise all results of testing each Hypothesis with different methods (see Table 4).

Firstly, this study contributes to the literature on the factors that affect cyberloafing behaviour by investigating the impact of responsible leadership on cyberloafing behaviour. Previous research has primarily explored factors related to employees’ characteristics (Lim & Teo, 2005), such as demographic variables (Jia et al., 2013), personal habits and individual self-control (Ugrin et al., 2007), or factors related to the organisational circumstances, such as role ambiguity and role conflict (Henle & Blanchard, 2008). However, little attention has been given to the influence of leadership style on cyberloafing behaviour, and previous studies have only examined the linear effects of leadership styles. Thus, this study bridges this gap by empirically investigating the U-shaped effect of responsible leadership on cyberloafing from a resource-based perspective, and expands our understanding of antecedents of cyberloafing behaviour.

Secondly, this study extends the research on the effects of responsible leadership. Existing studies on responsible leadership have predominantly concentrated on its positive impacts on employee behaviour (Wang et al., 2015; Wen & Xia, 2015), such as increasing job satisfaction (Voegtlin, 2011), organisational citizenship behaviour (Lord & Brown, 2001), and job performance (Wang et al., 2015), while its negative impacts have received less attention. Although some scholars have recognised the negative effects of responsible leadership (Wen & He, 2017), empirical evidence is lacking. Eisenbeiß and Boerner (2013) and Beyer (1999) have strongly advocated for exploring leadership styles from both positive and negative perspectives. This study combines the COR theory to theoretically and empirically examine the U-shaped effect of responsible leadership on cyberloafing behaviour. The study verifies the double-edged sword effect of responsible leadership, echoing Eisenbeiß and Boerner (2013) and Beyer (1999)’s call to deepen our understanding of the positive and negative effects of leadership styles.

Finally, this study contributes to the advancement and refinement of COR theory. Initially employed to explicate stress-related phenomena, the theory expounds on how resource depletion contributes to employee burnout, stress, and both physical and mental health issues (Halbesleben et al., 2014). However, previous research on COR theory has mainly concentrated on resource loss while giving scant attention to resource acquisition. Our study investigates the double-edged sword effect of responsible leadership by integrating the resource loss and resource acquisition, thereby adding depth to the COR theory. Responsible leadership helps employees to acquire resources by developing good relationship and crossover sense of being responsible to employees. At the same time, responsible leadership also consumes employees’ resources by expanding employees’ roles or tasks. Additionally, we advance COR theory by introducing self-efficacy as a pivotal boundary condition, which further enriches the scope of COR theory.

Firstly, it is worth noting that cyberloafing behaviour is characterised by instant gratification and a strong temptation (Wagner et al., 2012), making it challenging for employees to resist by themselves. Therefore, organisations should recognise the role of leaders in mitigating employees’ cyberloafing behaviour through effective supervision. Moreover, cyberloafing behaviour is often invisible and thus responsible leadership should be employed to help employees to reduce cyberloafing behaviour.

Secondly, organisations should recognise that responsible leadership has both positive and negative impacts, and careful consideration is needed to reduce cyberloafing behaviour. While responsible leadership can effectively reduce cyberloafing, it can also inadvertently increase it. To develop responsible leadership and mitigate negative effects, organisations should equip leaders with a comprehensive understanding of their impact, and support them in taking necessary measures that can actively enhance their leadership practices while also reducing their negative impact. For instance, providing employees with additional resources to support them in reducing work-related stress caused by responsible leadership can effectively curtail its negative outcomes.

Lastly, managers should take different management measures depending on different employees. Employees with low self-efficacy may have inadequate psychological resources to resist cyberloafing behaviour independently and may need external support, such as leadership supervision. Therefore, leaders should pay attention to such employees and help them to reduce cyberloafing behaviour through effective supervision. Conversely, employees with a high sense of self-efficacy expend their resources to resist cyberloafing behaviour actively. If responsible leaders pay much attention to these individuals, they may experience greater work pressure, resulting in more resource loss that could deplete their capacity to resist cyberloafing behaviour. To address this, leaders need to strike a balance in their attention to these employees without impeding their independent decision-making.

Firstly, all variables in this study are evaluated at the same time point through employee self-assessments, which may lead to common method bias. To address this limitation, this study employs both bias-corrected Bootstrap and Bayesian analysis to ensure robustness. The use of Bayesian analysis is a novel approach in management research, and its application in this study implies that the findings are highly reliable. However, future studies could opt for a more rigorous data collection method involving different points in time and participants. For instance, employees could evaluate responsible leadership and self-efficacy at time point 1, while leaders or colleagues assess employees’ cyberloafing behaviour at time point 2. This would enable the collection of different variables and reduce the effect of common method bias in the data analysis, resulting in more credible conclusions. Furthermore, it is also possible to conduct an experimental design to yield a causal relationship between the variables.

Secondly, this study primarily examines the double-edged sword effect of responsible leadership from a U-shaped perspective. Future research could explore the impact of responsible leadership from other aspects. For instance, Busse et al., (2016) proposed an antecedent-benefit-cost (ABC) model that suggests an independent variable affects the dependent variable in opposite directions through two distinct mediating variables. The mediating variable that enables the independent variable to positively affect the dependent variable is called the ‘benefit mediating variable’. In contrast, the mediating variable that allows the independent variable to negatively affect the dependent variable is called the ‘cost mediating variable’.

According to Busse et al., (2016)’s view, does responsible leadership have a similar dual effect on cyberloafing behaviour through separate mediating mechanisms? Exploring this question could provide deeper insights into the effects of responsible leadership on employee behaviour. It will help us to fully understand the reason why employees engage in cyberloafing. Furthermore, this study solely focusses on the impact of responsible leadership on cyberloafing behaviour; future research could delve into other leadership styles’ impact on cyberloafing behaviour, such as spiritual leadership. Spiritual leadership meets employees’ spiritual needs and motivates them intrinsically to work (Reave, 2005), which could encourage employees to consciously resist cyberloafing behaviour.

Finally, this study was conducted in a Chinese context but did not explore cultural variables unique to China that could impact cyberloafing behaviour. Future research could consider examining how specific cultural variables, such as traditionalism, influence cyberloafing behaviour. Traditional employees are more likely to adhere to rules and regulations set forth by their leaders and organisations, making them less likely to engage in cyberloafing behaviour. Further research could explore the impact of traditionalism on cyberloafing behaviour, and how it moderates the relationship between leadership style and cyberloafing behaviour.