Article body

Research studies examining the causes and management of absenteeism have been numerous, with more than 500 academic papers being published in the past 20 years (e.g., Brooke and Price 1989; Gellatly and Luchak 1998; Harrison, Johns, and Martoccio 2000; Harrison and Martocchio 1998; Johns 1997; Johns and Nicholson 1982; Price and Mueller 1986; Rhodes and Steers 1990; Xie and Johns 2000). In spite of the volume of research, Martocchio and Harrison (1993) argue that much of this work is plagued with conceptual and methodological deficiencies. This has exacerbated the problem of understanding and reaching agreement regarding the underlying processes that generate absenteeism. In particular, absenteeism has largely been considered to be an individual phenomenon that occurs exclusive of the employee’s social context (Johns 2001). More recent work has attempted to rectify this deficiency and has found that collective behaviour can have a significant impact on absence norms (Markham and McKee 1995; Mathieu and Kohler 1990; Martocchio 1994b). In other words, an absence culture may exist that potentially affects individual levels of absence. The salience of this culture has been argued to be dependent upon a number of personal and organizational factors such as trust and homogeneity within the work unit or organization. These factors, however, have been poorly understood (Gellatly and Luchak 1998).

A new approach to understanding the relationship between homogeneity and absence culture is relational demography. Tsui and O’Reilly define relational demography as the “comparative demographic characteristics of members of dyads or groups who are in a position to engage in regular interactions” (1989: 403). These researchers proposed that the comparative similarity or dissimilarity in attributes such as sex, age, tenure and so on of members who interact affect attitudes and behaviour. The basic premise of relational demography is that the greater the heterogeneity among group members, the greater the negative impact on group functioning (Tsui and O’Reilly 1989). We extend current relational demography research by examining how the integration of non-members influences the absence culture of union members within work groups.

In this paper we first examine Nicholson and Johns’ (1985) typology of absence culture salience (i.e., distinctiveness of beliefs regarding absence) and the degree of trust in the psychological contract (i.e., degree of vertical integration among organizational levels). Absence culture is then contextualized within our research settings of two hospitals. The paper proceeds with the issue of salience as determined by homogeneity within the work group (i.e., union membership status) and trust as operationalized by the perceived climate between the union and management (UM).[2] Finally, we report the results and discuss the theoretical policy implications of our analysis.

Nicholson and Johns’ Typology of Absence Cultures

Absence culture refers to the “beliefs and practices influencing the totality of absence frequency and duration- as they currently occur within an employee group or organization” (Chadwick-Jones, Nicholson, and Brown 1982: 7). It comprises the perceived legitimacy of absence taking (Nicholson and Johns 1985) and a normative philosophy held by employees based on the justification of absence (Deery et al. 1995). Hence, absence culture refers to the “extent to which there is homogeneity or mutual agreement among the group members about absence pattern” (Xie and Johns 2000: 32).

Nicholson and Johns (1985), building on earlier work by Hill and Trist (1953, 1955), argue that the nature of absence culture depends on the salience of the culture to members of the group (work unit or organization) and on the level of trust experienced by members (see Figure 1). In relation to the salience of culture, this refers to the homogeneity of norms regarding attendance behaviour. Cultural salience is determined by the organization’s absence control system, technology and the social ecology of the work setting. Thus cultural salience is expected to be high where there is a clear understanding of the formal and informal rules of absence, where work is interdependent and where there is opportunity for informal communication through networks and friendships. Current research (Aryee and Chay 2001; Tan and Aryee 2002) indicates that members engage in union citizenship behaviour when they positively evaluate the union, are loyal, and consider the union as supportive. From the quantitative and qualitative evidence, these pro-union norms, values, and behaviours are indicative of the cultural salience in our present setting. Therefore, we expect that cultural salience will be weakened by the integration of non-members with unionized employees (Kelly and Kelly 1994).

Trust, on the other hand, is embedded in the psychological contract and reflects the degree of vertical integration between employee and employer. Nicholson and Johns considered that absence beliefs differed by occupational status (i.e., high and low) and the level of trust associated with this status level. High trust is reflected by the congruence of employer-employee interests, while low trust is represented by the formal employment relationship. In the present setting, which is highly unionized and equal in status (non-supervisory staff), trust is represented by the perceived harmony in the UM climate (Lorenz 1992). UM climate measures the degree of fairness, cooperation and participation in the resolution of problems between management and the union (Dastmalchian, Blyton, and Adamson 1991; Gordon and Ladd 1990). Researchers have long argued that an adversarial UM climate is characterized by decreased trust, while a consensual UM climate is characterized by increased trust (Cooke 1992; Hammer, Currall, and Stern 1991; Huszczo and Hoyer 1994).

In their taxonomy, as shown in Figure 1, Nicholson and Johns propose four possible cultures with different expected outcomes on absenteeism. Although it is not our intention to test the four types, it is important to understand their differences: Type I (low salience, high trust) results in a dependent culture which is characterized by deviant absence (e.g., employees adhere to organizational rules rather than to the behavioural expectations of coworkers); Type II (high salience, high trust) engenders a moral culture which is reflected by a constructive absence (e.g., employees’ rights and obligations to the organization are met); Type III (low salience, low trust) promotes a fragmented culture which is embodied in calculative absence (e.g., employees weigh up the consequences in terms of satisfaction and pay); and Type IV (high salience, low trust) encourages a conflictual culture which denotes a defiant absence (e.g., employees’ culture is alienated from the organization’s). As will be discussed in the following section, the absence culture in our hospitals would be classified as a Type II culture.2

Figure 1

A Typology of Absence Cultures

A Typology of Absence Cultures

-> See the list of figures

Absence Culture in the Hospitals

In recent times, hospitals in Australia have been under increased pressure to improve the quality of patient care and to reduce costs. This primarily stems from changes in the formula by which Federal and State governments fund public hospitals. Casemix funding was introduced in July 1993 to promote a more flexible hospital system. Integral to this system is a measure known as diagnosis related groups, which enables the hospitals to record, value and charge the units of output to the state government. Absenteeism (i.e., sick leave) was considered to be a major cost to the hospitals, averaging around 6% or 13 days per year.[3] Based on this level, it was accepted by all parties (management and the union) that an absence culture existed at the hospitals. One hospital, for example, undertook an operational efficiency review (OER) of all services. This comprised the establishment of four task forces (i.e., food services, cleaning, operating theatres and anaesthesia, and radiology) to review, trial and monitor recommendations. The nursing union was actively involved in these task forces and was also instrumental in disseminating information about changes within the hospital. The union participated in a collective agreement with the hospitals to increase productivity and pay levels by reducing costs such as absenteeism. Hence, due to the gainsharing plans, there was a major incentive for both parties to improve the quality of the union-management relationship.[4]


Union Membership

Deriving from Nicholson and Johns (1985), we expect that the homogeneity of attendance norms for union members is likely to be diluted by the integration of non-union members. We examine this proposition using relational demography. As previously noted, relational demography refers to the differences between a target employee and her or his co-worker’s attributes. Social identity theory (SIT) (Tajfel 1972; Tajfel and Turner 1986) and self categorization theory (SCT) (Turner 1982, 1985) suggest that individuals use social characteristics to categorize themselves (and others) into psychological groups. A high level of attraction for individuals to members of a group exists where there is a “fit” on various attributes (e.g., Byrne 1971; Byrne, Clore, and Smeaton 1986). Conversely, individuals in work groups containing greater differences are expected to experience heterogeneity in work group norms (e.g., Jackson et al. 1991; Rosenbaum 1986).

Union membership is a discernible characteristic with which to differentiate the existence of at least two social groups (members and non-members). SIT asserts that “in-groups” (e.g., union members) and “out-groups” (non-union members) are formed by the degree to which individuals identify with the relevant group (Kelly and Kelly 1994). A positive social identity is promoted by the constant comparison between the groups. However, when comparisons indicate deficiencies on the part of the in-group (compared to the relevant out-groups), SIT suggests that individuals can either withdraw from their existing social group (psychologically or behaviourally) and become a member of the other distinct group or attempt to make their existing group more positively distinct. Allen and Stephenson (1983), studying the perceived differences between the personal values and out-group values of managers (i.e., non-union members) and union members, observed that both groups held traditional stereotypes about each other. Importantly, strong group identification was related to perceiving the out-group more stereotypically.

Kelly and Kelly (1994) propose that the collectivist orientation of union members is likely to be undermined by interaction with non-union members (e.g., free rider effect). This argument is reinforced when the status differential between the two groups is considered. Given that the average level of union membership coverage across our work groups was 76 percent, union members would be apportioned majority status and non-members, minority status. There is empirical support in the literature for the dysfunctional impact of the integration of these two status groups on the attitudes of the majority group. Race has been commonly used in the literature to illustrate these negative effects. Chattopadhyay (1999), for example, studying employees across four organizations (i.e., manufacturing, university, and transportation) observed that race dissimilarity negatively influenced peer relations of majority white employees. In an earlier study of unionized public school teachers in the U.S., Iverson and Kuruvilla (1995), reported race homogeneity (i.e., the racial difference between the teacher and her or his colleagues and students) to be related to organizational commitment, and union satisfaction. These researchers concluded that the “… integration of minority members into majority-dominated settings would be expected to lead to the lowering of majority members attitudes to the organization and union” (Iverson and Kuruvilla 1995: 577).

Based on the theoretical, empirical, and qualitative evidence, we therefore posit that the integration of non-union members with union members in work groups diminishes the bond between the union and its members, as well as the union’s role in acting as a disciplinary agent in enforcing the absence policy (Braithwaite 1989). As a consequence, the attendance norm of union members is expected to be weakened. We therefore assert that:

Hypothesis 1: The greater the difference in union membership, the higher the absence culture.

UM Climate

Social information processing theory (Salancik and Pfeffer 1978) asserts that contextual factors such as UM climate influence employees attitudes to work and the organization. Walton, Cutcher-Gershenfeld, and McKersie (1994) characterize a positive UM climate in terms of cooperation in achieving joint union-management goals, and a negative UM climate in terms of contentiousness and incompatible demands made by unions and management. In the context of a positive UM climate, this salience will result in a positive attendance culture. Nicholson and Johns note that “some cultures may contain norms that effectively dictate good attendance” (1985: 398). This is consistent with Deery, Erwin, and Iverson (1999) who, in a study of union members from a large automotive manufacturer, observed lower absence rates when UM climate was perceived to be positive. Newton and Shore (1992) posit that greater harmony in the workplace is associated with the union and organization being credited with improving lower level needs such as wage and working conditions. Another recent union study by Wagar (1997) found UM climate to be significantly and positively related to other organizational outcomes such as perceived productivity, quality of product or service, and customer or client satisfaction. Finally, Arthur and Jelf in their study of the effects of gainsharing concluded that there was “strong evidence for the ability of gainsharing to transform existing labor-management relations” (1999: 133). As the union in our setting had entered into a collective agreement to increase pay by reducing absenteeism, it is logical to propose that:

Hypothesis 2: The more positive the UM climate, the lower the absence culture.

Moderator Effect

In addition to the main effects for union membership and UM climate, based on Nicholson and Johns’ typology, it would be expected that a Type II culture (high salience, high trust) would create a moral culture. Specifically, in terms of our variables, we argue that homogeneity of membership will lead to a moral obligation for union members to internalize the attendance norms of their union and the organization, leading to an acceptable standard of absence behaviour. This is anticipated to occur when the UM climate is positive. Given that job and organizational factors are associated with absence culture (Harrison, Johns, and Martocchio 2000; Johns 1997), the motivation to attend work should be enhanced when these factors are improved. Although empirical evidence is rather scant, we expect similarity in union membership (i.e., high salience) and a positive UM climate (i.e., high trust) to interact in affecting members’ attendance attitudes.

Hypothesis 3: Similarity in union membership will be associated with a lower absence culture when the UM climate is positive (i.e., type II culture).

Control Variables

As can be seen in Figure 2, in order to more rigorously test our three main hypotheses, we also control for other relational, demographic and explanatory (i.e., job, organization, and union related) variables reported in the absence literature (e.g., Harrison and Martocchio 1998). Since absence culture is strongly related to actual absence behaviour (Gellatly and Luchak 1998; Harrison, Johns, and Martocchio 2000), we also draw on this latter body of research in specifying our control variables. In addition, Tsui, Egan, and O’Reilly III (1992) note that these types of control variables are essential in testing for the true net effects of the relational demography variables.

Relational demography variables. We sought to control for other relational demography variables in our analysis. The rationale for the inclusion of the relational variables of sex, age, education, and tenure derives from social identity (SIT) and self-categorization theory (SCT) of employees within work groups (Tsui, Egan, and O’Reilly iii 1992). As discussed in the Methodology section, a difference score for sex, age, education, and tenure is estimated using the O’Reilly iii, Caldwell, and Barnett (1989) formula. This score represents the difference between an individual and all other individuals in the work group in terms of these variables, with higher scores indicating greater dissimilarity. Previous research has linked these relational variables to organizational outcomes such as job satisfaction, performance and organizational citizenship behaviour (e.g., Manogram and Conlon 1993; Martocchio and Judge 1995; O’Reilly iii, Caldwell, and Barnett 1989; Tsui, Egan, and O’Reilly III 1992; Tsui, Porter, and Egan 2002; Tsui and O’Reilly iii 1989; Wayne and Green 1993). Although there is limited evidence in relation to absenteeism, apart from Tsui, Egan, and O’Reilly III (1992), we would hypothesize that the greater the difference between co-workers on sex, age, education, and tenure, the greater the absence culture. As absence culture is regarded as a negative outcome, it would be anticipated that the norms of attending work would be substantially weakened by the increased heterogeneity within work groups.

Figure 2

A Model of Absence Culture

A Model of Absence Culture

-> See the list of figures

Demographic variables. In addition to the relational demography control variables, we also controlled for the demographic variables of sex, age, education, and tenure. Due to there being little multivariate analysis of these variables with absence culture, we rely on both bivariate and multivariate results for our justification. Previous studies have linked sex, age, education, and tenure with absence culture and behaviour. For example, absence behaviour has been associated with younger (Farrell and Stamm 1988; Tsui, Egan, and O’Reilly III 1992), lower tenured (Chadwick-Jones, Nicholson, and Brown 1982), lower educated (Taylor 1979; Tsui, Egan, and O’Reilly III 1992) and female employees (Johns 1978; Tsui, Egan, and O’Reilly iii 1992).

Explanatory variables. The third set of control variables comprised job, organization, and union related determinants that are well established in the absence literature. Increased work group size has been observed to promote absence culture and behaviour (Martocchio 1994a). The variable of hospital (a and b) controlled for variations between the sites. Coworker support (defined as the degree of consideration expressed by co-workers) has been reported to have a negative relationship with absence culture (Deery et al. 1995). Absence permissiveness, the degree to which absenteeism is tolerated by the organization (Brooke and Price 1989), has been found to be positively associated with absence culture at the group level (Martocchio 1994a). The variable of external responsibility, the extent to which employees have responsibilities outside work (Deery et al. 1995) has been linked to absence. This is explained by increased responsibilities of caring for children and dependents (Deery et al. 1995). There is strong support in the literature for the differential impact of the job-related variables of routinization (defined as the degree to which employees jobs are repetitive) (Price and Mueller 1981) and promotional opportunity (defined as the degree of movement between different status levels in an organization) on absence. For example, employees who perceive their jobs to be boring and repetitive, as well as lacking career advancement opportunities, are more likely to display absence behaviour (Deery et al. 1995; Iverson, Deery, and Erwin 1995). Union instrumentality, defined as the degree to which the union achieves valued goals of employees, has been found to be negatively related to absence (Deery, Erwin, and Iverson 1999). External opportunity (defined as the availability of alternative jobs outside the organization) has been reported to have a positive impact on absence by employees re-evaluating their jobs in relation to current economic conditions (Mowday, Porter, and Steers 1982). That is, scarcity of alternative opportunities may increase the pressure to attend work, while greater opportunities may ease this pressure. The final variable of job satisfaction (defined as the overall degree to which an individual likes her/his job) has been found to be associated with decreased absence (Iverson and Deery 2001).

Hence, we have attempted to provide the theoretical and empirical rationale for the inclusion of union membership, UM climate and other control variables (i.e., relational, demographic, job, organization, and union related) used in testing these three hypotheses. In the following section we outline the methodology employed in the study.



The sample consisted of 43 nursing work groups (mean = 14.54, S.D. = 4.56) comprising 460 (14% male, 86% female) non-supervisory union members. Data were aggregated from hospital A with 227 employees and hospital B with 233 employees. The average age, tenure and education of employees were 34.73 years (S.D. = 9.98), 5.80 (S.D. = 4.63) and 13.80 years (S.D. = 1.65) respectively.

Data Collection

Two identical multiple-item surveys were conducted at both hospitals within a period of three months. In terms of hospital A, the survey was administered to a random sample of 740 employees, of which 473 questionnaires were returned. This represented a response rate of 64%. The data was confined to non-supervisory nursing staff who were union members and who could also be identified within specific work groups (hospital wards). Subsequent to the matching of respondents to wards and to the listwise deletion of missing data procedure, where all blue-collar employees were deleted from the analysis, 227 useable questionnaires were retained. In terms of hospital B, 1100 employees were surveyed. Eight hundred and twelve employees responded, representing a response rate of 74%. The process of matching surveys from hospital A was replicated in Hospital B. After deletion of all blue-collar employees and non-union members, 233 surveys were retained. The rationale for deleting all blue-collar employees from the analysis stems from their not being employed in work groups in both hospitals. From the two data sets of hospital A and B, a composite data set was constructed.


A five-point Likert scale was used to measure the perceived response of each employee to items in the questionnaire. The scale ranged from (5) strongly agree to (1) strongly disagree. The variables, where possible, were constructed from established scales and are specified in the sub-sections below. Cronbach’s alpha (1951) was calculated for all multiple measures. The descriptive statistics and correlations among measures are contained in Table 1.

Dependent variable. Absence culture was operationalized using Deery et al.’s (1995) scale and comprised the following three items: “my co-workers discourage others from using up their sick leave (R); my co-workers generally agree that you should use up your sick leave entitlement; my co-workers do not care if others are absent from work.” The questions were framed in terms of “co-workers” so as to capture the absence culture within the hospital wards. Consistent with our definition that highlighted the homogeneity or mutual agreement among the group members about absence patterns, we found the within-group interrater reliability (rwg) to be an acceptable .76 (Kozlowski and Hattrup 1992). Confirmatory factor analysis results (see Table 2) indicated that the factor loadings ranged from .57, p < .001 to .79, p < .001. These exceeded the minimum level required to form a construct as estimated using Norman and Streiner’s (1994) formula.[5] In addition, the loadings accounted for substantial total common variance (44%) of absence culture (Kelloway 1998). These results provided support for the convergent validity and, as the scale was newly developed, it was considered to display acceptable internal consistency (α = .64) (Nunnally 1978).[6] We correlated absence culture with a measure of prior absence based on individuals reporting how many days they were absent during the previous 3 months for hospital A and 6 months for hospital B. The correlation was significant and positive (r  = .19, p < .05).

Table 1

Means, Standard Deviations, Correlations, and Reliabilities

Means, Standard Deviations, Correlations, and Reliabilities

aN = 460.

b Reliabilities are reported along the diagonal.

c Correlations above [.08] are significant at p < .05, one-tailed test.

-> See the list of tables

Table 2

Confirmatory Factor Analysis Results of Absence Culture

Confirmatory Factor Analysis Results of Absence Culture

-> See the list of tables

Independent variables. Tsui, Egan, and O’Reilly iii (1992) define the relational demography score as the difference between an individual and all other individuals in the work group. For the difference in union membership between non-supervisory coworkers, O’Reilly III, Caldwell and Barnett (1989) formula was employed (see below).[7] It is the squared root of the summed squared difference between individual Si’s value on union membership and the value of every other individual in the work group (1 if different, 0 if no difference), divided by the total number of respondents in the group (n). In our study, we estimate the average person to person differences for union members. That is, a nurse who is a union member is initially compared with the union membership status (i.e., member and non-member) of each coworker in the group. The differences are then averaged across the total number of nurses within the group to obtain a relational score. For example, a union member in a work group of 5 union members and 10 non-union members would have a relational score of 0.82. This is calculated as 0 for being the same as the other union members and 10 for being different from each of the 10 non-union members. The score of 10 is divided by 15, and then the square root of the result is taken. The range for the relational score is between 0 and .99, with scores closer to zero indicating the individual is more similar to other members of the work group. In testing Nicholson and Johns’ (1985) type II absence culture, we focus only on the responses of union members in our analysis. This stems both from our theoretical framework and from the measurement of UM climate (see items below), which is only relevant for union members.

forme: forme pleine grandeur

UM climate was assessed using the 10-item (i.e., unions and management work together to make this a better place in which to work; unions and management have respect for each other’s goals; the parties in this company (unions and management) keep their word; in this company, joint management-union committees achieve definite results; there is a great deal of concern for the other party’s point of view in the union-management relationship; in this company negotiations take place in an atmosphere of good faith; employees have a positive view on joint union-management committees here; the enterprise agreement is regarded as fair by employees in this company; employees generally view the conditions of their employment here as fair; and a sense of fairness is associated with union-management dealings in this place) harmony component of Dastmalchian, Blyton and Adamson’s (1989) scale and has demonstrated acceptable levels of reliability and validity (Dastmalchian, Blyton and Adamson 1991; Deery, Erwin, and Iverson 1999).

Control Variables

Relational demography variables. Age, education, and tenure were measured in years, while sex was operationalized as 1 for female and 0 for male. The relational score for sex assigned a value of 1 if there was a difference and 0 if there was no difference. In terms of age, education, and tenure, a difference score was also calculated using the formula above. The larger the difference score, the greater the dissimilarity, while the smaller the difference score, the greater the similarity on these variables.

Demographic variables. As previously outlined, we also included the demographic variables of sex, age, education, and tenure.

Explanatory variables. Work group size and hospital (1 = Hospital A, 0 = Hospital B) controlled for the possible differentials between the data sets and between work group size. All work groups with less than four members were excluded from the analysis. Coworker support was measured by a modification of the scale by House (1981). Although coworker is a recently developed scale, it has displayed acceptable reliability and validity in recent research (Iverson and Erwin 1997). Absence permissiveness was measured using the scale by Brooke and Price (1989), while external responsibility was operationalized using the scale by Erwin and Iverson (1994).[8] Current research affirms the psychometric properties of these scales (Brooke and Price 1989; Erwin and Iverson 1994; Iverson, Deery, and Erwin 1995). The two measures of routinization and promotional opportunity were operationalized using Price and Mueller’s (1981, 1986) scale, and have demonstrated acceptable reliability and validity (Iverson and Erwin 1997; Iverson and Kuruvilla 1995).[9]Union instrumentality was measured by a scale developed by Deery, Erwin, and Iverson 1999) and focused on issues such as improving wages and working conditions, decision making, skills, safety and security. Although being recently developed, the scale has displayed acceptable psychometric properties in this research. External opportunity was operationalized using Price and Mueller’s (1981, 1986) scale and has displayed acceptable levels of reliability and validity (Dastmalchian, Blyton and Adamson 1991; Mueller et al. 1994). The final variable of job satisfaction was measured by six items from the Brayfield-Rothe (1951) scale. This scale has also been observed to be psychometrically sound (Price and Mueller 1981, 1986).


Following the computation of the relational scores of union membership and our control variables of sex, age, education and tenure using SPSS, we employed the program of PRELIS (Jöreskog and Sörbom 1996b) to transform the raw data and create a correlation matrix as the input to LISREL (Jöreskog and Sörbom 1996a). Our main aim was to test the absence culture model as displayed in Figure 1. Nevertheless, Anderson and Gerbing (1988) recommend that the measurement model be assessed independently and before that of the structural model. In terms of the measurement model, we supported the convergent validity (i.e., the degree of association between measures of a construct) and the discriminant validity (i.e., the degree to which measures of constructs are distinct) for constructs with multiple-items (i.e., absence culture, UM climate, coworker support, absence permissiveness, external responsibility, routinization, promotional opportunity, union instrumentality, external opportunity, and job satisfaction) using the procedures as recommended by Bagozzi and Yi (1988).[10] The measurement model was found to have a normed comparative fit index (CFI) (which avoids the underestimation of fit) (Bentler 1990) of .96 and root-mean-square error of approximation (RMSEA) of .05 (Browne and Cudeck 1993). In addition, these results demonstrate that the probability of common method variance occurring is minimized (i.e., inflating relationship between constructs) (Podsakoff and Organ 1986). This is affirmed by the better fit of the competing models as they increased in complexity (Iverson 1996; Korsgaard and Robertson 1995).

In terms of the structural model, a current issue in the literature is the effect of statistical power (Saris and Satorra 1993). Statistical power refers to the probability of rejecting a false null hypothesis (Type II (β) error). That is, meaningful effects may be negated when statistical power is low. Due to the potential power problems arising from the number of parameters in the model and from the estimation of interaction terms using our sample size, a manifest variables model was employed (Williams and Hazer 1986). Essentially, we use the variance-covariance matrix for input, where the “…latent-to-manifest parameter for each variables was fixed to the square root of the reliability (internal consistency coefficients) for each measure, and the value of one minus the reliability multiplied by a variable’s variance was used to represent residuals” (Carlson and Perrewè 1999: 526). This approach is commonly used in structural equation modelling (e.g., Frone, Russell, and Cooper 1992; Renn and Vandenberg 1995; Williams and Hazer 1986). Employing the SAS program by MacCallum, Browne, and Sugawara (1996), we calculated the statistical power of our model. This comprised inputting the null and alternative values of the RMSEA (ε0 and εa) (see Browne and Cudeck 1993 for discussion of ε0 and εa), the α level, degrees of freedom and sample size. This exceeded Cohen’s (1988) criterion of .80 (i.e., at least 80% probability), indicating the model had sufficient power to detect meaningful parameter estimates. The results of the structural model are contained in the following section.


We begin with a brief discussion of the bivariate results, followed by the multivariate analysis.

Correlational Data

The correlation matrix (see Table 1) indicates that although the general trend for the correlations was relatively low (and significant), the findings are consistent with other studies using a relational demography approach (e.g., Chattopadhyay 1999). Even though we find some high associations between the demographic variables of sex (r = –.74), education (r = –.62), and tenure (r = .67) with their relational counterparts (e.g., Tsui, Egan, and O’Reilly iii 1992), tests for multi-collinearity (see procedures as recommended by Berry and Feldman 1985) indicated this was not a problem.

As can be seen from Table 1 there are several significant bivariate relationships with absence culture. Differences in union membership (r = .10) and UM climate (r = –.18) were significantly related to absence culture. The greater the similarity in union membership in work groups and harmony between management and unions, the lower the absence culture. In addition, six control variables were found to have a significant relationship. In order of importance, the variables comprised external responsibility (r = .17); promotional opportunity (r = –.17); absence permissiveness (r = .16); job satisfaction (r = –.14); routinization (r = .13); and work group size (r = .11). Consequently, an increased absence culture is associated with substantial responsibilities outside work. If there is little opportunity for advancement up through the different status levels in the organization, if a high degree of tolerance of absenteeism characterizes the organization, if employees dislike their jobs, if the work itself is repetitive and the work group size is large, then employee belief in the legitimacy of absence taking is more probable.

Multivariate Data

Table 3 reports the LISREL findings. As shown, there is support for both hypothesis 1 and 2. In relation to hypothesis 1, the greater the differences in union membership (controlling for other relational, demographic, and explanatory variables), the higher the absence culture (β = .08). This result affirms the argument that the heterogeneity of union membership between co-workers in work groups significantly weakens the norm to attend work. In terms of hypothesis 2, the more positive the UM climate, the lower the absence culture (β = –.12). Having harmonious relations between management and the union is associated with an increase in attendance motivation. The independent variables added significant explained variance in absence culture (i.e., significant R2 increment of 2.0%, ΔF(2, 439) = 5.55, p < .05) above the control variables.

Six control variables also had net effects on absence culture. In order of significance, they comprised promotional opportunity (β = –.22) (Iverson, Deery, and Erwin 1995); absence permissiveness (β = .19) (Martocchio 1994a); routinization (β = .14) (Iverson, Deery, and Erwin 1995); hospital (b = –.13); external responsibility (β = .13) (Deery et al. 1995); and work group size (β = .12). Thus, an absence culture is associated with a lack of a career path, a high level of tolerance of absence by the organization, repetitive work, and the presence of greater personal obligations outside of work. In addition, hospital B and work group size were related to absence culture.

In relation to the interaction effects, we found support for hypothesis 3. We employed a nested goodness-of-fit strategy as recommended by Jaccard and Wan (1996). This procedure involves a multiple-group solution, whereby the fit of one model (coefficients unconstrained to be equal) is subtracted from the fit of a second model (i.e., coefficients constrained to be equal). If the resulting chi-square change is significant, then this indicates the presence of an interaction effect. Heterogeneity in membership status was found to have a significant positive relationship with absence culture when the UM climate is positive (β = .46, SE = .20, p < .05), and a non-significant effect when UM climate is negative (β = .04, SE = .07, p > .05). In other words, similarity in union membership status is associated with a lowering in the absence culture of members when the UM climate is positive.

Table 3

LISREL Results (Standardized Coefficients) for Absence Culture

LISREL Results (Standardized Coefficients) for Absence Culture

a LISREL coefficients for the full model are shown. Unique R2 refer to the unique variance accounted for by a set of variables.

* p < .05

-> See the list of tables

We undertook further analyses to examine the possible interaction between difference in membership status and perceived union instrumentality.[11] Although more exploratory, we expected that greater similarity in union membership would decrease absence culture when there was high union instrumentality (Deery, Erwin, and Iverson 1999). Homogeneity in union membership status was found to have a significant negative relationship with absence culture when union instrumentality is high (β = –.34, SE = .06, p < .05) and a significant positive relationship when union instrumentality is low (β = .11, SE = .06, p < .05).

The implications of these findings for understanding absence culture will be addressed in the Discussion section of the paper.


Drawing on Nicholson and Johns’ (1985) absence taxonomy, this paper attempted to address some of the major deficiencies in absence research by focusing on the social context in which the absence culture of union members is shaped. We found support for our three main hypotheses. That is, greater similarity in union membership status between co-workers was associated with a lowering of a member’s absence culture (supporting hypothesis 1), as did more harmonious UM relations (supporting hypothesis 2). These two variables were also observed to interact: greater similarity in union membership was linked to a lowered absence culture when the UM climate was perceived to be positive (supporting hypothesis 3). These results are particularly pertinent given that we controlled for other relational, demographic and explanatory variables in our analysis. The theoretical and practical implications of the findings are presented.

In relation to hypothesis 1, the results have important implications for work group theory. First, union membership appears to be an important category with which co-workers identify (Kelly and Kelly 1994). Second, our results show that work groups (hospital wards) where there are greater differences (i.e., fewer union members) have higher absence cultures or possess a greater perceived legitimacy in taking sick leave. This is consistent with our argument that unions can play a role in reducing the incidence of absence taking. Empirically there is support for this finding. Xie and Johns (2000), studying a sample of Chinese employees from state-owned manufacturing enterprises, reported that absence was increased when social control (i.e., group cohesion) was weak. In our study, the ability of unions to act as a disciplining agent via mechanisms such as shaming and reintegration (Braithwaite 1989) is dependent, however, upon the relationship between individuals and the union or the perceived instrumentality of the union. Although exploratory, we did observe union instrumentality to act as a moderator. If the union is credited with improving working conditions, then the motivation to attend is enhanced (Deery, Erwin, and Iverson 1999). Hence, even though we found main effects for union membership, researchers should not underestimate the moderating influence of the perceived instrumentality of the union in increasing or decreasing attitudes of attendance.

Turning to hypothesis 2, we affirmed the relationship between UM climate and absence culture. When the UM climate is positive, absence culture is lowered. Our results support those of Deery, Erwin, and Iverson (1999), who reported that members are motivated to assist the organization become more efficient and productive when a cooperative relationship exists between management and the union. Trust is increased via mechanisms such as improved communication and information sharing (Arthur and Jelf 1999). Moving from traditional forms of bargaining to more participative forms enhances the quality of the UM relationship, as well as in our case, the attendance motivation of union members. This “mutual gains” approach (Kochan and Osterman 1994) has implications for the current surge of interest in high-performance work systems (HPWS). There is now a growing body of evidence suggesting HPWS improve organizational effectiveness by reducing costs (Guthrie 2001) and by increasing productivity and profitability (Huselid 1995). An avenue for future research would be to examine the effects of UM climate for these various HPWS.

In terms of interaction effects, greater similarity in union membership was related to a lowered absence culture when the UM climate was perceived to be positive. We posited that this would be best characterized in terms of Nicholson and Johns’ (1985) Type II culture (high salience, high trust). In this situation, we anticipated that a moral norm to attend work would be elicited. The internalization of absence standards occurred when there was a high degree of trust between management and the nursing union. These findings are consistent with the socio-contextual explanation of Gellatly and Luchak (1998) who concluded that the perceived absence norm is influenced by a combination of personal, group, and organizational factors. Significantly, Gellatly and Luchak’s measure of absence norm predicted individual absence one year later. In the present study, we also observed promotional opportunity, absence permissiveness, routinization, and external responsibility to be linked to absence culture. In contrast to much of the organizational demography literature, neither the other relational or demographic control variables had a significant relationship with absence culture.

Our study is not without its limitations. This research was undertaken in a climate where absenteeism was targeted as part of a cost reduction program. This heightened awareness of the need to reduce absence levels by management and the union may account for the results. Further, generalizability of the results must be viewed with some caution given the sample comprised unionized nurses who were predominately female. Although the psychometric properties (i.e., reliability) of our dependent variable of absence culture and several control variables including absence permissiveness, external responsibility, and promotional opportunity were affirmed, the inclusion of additional items would improve these measures. For example, in terms of absence culture we employed a three-item attitudinal measure. This could be bolstered by either asking respondents to estimate the average number of days their group members were absent in the previous 12 months or by collecting these data from organizational records, as well as using qualitative methods such as structured interviews or focus groups to tease out attendance- related issues (Gellatly and Luchak 1998). This expanded measure would better capture the complexity of the absence-related beliefs and behavioural norms that comprise absence culture. Finally, we were only able to utilize a cross-sectional research design. Ideally, pre and post quasi-experimental designs (e.g., Gellatly and Luchak 1998; Mathieu and Kohler 1990) would provide for a better understanding of the temporal ordering of determinants.


Despite these limitations, our study contributes to the understanding of absence culture within work groups at the two hospitals. We found empirical support for Nicholson and Johns’ (1985) absence typology, specifically type II. This was contingent on the degree of dissimilarity in union membership within work groups, as well as the type of UM climate (i.e., trust) present. Future research must continue to disentangle the role of unions and their perceived instrumentality in fostering different types of absence culture. A natural corollary of this study would be to examine how these factors also influence actual absence behaviour (Gellatly and Luchak 1998; Xie and Johns 2000). Finally, our research extends that of Fuller and Hester (1998) by demonstrating that the social context of union members is an important determinant of their attitudes to attendance.