Work attendance anxiety.

An emerging problem during the early COVID-19 pandemic was work attendance anxiety. Workers in the food retail, food service, and hospitality industries experienced higher levels of anxiety because of the many uncertainties and personal safety issues during COVID-19 [23]. Available studies document the prevalence of general anxiety during COVID-19, but workplace anxiety specifically was not sufficiently addressed [22,41].

Investigating workplace anxiety is important as unmanaged anxiety can lead to long-term negative effects on work performance, motivation, health, and job satisfaction [12,42–45]. Despite its significant consequences, to date, workplace anxiety has not been adequately addressed as a health aspect in organizational psychology models. There have been few studies investigating workplace anxiety in organizational settings, and there are no unified terms or definitions. Muschalla and Linden (2009) [13] uses the term ‘workplace phobia’ and defines it as a “phobic anxiety reaction with symptoms of panic occurring when thinking of or approaching the workplace.” Jones et al. (2016) [12] measures job anxiety using three anxiety states by asking how much of the time the job made them feel tense, worried, or uneasy. Different types of stressors can lead to various dimensions of job anxiety.

Studies have shown that workplace phobia can be empirically distinguished from conventional anxiety disorders through psychological and physiological responses to work-related stress. One experimental study identified individuals with workplace phobia by their intense fear and severe anxiety symptoms when approaching the workplace, along with higher heart rate and subjective reports of fear [46]. Further research indicated that while some individuals experience both non-work-related anxiety disorders and workplace phobia, workplace phobia is distinct among anxiety sufferers, with some individuals experiencing only workplace phobia without other non-work-related anxiety disorders [13].

In this study, we focus on anxiety over going to work and coin this as “work attendance anxiety.” We define it as “a state of feeling nervous, anxious, or on edge to come into work.” We consider work attendance anxiety as a potential indicator of threats and problematic situations at work, which adversely affects workers’ judgments and their level of job satisfaction [42–44,47]. Also, the unpredictable and uncontrollable nature of precarious work schedules can lead to psychological impacts such as increased work attendance anxiety. Thus, given the limited elaboration on workplace anxiety in existing literature, we study the relationship between precarious work schedule and work attendance anxiety as well as the association between work attendance anxiety and job satisfaction. We hypothesize as the following:

1. H2a. Precarious work schedules – lack of control over work hours, unpredictability in work schedules, work hours instability, and non-standard work hours– will be associated with higher levels of attendance anxiety.
2. H2b: Work attendance anxiety will have a negative relationship with job satisfaction.

Employer support.

Many essential workers in frontline jobs were not able to access benefits such as personal protective equipment (PPE), paid sick leave, wage increases or flexible work schedules that can help them manage job stress and protect their health [10,11,23,48,49]. The Centers for Disease Control and Prevention guidelines for workplaces on preventing infections carried neither statutory nor regulatory power, leading to substantial variation in safety practices across states, sectors, and firms [5,50]. In such conditions, workers were even less likely to be able to satisfy their psychological and basic survival needs.

Studies demonstrate that employer support and policies can play a significant role in the psychological well-being of their workers. Essential workers who felt protected by employers or governments during COVID-19 experienced a decrease in general anxiety symptoms [5,22,29]. For instance, workers whose employers offered paid leave were less likely to work when sick and more satisfied with their jobs [51]. Under the JD-R framework, we expect that employer support as job resources will help reduce attendance anxiety and buffer the negative impact of attendance anxiety on job satisfaction. Thus, we hypothesize as the following:

1. H3a: Employer support – PPE, flexible work schedule, wage increase, paid leave – will be associated with lower levels of attendance anxiety.
2. H3b: Employer support – PPE, flexible work schedule, wage increase, paid leave – will be associated with higher levels of job satisfaction.
3. H3c: Employer support – PPE, flexible work schedule, wage increase, paid leave – will moderate the relationship between attendance anxiety and job satisfaction.

Sample inclusion criteria.

First, individuals were required to be part-time hourly workers who were currently employed, had been working since June 2020 or before, worked less than 35 hours a week, and were paid hourly wages. Workers needed to have worked for at least a month to answer the survey questions. Furthermore, workers had to be in non-managerial positions. This excluded self-employed workers and salaried workers (non-hourly workers). We chose to focus on part-time hourly workers because they often experienced more vulnerability related to precarious work schedules and may have received less employer support compared to full-time or salaried employees. This group was particularly relevant to our research objectives. Second, individuals were required to work in “food and beverage stores” within the retail sector. Food and Beverage Stores were classified as NAICS 445: 4971 Supermarkets and other grocery (except convenience) stores, 4972 Convenience Stores, 4980 Specialty food stores, and 4990 Beer, wine, and liquor stores. We focused on food and beverage stores since they had been less impacted by the immediate closures of non-essential businesses in many states. The U.S. government categorized these businesses as essential, so many food and beverage stores remained open. Lastly, we targeted the U.S. working-age population, but the minimum required age for MTurk workers was 18. So, participants were required to be between the ages of 18 and 64.

Sample recruitment.

We used Amazon Mechanical Turk (MTurk) platform linked with Qualtrics to recruit respondents in a timely manner. Despite MTurk not always being the optimal method for data collection, its use was particularly valuable given the constraints during the pandemic. It allowed us to reach specific workers effectively, considering their unique schedules, risks, and demands at that time. We conducted qualification tests twice to ensure that respondents met all the sample inclusion criteria. The main survey was conducted between August 6 and August 19, 2020, and 230 workers took the survey among 316 qualified workers (response rate: 72.7%). A small sample size can reduce statistical power and may underestimate the impact of the variables under investigation. However, it is worth noting that the part-time retail workers in food and beverage stores, meeting our inclusion criteria, constituted only a minuscule fraction of the population, representing only 0.04% of the July-August 2020 Current Population Survey (CPS) basic monthly sample. This limited representation underscores the challenge of obtaining a sizable sample from this specific target population.

Protection of human subjects.

The researcher protected the participants in the study by adhering to ethical and legal guidelines from the university institutional review board. Before the participants proceeded with the survey, written informed consent was presented through MTurk and obtained by checking the “I agree” box online. The informed consent indicated that participation in the study was voluntary and that participants had the right to withdraw from the study without incurring adverse consequences. The researcher did not collect any personally identifiable information from the participants. Their Worker IDs were only used to pay them correctly and were permanently deleted from the experimenter’s data after payment was complete. Upon completion of the survey, all participants were assigned unique identifying numbers for the purpose of analysis. Access to survey data during collections was restricted via a university-owned and password-protected Qualtrics account, accessible only to co-investigators of the study. Once collection rounds closed, exported data were stored in an encrypted folder on the researcher’s computer.

Sampling bias and re-weighting.

MTurk workers tend to be younger, make lower than average income, and be Caucasian [52]. To better approximate a nationally representative sample of individuals, we restricted the CPS Basic Monthly data from July to August based on our sample criteria and use weights based on gender, education, race, and age. All regression models were estimated with these survey weights.

Data.

The total sample in the original dataset consisted of 230 participants. Based on the eligibility criteria, the study excluded participants who were not retail workers, such as Instacart personal shoppers, provided inconsistent data (i.e., the greatest number of work hours was smaller than the smallest number of work hours), were below 18 or above 64 years of age, and worked 35 hours or more. The final analytic sample included 179 participants who provided information on all variables used in the analysis. Table 1 provides the individual demographic and socioeconomic characteristics of the sample with comparisons to the CPS in July-August 2020, where the CPS sample was restricted to those workers who met the same eligibility criteria as our study sample. Most participants in our sample were White (74.9%), not married or not partnered (62.6%), and non-immigrants (96.6%). Their mean age was 31.6, 50.8% had not completed college, and 80.4% had an annual household income below $75,000 (Table 1). Our sample was younger, more educated, and had lower household income than the CPS sample (In the CPS sample, college educated workers account for only 9.7%, which is much lower than in our study sample. We have verified that these CPS numbers are reliable by comparing our results with a study that closely aligns with our target industry and study period, utilizing an April 2020 CPS sample. Cho and Lee (2020) [53] reports that 14% of grocery store workers held bachelor’s degrees or higher. It is important to note that this percentage encompasses both part-time and full-time employees. In our CPS sample, we narrow the focus to part-time workers, which may explain why the educational attainment level is lower than in Cho and Lee (2020) [53]).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Demographic characteristics of workers in MTurk versus Current Population Survey (CPS) sample.

https://doi.org/10.1371/journal.pone.0318586.t001

Precarious work schedules.

Following the measurements of National Survey on Work Schedules, work-hour instability was calculated by dividing the hour range by the reported usual hours [(greatest number of hours worked in a week in the last month – smallest number of hours worked in a week in the last month) ÷ usual work hours in the last month].

For timing of hours, standard hours were a regular day shift and non-standard hours include an evening shift, a night shift, a rotating shift, a split shift, or irregular schedule. We created a dichotomous variable indicating workers who had (1) and had not (0) worked in nonstandard timing.

Predictability of hours was measured by the length of advance notice given to workers regarding their schedules and was categorized as the following: (1) A day or less in advance, (2) 2-6 days in advance, (3) 1-2 weeks in advance, (4) 3 or more weeks in advance, and (5) Schedule never changes. We created a dichotomous variable indicating workers who received their schedule notice in less than a week (1) and a week or more in advance (0).

For schedule control of hours, we asked workers if their total number of work hours each week and starting and finishing time were decided by (1) their employers, (2) their employer, but with employee’s input, (3) employees with certain limits, or (4) exogenous factors that are outside of employees’ and employers’ control. We created a dichotomous variable indicating workers who had no input (1) and some input (0) in their schedule decision.

Work attendance anxiety.

Attendance anxiety was measured by asking respondents if they feel nervous, anxious, or on edge to come into work on a 5-point scale. (1 = very slightly or not at all; 2 = a little; 3 = moderately; 4 = quite a bit, 5 = extremely).

Job satisfaction.

Job Satisfaction was measured by the Job Satisfaction Index (JSI) scale, which was developed by Brayfield and Rothe (1951) [54]. The JSI comprises 18 items to measure overall job satisfaction (e.g., ‘My job is usually interesting enough to keep me from getting bored’). Job satisfaction can be considered as a “global construct” by combining all feelings and cognitions toward the job. Job satisfaction can also be measured by various “facet items/scales” such as job task, payment, promotion, supervision, or coworkers [27]. The global construct is more appropriate when job satisfaction is considered for the well-being of employees. Facet scales may omit some areas that may be important to the individual. Moreover, simply adding facets for all people may not capture the unique individual methods of combining components to arrive at a summary feeling [55]. Therefore, the study employed a global construct approach to reflect the general job satisfaction of the individuals rather than adding various facets of jobs. Responses followed a five-point Likert scale for each statement ranging from a value of 1 for ‘strongly disagree’ to 5 for ‘strongly agree’. The Cronbach’s alpha equaled.94 in our study.

Control variables.

We controlled for gender, age, marital status, household income, and number of children which have been shown to influence job satisfaction [56–59]. Age was calculated by subtracting the birth year and month from the year and month of the survey. Gender was coded as male (0) and female (1). Marital status was categorized as workers who are (1) and are not (0) married or partnered. Number of children was a dichotomous variable indicating workers who have 1 and more children (1) and no children (0). Total household income was coded as below (1) and above (0) $75,000 which was the U.S. median household income.