I believe what helps researchers discover a dissertation topic is by engaging with his or her inner person. In my personal life, discovering ways to make my work environment enjoyable is a thought that resonates in my spirit. My ambition led me to explore the topic of work-related stress and provide senior leaders within the supermarket industry with evidence-based information to improve their employees' work environment. In this case study, I provide an overview of the research method I made use of to explore burnout and turnover intentions of frontline retail grocery employees.
By the end of this case study, you should
- Understand the basic principles and rationale for using the descriptive correlation design
- Understand the basic principles and rationale for using structural equation modeling
- Understand how structural equation modeling help researchers understand the relationships between factors
The topic of employee turnover has sparked interest in many organizations. Employee turnover has a negative impact on companies and is endemic in the retail industry. Because turnover is relatively high in the retail industry, is costly, and there is a dearth of evidence-based information regarding employee turnover in the retail industry in the literature, I decided to explore turnover in this industry.
After a more exhaustive literature review, I discovered that researchers Michael Leiter and Christina Maslach (2009) described burnout as emotional exhaustion, cynicism, and lack of accomplishment. Emotional exhaustion is mental or emotional fatigue that one experiences on the job. Cynicism is pent-up anger against or a lack of trust in the company and lack of accomplishment is the belief that one is not able to make a meaningful contribution toward the organization's mission. The six salient areas of the work environment that could cause burnout are the employees' perception of their workload, community, control, reward, fairness, and values. Workload is the amount of work that an employee can carry out in a set amount of time. Community is the relationship that employees have on the job with their manager, fellow co-workers, vendors, and customers. Control is the ability to make choices about one's job tasks. Reward is not only the financial compensation for fulfilling one's job duties but also the social recognition and appreciation for one's job efforts. Fairness is the perception that management treats everyone equally and that workplace policies are the same for everyone. Values are principles that employees hold to be true, such as compassion, trust, and respect (Leiter & Maslach, 2011).
The literature review also revealed that turnover in the retail industry was significantly higher among major industries in the United States. According the Bureau of Labor Statistics (2013), the annual quit rate for the retail industry was 26.8% compared to the national average of 18.8%.
Job burnout is an occupational condition that can have negative implications for any organization as well as workers' well-being. Burnout is characterized by chronic stress and manifests itself as emotional exhaustion, cynicism, or inefficacy (i.e. lack of personal accomplishment). For organizations, job burnout is problematic because of work-related outcomes such as low productivity, high absenteeism, low morale, more mistakes made on the job, increased health care claims, more on-the-job accidents, and job turnover. Employees are likely to experience burnout when there are negative perceptions of the work environment. Leiter and Maslach (2005) estimated that the combined cost of burnout for all organizations within the United States is US$300b each year.
The negative implications of burnout on the well-being of workers are chronic fatigue, insomnia, forgetfulness, susceptibility to illness, loss of appetite, increased anxiety, anger, and depression. Research suggests that burnout is particularly high in occupations with strong interpersonal interaction and for individuals who work in client-centered professions such as retail. Although retail is one of the most client-centered professions, burnout research in this industry is scarce.
I explored the topic of burnout in the retail industry because within the sales literature, researchers have overlooked burnout almost completely (Lewin & Sager, 2007). To reduce turnover and build a more engaging work environment, senior leadership needs an understanding of burnout beyond the emotional exhaustion dimension (Rutherford, Hamwi, Friend, & Hartmann, 2011).
I carried out this research from January 2011 to September 2013. I provided funding for my own research, and during this time span, I completed the following:
- found a retail grocery chain that was willing to participate in this research
- determined the appropriate sample size in an effort to obtain statistically significant results
- received Institutional Review Board (IRB) approval to carry out this research
- captured survey responses from participants
- analyzed data and report findings
- interpreted findings and provided recommendations
- submitted final study to Quality Review Board (QRB) for approval
This study made use of a descriptive correlational design, which researchers use to describe the relationship between variables under study. Paul D. Leedy and Jeanne Ellis Ormod (2010) asserted that researchers make use of this type of design to analyze and report existing phenomena. Based on the findings about particular phenomena, researchers can formulate meaningful conclusions in a timely manner. For my research, I used this particular design to describe the relationship between the work environment, burnout, and turnover intentions. The statistical tool I used to discover the relationship among these previously mentioned variables was structural equation modeling (SEM). The advantage of SEM is that this research tool enables researchers to understand the strength of association between factors and the direct and indirect effects. This statistical tool also enables researchers to understand how each factor influences work-related outcomes, such as job turnover. By understanding the direct and indirect effects, researchers can draw more sound conclusions. Researchers refer to the strength of association between variables as regression weights. Each regression weight in the SEM has probability values (p values) that help readers determine the statistical significance between variables (Vogt, 2007); p values less than .05 are statistically significant, whereas p values that are greater than .05 are not statistically significant.
I used three survey instruments to capture the frontline retail grocery employees' perception of their work environment, burnout, and turnover intentions. First, I used the Areas of Worklife Survey (AWS), 6th edition, to assess how the employees perceive their work environment. The AWS contains 28 items that produce scores for the six areas of the work environment (Leiter & Maslach, 2009). Second, I made use of the Maslach Burnout Inventory (MBI) to ascertain the employees' perceptions of burnout (i.e. emotional exhaustion, cynicism, and inefficacy). Last, I used E. Kelloway, Benjamin Gottlieb, and Lisa Barham's (1999) Turnover Intentions Survey (KGB-TIS) to assess the employees' turnover intentions.
According to John Creswell (2008), the aim of sound research is to have reliable measures. The MBI General Survey (MBI-GS), AWS, and KGB-TIS are all reliable because researchers established reliability of each measure in previous studies. These researchers made use of Cronbach's alpha, which predicts the likelihood for the scores in one item in a particular subscale to be similar to the scores of another item in the same subscale. Cronbach's alpha scores that range from .70 to .80 are satisfactory, and scores of .90–.95 are highly reliable (Connelly, 2011).
In a study of 667 nurses in Canada, Leiter and Maslach (2009) established reliability for the MBI-GS and AWS by reporting acceptable Cronbach's alpha scores: exhaustion (α = 0.89), cynicism (α = 0.89), inefficacy (α = 0.92), workload (α = 0.85), control (α = 0.70), reward (α = 0.82), community (α = 0.80), fairness (α = 0.77), and values (α = .82). In a study to assess turnover intentions among 236 employees from various industries in Western Ontario, Canada, Kelloway et al. (1999) used Cronbach's alpha to establish reliability for the KGB-TIS (α = 0.92).
This research used analysis of moment structures software to analyze the data. Researchers can also use this software to transform a graphic model into an equation. I also made use of predictive analytic software to compute the descriptive statistics (i.e. mean, median, and mode). Descriptive statistics help researchers understand trends in their data and provide insight on where one score stands in comparison with others (Creswell, 2008).
I used a non-probability sampling approach because this approach allows researchers to select participants because they are (a) willing and able to participate and (b) represent some characteristics of what the study explores.
Participation in this study was voluntary and allowed participants the opportunity to withdraw at any moment without penalty to themselves. The test site's human resources department provided permission to recruit participants. Moreover, the human resources department did not have access to the participants' responses. The demographic survey requested only general information such as age, gender, years of experience, and job title. The participants had the option to access the survey link at their place of employment or from home via the http://mindgarden.com website.
I provided participants with an informed consent form that explained the purpose of the study and how the study would protect their identities and personal information. In that form, I explained that participation was voluntary and that they could withdraw at any time without penalty. This form also contained my e-mail address and contact number so that participants could reach me to ask any questions about the study.
A human resources administrator provided me with a list of employees for each of the store locations from the test site. This list contained the employees' hire dates and job titles. From this list, I sought out employees who were 18 years of age or older and currently work for the organization. From the eligible pool, I chose 100 candidates per store and provided them with printed copies of the informed consent form.
I provided each participant with a unique pass code to access the surveys from http://www.mindgarden.com. This website provided participants with confirmation when they complete the surveys. The data collection process began on 18 February 2013, and ended on 28 March 2013. After the data collection period, I reached out to the participants by letter informing them that data collection was over and thanked them for their participation.
I maintained the participants' data on a USB flash drive, and in 5 years after the study's completion, I will delete the information and shred the informed consent forms.
The hypothesis used to guide this study was that the work environment, burnout, and turnover intentions are all statistically related. Leiter and Maslach (2009) outlined two specific paths for employees' intention to leave their organization and I made use of SEM to test both paths. In the first path for turnover, Leiter and Maslach (2009) posited that the employees' perception of control shapes their perception of community, fairness, and reward. The employees' perception of community, fairness, and reward shapes their perception of values. The employees' perception of values shapes their perception of cynicism, and the employees' perception of cynicism has a positive relationship with turnover. For the second path for turnover, Leiter and Maslach (2009) asserted that control shapes the employees' perception of their workload. Work overload and emotional exhaustion are statistically related. Emotional exhaustion leads to cynicism, which in turn increases the likelihood for employees to leave their organization. Figure 1 is an illustration of Leiter and Maslach (2009) combined hypothesis.
In this model, there are 10 variables, and 9 of these variables are exogenous. The ‘e’ above each variable represents an error term. Error terms take into consideration any unspecified error and variance due to measurement (Vogt, 2007). Arrows between variables indicate that one variable influences the other. After I made use of analysis of moment structures to analyze this model, each factor contained a squared multiple correlation (R2) coefficient that displays the strength of the relationship between variables (Polit, 2010). Each arrow has an assigned path coefficient.
Both paths to turnover were statistically significant. For the first path, the employees' perception of control had a positive relationship with fairness (.417, p < .01), community (.558, p < .01), and reward (.370, p < .01). This means that if employees believe that they do not have control of their job, they are more likely to believe that the organization does not show enough fairness, does not show community, and lacks rewards. Fairness, community, and rewards are statistically related to values. Values had a positive, statistically significant relationship with fairness (.417, p < .01), community (.265, p < .05), and reward (.205, p < .05). A negative perception of fairness, community, and rewards may cause value incongruence. Value incongruence is when individuals believe they are not respected by their organization. Values have an inverse relationship with cynicism (−.317, p < .05). This means that if employees believe that the company is not honoring their values, the employees are more susceptible to experience cynicism. Cynicism had a positive relationship with turnover intentions (.399, p < .01). This means that the higher level of cynicism, the higher the likelihood for turnover.
With regard to the second path of turnover, control has an inverse relationship with workload (−.217, p = .01). This means a decrease in control may cause the employees to believe that there is work overload. Workload has a positive relationship with emotional exhaustion (.356, p < .01). This means that as employees' level of work increases, their level of emotional exhaustion increases. Emotional exhaustion has a statistical relationship with cynicism (.593, p < .01). This means that the higher the employees' level of emotional exhaustion, the more likely they are to experience cynicism. Cynicism has a positive relationship with turnover (.399, p < .01). This means as the employees' level of cynicism rises, their intent to leave their company may increase.
I made use of SEM because Leiter and Maslach (2009) used this statistical tool in their research to determine the relationships between the work environment, burnout, and turnover among nurses in Canada. I wanted to use their hypothesized model to confirm whether the employees' work environment and perceptions of burnout could predict turnover intentions in a retail grocery chain.
SEM is useful for describing the relationships between variables. Moreover, SEM shows the direct and indirect effects of factors on the outcome variable. The factors in this research were control, community, reward, fairness, workload, values, emotional exhaustion, cynicism, and inefficacy. The outcome variable was turnover. Another benefit of SEM is that this statistical tool addresses multi-collinearity, which happens when factors within the model are too highly correlated.
Novice users of a SEM should bear in mind to use a model that was used in a previous study. One could use SEM that was not previously tested; however, determining a good fit could be very time consuming.
Although SEM is a very effective statistical tool, some users may have difficulty mastering and understanding it. Initially, I tried to use multiple regression to determine the relationships between variables and later discovered that this was not plausible because of multi-collinearity. I lost valuable time and money trying to use multiple regression. If I could do everything all over again, I would consult with a statistician early in the research so that he or she can explain to me the components of SEM and how this tool addresses the relationships between variables.
This case provided an insight into the research method I employed to describe the relationship between frontline retail employees' perception of their work environment, burnout, and turnover intentions. Through SEM, researchers can better understand the statistical significance between variables. By understanding the statistical significance between variables, researchers can formulate sound conclusions.
- What is multi-collinearity and how does structural equation modeling deal with this issue?
- Why is it important to establish reliability of survey instruments?
- Protecting the participants' information is important and failing to do so is unethical. What methods can you employ to ensure that you fully protect your participants' personal information and survey responses?
- Why are probability values (p values) important?
- What are the pros and cons of using a structural equation modeling?
- Explain why or why not you believe that saving and destroying participants' information 5 years after completion of a study is important.