Job Insecurity and Emotional Exhaustion

Job insecurity is a perceptual phenomenon that focuses on the potential threat to the stability of an individual’s current job (Greenhalgh & Rosenblatt, Reference Greenhalgh and Rosenblatt1984). The literature describes job insecurity as either objective (Pearce, Reference Pearce, Cooper and Rousseau1998) or subjective (DeWitte, Reference DeWitte1999). Objective job insecurity focuses on the anticipation of unemployment generated from organizational cues, such as intended organizational messages, unintended organizational clues, or rumors (Greenhalgh & Rosenblatt, Reference Greenhalgh and Rosenblatt1984). Subjective job insecurity is derived from a previously signaled threat and highlights an individual’s perceived fear or worry about the future of his or her job based on their assumed severity of the threat (Greenhalgh & Rosenblatt, Reference Greenhalgh and Rosenblatt1984). We focus on subjective job insecurity because previous research suggests that subjective insecurity is more consistently related to psychological well-being than objective job insecurity (Ferrie, Reference Ferrie2001; Sverke, Hellgren, & Naswell, Reference Sverke, Hellgren and Naswell2002).

Previous research suggests that job insecurity is positively related to emotional distress and negatively related to physical and mental health (DeWitte, Reference DeWitte1999; Cheng & Chan, Reference Cheng and Chan2008; Sverke et al., Reference Sverke, Hellgren and Naswell2002). From a self-regulatory theory perspective, the perception of job insecurity creates a threat to several types of resources. Initial reactions to job insecurity may include an impaired mood. In an attempt to maintain balance, employees will evoke self-regulatory mechanisms by either suppressing or altering their negative thoughts or emotions (Eisenberg, Fabes, & Guthrie, Reference Eisenberg, Fabes, Guthrie, Wolchik and Sandler1997; Hancock & Warm, Reference Hancock and Warm1989). For instance, in order to suppress the fear of job loss, an employee might choose to remain calm and work as usual.

Using energy to maintain a positive attitude in the face of the threat of losing one’s job depletes self-regulatory resources and modifies individual’s psychological processing (Cheng & Chan, Reference Cheng and Chan2008; Sverke et al., Reference Sverke, Hellgren and Naswell2002). For instance, the uncontrollability of the situation and constant monitoring of the uncertain work environment wears on an individual and creates feelings of emotional exhaustion. With the threat of job loss hanging over one’s head, valued resources such as a stable income and a feeling of control over the one’s life are under constant threat (Sverke et al., Reference Sverke, Hellgren and Naswell2002). This threat to valuable resources presents a need for employees to monitor threatening stimuli, such as layoffs, and creates a continuous focus on the uncertainty (Baumeister, Reference Baumeister, Gilbert, Fiske and Lindzey1988). The constant focus on the uncertainty of the situation creates emotional exhaustion for the individual (Cheng & Chan, Reference Cheng and Chan2008; DeCuyper, DeWitte, Elst, & Handaja, Reference DeCuyper, DeWitte, Elst and Handaja2010; Sverke et al., Reference Sverke, Hellgren and Naswell2002). Thus, we predict:

Moderating Effect of Adaptability on the Job Insecurity - Emotional Exhaustion Relationship

Personality is an individual difference characteristic that can influence one’s ability to identify unethical issues, make assessments about the related issues, and develop appropriate responses (Bass, Barnett, & Brown, Reference Bass, Barnett and Brown1999; Christie & Geis, Reference Christie and Geis1970; Jones, Reference Jones1991; Treviño & Youngblood, Reference Treviño and Youngblood1990). The literature has recognized personality traits such as locus of control, self-esteem, (Detert, Treviño, & Sweitzer, Reference Detert, Treviño and Sweitzer2008; Treviño, Reference Treviño1986; Treviño & Youngblood, Reference Treviño and Youngblood1990), and self-efficacy (Belar et al., Reference Belar, Brown, Hersch, Hornyak, Rozensky, Sheridan, Brown and Reed2001) as predictors of unethical behaviors. In an effort to advance our understanding of dispositional factors on ethical decision-making, we examine personal adaptability. Personal adaptability is an individual’s willingness and ability to alter one’s behaviors, feelings, and thoughts when encountering situations (Chan, Reference Chan and Ferris2000; Fugate et al., Reference Fugate, Kinicki and Ashforth2004). The ability to adapt to changing circumstances is an individual difference that prompts persons to engage in proactively adaptive behaviors (Chan, Reference Chan and Ferris2000; Crant, Reference Crant2000; Siebert, Kraimer, & Crant, Reference Seibert, Kraimer and Crant2001). Previous research suggests the dispositional characteristic personal adaptability can reduce strain when workplace stressors are experienced (Chan, Reference Chan and Ferris2000; Fugate et al., Reference Fugate, Kinicki and Ashforth2004).

According to self-regulation theory, individuals exercise a conscious effort to align their behavior with recognized or desired standards (Baumeister, 1988; Baumeister & Bushman, Reference Baumeister and Bushman2010). Self-regulation requires individuals to alter their standard thoughts, feelings, and behavior in order to deter prohibited reactions and to encourage more desirable responses (Muraven & Baumeister, Reference Muraven and Baumeister2000). Adaptable individuals are willing and have the capability to modify their knowledge, skills, abilities, and behavior to meet the demands of situations they encounter (Ashford & Taylor, Reference Ashford, Taylor, Ferris and Rowland1990; Chan, Reference Chan and Ferris2000). Adaptive employees have a high tolerance for uncertainty and ambiguity and are comfortable when faced with unique organizational situations (O’Connell, McNeely, & Hall, Reference O’Connell, McNeely and Hall2008). These qualities suggest that highly adaptive individuals have a positive assessment of their job situation that will likely result in lower perceptions of resources being threatened. This ability to view situations in a positive manner will likely serve as a coping resource for these individuals resulting in more resources available for dealing with job insecurity. These resources are likely to equip individuals with a sense of control, which enables them to react more calmly to the situation and reinforce the belief that they can resolve the situation with their own skills and abilities. Their positive reaction to the uncertain situation will result in fewer resources being depleted when faced with job insecurity (May, Schwoerer, Reed, & Potter, Reference May, Schwoerer, Reed and Potter1997).

In contrast, individuals with low levels of adaptability view themselves and their environment more negatively and feel as if there is a deficiency in their skills and capabilities necessary to obtain positive outcomes (Ashford & Taylor, Reference Ashford, Taylor, Ferris and Rowland1990; Chan, Reference Chan and Ferris2000). Individuals who have difficulty in adapting to situations are likely to experience high levels of stress when faced with job insecurity because they are likely to feel a sense of helplessness and lack of control over the situation. Feelings of inadequacy in their capabilities result in these individuals continually focusing on job insecurity as a problem rather than an opportunity. Less adaptable employees are also unlikely to exhibit flexibility and openness to new opportunities when faced with job insecurity. The continual negative focus depletes their psychological resources and energy (Baumeister, Heatherton, & Tice, Reference Baumeister, Heatherton and Tice1993). Therefore, we propose the following:

Moderating Effect of Embeddedness on the Job Insecurity - Emotional Exhaustion Relationship

When an employee encounters job insecurity, the threat to financial resources can be overwhelming. The risk of losing employment also causes employees to consider the threat to their feelings of belongingness and fit within their present organization. Job embeddedness has been described as a sticky web or net through which people at work are connected to one another (Mitchell et al., Reference Mitchell, Holtom, Lee, Sablynski and Erez2001). The more connections at work one has, the more embedded one is. The literature on job embeddedness suggests a negative relationship with employee turnover (Eberly, Holtmon, Lee, & Mitchell, Reference Eberly, Holtmon, Lee, Mitchell and Locke2009) as employees who are deeply attached to their jobs feel a strong fit, or the extent to which they think they belong or are compatible with the work environment. Embedded employees also recognize the benefits they gain from the attachment to their job.

Individuals who have a high level of job embeddedness feel a level of comfort and compatibility that results in increased levels of attachment to the organization (Eberly et al., Reference Eberly, Holtmon, Lee, Mitchell and Locke2009). As it relates to job insecurity, while job embeddedness creates increased levels of attachment to the organization it also creates an inherent level of dependence on the organization (Eberly et al., Reference Eberly, Holtmon, Lee, Mitchell and Locke2009). Self-regulation theory suggests that individuals enact a conscious effort to align their behaviors with recognized standards (Baumeister, 1988; Baumeister & Bushman, Reference Baumeister and Bushman2010). As a result of increased levels of fit and attachment, employees with higher levels of embeddedness will exhibit behaviors that are aligned with the organization. The threat of the loss of a job with an organization in which they have a high level of attachment, congruent goals, and a strong fit is overwhelming and potentially detrimental. Feelings of ambiguity, betrayal, and the potential loss of organizational fit will deplete their psychological capabilities and may result in emotional exhaustion.

Conversely, individuals with low levels of embeddedness have not established a strong attachment or endearment to the organization (Mitchell et al., Reference Mitchell, Holtom, Lee, Sablynski and Erez2001). This lack of fit and interconnectedness to the organization and its members often result in increased levels of turnover. Employees who don’t feel a strong level of obligation to the organization will view the potential loss of employment as an avenue to pursue other opportunities. While these individuals may potentially experience some levels of concern when faced with job insecurity, they will be able to cope with the situation by regulating their emotions and making the decision to utilize their skills and abilities within another organization or in becoming entrepreneurs. Therefore, we propose:

Mediating Effects of Emotional Exhaustion

Perceived job insecurity usually triggers a stress reaction within individuals. A stress reaction often drives an individual to engage in behaviors that assist him or her in coping with the perceived threat. Research examining regulatory depletion suggests that individuals, who utilize self-regulation to manage their stress, focus on their feelings of fatigue rather than on the task (Vohs & Schmeichel, Reference Vohs and Schmeichel2003). Employees who experience emotional exhaustion are likely candidates for focusing on their fatigue rather than on the task.

Self-regulation theory (Baumeister, Reference Baumeister, Gilbert, Fiske and Lindzey1998, Reference Baumeister, Roberts and Hogan2001) suggests that emotional exhaustion may serve as a mediating step, whereby individuals, who perceive job insecurity, will have increased levels of emotional exhaustion. Emotionally exhausted individuals have lower self-regulatory resources, which may tempt them to ignore their internal moral compass and engage in unethical behaviors. Individuals can engage in unethical workplace behaviors by either acting in their own self-interest or in a manner that benefits the organization. Although self-interested unethical behavior deviates from one’s personal moral standards or from established organizational norms, individuals engage in them for positive personal gain. Examples of these behaviors include padding an expense report to receive a higher reimbursement than one is eligible for, claiming credit for someone else’s work, or stealing company materials and supplies.

Employees also may engage in unethical behavior in an effort to help the organization. Unethical pro-organization behavior (UPB) is a form of unethical behavior designed to benefit the organization rather than the actor (Umphress, Bingham, & Mitchell, Reference Umphress, Bingham and Mitchell2010). Such activities might include providing false information to clients to ensure the organization meets its quarterly goals. While, individuals engage in UPB to help the organization, it is important to note that these acts are unethical and include both direct (e.g., inflating costs incurred while working on a project) and inadvertent (e.g., withholding information about organization practices that are harmful to the public) behavior. Additionally, UPB is neither specified in an individual’s job description nor directed by supervisors (Umphress et al., Reference Umphress, Bingham and Mitchell2010).

As posited by self-regulation theory (Baumeister, Reference Baumeister, Gilbert, Fiske and Lindzey1998, Reference Baumeister, Roberts and Hogan2001), when individuals’ moral regulatory resources have been depleted, their cognitive resources are strained, and their subsequent ability to self-regulate is hindered. Depleted self-regulatory resources allow individuals to make a conscious decision to engage in unethical behavior as a coping mechanism with the thought that these behaviors will either benefit the organization or themselves. For instance, an employee may inflate the cost of a project to clients because this behavior will translate into increased financial rewards for the organization. Employees facing job insecurity may engage in these behaviors under the premise that their loyalty will be reciprocated with continued employment with the organization. However, there is little to no consideration given to the fact that the acts are unethical and can produce subsequent harm to the organization or the public. On the other end of the spectrum, employees also may decide to engage in unethical behavior such as padding their overtime hours in an effort to release some of the frustration they are experiencing regarding their potential job loss. We propose the following:

Moderating Effect of Adaptability on the Mediating Effects of Emotional Exhaustion

As previously discussed, employees with higher levels of adaptability experience lower levels of emotional exhaustion as a result of their ability to see the perceived threat to their job as an opportunity rather than as a challenge (Chan, Reference Chan and Ferris2000; Fugate et al., Reference Fugate, Kinicki and Ashforth2004). Highly adaptive individuals tend to view personal outcomes as dependent on their own actions and are likely to neutralize the potential threat of job loss as an opportunity to scan the environment to seek employment with a different organization or potentially establish their own entrepreneurial venture (O’Connell et al., Reference O’Connell, McNeely and Hall2008). Because highly adaptive people generally believe they can obtain their desired results, they do not need to deactivate their moral self-regulatory compass in order to engage in unethical acts, even if they benefit the organization. Additionally, highly adaptable employees are likely to perceive engaging in unethical acts as potentially damaging to their perception of themselves and professional reputation.

In contrast, individuals with low levels of adaptability have a tendency to be less confident in the efficacy of their skills and abilities and may experience uncertainty and self-doubt when facing ethical dilemmas (Ashford & Taylor Reference Ashford, Taylor, Ferris and Rowland1990; Fugate et al., Reference Fugate, Kinicki and Ashforth2004). These feelings of uncertainty require the use of self-regulatory resources to remain balanced. People who are unable to adapt to uncertainty will have an inclination to use self-regulatory resources to overcome their self-doubt and negative self-views (Fugate et al., Reference Fugate, Kinicki and Ashforth2004). They also will exert energy by not being flexible and not considering new opportunities. The continual focus on uncertainty and personal inadequacies depletes the subsequent energy needed to maintain focus and make decisions that align with moral and organizational standards. This diminished capacity to regulate decision-making will result in employees with low levels of adaptability engaging in unethical behaviors. Because we expect personal adaptability to account for differences in the ability to regulate behavior in stressful situations, we propose that the influence of job insecurity on unethical behaviors through emotional exhaustion is more evident among employees with low adaptability as compared to highly adaptable employees.

Moderating Effect of Job Embeddedness on the Mediating Effects of Emotional Exhaustion

As argued above, employees who are highly embedded within their organizations and who face the potential loss of their jobs, experience higher levels of emotional exhaustion as a result of their fear of losing the attachment and interconnectedness they enjoy. Embedded employees develop a level of fit within the organization that often translate into increased organizational citizenship behaviors and job performance and decreased absences and turnover (Lee, Mitchell, Sablynski, Burton, & Holtom, Reference Lee, Mitchell, Sablynski, Burton and Holtom2004). This level of fit and connectedness to the organization can translate into a strong level of employee dependence on the organization for financial, psychological, and interpersonal needs. The continual focus on their potential departure from the organization becomes overwhelming for these individuals and decreases the self-regulatory resources needed to regulate decision-making. Thus, the link between job insecurity and emotional exhaustion due to the threat of potential job loss may pave the way for them to engage in unethical behaviors to compensate for the perceived loss. Additionally, the loss of their current job would result in the potential sacrifice or loss of benefits and promotional opportunities.

In contrast, employees with low levels of embeddedness have a reduced level of attachment to and dependence on the organization (Lee et al., Reference Lee, Mitchell, Sablynski, Burton and Holtom2004; Mitchell et al., Reference Mitchell, Holtom, Lee, Sablynski and Erez2001). These employees may experience levels of uncertainty when faced with job insecurity. However, the decreased fit and attachment will allow these individuals to regulate their feelings and redirect their attention to capitalize on the opportunities that can result from the situation. The lack of dependence does not produce the sense of betrayal felt by those more deeply embedded in the organization reducing the need to pay back the organization through unethical behaviors such as employee theft. Further, employees who are less embedded in their organization will be less likely to engage in unethical behaviors as doing so can potentially tarnish their reputations as they move forward and pursue other opportunities.

Sample and Procedure

To test Hypotheses 1, 2, 4, and 5, we designed a time-lagged survey study. We elected to use this methodology because self-regulatory theory, which we used to develop these hypotheses, focuses specifically on an individual’s ability to regulate his or her emotions, thinking, and behaviors. To gain access to data about how an individual thinks, feels, or behaves, asking the individual directly is an appropriate approach. We surveyed currently licensed architects, engineers, and accountants who resided in the southern United States. Because we targeted professionals who served clients, we focused on unethical behaviors designed to help the organization, specifically unethical pro-organizational behavior (UPB).

We began by sending an invitation email to all individuals on the active licensed email lists for these three professions (N = 2000). The introductory email included information about their rights as participants, assured them of confidentiality, explained that there were no right or wrong answers to the questions found on the survey, and indicated the deadline for completing the survey was in two weeks. Respondents who wished to participate indicated their agreement by clicking on a link to the survey that included our predictor scales of job insecurity and adaptability. Those who elected not to participate were encouraged to click the opt out button. A reminder email was sent to individuals who had not completed the survey or not opted out after one week. The sample was reduced to 1599 by removing those with invalid email addresses or who opted out. Four hundred and thirty-one individuals started the survey (27%) and 243 (15% response rate) completed the survey. Data from 31 respondents were not included in the analysis because of significantly missing data; this left a usable sample of 212 respondents with complete data at time 1.

One month later we sent the 212 respondents from time 1 a second survey to capture our outcome variables of emotional exhaustion and UPB and our control variables. Of the 212, 119 started the survey (56%) and 117 (55%) finished it. Data from 10 respondents were eliminated from the dataset due to missing data resulting in a usable sample size of 107 (50%). The sample consisted of 91 (85%) males and 16 (14%) females with a mean age of 50.5 years. The majority of the sample (103; 96%) was Caucasian. Respondents indicated that they had worked for their organization an average of 11.5 years and for their immediate supervisor an average of 4.1 years.

Measures at Time 1

Job insecurity. We measured job insecurity with the 7-item scale developed by Hellgren, Sverke, and Isaksson (Reference Hellgren, Sverke and Isaksson1999). Responses were made on a 5-point scale (1= strongly disagree to 5= strongly agree). A sample item includes “I am worried about having to leave my job before I would like to.” Cronbach’s (Reference Cronbach1951) alpha for this scale was .83.

Adaptability. Participants provided self-ratings of adaptability using the 6-item scale developed by Morrison and Hall (Reference Morrison and Hall2001). Responses were made on a 5-point scale (1=strongly disagree to 5=strongly agree). Sample items from this measure include “I find it discouraging when the work I do in my job changes” (R), and “It is easy for me to adapt when I change jobs.” The Cronbach’s (Reference Cronbach1951) alpha for this scale was .76.

Measures at Time 2

Emotional exhaustion. Emotional exhaustion was measured with five items from Schaufeli, Leiter, Maslach, and Jackson (Reference Schaufeli, Leiter, Maslach, Jackson, Maslach, Jackson and Leiter1996). A sample item is “I have felt emotionally drained from my work.” Responses were made on a 5-point scale (1= never to 5 = very often). Cronbach’s (Reference Cronbach1951) alpha for this scale was .88.

Unethical pro-organizational behavior (UPB). UPB was measured with seven items developed by Umphress et al. (Reference Umphress, Bingham and Mitchell2010). A sample item is “To benefit my organization, I have withheld negative information about my company or its practices from clients.” Responses were measured on a 5-point scale (1 = strongly disagree and 5 = strongly agree). The Cronbach’s (Reference Cronbach1951) alpha for this scale was .87.

Control variables. In an effort to control for alterative explanations for variance in our dependent variable, UPB, we explored the usefulness of including several variables in our analyses as controls. These variables included age, gender, and social desirability. Extant research has demonstrated a relationship between ethical behavior and each of these variables (Borkowski & Ugras, Reference Borkowski and Ugras1998; Franke, Crown, & Spake, Reference Franke, Crown and Spake1997; Kish-Gephart et al., Reference Kish-Gephart, Harrison and Treviño2010; Tenbrunsel & Smith-Crowe, Reference Tenbrunsel and Smith-Crowe2008). However, since none of these variables were significantly correlated with our dependent variable (see Table 1), we followed the advice of Becker (Reference Becker2005) and did not include them in our analyses. Finally, we used Liden and Maslyn’s (Reference Liden and Maslyn1998) 3-item leader-member exchange contribution scale (α = .98) as our marker variable. A marker variable is a variable that is collected at the same time and in the same manner as the variables of interest. However, a marker variable must not be theoretically relevant to the study being conducted. If these conditions are met, then the only connection between the marker variable and the variables of interest is how and when it was collected. According to Lindell and Whitney (Reference Lindell and Whitney2001) a marker variable that adheres to these conditions can serve as a proxy for method variance. The variance due to the marker variable can be isolated and removed as a means of addressing common method variance (CMV) in the model. CMV is variance that is shared among the variables of interest that can be attributed to the way in which the data were collected and, when present, can inflate or deflate the actual relationships between the variables of interest.

Table 1: Means, Standard Deviations, and Correlations for Study 1 Variables

N = 107. UPB = Unethical pro-organizational behavior. LMX = Leader-member exchange.

* p < .05. ** p < .01. *** p < .001.

Note. Values in bold on the diagonal are the square root of the average variance explained which must be larger than all zero-order correlations in the row and column in which they appear to demonstrate discriminant validity (Fornell & Larcker, Reference Fornell and Larcker1981). Values above the diagonal are correlations after controlling for the marker variable (Lindell & Whitney, Reference Lindell and Whitney2001).

Analysis

We began by estimating a measurement model using LISREL 8.8 to examine the discriminant validity of our scales. Establishing discriminant validity allows us to infer that the scales we used in our study are separate and distinct from one another. This is an important first step as it provides evidence that the results from our analyses are not due to overlapping relationships (i.e., shared variance) among the variables studies. To examine the fit of our measurement model we used three statistics: comparative fit index (CFI), non-normed fit index (NNFI), and the root mean square error of approximation (RMSEA). The lower threshold for acceptable values for CFI and NNFI is .90 while the RMSEA should be .08 or less (Hu & Bentler, Reference Hu and Bentler1999). Next, we used hierarchical moderated regression analyses via IBM SPSS 22 and the PROCESS Macro written by Hayes (Reference Hayes2013) to test our hypotheses. The PROCESS macro estimates the bootstrapped conditional indirect effect of an independent variable on a dependent variable through a mediating variable at varying levels of the moderator. Bootstrapping is a process that uses sampling with replacement from the original sample to create a large number of bootstrap samples that have the same properties of the original sample. The process is designed to replicate the population from which the sample was drawn. Results from the analyses conducted on the bootstrap samples are used to create a 95% confidence interval. A confidence interval consists of an upper and lower bound into which a conditional indirect effect, calculated from the same sample, would fall 95% of the time.

Preliminary Analysis

Table 1 displays the descriptive statistics for and the correlations among the variables of interest for Study 1. Prior to testing our hypotheses, we conducted a confirmatory factor analysis (CFA) using LISREL 8.8 on the scales used in our study to ensure that they were independent and that the items produced the expected factor structures. Our measurement model consisted of four factors, one for each of our scales (i.e., job insecurity, adaptability, emotional exhaustion, and UPB). After allowing two pairs of items within the same scale to correlate, the fit indices showed that the 4-factor model adequately fit the data (X ² (267) = 479, p < .01; CFI = .90; NNFI = .89; RMSEA = .08) and all of the standardized path loadings were significant (p < .01).

Although our CFA results were encouraging, we also conducted a discriminant validity test described by Fornell and Larcker (Reference Fornell and Larcker1981) due to the moderate correlation between job insecurity and emotional exhaustion. Specifically, we calculated the square root of the average variance explained (AVE) for each of the scales in our study and placed these values on the diagonal in Table 1. According to Fornell and Lackner, discriminant validity is demonstrated when the square root of the AVE exceeds all of the corresponding correlations in the same row and column. For example, in Table 1, the square root of the AVE for emotional exhaustion is .77. This value is larger than the correlations in the row and column in which it resides. This same conclusion is reached for all comparisons in Table 1. The CFA results combined with the Fornell and Lackner discriminant validity test demonstrate support for the independence of our scales.

To minimize the threat of CMV, we followed several procedural and statistical steps suggested by Podsakoff, MacKenzie, Lee, and Podsakoff (Reference Podsakoff, MacKenzie, Lee and Podsakoff2003). On the procedural side, we collected our independent and dependent variables at two points in time. Introducing a time lag in the data collection effort was designed to help minimize consistency response patterns and demand characteristics. Second, in both survey administrations we guaranteed respondents confidentiality, which should decrease socially desirable responding. Third, we assured participants that there were no correct answers to reduce evaluation apprehension. Finally, we counterbalanced the order of the questions by randomly presenting them to our respondents.

To statistically explore the degree of CMV in our data, we implemented the approach suggested by Lindell and Whitney (Reference Lindell and Whitney2001). Specifically, we included a theoretically unrelated marker variable in our data collection. Any correlation between the marker variable and our variables of interest serves as a proxy for CMV as the only relationship between the theoretically unrelated marker variable and the constructs of interest are how the data were collected. We selected leader-member exchange contribution (Liden & Maslyn, Reference Liden and Maslyn1998) as our marker variable. As can be seen in Table 1, our marker variable was not significantly correlated with any of our constructs of interest, and there was minimal change between the zero-order correlations and those with the marker variable partialled out. Thus, we submit that the steps we took to avoid CMV in Study 1 were effective.

Hypotheses Testing

To test the first-stage moderated mediated model shown in the top half of Figure 1, we used IBM SPSS 22 to run the PROCESS macro (Hayes, Reference Hayes2013). The variables included in the interaction analyses were centered around the mean by subtracting the mean from the variables (Cohen, Cohen, West, & Aiken, Reference Cohen, Cohen, West and Aiken2003). Centered variables have a mean of 0 but retain their frequency distribution. Centering variables used in interaction analyses reduces the degree of collinearity among the variables that make up the interaction term and aids in the interpretation of the interaction coefficients. We also employed the bootstrapping option in PROCESS that uses sampling with replacement to create confidence intervals for the indirect effects. Results for these analyses are shown in Table 2.

Table 2: Moderated Mediation Regression Results

N = 107. DV = dependent variable. UPB = Unethical pro-organizational behavior.

Hypothesis 1 predicted that job insecurity at time 1 would predict emotional exhaustion at time 2. As can be seen in Table 2, Hypothesis 1 was supported as the relationship between these two variables was positive and significant. Hypothesis 2 predicted that the relationship between job insecurity and emotional exhaustion would be moderated by adaptability such that when adaptability was low, the positive relationship between job insecurity and emotional exhaustion be stronger. While the interaction was significant, prior to confirming support for Hypothesis 2 we first needed to graph the interaction so that we could confirm that it matched the form of our prediction. The graph of the significant interaction is shown in Figure 2.

Figure 2: Interaction Between Job Insecurity and Adaptability on Emotional Exhaustion.

The graph in Figure 2 shows that, as predicted, the relationship between job insecurity and emotional exhaustion was positive when adaptability was low and flat when adaptability was high. Simple slope analysis revealed that the slope of the low adaptability line was significantly different than zero (t = 4.97, p < .000) while the slope of the high adaptability line was not (t = 1.07, p = .29). These results offer support for Hypothesis 2.

Hypothesis 4 predicted that emotional exhaustion at time 2 would mediate the relationship between job insecurity at time 1 and UPB at time 2. As can be seen in Table 2, full mediation was found. Specifically, the link between job insecurity at time 1 and emotional exhaustion at time 2 was significant (.36, p < .00), the link between emotional exhaustion at time 2 and UPB at time 2 was significant (.16, p < .05), but the link between job insecurity at time 1 and UPB at time 2 was not significant (.08, p = .28). The indirect effect was .06, and this effect was significant (t = 2.00, p < .05). These results offer support for Hypothesis 4.

Finally, to test Hypothesis 5, we explored the conditional indirect effects of our model. Hypothesis 5 predicted that the indirect effect of job insecurity on UPB through emotional exhaustion would be stronger when adaptability was low. Results showed that the conditional indirect effects when adaptability was low (.09) or moderate (.06) were significant because the 95% confidence interval did not contain zero (.01, .22; .01, .16, respectively). However, when adaptability was high, the conditional indirect effect (.02) was not significant (95% confidence interval -.03, .14). This finding offers support for Hypothesis 5.

Sample and Procedure

We invited 169 undergraduate business students from two universities in the southern United States to take part in the study as organizational contacts and 52 participated. We used a “snowball method” where undergraduate students recruited up to 10 working adults (i.e., friend, relative, colleague) to complete the survey. Recruits provided their names and email addresses to the student who passed them on to the researchers. The names and email addresses of these individuals were compiled by the research team and an invitation to complete the survey was emailed to respondents from the research team. The snowball method of data collection has been successfully used in the management literature (e.g., Ambrose, Schminke, & Mayer, Reference Ambrose, Schminke and Mayer2013; Grant & Mayer, Reference Grant and Mayer2009; Marwitz, Mayer, Hoobler, Wayne, & Marinova, Reference Marwitz, Mayer, Hoobler, Wayne and Marinova2012; Moore, Detert, Treviño, Baker, & Mayer, Reference Moore, Detert, Treviño, Baker and Mayer2012). One week after the initial invitation a reminder email was sent to invited participants.

We recruited a total of 275 respondents and 213 individuals completed the survey (77% response rate). The surveys were administered online and each respondent created a unique identification code to ensure anonymity. We took a number of steps to ensure that the surveys were completed by the correct sources. First, in introducing the study, we highlighted the importance of integrity in the scientific process and articulated to the students that it was critical for the recruits to complete the surveys and that they could only complete one survey. Second, participants were asked to provide their telephone numbers so that they could be contacted to verify their completion of the survey. Participants were randomly called to verify their participation in the study. Two responses were excluded from the study because respondents indicated they did not complete the survey. When respondents completed their online surveys, time stamps and IP addresses were recorded to ensure that the surveys were not submitted at the same time from the same IP addresses. If two surveys were sent from the same IP address within a 5-minute time frame, we contacted the respondents via telephone to verify their completion of only one survey. Two responses were eliminated because they were submitted from the same IP address within a 5-minute time frame. After the completion of the verification process participant phone numbers were deleted from the data used for analysis.

Responses from 209 individuals were included in the data analysis. The participants (66% female) were an average of 37.2 years old. As for their ethnic background, 7% were African American, 2% Asian American, 81% Caucasian, 7% Hispanic, and 2% of the respondents listed “other.” They reported an average of 7.93 years of organizational tenure. All participants worked in the United States and in a variety of industries including academia, technology, government, insurance, finance, food service, retail, manufacturing, and healthcare.

Measures

All responses in Study 2 were measured on a 5-point scale with strongly disagree the anchor on the low end of the scale and strongly agree the anchor for the high end of the scale. We coded all items so that high values reflect high levels of the concepts studied. We used the same measures for job insecurity and emotional exhaustion that we used in Study 1. The unique scales for Study 2 are described below.

Embeddedness. Embeddedness was measured with six items from Mitchell et al.’s embeddedness scale (2001) that focused on organizational fit. A sample item is “I fit with the company’s culture.” The Cronbach’s (Reference Cronbach1951) alpha for this scale was .87.

Unethical behavior. Unethical behavior was measured with seventeen items utilized by Akaah (Reference Akaah1992), which were adapted from Treviño, Butterfield, and McCabe (1998). Respondents were asked if they had engaged in certain behaviors over the past 6 months. Sample items include “Padded an expense account up to 10%,” “Falsified time/quality/quantity reports,” and “pilfered company materials and supplies.” The Cronbach’s (Reference Cronbach1951) alpha for this scale was .90.

Control variables. In an effort to control for alterative explanations for variance in our dependent variable, unethical behavior, we explored the usefulness of including age and gender as control variables. Extant research has demonstrated a relationship between ethical behavior and these variables (Borkowski & Ugras, Reference Borkowski and Ugras1998; Franke et al., Reference Franke, Crown and Spake1997; Kish-Gephart et al., Reference Kish-Gephart, Harrison and Treviño2010; Tenbrunsel & Smith-Crowe, Reference Tenbrunsel and Smith-Crowe2008). However, since these variables were not significantly correlated with our dependent variable (see Table 3), we followed the advice of Becker (Reference Becker2005) and did not include them in our analyses. Finally, we used status striving as our marker variable. We measured this construct with 11 items from Barrick, Stewart, and Piotrowski (Reference Barrick, Stewart and Piotrowski2002). A sample item is “I focus my attention on being the best employee in the office.” The Cronbach’s (Reference Cronbach1951) alpha for this scale was .90.

Table 3: Means, Standard Deviations, and Correlations for Study 2 Variables

Listwise N = 205.

** p < 01. *** p < .001.

Note. Values in bold on the diagonal are the square root of the average variance explained which must be larger than all zero-order correlations in the row and column in which they appear to demonstrate discriminant validity (Fornell & Larcker, Reference Fornell and Larcker1981). Values above the diagonal are correlations after controlling for the marker variable (Lindell & Whitney, Reference Lindell and Whitney2001).

Analysis

We followed the same analysis procedures that we used in Study 1. We began by estimating a measurement model using LISREL 8.8 to examine the discriminant validity of our scales. Next, we used hierarchical moderated regression analyses via IBM SPSS 22 and the PROCESS Macro written by Hayes (Reference Hayes2013) to test our hypotheses. The PROCESS macro estimates the bootstrapped conditional indirect effect of an independent variable on a dependent variable through a mediating variable at varying levels of the moderator, making it an appropriate analytic technique for our study.

Preliminary Analysis

Table 3 provides the descriptive statistics and correlations for the variables of interest in Study 2. We followed the same procedure as in Study 1 and began our analyses by conducting a confirmatory factor analysis (CFA) using LISREL 8.8 on the scales used in Study 2 to ensure that they were independent and that the items produced the expected factor structures. Our four-factor measurement model included a factor each of our scales (i.e., job insecurity, embeddedness, emotional exhaustion, and unethical behavior). After allowing two pairs of items within the same scale to correlate, we found that the 4-factor model adequately fit the data (X ² (552) = 1225, p < .01; CFI = .93; NNFI = .92; RMSEA = .08) and all of the standardized path loadings were significant (p < .01).

Once again we conducted a discriminant validity test described by Fornell and Larcker (Reference Fornell and Larcker1981) by calculating the square root of the average variance explained (AVE) for each of the scales in our study. These values can be found on the diagonal in Table 3. Discriminant validity is demonstrated when the square root of the AVE exceeds all of the corresponding latent variable correlations in the same row and column (Fornell & Lackner, Reference Fornell and Larcker1981). This condition was met for all of our scales. Taken together, the CFA results and the Fornell and Lackner discriminant validity test demonstrate support for the independence of our scales.

Just as in Study 1, we followed several steps suggested by Podsakoff et al. (Reference Podsakoff, MacKenzie, Lee and Podsakoff2003) to minimize the threat of CMV. For instance, we guaranteed respondent confidentiality, which should decrease socially desirable responding, and we assured participants that there were no correct answers to reduce evaluation apprehension. Finally, we counterbalanced the order of the questions by randomly presenting them to our respondents.

Paralleling Study 1, to statistically explore the degree of CMV in our data, we implemented the approach suggested by Lindell and Whitney (Reference Lindell and Whitney2001) by including a theoretically unrelated marker variable in our data collection. Any correlation between the marker variable and our variables of interest serves as a proxy for CMV as the only relationship between the theoretically unrelated marker variable and the constructs of interest are how the data were collected. We selected status striving (Barrick et al., Reference Barrick, Stewart and Piotrowski2002) as our marker variable. As can be seen in Table 3, our marker variable was not significantly correlated with any of our constructs of interest, and there was minimal change between the zero-order correlations shown on the bottom half of Table 3 and those with the marker variable partialled out on the top half of Table 3. Thus, we submit that the steps we took to avoid CMV in our study were effective.

Hypotheses Testing

Similar to Study 1, to test the first-stage moderated mediated model shown in the bottom half of Figure 1, we used IBM SPSS 22 to run the PROCESS macro (Hayes, Reference Hayes2013). The variables included in the interaction analyses were centered (Cohen et al., Reference Cohen, Cohen, West and Aiken2003). We also employed the bootstrapping option in PROCESS that uses sampling with replacement to create confidence intervals for the indirect effects. Results for these analyses are shown in Table 4.

Table 4: Moderated Mediation Regression Results for Study 2

N = 205. DV = dependent variable.

Hypothesis 1 predicted that job insecurity would be positively related to emotional exhaustion. As can be seen in Table 4, Hypothesis 1 was once again supported as the relationship between these two variables was positive and significant. Hypothesis 3 predicted that the relationship between job insecurity and emotional exhaustion would be moderated by embeddedness such that when embeddedness was high, the positive relationship between job insecurity and emotional exhaustion would be stronger. While the interaction was significant, prior to confirming support for Hypothesis 3 we graphed the interaction so that we could confirm that it matched the form of our prediction. The graph of the significant interaction is shown in Figure 3.

Figure 3: Interaction Between Job Insecurity and Embeddedness on Emotional Exhaustion.

The graph in Figure 3 shows that, as predicted, the relationship between job insecurity and emotional exhaustion was positive when embeddedness was high and flat when embeddedness was low. Simple slope analysis revealed that the slope of the high embeddedness line was significantly different than zero (t = 3.88, p < .00) while the slope of the low embeddedness line was not (t = .72, p = .47). These results offer support for Hypothesis 3.

We once again tested Hypothesis 4, which predicted that emotional exhaustion would mediate the relationship between job insecurity and unethical behavior, but substituted a generic measure of unethical behavior for UPB. As can be seen in Table 4, partial mediation was found. Specifically, the link between job insecurity and emotional exhaustion was significant (.37, p < .00), the link between emotional exhaustion and unethical behavior was significant (.10, p < .00), and the link between job insecurity and unethical behavior was significant (.17, p < .00). The indirect effect was .04, and this effect was significant (t = 3.08, p < .00). These results offer additional support for Hypothesis 4.

Finally, to test Hypothesis 6, we explored the conditional indirect effects of our model. Hypothesis 6 predicted that the indirect effect of job insecurity on unethical behavior through emotional exhaustion would be stronger when embeddedness was high. Results showed that the conditional indirect effect when embeddedness was low (.01) was not significant because the 95% confidence interval contained zero (-.03, .06). However, when embeddedness was high (.06) or moderate (.04), the conditional indirect effects were significant (95% confidence interval .01, .09 and .01, .15 respectively). This finding offers support for Hypothesis 6.

Theoretical and Practical Implications

Our article contributes to the ethics literature in several ways. First, it highlights the role of the self-regulation process in promoting unethical behavior by emphasizing the role of stressors (i.e. emotional exhaustion) on various unethical behaviors. Previous research has identified important mechanisms such as moral awareness in the self-control depletion and unethical behavior process (Christian & Ellis, Reference Christian and Ellis2011; Gino et al., Reference Gino, Schweitzer, Mead and Ariely2011; Mead, Baumeister, Gino, Schweitzer, & Ariely, Reference Mead, Baumeister, Gino, Schweitzer and Ariely2009; Treviño et al., Reference Treviño, den Nieuwenboear and Kish-Gephart2014). We extend this research by showing that feelings of emotional exhaustion result in the depletion of an individual’s self-regulatory capabilities. This depletion results in the employee’s subsequent engagement in unethical behaviors that are either self-serving or benefit the organization.

Second, a number of studies have examined the influence of various dispositional factors and moral reasoning capabilities on ethical decision-making (Detert et al., Reference Detert, Treviño and Sweitzer2008; Moore, Reference Moore2008; Moore et al., Reference Moore, Detert, Treviño, Baker and Mayer2012). This article extends this line of research by examining how a situational factor, job insecurity, influences engagement in unethical behavior.

Third, we examine the role of the dispositional characteristic adaptability as a factor that diminishes the likelihood that individuals experiencing job insecurity would become emotionally exhausted and subsequently engage in unethical behavior in the form of UPB. Based on our findings it appears that highly adaptable individuals have skills and abilities needed to cope with the uncertainty presented by job insecurity by being open to scanning the environment for new opportunities. On the other hand, individuals with low levels of adaptability found the uncertainty of job insecurity more challenging and exerted energy in performing unethical behaviors to benefit the organization instead of adjusting and seeking alternative solutions or opportunities.

Last, we examine the role of embeddedness as a factor that can exacerbate the likelihood that individuals, who are faced with job insecurity, would engage in unethical workplace behavior. While a majority of the literature on job embeddedness has highlighted multiple benefits of job embeddedness, we found that highly embedded individuals viewed the uncertainty of job insecurity to be overwhelming and were more likely to engage in unethical behaviors as they consider the devastation of the loss of their position with the organization.

The findings of this study emphasize the importance of organizations creating an environment that reduces levels of emotional exhaustion in their employees since it is a pathway to unethical behaviors. To address this issue, organizations should consider implementing and enforcing policies that can reduce levels of emotional exhaustion in employees. Monitoring employee workloads to reduce excessive hours, discouraging electronic communication during certain hours, and requiring individuals to take accumulated vacation are policies organizations can implement to decrease levels of emotional exhaustion in employees.

Prescriptively, previous research also has suggested that organizations can decrease feelings of job insecurity by improving perceptions of supervisor and organizational support (Bussing, Reference Bussing1999; Eby, Casper, Lockwood, Bordeaux, & Brinley, Reference Eby, Casper, Lockwood, Bordeaux and Brinley2005; Halbesleben, Reference Halbesleben2006). Therefore, organizations should seek to provide an environment that supports, not excludes, employees who are experiencing job insecurity. One way of accomplishing this is by being transparent in communications with employees when considering layoffs. Organizations also can communicate the procedures that are being undertaken in the organization decision-making process if restructuring is being considered.

Strengths, Limitations, and Directions for Future Research

This article contains several strengths that should be highlighted. First, the samples were appropriate for the questions studied. The respondents in Study 1 were members of professions in which UPB could be undertaken as the professions studied deal with clients and the sample for Study 2 incorporated broader ranges of jobs which mapped well to the more generic unethical behavior measure we employed. Second, we extend previous research by examining the impact of an additional situational factor that is prevalent in today’s society, job insecurity, on an individual’s willingness to engage in unethical behavior. We highlight how perceptions of job insecurity can diminish an individual’s capacity to regulate their behaviors of engaging in acceptable organizational norms. Third, we offer the dispositional characteristic, adaptability, as a coping mechanism that reduces the level of emotional exhaustion employees facing job insecurity experience and the subsequent likelihood that these individuals will engage in unethical behaviors in the workplace. Last, we highlight the increased possibility of individuals, who are highly attached to their organizations and are placed in a situation that potentially threatens their livelihood, engaging in self-serving unethical behavior.

There are also a number of limitations to our research that are worth noting. First, Study 1 examined the proposed relationships using a sample of architects, engineers, and accountants, which may limit the external validity of our findings. In order to increase the generalizability of our findings, we replicated our study using a more generalizable sample. Additionally, the sample used in Study 1 was predominantly male, which is common in these professions. We offset this limitation by employing a more gender-balanced sample in Study 2. Last, the response rate for Study 1 was low at 15%. However, we addressed this issue by conducted a second study that had a higher participation rate of 77%. What remains to be seen is whether our results hold with a sample from a single organization. We leave this avenue for future researchers.

While this article highlights the role of emotional exhaustion in depleting self-regulatory resources and the role of adaptability and job embeddedness as boundary conditions of this relationship, future research should investigate additional ways to minimize the loss of or restore self-control resources. Conservation of resources theory suggests that individuals strive to acquire and maintain a balance of resources and that they experience stress when they perceive threats to their current resources (Hobfoll, Reference Hobfoll1988, Reference Hobfoll1998). To cope with the stress they experience in the work domain due to threatened resources, individuals often pull resources from other life domains (Hobfoll, Reference Hobfoll2001). As this relates to the current study, social support (i.e., coworker and supervisor support) is a resource that may mitigate the relationship between job insecurity and emotional exhaustion. Specifically, if an individual perceives that he or she has the support of others while facing job insecurity, the amount of emotional exhaustion experienced may decrease, even for individuals with low levels of adaptability or high levels of embeddedness. Moral identity also should be examined as a potential boundary condition of the emotional exhaustion and unethical behavior relationship as previous research has demonstrated that moral identity weakens the strength of the self-control depletion and dishonesty relationship (Gino et al., Reference Gino, Schweitzer, Mead and Ariely2011).

Additionally, future research should examine the role of moral disengagement in the decision-making process for individuals facing job insecurity and emotional exhaustion. Moral disengagement is the self-regulatory process through which individuals free themselves from guilt and self-sanctions allowing them to engage in unethical behaviors (Bandura, Reference Bandura, Kurtines and Gewirtz1991, Reference Bandura1999). A stream of literature has identified factors that influence an individual’s decision to engage in unethical behavior (Caruso & Gino, Reference Caruso and Gino2011; Chugh et al., Reference Chugh, Bazerman, Banaji, Moore, Cain, Lowenstein and Bazerman2005; Gino et al., Reference Gino, Schweitzer, Mead and Ariely2011; Kish-Gephart et al., Reference Kish-Gephart, Harrison and Treviño2010; Tenbrunsel & Messick, Reference Tenbrunsel and Messick2004; Treviño & Youngblood, Reference Treviño and Youngblood1990). The inclusion of moral disengagement as a potential mediator of the emotional distress and unethical decision-making process will further the understanding of these complex relationships.