Abstract
Although considerable research has examined various antecedents of employee absence, relatively little research has examined the effects of organizational absence control initiatives on different types of absence, or how differences in employee perceptions of these initiatives may influence employee absences. Through the lens of organizational justice, the authors address this gap in the literature by tracking absenteeism in two manufacturing plants that implemented, at different times, absence control initiatives designed to increase the salience of absence outcomes via explicit rewards and punishments. Results showed that the policy changes had the intended effect of reducing casual absence, but also the unintended effect of increasing Family and Medical Leave Act (FMLA) absence. The results for disability-based absence were mixed. Individual employee perceptions of both the salience of absence outcomes and the fairness of the absence policies showed differential effects on casual and FMLA absence. These perceptions interacted such that employees who perceived the policies to be more salient and unfair had the highest instances of FMLA absence. Implications for absence and justice research and practice are discussed.
Absenteeism is a costly problem for many organizations. According to the Bureau of Labor Statistics (2010), the absence rate for 2009 in the U.S. private sector was 3.0% (for the public sector, it was 4.0%). For a private company of 25,000 employees, this adds up to 200,000 lost workdays per year; with an average salary of $40,000, this average absence rate would translate into estimated annual direct costs of $43,200,000 (Nucleus Solutions, 2008). As large as this figure is, it does not include indirect costs such as the lost revenue opportunity, decreased customer satisfaction, and lower employee morale that may accompany employee absence. Because of these potential costs, many organizations engage in initiatives such as implementing flextime (Baltes, Briggs, Huff, Wright, & Neuman, 1999), employee recognition programs (Markham, Scott, & McKee, 2002), and group therapy sessions (van Dierendonck, Schaufeli, & Buunk, 1998), all of which are designed to reduce unscheduled employee absences.
Despite these efforts, relatively little empirical research has examined the outcomes associated with organizational absence control policies (the organizational policies that specify the penalties and rewards associated with absence and attendance). A few early studies found that absence control policies generally reduced employee absence (Baum, 1978; Baum & Youngblood, 1975; Dalton & Mesch, 1991; Morgan & Herman, 1976), but they may have also led to oversimplified conclusions regarding the positive impact of absence control initiatives. Specifically, at the time of these early studies, it was suggested that absences resulting from medical issues were too unreliable to use as an absence criterion (Huse & Taylor, 1962); thus, the majority of research on the effects of absence control initiatives has ignored medical absences in their analyses. We argue that to understand the full impact of absence control initiatives on employee behavior, it is important to study a broad range of employee absence, including absence attributed to medical reasons.
The Family and Medical Leave Act (FMLA), enacted in the United States in 1993, allows employees to take nonpenalized time off work for either personal medical reasons or the medical needs of close family members. Many organizations consider absences reported as FMLA to be involuntary absences that should not be influenced by the implementation of absence control initiatives. However, we suggest that because the FMLA is federally mandated (i.e., employers may not legally take any punitive action against employees taking absences under the FMLA), and because FMLA absences can be difficult to verify, it may also function as an attractive reporting category for employees who wish to circumvent the punishments and preserve the potential rewards associated with absence control policies. In other words, FMLA absences may be a potential indicator of unintended, and largely unaccounted for, deviant behavior that employees engage in when the outcomes of organizational absence policies are salient. Thus, we ground our article in the organizational justice literature to (a) examine the effects of an absence policy change—implemented in two manufacturing plants at different points in time—on casual, FMLA, and disability-based employee absence and (b) examine how differences in employee perceptions of the salience of absence outcomes and the fairness of the absence policy may ultimately influence reported employee absences.
Nature of Absences and Absence Policy Implementation
Although absences occur for many reasons, two of the most common types of unscheduled absences recorded in organizations are casual and FMLA absence. Casual absence comprises absences because of mild illness, non-work-related injury, and other personal reasons. This conceptualization is closest to the way voluntary absence has been treated in the extant absence literature (Steers & Rhodes, 1978). In addition, casual absence is the type of absence most often targeted by organizations in their implementation of absence control initiatives.
Absences related to the FMLA have been excluded from traditional research on voluntary absence. Since it was signed into law by President Clinton in 1993, the FMLA has allowed employees to take “no-fault” unpaid leave for up to 12 weeks in a 12-month period for family-related medical issues, such as the birth and care of a newborn child, the placement of a son or daughter for adoption or foster care, to care for an immediate family member with a serious health condition, or the inability to work because of a serious health condition. Because the act is federally mandated, employers may not legally take any punitive action against employees taking absences under the FMLA.
No published study has examined the effects of organizational absence control initiatives on FMLA absence. Indeed, the only published studies on FMLA absence have simply examined the effect of demographic variables (Gerstel & McGonagle, 1999; Lee & Sanford, 2004). These studies appeared to replicate earlier findings by Blau (1985), who separated “excused sick family” absences from other excused absences. He found that the best predictors of this kind of absence were demographic: Both marital status and the number of dependents were significantly related to excused sick family absence. Notably, this study was conducted prior to the enactment of the FMLA and thus could not have examined absences attributed specifically to the FMLA.
Absence Policies and Organizational Justice
Organizational justice research focuses on the role and importance of fairness perceptions in the workplace (Greenberg, 1990). Two justice dimensions—distributive and procedural justice—provide the theoretical foundation for exploring the outcomes associated with the implementation of absence control initiatives. Distributive justice is the perception of fairness regarding outcomes, or goods, received by employees, resulting from an allocation decision (Greenberg, 1987). Procedural justice is the perceived fairness of the policies and procedures used to decide outcome allocation (Leventhal, 1980; Thibaut & Walker, 1975). Thus, procedural justice is focused on how the decision is made, and distributive justice is focused on the fairness of the decision outcome itself (Konovsky, 2000).
Using this justice framework, both absences themselves and the new punishments and rewards in absence control initiatives can be conceptualized as goods, or outcomes, that may be “consumed” by employees. That is, employees may view days off as something they deserve or earn, and absence control policies that create punishments for absence and rewards for attendance may be perceived as both distributively and procedurally unfair. Employees use their current asset position as the reference outcome to which they compare a new set of outcomes to determine if they have suffered a loss or experienced a gain (Kahneman & Tversky, 1984). The change, or reduction, in goods or outcomes associated with casual absence under the new policy can create a sense of loss when employees compare the outcomes or goods they had before the initiative with the outcomes or goods they have after the initiative. Because a newly implemented absence policy acts as a set of rules that determine the type and number of allowed absences—as well as the punishments and rewards associated with those absences—employees are likely to avoid taking absences that would result in a loss of goods (e.g., experiencing punishments or losing rewards).
It is interesting that the use of punishments and rewards in absence control initiatives may also influence the frames employees use when deciding how to report their absences. Tenbrunsel and Messick (1999) found that the use of sanctions shifted how people approached a social dilemma. Subjects initially saw the decision as an ethical dilemma, but when sanctions were implemented, their perceptions and motives moved from ethical to calculative. That is, people evaluated the dilemma in terms of gains and losses rather than in terms of ethical responsibilities. In response to this shift in both outcomes and procedures, we argue that employees will frame their reporting decisions such that they will try to avoid the negative consequences associated with taking casual absences.
Hypothesis 1: Absence policies designed to increase the salience of absence outcomes lead to reductions in reported casual absence.
When faced with an inequity or a perceived loss in outcomes or goods, employees tend to change their inputs (e.g., decrease effort), change their outcomes (e.g., ask for a raise), distort their inputs or outcomes, select a different comparison other, or leave the organization (Adams, 1965). Thus, if employees perceive that a policy has resulted in unfair allocations of goods, they may choose to retaliate or engage in deviant behavior against the changes in the absence policy and the organization that implemented it as a means of restoring equity (Bies & Tripp, 2002; Robinson & Bennett, 1995). Although equity theory suggests that employees can restore equity using a variety of strategies such as the ones mentioned above, very few strategies can “replace” or “recapture” what was perceived as lost in an absence policy change (decreased freedom to take casual absence). In other words, to take absences without the risk of punishment and without forfeiting potential rewards, employees may choose to restore equity through reporting casual absences under categories that are protected from punishment and that do not sacrifice the opportunity to be rewarded for high levels of attendance.
In his early research on distributive justice, Homans (1961) argued that when employees perceive that they are less powerful than the source of the perceived injustice (e.g., the organization, their supervisor, etc.), their actions to restore equity will be largely indirect. Jermier, Knights, and Nord (1994) revisited this concept of indirect action by arguing that employees may choose to engage in covert retaliation, “such as the withdrawal of citizenship behaviors, psychological withdrawal, and resistance behaviors” in response to perceived injustice (Skarlicki & Folger, 1997: 434). Thus, if a direct path to retaliation against the organization is blocked, covert retaliation against the organization using another, more indirect method may become an attractive way for the employee to restore perceptions of unfairness regarding the absence control initiative.
We suggest that FMLA absences might provide just such an attractive opportunity for employees to retaliate against a new absence policy. Because the act specifically states that employers may not “interfere with, restrain, or deny the exercise of, the attempt to exercise any right” (U.S. Department of Labor, 1993: section 105a1) under the FMLA, supervisors may be reluctant to question the validity of an employee’s claim of FMLA absence. Although employees may need to provide documentation for taking FMLA absence, this may not always be true. The act states, “An employing agency may require that a request for leave under subparagraph (C) or (D) of section 6382(a)(1) be supported by certification issued by the health care provider of the employee or of the son, daughter, spouse, or parent of the employee, as appropriate” (U.S. Department of Labor, 1993: section 103a, emphasis added). Thus, companies are not required under the act to obtain medical certification from their employees who take FMLA absence. Moreover, for companies that do require documentation, the exact nature of that documentation is not entirely clear. In Schaar v. Lehigh Valley Health Services (2010), the 3rd Circuit Court of Appeals noted that a significant number of lower courts viewed “lay testimony, on its own, is sufficient” for establishing cause for FMLA leave. 1
Moreover, many companies may hesitate to ask for documentation, either because they are unfamiliar with the FMLA or because they fear overstepping the bounds of the act. The FMLA allows for “liquidated damages,” meaning that employees whose rights have been violated under the FMLA are eligible for double the damages they would normally receive. In addition, in Narodetsky v. Cardone Industries, Inc., et al. (2010), the court ruled that individual HR managers could be held liable for damages pursuant to the FMLA. In sum, FMLA could actually be seen as one of the least risky ways to regain perceived lost absences because employers cannot retaliate or easily verify the FMLA absence.
The ambiguous nature of the FMLA, its reporting policies, and the possible ramifications to the organization and individual managers for challenging employees’ usage of the FMLA for absences combine to make the FMLA a particularly attractive option for taking absences as a form of indirect retaliation. Accordingly, we offer the following hypothesis:
Hypothesis 2: Absence policies designed to increase the salience of absence outcomes lead to increases in reported FMLA absence.
Disability absences provide an interesting and important comparison to FMLA absences. Both categories have been widely considered to be involuntary medical forms of absence, and both have been excluded from research on the impact of absence control initiatives. However, both categories also require different amounts of documentation and verification and carry with them different risks for employees. Specifically, the focal company of this study had a long history of managing disabilities, dating back to a 1949 union initiative to allow for paid disability retirement. The FMLA, on the other hand, was fairly new and, as a federal mandate, was not included in union contracts. As a result, the company did not have as clear of a procedure for reporting FMLA absences as it did for reporting disability absences. In addition, under the union contract, disability absences were instances of paid leave, whereas FMLA absences were unpaid. Although this may encourage employees to try to take disability absences (and thus get paid for the absences), the company was understandably quite stringent in granting disability absences because of the financial cost the company would bear for these absences. Thus, we argue that the ability and willingness of employees to use disability absences as a means of retaliation against the organization for changes to the absence policy should be relatively low compared with that of FMLA. Accordingly, we offer the following hypothesis:
Hypothesis 3: Absence policies designed to increase the salience of absence outcomes should cause neither reductions nor increases in reported disability absence.
Differences in Employee Policy Perceptions
In the above sections, we argued that newly implemented absence policies change the outcomes associated with employee casual absences and that this shift creates a general sense of loss, or inequity, that shifts the decision frame employees use when reporting absences. Employees’ perceived loss of equity and their shift to a calculative mind-set suggest an increase in FMLA absences as well as the expected decrease in casual absence. Differences in individual perceptions of absence policies, however, may also influence the behavior of employees and their reactions to the policy change. Although absence policies—like all human resource policies—are objective realities in organizations, a great deal of empirical evidence suggests that employees perceive the same policies in different ways. These different perceptions may occur for two reasons. First, human resource policies may be differentially enforced throughout the organization; some supervisors may be vigilant in enforcing policies, but others may be more lax in their enforcement (Bacharach, Bamberger, & Sonnenstuhl, 2002). Second, individual differences, such as demographic characteristics, may cause different employees to perceive the same policies differently. For example, differences in ethnicity have been linked to different perceptions of affirmative action policies (Kravitz et al., 2000). If employee perceptions of absence policies differ, they may have an effect on absence behavior that could vary across employees despite the fact that they are operating under the same absence policies. Moreover, if employees shift the reporting of their absences from casual to FMLA absence as we suggest above, employee differences in their perceptions of organizational absence policies may explain this shift.
Specifically, we define perceived salience of absence outcomes (or policy salience) as being the extent to which individual employees perceive that rewards and/or punishments are tied to absence behavior. As noted above, absence policies may be differentially enforced throughout an organization, and thus some individuals may perceive an absence policy to be more salient than others. Furthermore, individuals may “differ in the extent to which they perceive issue-related costs versus benefits” (Darke & Chaiken, 2005: 878). That is, employees are likely to differ in terms of how attractive the rewards and how onerous the punishments are with respect to absence behavior. As a result, when employees weigh the rewards and/or punishments they may face by choosing whether to adhere to the policy, individuals may perceive what would appear to be the same costs and benefits quite differently.
By reflecting on the perceived salience of outcomes associated with reporting an absence, individuals choose a course of action regarding absence in their organization. The stronger their perceptions of the salience of absence outcomes, the greater will be their perception that negative consequences will result from taking a casual absence. Because these consequences under the new policy can be applied only to casual absences (per legal restrictions associated with the other two categories), if employees need to report an absence (e.g., they refuse to miss their child’s kindergarten graduation), we argue that how they go about reporting it will be largely guided by their perceptions of the strength of the consequences associated with casual absences. In other words, when taking an absence, individuals with stronger policy salience perceptions may, then, report it as a type of absence that is unaffected by the policy; specifically, as stated above, we believe that they would be more likely to report their absence as FMLA. Reporting casual absences as FMLA absences would circumvent the reward and punishment process associated with casual absence in the new absence control policy. As a result, we suggest that individuals with stronger perceptions of the salience of absence outcomes should be likely to take more FMLA absences and fewer casual absences.
Hypothesis 4: Individual perceptions of the salience of absence outcomes will be associated with (a) reductions in reported casual absence and (b) increases in reported FMLA absence after the policy change.
Differences in employee perceptions of the fairness of the absence policy are also likely to influence how employees report absences. Previous research has linked general fairness perceptions to absence behavior (Colquitt, Conlon, Wesson, Porter, & Ng, 2001) but has not specifically examined how perceptions of the fairness of an absence policy affected absence behavior. This is unfortunate because according to fairness heuristic theory (Lind, 2001), changes in organizational policies represent the kind of “phase-shifting events” that cause individuals to reexamine their judgments of organizational justice. As employees form judgments regarding the fairness of the new absence policy, they may subsequently change their absence behavior.
We suggest that employee fairness perceptions of new absence policies should affect both their casual absence and their FMLA absence. First, in relation to casual absence, we draw on the group value model of procedural justice (Lind & Tyler, 1988) to predict that the fairer employees perceive the policy to be, the fewer casual absences they will take. 2 This model suggests that people’s perceptions of the fairness of group authorities affect their feelings of pride and respect in the group, which in turn affect their group-oriented behaviors. The more fair people perceive the authorities to be, the more they identify with the group and act in ways that are consistent with group expectations. In the case of organizational policies, employees who perceive that the policies are fair should be more likely to comply with them. Thus, we would expect perceptions of policy fairness to be negatively related to casual absence.
Second, in relation to FMLA absence, equity theory (Adams, 1965) predicts that the more fair employees perceive the new policy to be, the less employees will be inclined to retaliate against the organization. Indeed, empirical research has found that perceptions of inequity can lead to resentment and absenteeism (Cropanzano & Greenberg, 1997; Geurts, Buunk, & Schaufeli, 1994). Therefore, employees who perceive that the new absence policy is unfair are likely to attempt to restore equity by taking more FMLA absences. Although they may react in multiple ways, FMLA could be seen as one of the least risky ways to regain their lost absences because employers cannot retaliate or easily verify FMLA absences. If employees perceive the new absence policy as fair, however, they will be less likely to take more FMLA absences and also less likely to take more casual absences.
Hypothesis 5: Perceptions of policy fairness will be associated with reductions in both (a) reported casual absence and (b) reported FMLA absence after the policy change.
Finally, we expect that individual perceptions of the salience of absence outcomes and policy fairness perceptions may interact in predicting FMLA absence. If certain employees merely shift their absences from casual to FMLA, the ones who would be most likely to do this would be the employees who perceive the new policy is both salient and unfair. Because they perceive the salience of absence outcomes to be high, they may seek to avoid the negative consequences, and preserve their opportunity to be rewarded, by attributing their casual absence to FMLA. We suggest, however, that this type of behavior is most likely to take place among employees who perceive that the policy is unfair. As noted above, the group value model of procedural justice predicts that people who perceive the authorities to be fair will be more likely to comply with organizational policies, regardless of how salient the outcomes are. Employees who believe that the salience of absence outcomes are strong—but also believe it to be fair—would be less likely to shift their absences from casual to FMLA. Fairness perceptions, then, should attenuate the effect of salience of absence outcomes on FMLA absence.
Hypothesis 6: Perceived salience of absence outcomes and fairness will interact to predict FMLA absence, such that the relationship between salience of absence outcomes and FMLA absence is attenuated when fairness perceptions are high.
Method
Participants and Procedure
In this study, we used a variation of an “interrupted time series with switching replications” quasi-experimental design (Cook & Campbell, 1979). This particular design involves taking measures of the dependent variable (i.e., absences) over multiple periods with two groups, and the treatment (in our case, policy change) takes place at different times. According to Cook and Campbell (1979: 223), “[T]he power of [this] design derives from its control for most threats to internal validity and from its potential in extending external and construct validity. External validity is enhanced because an effect can be demonstrated in at least two settings at different moments in history.” It is important to note that the interventions were similar but not identical across plants (details below). Table 1 shows a visual representation of our experimental design and data collection.
Experimental Design and Data Collection
Note: O = observation. Absence data were collected on a quarterly basis over 24 months.
We tracked the absences of 1,019 employees in two plants of a large unionized Midwestern automobile parts manufacturer over a 24-month period. A total of 733 employees worked in Plant A; of these, 74% were male, with a mean age of 48.78 years (SD = 9.21) and a mean tenure with the organization of 21.47 years (SD = 9.83) at the beginning of the absence data collection. A total of 286 worked in Plant B; demographic data were not available for this sample. We obtained permission to administer a survey to 250 randomly selected employees in Plant A 18 months after we began tracking absences (6 months after the implementation of the policy change) and were able to match 223 of these surveys to the employee’s absence data. This subsample was 75% male, with a mean age of 49.84 (SD = 9.14) and a mean tenure with the organization of 21.88 years (SD = 9.77). A comparison of this subsample to the full sample of employees revealed that there were no significant differences in gender, age, tenure, or any of the absence behaviors either before or after the policy change. Thus, this subsample appears to be representative of the population of employees.
Twelve months after we began tracking absences, the company initiated an organizational policy change at Plant A to clarify and strengthen their absence policies. Specifically, they implemented a “No Fault” absence policy that communicated new policies to the employees regarding the rewards that could be gained for good attendance as well as the punitive measures associated with poor attendance. Among other provisions, employees received written reprimands for their first two absences and in-house disciplinary layoffs of increasing severity for their next three absences and would be discharged on their sixth absence. In addition, employees could receive a “Good Attendance Bonus” that involved preferred parking spaces and more control over their vacation leave. Nine months later, Plant B also implemented a policy change. The policy change was not as extensive as it was at Plant A; the major change was in the implementation of in-house disciplinary layoffs like those implemented at Plant A. Although the policy changes were not identical at the two plants, they appeared to be similar enough in intent (they were both focused on increasing the salience of absence outcomes) and in application (they both involved similar punitive measures).
Measures
Absence
We obtained absence data through company records. We operationalized absence as total unscheduled absence, which included both full and partial days off (we did not track scheduled absences, such as holidays and vacation leave). The absence category was recorded by the supervisors after the employees reported the reasons for their absences. The company recorded unscheduled absences in 16 subcategories, and after conferring with company personnel as to the definition and intention of each subcategory, we assigned them to our three superordinate categories. We operationalized an absence as casual if it had been attributed to 1 of 10 subcategories, including “informal personal leave” and “unexcused.” We operationalized an absence as FMLA if it had been attributed to one of two subcategories: FMLA (full day) or FMLA (partial day). We operationalized an absence as disability if it had been attributed to one of four subcategories, including “compensable leave” and “sick leave.” At Plant A, we were able to compare a full year of absence prior to the policy change to a year after the change. At Plant B, however, we were able to gather 21 months of absence data prior to the change but only 3 months of absence data after the change. 3
Policy perceptions
Because no extant scales specifically examine perceptions of the salience of absence outcomes or perceptions of absence policy fairness, we measured policy perceptions with items developed for this study. Perceived salience of absence outcomes was measured with seven items that captured employee perceptions of the rewards and punishments associated with absence. The items were “I am appreciated for good attendance,” “Formal attendance reward programs reward or recognize employees for good attendance,” “Management makes it a priority to recognize employees for good attendance,” “Attendance rewards motivate me to come to work,” “Employees who are frequently absent receive appropriate discipline,” “Employees with poor attendance fail to succeed at this company,” and “There are serious consequences associated with poor attendance.” Perceptions of policy fairness were measured with four items, which were “Attendance policies are fair and reasonable,” “Employees are provided with sufficient time off to attend to personal matters,” “Attendance policies are not sensitive to employee personal needs and obligations,” and “Limitations regarding when employees can take time off are unfairly restrictive.” These last two items were reverse coded. Responses were provided on a 5-point scale ranging from 1 = strongly disagree to 5 = strongly agree.
Because these were new scales, we examined their psychometric properties before testing our hypotheses. The items were randomly ordered on the survey, which allowed us a rigorous test of the discriminant validity of the two scales. We subjected the items to a confirmatory factor analysis and compared the hypothesized two-factor structure to a one-factor model. The two-factor model provided a much better fit to the covariance structure of the data (χ2 = 127.54, 43 df, root mean square error of approximation [RMSEA] = .05) compared to the one-factor model (χ2 = 261.48, 44 df, RMSEA = .08). Coefficient alphas were .77 for policy fairness and .71 for policy salience, and the scales showed a small but significant correlation with each other (r = .24, p < .05). With reasonable confidence that the scales were psychometrically sound, we proceeded with our hypothesis testing.
Results
Table 2 provides the means, standard deviations, and intercorrelations of all of the variables included in the study for both manufacturing plants. In this matrix, the absence data were aggregated to pre- and post-policy-change data. The table shows similar results for the demographic variables as have been found in previous studies: Women were absent more than men in all absence categories, and younger workers tended to be absent more than older workers. Inspection of this table also reveals that the absence categories tended to be positively correlated; that is, if employees took time off work for one type of absence, they were likely to have taken time off for another type of absence as well. It is interesting that the largest correlation among the absence categories is between prechange casual absence and postchange FMLA absence (r = .53, p < .01), suggesting that employees who took casual absences before the policy change may have shifted to FMLA after the change.
Means, Standard Deviations, and Intercorrelations
Note: FMLA = Family and Medical Leave Act. Correlations below the diagonal are for Plant A; N = 733 for all correlations except those with policy salience and policy fairness, where N = 223. Correlations above the diagonal are for Plant B, where N = 285. Absences are recorded as the mean number of absences per quarter. Gender is coded as 1 = female, 2 = male.
p < .05. **p < .01.
Before testing our hypotheses, we ran an omnibus analysis to determine the overall effectiveness of the policy change on reducing absence across all categories. This test revealed that the policy change had the intended effect, in that absences across the two plants reduced an average of 1.14 days per employee per quarter, t(1017) = 8.66, p < .01. The effect was similar across the two plants, with a reduction of total absence per employee per quarter of 1.06 days at Plant A and 1.38 days at Plant B.
Hypotheses 1 and 2 proposed that the policy change would reduce casual absence but increase FMLA absence. We tested these with separate dependent-sample t tests for each plant by comparing employee absence before the policy change with absence after the policy change. We note that dependent-sample t tests are within-person tests, where the employees act as their own control (i.e., each employee’s post-policy-change absence is being compared to that same employee’s pre-policy-change absence), and thus interindividual control variables are unnecessary. These tests revealed that at both plants, casual absence was significantly lower following the policy change. At Plant A, casual absence dropped from an average of 1.92 days per employee per quarter (SD = 3.37) to 0.63 days per quarter (SD = 1.25), t(732) = 11.77, p < .01, and at Plant B, casual absence dropped from 2.86 days per quarter (SD = 2.45) to 1.34 days per quarter (SD = 2.17), t(285) = 10.07, p < .01. Because we only had 3 months of post-policy-change absence data at Plant B, we also tested whether casual absence in those 3 months was significantly lower than casual absence for the same 3 months in the previous year, in case there were any seasonal effects associated with time of year. This t test was also significant, t(285) = 6.50, p < .01; casual absence dropped from 3.58 days for the quarter in the previous year (SD = 5.77) to the 1.34 days (SD = 2.17) in the quarter immediately following the policy change. Thus, Hypothesis 1 was supported. We also analyzed effect sizes using the d statistic, which is computed by dividing the mean difference by the pooled standard deviation. Cohen (1988) suggested that a d of 0.20 generally reflects a small effect, 0.50 represents a moderate effect, and 0.80 represents a large effect. In the present research, the d was 0.56 at Plant A and 0.66 at Plant B, suggesting that the policy change had a moderate effect on casual absence.
Hypothesis 2 was also supported as the t tests revealed that FMLA absence increased at both plants following the policy change. At Plant A, FMLA absence increased from an average of 0.36 days per employee per quarter before the policy change (SD = 2.06) to 0.66 days per quarter (SD = 2.00), t(732) = 3.81, p < .01, and at Plant B, FMLA absence increased from 0.67 days per quarter (SD = 1.68) to 1.84 days per quarter (SD = 2.19), t(285) = 9.29, p < .01. A comparison at Plant B for the same 3 months in the previous year was also significant, t(285) = 10.89, p < .01, with FMLA absence increasing from 0.44 days (SD = 1.52) to the aforementioned 1.84 days. The d statistic for these differences was 0.15 in Plant A and 0.60 in Plant B. Thus, the policy change had a small effect at Plant A and a moderate effect at Plant B.
In contrast to Hypotheses 1 and 2, Hypothesis 3 predicted that disability absence would not be affected by the policy change. Although it is often difficult to draw strong conclusions from null hypotheses like this one, we proceeded with confidence because we had high statistical power to conduct the test, even to detect a small effect size (more than 0.90 for all tests). Thus, the probability of committing a Type II error (failing to reject the null hypothesis when it really should be rejected) was quite small. Following the procedure outlined by Rogers, Howard, and Vessey (1993), we tested this hypothesis by examining the confidence intervals surrounding the mean difference. If the confidence interval does not include the value that would be considered a small effect size, the null hypothesis is supported. Using Cohen’s (1988) value of 0.20 as a guide, a small effect size, then, would be 20% of the pooled standard deviation.
At Plant A, the t test revealed that disability absence before the policy change (M = 0.30, SD = 1.08) was not significantly different from disability absence after the change (M = 0.23, SD = 0.75), t(732) = 1.64, ns. The 95% confidence interval (–0.01, 0.15) did not include 0.18, the value that would be considered a small effect size (20% of 0.92, the pooled standard deviation). Thus, Hypothesis 4 was supported for Plant A. At Plant B, however, disability absence dropped from an average of 1.80 days per quarter (SD = 4.97) to 0.76 days per quarter (SD = 2.92), t(285) = 3.43, p < .01. This difference resulted in a d statistic of 0.26, indicating a small effect. A comparison at Plant B for the same 3 months in the previous year was not significant, however, when the average amount of disability absence was 0.98 days for the quarter (SD = 6.42), t(285) = .55, ns. The 95% confidence interval for this test (–0.57, 1.02), however, did include 0.93, the value that would be considered a small effect size. Therefore, for Plant B, although our test that attempted to remove possible seasonal effects on disability absence indicated that disability absence after the policy change was not significantly different from disability absence before the policy change, the stricter test using the confidence interval does not allow us to conclude that Hypothesis 3 was supported. Thus, Hypothesis 3 was only partially supported. One possible explanation for the anomalous results in the number of disability absences at Plant B is that the company tightened up its absence reporting at the time of the policy change to more effectively levy the punishments and rewards associated with casual absences. Thus, the decrease in reported disability absences may have been the result of stricter enforcement of overall absence reporting.
Hypothesis 4 predicted that salience of absence outcomes would lead to (a) reductions in casual absence and (b) increases in FMLA absence after the policy change. Hypothesis 5 predicted that policy fairness would lead to reductions in both (a) casual and (b) FMLA absence after the policy change. Hypothesis 6 predicted that these two perceptions would interact in their prediction of FMLA absence, such that the highest amount of FMLA absence would be taken by those who perceived the policy to have salient absence outcomes and also to be unfair. We tested these hypotheses with two hierarchical regression equations (one for casual and one for FMLA absence); the results of these equations can be found in Table 3. Because these were between-person tests, we controlled for employee absence prior to the policy change, gender, age, and organizational tenure. As noted above, we were able to capture these perceptions only for 223 randomly selected employees in Plant A.
Hierarchical Regression of Absence on Policy Perceptions
Note: N = 223. FMLA = Family and Medical Leave Act. Coefficients are standardized regression coefficients. Gender is coded as 1 = female, 2 = male. Rounding error caused the R2 for casual absence to appear not to total correctly.
p < .05. **p < .01.
Note: The figure is plotted at one standard deviation above and below the mean for both policy salience and policy fairness.
As can be seen in Table 3, after controlling for absence prior to the policy change, gender, age, and organizational tenure, salience of absence outcomes did not have a significant effect on casual absence (β = .02, ns) but did have a significant positive effect on FMLA absence (β = .17, p < .01). This indicates that the stronger that the employees perceived the salience of absence outcomes to be, the more FMLA absence they were likely to take in the year following the policy change. Thus, Hypothesis 4a was not supported, but 4b was supported. Perceptions of policy fairness had a significant negative effect on casual absence (β = –.14, p < .05), but did not have a significant effect on FMLA absence (β = –.08, ns). This indicates that the more fair employees perceived the policy to be, the less casual absence they were likely to take in the year following the policy change. Thus, Hypothesis 5a was supported, but 5b was not.
The interaction term for salience of absence outcomes and policy fairness was significant in predicting FMLA absence (β = –.69, p < .01, ΔR2 = .02). The nature of this interaction is plotted in Figure 1 at one standard deviation above and below the mean for both variables. This figure shows that the relationship between policy salience and FMLA absence was positive only for employees who also perceived the policy to be unfair. Indeed, the average number of FMLA absences taken in the year following the policy change was approximately the same for all employees except those who were high in policy salience and low in policy fairness. These employees took an average of approximately 5 days of FMLA absence in the year, compared to 1–2 days for those who were high in policy fairness or low in both salience and fairness. Thus, Hypothesis 6 was supported.
Interactive Effect of Policy Salience and Fairness on Family and Medical Leave Act Absence
Additional Post Hoc Analyses
One of our main contentions is that employees switched their absences from casual to FMLA following the policy change. We cannot address this question directly because that would entail having employees report that they falsified their absence reporting. Instead, we address the question indirectly in two ways: (a) by examining how much variance in post-policy-change absences is explained by each type of absence before the policy change and (b) examining the absence patterns of employees who thought the policy was both strong and unfair compared to the other employees.
For the first analysis, we ran three separate regression equations where post-policy-change absence for each category was regressed on the three pre-policy-change absence variables. We used stepwise regression to parse out the strongest effects from the others. Stepwise regression finds the best predictor of the dependent variable, enters it into the equation and parses out its variance, and then looks for the next strongest predictor. For post-policy-change casual absence, pre-policy casual absence was the best predictor (β = .44, p < .01, R2 = .19), followed by pre-policy-change FMLA absence (β = .15, p < .01, ΔR2 = .02); pre-policy-change disability absence was not a significant predictor. For post-policy-change FMLA absence, pre-policy-change casual absence was the best predictor (β = .53, p < .01, R2 = .28), followed by pre-policy-change FMLA absence (β = .36, p < .01, R2 = .13), followed by pre-policy-change disability absence (β = .06, p < .05, R2 = .004). The fact that pre-policy-change casual absence (rather than pre-policy-change FMLA absence) was the best predictor of post-policy-change FMLA absence provides strong evidence for ordering effects. This shows that the employees who took FMLA absences after the policy change were largely the ones who took casual absences before the policy change. It is interesting that pre-policy-change casual absence was also the best predictor of post-policy-change disability absence (β = .26, p < .01, R2 = .07), followed by pre-policy-change disability absence (β = .20, p < .01, R2 = .04). This suggests that some of the people who took casual absences may have switched to disability absences after the policy change.
For the second analysis, we compared the absence patterns of employees who believed the policy was unfair and that the outcomes were salient versus the rest of the employees. We did this by computing the levels of each type of absence (pre– and post–policy change) at one standard deviation above and below the mean for both salience and fairness perceptions. Then we conducted simple slopes analysis for salience at high and low levels of fairness for each type of absence. The results of this post hoc analysis can be found in Table 4. As expected, the only two simple slopes that were significant were for pre-policy-change casual absence and post-policy-change FMLA absence (both p < .05). These results indicate that for employees who perceived the new policy to be unfair, those who also perceived it to be salient had higher levels of prechange casual absence and postchange FMLA absence (relative to those who believed it was unfair but not salient). Employees who thought the new policy was both salient and unfair took 14.18 days of casual absence in the year before the policy change. This compares to only 3.89 days of casual absence for employees who thought the policy was unfair but not salient. Similarly, employees who thought the new policy was both salient and unfair took 5.33 days of FMLA absence in the year after the policy change. This compares to only 0.57 days of FMLA absence for employees who thought the policy was unfair but not salient. This provides further support for our contention that the employees who increased their FMLA absences after the policy change were those who had high levels of casual absence before the policy change.
Comparison of Employees on the Basis of Salience and Fairness
Note: FMLA = Family and Medical Leave Act. Numbers indicate the predicted number of days employees were absent at one standard deviation below and above the mean for both salience and fairness. Simple slopes were tested for salience at low and high levels of fairness. Values in bold indicate a significant simple slope for salience at that level of fairness.
p < .01.
An anonymous reviewer suggested that we examine possible gender effects. We reanalyzed our data to see whether women may have been disproportionately affected by the policy change and thus increased their use of FMLA absences more than men. As can be seen in Table 2, regression results indicated that women were indeed significantly more likely to take FMLA absences after the policy change; specifically, controlling for pre-policy-change FMLA absences, gender accounted for 3.4% of the variance in post-policy-change FMLA absences. This gender effect did not hold, however, for casual absences (ΔR2 = .1%, ns).
In addition, we found that women’s perceptions of the salience of absence outcomes and their perceptions of the fairness of the new policy affected their FMLA absences much more than was the case for men. The main effects for perceptions of the salience of absence outcomes and fairness accounted for 13.2% of the variance in FMLA absences for women (controlling for previous FMLA absence) compared to only 0.7% for men. Similarly, the interaction between perceptions of policy salience and fairness accounted for 6.1% in FMLA absences for women compared to 0.2% for men. Therefore, it would appear that the attitudes of women toward the new policy mattered more than the attitudes of men regarding the types of absences they chose to report.
Discussion
This study examined the effects of a change in absence policy and employee perceptions of the new policy on their subsequent absence behavior. Employees took fewer casual absences after the policy change, and thus the policy change did have the expected effect of reducing casual absence. The policy change also, however, had the unintended (but hypothesized) effect of increasing absence attributed to the FMLA. This suggests that although organizational policy changes may be effective in reducing certain types of absence, employees may retaliate against the organization by reporting their absences under a different name.
Employee perceptions of the salience of absence outcomes affected only FMLA absence after the policy change, whereas employee perceptions of policy fairness affected only casual absence. The interactive effect of these two perceptions indicated which employees were the most likely to have taken FMLA absences. Employees with high perceived salience of absence outcomes and low levels of fairness took the highest level of FMLA absences in the year subsequent to the policy change. Therefore, when the absence policy was seen as both salient and unfair, it appears that employees may have “gamed the system” by shifting the reporting of their absences from casual to FMLA absences.
Theoretical Implications
The results of this study have implications for both the literature on absenteeism and the literature on organizational justice. In terms of the absenteeism literature, because past research on the impact of absence policy implementation measured only casual absence, the impact of the implementation of absence control initiatives on employee behavior was never fully captured or understood. In contrast, this article takes a more comprehensive look at the impact of absence policies by examining a broader range of reported absences and also by identifying possible forms of covert retaliation in which employees may engage in response to absence policy changes.
In terms of the literature on organizational justice, we found that employee perceptions of both the salience of absence outcomes and the fairness of absence policies can account for variance in absence behavior. Although previous research has shown that justice perceptions are related to absence (and other forms of work withdrawal), no study had examined how perceptions of the policy itself affected absence behavior. We suggest that our findings may speak not only to the effects of justice perceptions on withdrawal, but also on counterproductive work behavior (CWB) and organizational deviance. Previous research has found that low justice perceptions have been associated with higher levels of CWBs, such as rude behavior to coworkers, theft, and sabotage (Ambrose, Seabright, & Schminke, 2002; Fox, Spector, & Miles, 2001; Judge, Scott, & Ilies, 2006). In this study, employees who perceived the new absence policy as salient and unfair had the highest levels of subsequent FMLA absence. This may be interpreted as a form of CWB, as these employees may be intentionally obfuscating the cause of their absence to avoid the negative consequences associated with taking casual absence.
Practical Applications
This study also has at least three practical implications for organizations. First, our results show that policy changes can be effective in reducing casual absences. Implementing a policy designed to increase the salience of absence outcomes may be a more cost-effective solution to the problem of high absence rates than more labor- or resource-intensive absence control initiatives. Indeed, it has been suggested that the cost of many absence control initiatives designed to reduce employee absence can equal or even exceed the cost of the absences they are intended to reduce (Labour Research, 2006). As the present policy change was not particularly costly (e.g., providing preferred parking spaces and more control over vacation), it represents a viable strategy for organizations seeking to reduce absence levels. Second, however, despite the fact that highly voluntary absences decreased, FMLA absences increased. Thus, organizations should monitor all forms of absence to determine whether other absence is merely being shifted because of the changes in organizational policies. It is important to note that FMLA absences are no less costly than casual absences, as well as the fact that FMLA absences may not be as “fixed” as many organizations perceive them to be. Organizations would be well served to clarify their FMLA policies to minimize the kinds of unintended consequences observed in this study.
Third, organizations should monitor their employees’ perceptions of the new policies because perceptions of policy salience and fairness may have additional effects on absence rates and on employee acceptance and adherence to new absence policies. Careful attention to how an absence policy is developed and communicated is likely to have a major effect on policy perceptions. The organizational justice literature would suggest that the potential negative reactions to changes in important outcomes and goods can be moderated by providing adequate information regarding the reasons for the policy change as well as by giving employees voice and control over the policy change itself (Leventhal, 1980; Skarlicki & Folger, 1997; Thibaut & Walker, 1975).
Limitations
At least six factors potentially limit the findings of this study. First, although we were able to gather a full year of absence both before and after the policy change at one plant, we were able to gather only 3 months of post-policy-change data at the second plant (although we did gather 21 months of pre-policy-change absence at this plant). A period of 3 months is a short time over which to assess the effects of some causes of absence (Harrison & Martocchio, 1998) but has been shown to be a reasonable amount of time to assess the effects of absence outcomes and job attitudes (Martocchio, 1992). In addition, the fact that absences at this plant mirrored the results at the plant for which we had longer post-policy-change absence data strengthens our confidence in the results.
Second, as mentioned in the method section, the design used in this article is a variation of what Cook and Campbell (1979) refer to as an “interrupted time series with switching replications.” This type of quasi-experimental design involves taking measures of the dependent variable over multiple periods with two groups, and the treatment (in our case, policy change) takes place at different times. This design also strengthens external validity because an effect can be demonstrated in at least two settings at different moments in history. Although Cook and Campbell suggest this design controls most threats to internal validity, one threat that is not controlled is a selection–history interaction. In our case, this would mean that there could potentially have been some local history effects in each plant; that is, other organizational factors may have occurred simultaneously with the policy change at each plant that account for the results. It seems unlikely to us that other organizational events would have more of an effect on absences than policy changes that were directly intended to affect absences, but because we cannot completely rule it out, we note this as a potential limitation.
Third, we were able to gather data on employee perceptions of the absence policy at only one point in time. Thus, we have no information on whether perceptions changed from what they were under the old policy. It is possible, for example, that employees who perceived the new policy as salient and unfair also perceived the old policy as salient and unfair, reflecting the effect of dispositions or stable work characteristics. Our confidence that this is not the case, however, is strengthened because we controlled for previous absence when examining the effects of the policy perceptions. If employee perceptions had not changed, it is unlikely that FMLA absence would have been affected by employee perceptions subsequent to the policy change.
Fourth, we were able to gather perceptions from only 250 employees at one of the plants we studied. Therefore, we cannot be certain that their responses are reflective of the entire population of employees at the two plants. Two factors, however, suggest that they are indeed representative. First, these employees were randomly selected from all of the employees at their plant; random selection from a population is the strongest method of inferring generalization of the results of a sample to the population from which it was drawn (Cook & Campbell, 1979). Second, we found no significant differences between those who completed the survey and those who did not in terms of gender, age, or organizational tenure. Thus, we are reasonably confident that those who completed the survey are representative of the population of employees in the plant.
Fifth, it is possible that some casual absences reported before the change in policy were actually FMLA absences that were misreported. Unfortunately, we do not have the ability to determine whether absences were misreported before (or after) the policy change. The reporting procedure for absences did not change when the policy changed: Employees reported the reason for their absences to their supervisors, who then recorded the absence in one of 49 company categories (these categories include both scheduled and unscheduled absence; we limited our analysis to unscheduled absence). And although it may be that employees were simply misreporting their FMLA absences as casual absences before the policy changed, we argue that our data seem to suggest a different explanation. Only employees who perceived the policy to have salient absence outcomes and perceived the policy to be unfair showed high levels of FMLA absence after the policy change. Therefore, it seems unlikely that only this group of employees had misreported their absences prior to the change.
Finally, because the FMLA is law only in the United States, our findings may have limited international applications. The United States is certainly not a leader, however, in providing leave for family medical issues. Indeed, 75 countries established leave for childbirth seven years before the FMLA was established (Crampton & Mishra, 1995). Most industrialized countries now offer family leave, and unlike in the United States, employees receive pay during the leave. For example, in Germany, employees can take 14–19 weeks of fully paid leave for family medical emergencies, and parents can take an additional 5 days of fully paid leave per year to care for a sick child. Other examples include Venezuela, which provides 18 weeks of fully paid leave; Sweden, which provides up to 38 weeks with 90% pay; and Japan, which provides 12 weeks with 60% pay. It would seem, then, that employees in these countries would have an even greater incentive to apply their absences to family leave than to casual leave. Thus, our results may actually understate the effects of similar absence policy changes in other countries.
We also note that although the FMLA is federal law in the United States, states also may establish family leave laws. In 2004, California became the first state to establish a comprehensive paid family leave law, and others are considering similar laws (Gotbaum & Rankin, 2006). In addition, federal laws can and do change, and Congress may choose to alter the provisions of the FMLA in the future. If this were to happen, it would be interesting to study whether employees would find another avenue for attributing their absences (other than FMLA) that would avoid the negative consequences associated with absence control initiatives designed to increase the saliency of absence outcomes.
Directions for Future Research
As noted by an anonymous reviewer, at Plant B (where only disciplinary action was used), there was a larger shift (relative to Plant A) in employees reporting more FMLA absences post–policy implementation (see Table 2). Using the theoretical reasoning proposed in the article, it is possible that a policy with increased disciplinary consequences (without rewards) may have created a greater perceived loss and increased fear of punishment for employees in Plant B relative to what was experienced by employees in Plant A. As such, employees in Plant B may have had a stronger overall motivation to shift their reported absences to a category of protected absences. Although we have policy perceptions only for employees in Plant A, it may have been the case that employees in Plant B experienced higher perceived levels of injustice following the change and, accordingly, may have been more motivated to retaliate against the new policy through unexpected channels (e.g., FMLA absences). The differences found between the two plants demonstrate the need for future research that examines employee perceptions during the creation and the implementation of new absence policy.
We suggest six fruitful avenues for such future research. First, future studies could incorporate measurements of policy fairness and salience both before and after absence policy changes. This would determine whether it is simply perceptions of a new policy or changes in perceptions of policies that drive absence behavior. Second, future research could examine the role of interactional justice in mitigating the potentially negative impact of a new absence control initiative designed to increase the saliency of absence outcomes. Previous research has found that interactional justice can buffer the negative effects of low distributive justice on CWBs (Skarlicki & Folger, 1997), and similar effects have been found for stress-related symptoms (Greenberg, 2006). Managers may be able to reduce absence behavior by explaining the need for the new absence policy and the ways in which it can be beneficial to the organization.
Third, it may be useful to explore the antecedents of policy salience and fairness. Our data revealed small but significant effects of gender, age, and organizational tenure on policy salience and similar effects of gender and age on policy fairness. An examination of individual difference predictors of these attitudes (e.g., Big Five personality factors, dispositional affect) could highlight which types of employees are most likely to react negatively to policy changes. Fourth, future research could involve a similar design but track absence over a longer period of time. It may be that some of the changes found in employee absences may stabilize, go back up, or go back down over time either as employees simply grow accustomed to the policy or as new employees enter the organization. In addition, employees may react differently to the new absence policy if it is implemented over a period of years versus overnight. Employees may find the policy to be more fair if the changes are done incrementally; conversely, they could also become more distraught by the many times that the policy would have to be changed over an extended period of time.
Fifth, research could investigate whether employees have less resistance to changes in absence policies if the old absence policy had only been in place for a short period of time. For example, a start-up company that is 4 years old may change its absence policy with relatively less resistance or with higher perceptions of policy fairness than might a company that has had the same absence policy for 20 years. Finally, future research could examine the effect of changes in absence policies on “presenteeism”: when employees attend work, but are not fully functional because of illness or other medical conditions (Hemp, 2004). If organizations enact absence control initiatives like the ones in this study, employees may not only be more likely to take absences under the FMLA but also be more likely to show up at work sick. This could mean that although absences are reduced, the reduced productivity associated with sick—but present—employees may to some extent counteract the positive effects of the policy.