Introduction

Absenteeism, or not attending work when one is expected to, in the public sector has received considerable attention over the last decades due to the costs absenteeism imposes on a sector in constant search for increasing efficiency and reducing costs (McGrandle and Ohemeng 2017). At the same time, the occurrence of part-time employment (PTE) has increased significantly in both public and private sector organizations in most Western countries (Anxo et al. 2012; Buddelmeyer et al. 2008; EU 2016; Martin and Sinclair 2007; OECD 2016; Sandor 2013; Wittmer and Martin 2011). Part time work is particularly common in organizations dependent on 24/7 functioning, and thus includes most large health organizations like hospitals and other health institutions, most of them – at least in Europe - public organizations (Cooke and Bartram 2015).

Still, there are very few studies focusing on the possible relationship between part-time employment (PTE) and absenteeism, and the results from these studies are inconclusive. Some report PTE to be associated with higher absenteeism (Flach et al. 2012; Gjesdal and Bratberg 2002, 2003; Kaerlev et al. 2004), while other report no significant association between the two (Flach et al. 2008; Niedhammer et al. 2008; Smulders 1993). A few studies come to the conclusion that PTE is associated with lower absenteeism (Dionne and Dostie 2007; Grobe 2016).

Building on role theory (Katz and Kahn 1978) it is argued that PTE may have dual indirect effects that at the same time have positive and negative effects on absenteeism. On the one hand, less hours spent in the organization may result in lower integration into the organization, in turn resulting in lower affective organizational commitment (AOC). As several studies conclude with a negative relationship between AOC and absenteeism (Clausen et al. 2014a; Clausen et al. 2012; Knapstad et al. 2014; Rongen et al. 2014; Sagie 1998; Schaufeli et al. 2009), it is argued that a positive indirect effect of hours worked will be mediated to absenteeism by AOC.

On the other hand, working part time may decrease tensions between family obligations and the organization. Spending less time in the organization might create less work-family conflict (WFC). As several studies indicate that WFC is positively related to absenteeism (Clays et al. 2009; Darr and Johns 2008; Holmgren et al. 2009), it is argued that a negative indirect effect of hours worked will be mediated to absenteeism by WFC. In sum, the two diverging effects might explain the multiple studies reporting no total effect of PTE on absenteeism.

Empirically, data stems from a large survey in a Norwegian public hospital in 2013 where all employees working more than 30% of a full-time position (i.e. 100%) received a questionnaire. Survey data from 2013 was coupled with register (objective) data on actual absenteeism in 2014 constituting a database combining subjective and objective data as well as a time lag between the survey and the registration of absenteeism.

Linking Part-Time Work and Absenteeism

The arguments proposed on the links between PTE and absenteeism rely on the basic assumption that humans enact multiple roles, thus encountering different and sometimes conflicting demands (Katz and Kahn 1978; Mowday et al. 1982; Pfeffer and Baron 1988). The hypotheses outlined in the following all share the same basic assumption: the number of hours one spends at work will directly affect where individuals spend time, in turn forming the links between employees and the organization.

Time spent in the organization will probably affect the strength of emotional bonds between an employee and her/his organization. In general, studies support the notion that part time workers are somewhat less emotionally involved in the job and the organization than full time workers (Eberhardt and Shani 1984; Feldman and Doerpinghaus 1992; Thorsteinson 2003). The more time one spends in a social setting like an organization, the stronger the propensity to identify with that social setting as time represents “a form of accumulated history” (Nahapiet and Ghoshal 1998; Schaubroeck et al. 2013). Emotional attachment, identification, or affective organizational commitment (AOC) is central in a large body of research over the last two decades (Kooij et al. 2010; Meyer and Allen 1991; Meyer et al. 2002), and some empirical studies indicate PTE and AOC to be associated. Jacobsen (2000) argues that part-time workers are less integrated, feel less as “a part of the organization”, both operationally and socially. Following this line of thinking, AOC should be lower for part time that full time workers, and should decrease with decreasing size of post, an assumption supported in most existing empirical studies on the topic (i.e. Chang and Chelladurai 2003; Conway and Briner 2002; Pfeffer 1994; Thorsteinson 2003).

As most studies link high AOC to low absenteeism (Clausen et al. 2010, 2014b; Mathieu and Zajac 1990; Meyer et al. 2002), it is assumed that strong affective commitment to the organization binds an employee to the workplace, concurrently reducing absenteeism. The following hypotheses on how the effect of PTE on absenteeism is mediated through AOC is thus presented:

  • H1a: Full time employees will have stronger affective organizational commitment (AOC) than part time employees.

  • H1b: The stronger the affective organizational commitment (AOC), the lower the absenteeism.

  • H1c: A negative indirect effect of full time employment on absenteeism is mediated through AOC.

Time spent fulfilling demands from work related roles comes at the expense of time spent with the family or at home. This simple, but powerful insight from role theory links PTE clearly to work-family conflict (WFC) (Sandor 2013). WFC is an inter-role conflict in which engaging in one role interferes with engaging in another role (Greenhaus and Beutell 1985). In this study, the focus is set on work interferences with family, defined as “a form of inter-role conflict in which the general demands of, time devoted to, and strain created by the job interfere with performing family-related responsibilities” (Netemeyer et al. 1996, p.401). As WFC is a potential source of stress, both psychological well-being and behavior can be affected negatively (Amstad et al. 2011; Geurts et al. 2003). Absenteeism may thus be due to coping with the inter-role conflict situation, as “taking some days off” to manage these conflicts (Allen et al. 2000; Jansen et al. 2006). Part-time work might thus reduce WFC, and earlier research has found that part-time workers, compared to full-time workers, have lower WFC (for example Higgins et al. 2000; Olsen and Dahl 2010). This possible association between absenteeism and WFC, as well as between PTE and WFC leads the following hypotheses:

  • H2a: Full time employees will experience more intense work-family conflict (WFC) than part time employees.

  • H2b: The more intense the work-family conflict (WFC), the higher the absenteeism.

  • H2c: A positive indirect effect of full-time employment on absenteeism is mediated through WFC.

Control Variables

For some, PTE may be a rational choice, an adjustment to individual wants and needs (Feldman 1990; Iseke 2014; Krausz et al. 2000; Van Emmerik and Sanders 2005). For others, PTE is involuntary and represents a sort of job discrepancy. Working less than wanted might conduce to economic challenges and stress (Iseke 2014), as well as poorer job quality and worker well-being (Kauhanen and Nätti 2015). Employees working less than wanted may be more hesitant to take some days off as this might impair future possibilities of getting a higher percentage of work. On the other hand, there are those experiencing the current position as too demanding, and want to reduce it. Working too much may also increase absenteeism.

PTE is in danger of getting confounded with shift work, a work schedule much used in health organizations. Part-time workers tend to work more hours outside standard schedules (Monday through Friday on daytime) compared to full-time workers (Wittmer and Martin 2011). PTE is also associated with both specific occupations (Olberg and Nicolaisen 2013) and educational level (Division for Labour Market Statistics 2015). PTE is much more common among employees with low education (Kauhanen 2008). In hospitals, working part-time is much more common among nurses than doctors (Køber and Vigran 2011). Regarding absenteeism, there are significant differences in absenteeism patterns related to occupational positions and educational level (Clausen et al. 2014a, b; Hammig and Bauer 2013; Kristensen et al. 2010), which underline the importance of controlling for these variables.

Finally, as WFC often is closely linked to obligations at home (Rose et al. 2013), controls for whether the employee has small children at home, and whether he/she is living in a partnership or alone are included. Further, it is controlled for the individual variables gender and age, factors well known to be associated with absenteeism (Allebeck and Mastekaasa 2004). Finally, self-perceived health, both physical and psychic, another factor closely associated with absenteeism, is included (Darr and Johns 2008; Eriksson et al. 2008).

Method and Data

This study uses two sets of data from a Norwegian public hospital. All employees in more than 30% position received an electronic questionnaire in January 2013. The survey closed at the end of February 2013. 3462 valid responses were returned (total response rate = 66). The survey gives self-reported information on demographics, work arrangements, family situation, as well as questions about self-perceived health, AOC and WFC. The second set of data is the employee’s registered sick leave data from 2014. 1864, or 54% of the respondents and 36% of the population, gave consent to connect survey responses with registered data on sick leave. The sample is highly representative of the population on demographic variables and days of absenteeism.

Having two different and independent data sources minimizes the problem of common source bias. The problem of under-recording absenteeism often encountered in self-reported studies is thus avoided (Johns and Miraglia 2015; Murphy 2014). The time lag (1 year) between the survey and the measurement of absenteeism also reduces problems of endogeneity.

Norway is somewhat a special case as the social benefits in Norway related to absenteeism are generous compared to other countries (Gleinsvik et al. 2014). Employees in Norway are entitled to full sickness benefits from the first day of the absenteeism, and may receive sickness benefits for 1 year (NAV 2015). A personal declaration of absenteeism can be used up to 8 consecutive calendar days, and 24 calendar days in a 12-month period. Beyond this it is required a sickness declaration from a doctor. In the current study, only absenteeism falling in the category self-declared sickness absence is included, and all absence for other reasons (doctor ordinated sickness absence, pregnancy, care for sick children, holidays) have been filtered out. The category self-declared sickness absence is most likely to also include other reasons for absence, like taking an extra day of vacation, staying in bed after partying too hard, or just for staying home because one does not want to go to work. Such absence is probably particularly sensitive to variance in AOC and WFC. As Norway has a more liberal system for absence from the workplace than most other countries, including neighboring Scandinavian countries, it constitutes a critical case concerning the mediating effects of AOC and WFC on absenteeism.

Absenteeism was measured as total number of days during a year a person had been away from work due to own illness. A recent meta-study indicates that number of days away from work (“time lost”) measures approximately the same type of absence as number of times away from work (“frequency”). The same study argues that “time lost” (days) should be used as “default measure” when studying absenteeism (Johns and Al Hajj 2016:471). In addition, these data are objective, originating from the hospital’s own register, thus avoiding the possible validity problem of self-reported absenteeism (Jenkins 2014). The mean number of days absent in the sample is 2.0 in 2014 (STD = 2.8, Median = 1, Mode = 0).

As number of days of absence on the individual level must be measured as relative to the total number of days a person works during a year, the relative number of days of absenteeism (both self-declared and doctor-certified) were estimated. This was done by taking the total number of expected working days in a “normal” working year (currently 230 working days), and multiplying it with the percentage position. Days of absenteeism for a person for a year was divided by this product, resulting in the measure relative days of absenteeism of a person’s working year. For an employee, a proportion of, for instance, 0.10 indicates that this person has been absent 10% of his or her total working year. For a person working 50% position, this would mean a total of 11.5 days of absenteeism [(230 × 0.10) × 0.50], while the same would mean 23 days for a person working full-time (100%).

OECD defines part-time employment (PTE) as working less than 30 h per week in the main job (OECD 2016). Empirical studies on PTE and absenteeism mostly defines PTE as a dichotomy, but with rather diverging cut-off points (Flach et al. 2008; Gjesdal and Bratberg 2002, 2003; Kaerlev et al. 2004; Niedhammer et al. 2008). Many of these studies lump together persons working between 30% and 80% of a full-time position in the same group, thus missing the potentially important difference between a part-time worker employed in an 80% percent position and a person employed 30% (Krausz et al. 2000). A person working 80% of a full-time position will probably have more in common with a full-time employee than with the person working in a 30% position. In this study, PTE is measured as a continuous variable, percentage of a full position (PFP), ranging from 30 to 100. In Norway, it is very uncommon for hospital employees working part-time to have second jobs.

Involuntary size of post was computed by deducting the wanted percent position from the current percent position. 81% wanted the percentage currently held, 9% wanted a reduced position and 10% wanted an increase. Involuntary position includes also employees presently working full-time, but wanting a lower percentage position. Two dummy variables were computed: wanting reduced percentage, and wanting higher percentage.

Work-family conflict (WFC) is measured by the scheme from Netemeyer et al. (1996) (5 items), and affective organizational commitment is measured by Meyer and Allen’s scheme (Meyer and Allen 1997). Items measuring the two phenomena used a five-point Likert scale. Four items measured AOC, displaying an alpha of .92, while five items measured WFC (alpha .88). Items were collapsed into two summative indexes with high values indicating high AOC and high WFC.

Shift work

Several different measures were used to tap into different dimensions of the phenomenon: a) the total number of shifts (ranging from one to three), b) whether shift includes night (1 = yes), and c) whether the respondents worked only weekdays, or if also weekends (Saturdays and Sundays) were included (the latter was given the value 1).

Occupation

In Norwegian hospitals, there are more than 80 different occupational positions doing work that differs along several dimensions. In this study, the focus is narrowed to the two dominating professions in a hospital: doctors and nurses.

Position

Those having a formal leadership position were assigned the value of 1, while other employees were assigned the value of 0.

Gender and family situation

Family situation was measured by a question on having children at home (1 = yes, 0 = no), and marital status (1 = living in a partnership). Concerning gender, women were given high value (1).

Health

Two variables on self- reported physical and psychical health (on a scale from 1 = very poor health to 5 = very good health) were included.

Additional controls

The analysis also controls for age (years), and level of education (6 = highest level).

Table 1 displays the descriptive statistics for the variables used in the analysis as well as the zero order correlations between them.

Table 1 Bivariate correlation (Pearson’s r) between variables
Full size table

Only the three aspects of shift work correlated to the degree to fear (multi)collinearity. As these are only control variables, not the focus of this study, all were included in the multivariate analysis (Allison 2012).

Analysis

Multiple mediation analysis based on the Preacher and Hayes (2008) frameworkFootnote 1 was used, applying the PROCESS procedure for SPSS (version 2.24) using OLS regression. PTE (independent variable), WFC and AOC (mediators) were included simultaneously, controlling for the variables gender, age, educational level, position, job category, shift work and wanting to work more/less. PROCESS applies a bootstrapping procedure (5000 bootstrap samples) minimizing problems of non-normal, zero inflated dependent variables, and is thus better suited than for instance structural equation modelling (SEM). Figure 1 shows graphically the model to be tested:

Fig. 1
figure 1

The variables mediating the effect of percentage of a full position (PFP) on absenteeism (ABSENCE)

Full size image

Path a1 and a2 indicate the direct effects of PTE on AOC (MODEL 1) and WFC (MODEL 2), while path b1 and b2 indicate the direct effects of OAC and WFC on absenteeism controlling for PTE. Path c denotes the total effect of PTE on absenteeism (controlling for all variables except AOC and WFC) (MODEL 3), while c’ denotes the direct effect of PTE on absenteeism (controlling for all variables including AOC and WFC) (MODEL 4). The indirect effect of PTE on absenteeism will be mediated through WFC (a1 b1) and AOC (a2 b2), while the total indirect effect of PTE will be ((a1 b1) + (a2 b2)).

Table 2 shows the full results, including the effects of control variables, of the four models described above with self-declared absenteeism as the dependent variable and AOC and WFC as mediating variables. Table 3 displays the total, direct and indirect effect of percentage position (PTE) on self-declared absenteeism.

Table 2 Outcomes of PROCESS, regression analysis using OLS
Full size table
Table 3 Total, direct and indirect effects of Percentage of Full Position (PFP) on absenteeism
Full size table

There is a significant, negative effect of percentage position on self-declared absenteeism, even when controlling for all other factors. Percentage position has no significant effect on AOC, while it has a positive effect on WFC, supporting hypothesis 2a that work-family conflict increases as the percentage position increases. As WFC also has a significant effect on self-declared absenteeism, one should expect the indirect effect of path b1b2 to be significant. The bootstrapping procedure indicates otherwise, as the 95% confidence interval (CI) of this indirect effect includes zero. Still, this indirect effect is significant at the 8%-level.

Discussion

The argument put forward in this study was that PTE would be linked to absenteeism through two organizational phenomena: affective organizational commitment (AOC) and work-family conflict (WFC). On the one hand, it was assumed that less time spent in the organization would decrease the employee’s commitment to the organization, thus working part time would be negatively linked to AOC. Neither the bivariate correlations nor the multivariate analysis supported this assumption. Affective commitment thus seems to be rather independent of the objective amount of time an employee spends in the organization. It is also highly interesting to note that AOC does not have any significant effects on absenteeism, as assumed in hypothesis 1b, even in the (critical) Norwegian case with very liberal rules for staying away from work based on self-declaration. This goes counter to most other studies focusing on the relationship between AOC and absenteeism. One reason for this may be that this study controls for variables, for instance perceived mental health, connected to both AOC and absenteeism, that other studies do not include (Clausen et al. 2010, 2014b). Another reason may be that this study uses objective data on absenteeism, and thus avoids problems associated with self-reported absenteeism and common instrument effects (Johns and Miraglia 2015; Mathieu and Zajac 1990; Meyer et al. 2002; Murphy 2014). The findings in this study indicates that AOC is probably more important for understanding other organizational behavioral phenomena than absenteeism, for instance turnover or performance, or that the connection is defined to specific occupational groups not including hospital workers being the focus of this study (Somers 2010).

PTE is, however, significantly linked to work-family conflict as hypothesized in hypothesis 2a. WFC also has a significant direct effect on absenteeism. The more intense the perception of WFC, the higher the absenteeism. This study corroborates only partially the assumption that PTE will have a negative effect on absenteeism through reduced WFC, as the indirect effect is only marginally significant. Still, this study gives some support to the idea that part time work may be – at least for some and to a certain extent – a way to relieve conflicts between work and home (Beham et al. 2012; Higgins et al. 2000), and that this in turn may affect absenteeism.

As AOC has no effect on absenteeism, one could also question the links between this psychological phenomenon and certain types of factual behavior in healthcare organizations (Riketta 2002). A possible alternative explanation for the missing link between AOC and absenteeism might be the personal working ethos of employees within the healthcare sector (Ellershaw et al. 2012). Hospitals employees are mostly highly skilled professionals who seem to be more committed to the profession and professional skills, and thus to the patient, than to the organization (Fjeldbraaten 2010; Meyer et al. 1993). Commitment may thus be more directed towards the profession and the patients than the organization. Knowing that patients may suffer, or that absenteeism will put a heavier workload on close colleagues, makes employees attending work no matter how low the commitment to the organization is (De Clercq et al. 2015).

Still, the study reports a significant direct effect of PTE on both types of absenteeism. The higher percentage an employee has, the lower the (relative) absence. Even though this finding support other research indicating that PTE leads to higher absenteeism, it is still a puzzle how PTE affects absenteeism. This study has made it less probable that this link has anything to do with affective commitment. In addition, only a small fraction of the effect of PTE is mediated through the WFC “mechanism”. This study also makes it less probable that there is a “sickness mechanism” mediating between PTE and absence (Køber and Vigran 2011). As PTE is uncorrelated to both types of perceived health in this study, such a mechanism does not seem very likely.

Future studies should focus on other “mechanisms” explaining more in detail how PTE works on absenteeism by linking PTE to organizational phenomena other studies have linked to absence from work. One such possibility is that PTE may be linked to engagement to the task and to the meaningfulness of work (Soane et al. 2013), for instance that part time workers will be assigned to less challenging, more routinized and less autonomous tasks. And as shown in meta-analyses, low control over the job is significantly connected to absenteeism (Beemsterboer et al. 2009). A related mechanism could be differences in leader support between part-time and full-time employees as full-time workers may interact more with managers, in turn increasing trust, finally resulting in a better relationship between leader and employee (Clausen et al. 2012). A completely different mechanism could be related to union membership (Deery et al. 2014). Union membership implies having a “voice” within the organization, thus reducing the probability of “exit” (absence). It seems plausible that part time employees may be less union organized than personnel working full time. The possibilities outlined above are not intended to be exhaustive, but rather as a list of ideas on how it could be possible to further investigate the relationship between PTE and absenteeism.

As a general conclusion, this study supports previous studies linking part time work to higher absenteeism (Flach et al. 2012; Gjesdal and Bratberg 2002, 2003; Kaerlev et al. 2004). It also provides a possible explanation why some studies report no significant association between the two (Flach et al. 2008; Niedhammer et al. 2008; Smulders 1993), as part time work may have both a positive direct effect and a negative indirect effect through decreased work-family conflict. In sum, the two effects may equal each other out. Finally, the study goes counter to most other studies in finding no significant relation between affective commitment and absenteeism, a finding probably due to the organization context of the study. In highly professionalized organizations like a hospital, as well as most other public service providing organizations, it would probably be more fruitful to focus on commitment to the profession or the client rather than the organization, and how this affects absenteeism.