Work–Life Program Participation and Employee Work Attitudes: A Quasi-Experimental Analysis Using Matching Methods

Two Types of WLPs

The primary purpose of WLPs is to create a supportive work environment for employees by providing them with greater choices and flexibility in arranging work and life responsibilities. As briefly mentioned above, Hoyman and Duer (2004) suggested workplace policies may have distinct goals and different target beneficiary groups. For instance, work-oriented programs such as telework and alternative work schedules are aimed at supporting eligible workers, including those with and without dependent care responsibilities, in performing their job duties by providing flexibility in work location and time. Meanwhile, life-oriented programs such as child and elder care programs support employees’ family responsibilities by providing benefits to those who have caregiving obligations to their dependents. Although each type of program has a disparate goal, focus, and target group, it should be noted that the benefits from these two programs may overlap to some extent and may have spillover effects on each other. For example, employees may use work-oriented programs to handle their nonwork obligations. Also, those participating in life-oriented programs may ultimately become more productive at work because their life burdens and pressures are lessened.

Among the WLPs offered by U.S. federal agencies, telework and alternative work schedules are the programs most widely used. These programs focus on removing barriers to work by providing employees with greater flexibility in work location and time, thereby “blending spheres between home and work” (Hoyman & Duer, 2004, p. 121). Specifically, telework allows employees to choose from where they work (e.g., home, regional, or satellite offices). The OPM (2013) defines telework as “a work flexibility arrangement under which an employee performs the duties and responsibilities of such employee’s position, and other authorized activities, from an approved worksite other than the location from which the employee would otherwise work” (pp. 17-18). Alternative work schedules include such programs as flexible work schedules, compressed workweeks, and part-time schedules in which employees can complete their jobs during times other than regular work hours (e.g., 9 to 5).

Previous findings about the effects of telework and alternative work schedules are somewhat mixed. Caillier (2012, 2013a, 2013b) found teleworkers neither had significantly higher levels of work motivation nor reported lower turnover intentions than nonteleworkers; nevertheless, satisfaction with telework was positively related to organizational commitment. In addition, alternative work schedules have been found not to be significantly related to either job satisfaction and involvement (Facer & Wadsworth, 2008) or organizational commitment (Caillier, 2013b), but they have been associated with greater work productivity and improved service quality in federal employees (Facer & Wadsworth, 2008). However, Glass and Estes (1997) found a lack of workplace flexibility and rigid work schedules are associated with lower job satisfaction and higher turnover intentions.

With a focus distinct from work-oriented programs, life-oriented programs such as child care and elder care are provided to support employees who have qualified dependents. These programs include monetary benefits (subsidies), access to on-site day care facilities and employer-sponsored dependent care centers, referral services, seminars, workshops, support groups, and information on health insurance for older individuals and community resources for employees responsible for taking care of young children, elderly relatives, and/or friends (OPM, 2012, 2015). They are intended to provide benefits in nonwork domains to address employees’ life responsibilities. These programs are provided to not only employees who have an immediate child or an elderly parent but also those who claim responsibility for a relative or a loved one, specified by the OPM as “any individual related by blood or affinity whose close association with the employee is the equivalent of a family relationship” (OPM, 2012, p. 6).

Both child care and elder care programs have been shown to be positively related to the organizational commitment of federal employees (Caillier, 2013b). Also, child care subsidies have been associated with reduced turnover and improved organizational performance at the federal agency level (S.-Y. Lee & Hong, 2011). However, Caillier (2013b) found that child and elder care programs did not have significant relationships with employees’ job involvement. Taken together, these inconsistent findings necessitate development of a theoretical framework and empirical research on the distinct forms and outcomes of various WLPs.

Theoretical Background

Social exchange theory provides theoretical foundations for explaining the relationship between WLPs and employee attitudes. Unlike economic exchange, which is based on formal contracts that entail relatively explicit and instantaneous interchange, social exchange involves a broad and nonspecific obligation between exchange partners (e.g., between employee and organization) in which the nature of the exchange is at their discretion (Blau, 1964). Exchange partners develop long-term interdependent relationships with each other on the basis of the norm of reciprocity (Gouldner, 1960) because the action of one party leads to a response by the other (Cropanzano & Mitchell, 2005). One offers benefits to the other in expectation of a return in the future, and the other feels obligated to reciprocate. However, the return does not need to be in the same form as the benefits, and there may exist a “temporal gap between what is given and what is returned” (Mostafa, Gould-Williams, & Bottomley, 2015, p. 748).

Organizations provide a broad range of WLPs to accommodate employees’ diverse needs in dealing with the conflicting demands of work and life situations. As the boundary between work and nonwork domains has increasingly blurred, employees have had a greater need to balance work and personal responsibilities. Investment in human resource (HR) practices may lead employees to believe that the organization and the supervisor value their contributions and are committed to their well-being (Mostafa et al., 2015). Employees would view WLPs as beneficial if these programs helped resolve such tensions and improve working conditions and personal well-being. According to social exchange theory, employees who participate in WLPs are likely to feel obligated to return the favorable treatment by exhibiting positive work attitudes in the form of improved satisfaction and commitment.

However, different types of WLPs may generate dissimilar levels of social exchanges between employee and organization because employees may not perceive and respond to all types of WLPs in the same way (Nishii & Wright, 2007). Although both work- and life-oriented programs are primarily intended to create a supportive work environment in which employees can seek a balance between job duties and personal circumstances, individuals may see different levels of organizational benefits associated with each of these programs, which in turn may lead them to develop a varying sense of obligation to reciprocate. Work-oriented programs are directed at removing obstacles and providing greater flexibility to support the successful completion of work duties, but they can also be used to meet personal and family needs in nonwork domains. Therefore, both employees and organizations receive direct benefits from work-oriented programs (Hoyman & Duer, 2004). In the case of life-oriented programs, employees who have caregiving obligations are the primary beneficiary group. Organizations can also benefit from these programs indirectly through improved work motivation and recruitment and retention of key employees (Hoyman & Duer, 2004).

We therefore speculate that employees who receive life-oriented benefits feel stronger obligations to return the investment. For organizations, the advantages of work-oriented programs can be direct and immediate, realized by employees performing tasks more efficiently and effectively, whereas the advantages of life-oriented programs are rather indirect and often require a certain period of time to be realized. In a similar vein, Caillier (2013b) stated that providing child care and elder care programs “may be viewed as an act of goodwill rather than a way to get employees to work harder, which in turn may make employees feel more obligated to reciprocate by demonstrating loyalty to the organizations” (p. 359).

To examine the effects of work- and life-oriented programs, we focus on three outcome variables: job satisfaction, organizational satisfaction, and affective commitment. First, job satisfaction refers to an individual’s affective reactions to a job, which result from “the appraisal of one’s job or job experiences” (Locke, 1976, p. 1304). Job satisfaction is an important consequence of WLPs because participating in WLPs can significantly influence individuals’ work environments and experiences, including schedules, locations, and stress levels. In addition, job satisfaction has been widely used as an outcome variable in previous research on WLPs (T. D. Allen, 2001; Caillier, 2012; Facer & Wadsworth, 2008; J. Kim & Wiggins, 2011; Saltzstein et al., 2001). Second, organizational satisfaction is similarly defined as an employee’s subjective response to an organization. Based on social exchange theory, Lambert (2000) argued that “workers differentiate between multiple partners of exchange in the workplace—coworkers, supervisor, organization—and aim their efforts to reciprocate toward a particular partner” (p. 802). Given that WLPs are part of the HR practices provided by an organization, employees’ obligations to repay are likely directed to the organization and fulfilled in the form of organizational satisfaction. Last, affective commitment is defined as “employees’ emotional attachment to, identification with, and involvement in the organization” (N. J. Allen & Meyer, 1990). As with organizational satisfaction, when employees receive additional support from the organization, one way they reciprocate is to increase the level of their commitment toward the organization.

In sum, this study attempts to explore whether and how employees respond to work- and life-oriented programs in similar or different ways. Drawing on social exchange theory, we examine to what extent and in which ways disparate types of WLPs lead to social exchange relationships between employee and organization. Specifically, we expect life-oriented programs to be positively related to job satisfaction, organizational satisfaction, and affective commitment, while such relationships are less positive for work-oriented programs.

MDM

A quasi-experimental method called MDM is utilized in this study. Quasi-experimental methods are particularly useful for examining the effect of a particular program or intervention (i.e., “treatment”) when a randomized experiment is not feasible (Heinrich, Maffioli, & Vázquez, 2010; Morgan & Winship, 2007; Stuart, 2010), especially in a study such as this in which federal employees receive work–life benefits without first being randomly assigned. Even if random assignments were possible, it would be ethically undesirable to assign people into different WLP conditions while disregarding their personal attributes, needs, and motivations. Instead, quasi-experimental methods can be used to produce a plausible evaluation of the impacts of WLPs on work outcomes when individuals are not randomly assigned into conditions, because a reasonable comparison group can be identified that is not affected by the program. Specifically, matching methods reduce selection bias in nonexperimental data by finding nonparticipating individuals who are most similar to each program participant based on observable characteristics (Stuart, 2010).

The fundamental idea of matching is to find hypothetical counterfactuals (i.e., comparison units) by selecting untreated units that are most comparable to treated units through a strict matching process using observable, pretreatment attributes. Different matching methods rely on different metrics to identify the closest untreated units for each treated unit. For example, propensity score matching (PSM), one of the most commonly used matching methods, computes the probability of receiving a binary treatment (i.e., the propensity score) conditional on observable, pretreatment characteristics (i.e., matching covariates), which is defined as Pr ( T i = 1 | X ) , where T denotes treatment status and X is a vector of matching covariates (Dehejia & Wahba, 2002). The distance between treated and untreated units is calculated as the absolute difference between propensity scores. MDM, another long-standing matching method, uses a metric called the Mahalanobis distance that measures the distance between treated unit i and untreated unit j as ( X i − X j ) T S − 1 ( X i − X j ) , where S is the sample covariance matrix of the original data matrix X (Stuart, 2010).

Recently, scholars have suggested that MDM has advantages over PSM (King & Nielsen, 2015). The primary benefit of PSM is the reduction of dimensionality as it uses a single scalar, the propensity score, to measure the distance between treated and untreated units. However, matching of propensity scores does not guarantee zero, or nearly zero, imbalance in all of the covariates, which may increase selection bias. MDM measures the distance between the two units using the full vector of all matching covariates so that the imbalance is always zero. Based on a series of simulations, King and Nielsen (2015) concluded that “the potential for bias with PSM dramatically grows even while the bias under MDM monotonically declines as we would expect and desire” (p. 19).

In the matching process, untreated units are selected for each treated unit so that the distance between them is as small as possible. Once units are matched and checked for balance in terms of observable characteristics, the selected treated and untreated units serve as treatment and comparison groups, respectively, and the observed outcomes of the comparison group are used to estimate the potential outcomes of the treatment group under the comparison condition. Thus, the average difference in the outcome of interest between the two groups is an estimate of the impact of being treated by the program, which is referred to as the average treatment effect on the treated (ATT).

In our analysis, to ensure equivalence between the treatment and comparison groups after matching, we checked whether the two groups were balanced in regard to the matching covariates. The mean scores of each of the covariates for the treatment and comparison groups were compared. We found that, before matching, the unmatched treatment and comparison groups were significantly different from each other on most of the covariates. After matching, however, the differences in the mean scores were not significantly different from zero for all of the covariates. These results suggest that the matching process was completed successfully, in that the treatment and comparison groups were balanced on all matching covariates.

Data and Variables

Variables

Survey items used in the analysis are presented in the “Appendix” section. Descriptive statistics of the variables are reported in Table 1.

OPEN IN VIEWER

Table 1.

Descriptive Statistics.

Variable	M	Minimum	Maximum
Telework	0.40	0	1
Alternative work schedules	0.43	0	1
Child care programs	0.03	0	1
Elder care programs	0.02	0	1
Job satisfaction	3.89	1	5
Organizational satisfaction	3.73	1	5
Affective commitment	4.04	1	5
Supervisory support	4.13	1	5
Gender (1 = female; 0 = male)	0.46	0	1
Minority status (1 = minority; 0 = nonminority)	0.33	0	1
Age group
29 and under	0.06	0	1
30-39	0.15	0	1
40-49	0.29	0	1
50-59	0.37	0	1
60 or older	0.13	0	1
Supervisory level
Nonsupervisors and team leaders	0.72	0	1
Supervisors	0.17	0	1
Managers and executives	0.11	0	1
Work location (1 = field; 0 = headquarters)	0.54	0	1
Pay category
Federal wage system	0.03	0	1
GS 1-6	0.04	0	1
GS 7-12	0.37	0	1
GS 13-15	0.48	0	1
SES/SL/ST/Other	0.08	0	1
Federal tenure, years
Up to 3	0.15	0	1
4-5	0.08	0	1
6-10	0.16	0	1
11-14	0.10	0	1
15-20	0.11	0	1
More than 20	0.39	0	1
Agency tenure, years
Up to 3	0.21	0	1
4-5	0.10	0	1
6-10	0.19	0	1
11-20	0.21	0	1
More than 20	0.29	0	1

Note. WLPs are measured as a binary variable indicating participation status (1 = participated; 0 = not participated). Supervisory support, job satisfaction, and organizational satisfaction are measured using a 5-point ordinal scale. Affective commitment is measured using a summated rating scale of three items. All other variables are dummy-coded. N = 153,702; GS = General Schedule; SES = Senior Executive Service; SL = Senior Level; ST = Scientific or Professional; WLP = work–life programs.

Treatment variables

The following four variables were used as the treatment variables: telework, alternative work schedules, child care, and elder care. The treatment status refers to whether the respondent participated in a particular WLP and is coded as a binary variable in which 1 indicates participation and 0 refers to nonparticipation. The response categories for the treatment variables, except telework, included “yes,” “no,” and “not available to me.” “Yes” was coded as 1 and “no” was coded as 0. “Not available to me” was omitted. To check the robustness of the findings, we also used another operationalization of these variables by coding “yes” as 1 and “no” and “not available to me” as 0.

For telework participation, the respondents were asked to choose one of eight response categories to best describe their current telework situations (see Appendix). Those who did not telework were coded as 0 and those who teleworked, whether frequently or infrequently, were coded as 1. We also conducted a robustness check using multiple operationalizations of telework participation to examine whether participation had varying effects depending on its frequency. This result is discussed in detail in the “Results” section.

Outcome variables

The three outcome variables used in this study were job satisfaction, organizational satisfaction, and affective commitment. For job satisfaction, we used a single global measure that refers to overall satisfaction with one’s job. Another single global measure was used for organizational satisfaction to represent overall satisfaction with one’s organization. Single-item measures of satisfaction have been shown to be highly and significantly correlated with multiitem measures and have proper reliability as well as face and construct validity (Wanous, Reichers, & Hudy, 1997). Affective commitment was measured as a summated rating scale of three items that measured “personal involvement,” “shared values,” and “identity-relevance” (Moldogaziev & Silvia, 2015, p. 562). Cronbach’s alpha for affective commitment was .792. All outcome variables were standardized before analysis.

Matching covariates

The respondents’ demographic and work characteristics were used as covariates in the matching process. The matching covariates included gender, minority status, age group, supervisory level, work location, pay category, and federal and agency tenure, all of which had categorical responses. Dummy variables were created if the response categories had more than two options. In addition, supervisory support for work–life balance was used as a matching covariate. These variables have been commonly used in the WLP literature as antecedents of or control variables for employee perceptions and use of WLPs (T. D. Allen, 2001; Caillier, 2013b; Facer & Wadsworth, 2008; J. Kim & Wiggins, 2011; Saltzstein et al., 2001).

Results

The results from MDM are presented in Table 2. First, we focus on the effects of work-oriented programs. Participating in telework does not have a significant impact on job satisfaction (ATT = −0.023, ns), organizational satisfaction (ATT = −0.034, ns), or affective commitment (ATT = −0.004, ns), which indicates that individuals who participated in telework do not show significantly different levels of job satisfaction, organizational satisfaction, or affective commitment from those who did not participate. Similarly, the effects of participating in alternative work schedules are not statistically significant. Participating in alternative work schedules does not have a significant impact on job satisfaction (ATT = −0.017, ns), organizational satisfaction (ATT = −0.016, ns), or affective commitment (ATT = −0.020, ns). The findings demonstrate that employees who participated in alternative work schedules do not report significantly different levels of job satisfaction, organizational satisfaction, or affective commitment from employees who did not participate.

OPEN IN VIEWER

Table 2.

Results From Mahalanobis Distance Matching: WLP Participants Versus Nonparticipants.

	ATT
	Job satisfaction	Organizational satisfaction	Affective commitment
Participating in work-oriented programs
Telework	−0.023	−0.034	−0.004
	(0.024)	(0.024)	(0.029)
Alternative work schedules	−0.017	−0.016	−0.020
	(0.022)	(0.021)	(0.022)
Participating in life-oriented programs
Child care programs	0.098***	0.121***	0.122***
	(0.026)	(0.027)	(0.027)
Elder care programs	0.094*	0.140***	0.070*
	(0.037)	(0.034)	(0.032)

Note. Standard errors are in parentheses. All outcome variables are standardized. WLP = work–life programs; ATT = average treatment effect on the treated.

*

p < .05. **p < .01. ***p < .001.

Next, we examine the effects of life-oriented programs on the outcome variables (see Table 2). The effects of participating in child care programs are positive and significant for job satisfaction (ATT = 0.098, p < .001), organizational satisfaction (ATT = 0.121, p < .001), and affective commitment (ATT = 0.122, p < .001), which suggests that employees who participated in child care programs are 0.098 standard deviations (SDs) more likely to be satisfied with their jobs, 0.121 SDs more likely to be satisfied with their organizations, and 0.122 SDs more likely to be committed to their organizations than those who did not participate. Finally, the use of elder care programs also leads to greater job satisfaction (ATT = 0.094, p < .05), organizational satisfaction (ATT = 0.140, p < .001), and affective commitment (ATT = 0.070, p < .05). The levels of job satisfaction, organizational satisfaction, and affective commitment for individuals who participated in elder care programs are 0.094, 0.140, and 0.070 SDs higher, respectively, than for those who did not participate. Unlike work-oriented programs, all life-oriented programs significantly increased the levels of employees’ job satisfaction, organizational satisfaction, and affective commitment, congruent with our expectations. The results are consistent when operationalization of the treatment variables is modified by coding the response “not available to me” as 0. For an additional robustness check, agency dummies are included in the analysis as matching covariates, and the results remain consistent.

As mentioned earlier, we initially operationalized telework participation as a binary treatment variable that indicated whether or not employees participated in telework. However, telework participation had several response categories representing different frequencies, ranging from infrequently or on a short-term basis (e.g., less than once a month) to frequently (e.g., more than 3 times per week). As such, employees who teleworked more frequently might have had different attitudes and behaviors from those who teleworked infrequently. Thus, it would have been problematic to treat all of these responses as one group labeled “telework participants” and compare them with nonparticipants. As a robustness check, we conducted additional matching to examine the effects of different degrees of telework participation on the outcome variables.

Table 3 shows that participating in telework programs has an insignificant relationship with job satisfaction, organizational satisfaction, and affective commitment, regardless of frequency, with the exception being employees who teleworked 3 days or more per week, who had significantly higher levels of job satisfaction than those who did not telework. Golden and Veiga (2005) argued that the positive impact of teleworking could be reduced by extensive participation, due to feelings of isolation and lack of face-to-face interactions with colleagues when performing tasks. In other words, the relationship between teleworking and employee work outcomes would have an inverted U-shape in which the positive impact of teleworking peaks at a certain amount of teleworking and becomes smaller or even negative. However, we do not find supportive evidence for this argument, and our results suggest that only those who telework very frequently may benefit from teleworking, in the sense that it increases levels of satisfaction with their jobs.

OPEN IN VIEWER

Table 3.

Results From Mahalanobis Distance Matching: Teleworkers Versus Nonteleworkers.

	ATT
	Job satisfaction	Organizational satisfaction	Affective commitment
Telework at least 3 days per week	0.084**	0.062	0.004
	(0.030)	(0.033)	(0.034)
Telework at least 1 day per week	−0.027	−0.006	−0.053
	(0.028)	(0.027)	(0.028)
Telework at least 1 day per month	−0.017	−0.025	−0.045
	(0.030)	(0.039)	(0.031)
Telework at least on a short-term basis	−0.023	−0.034	−0.004
	(0.024)	(0.024)	(0.029)

Note. Standard errors are in parentheses. All outcome variables are standardized. ATT = average treatment effect on the treated.

*

p < .05. **p < .01. ***p < .001.

Discussion and Conclusion

In this study, we examine the effects of WLP participation on employee satisfaction and commitment using matching methods. Grounding our theory in Hoyman and Duer’s (2004) typology of workplace policies, we propose that WLPs can be categorized into work- and life-oriented programs and that the effects of WLPs on employee attitudes vary by type of program. Drawing on social exchange theory, we investigate how different types of WLPs may generate dissimilar levels of social exchange between employee and organization. Our findings show participating in work-oriented programs does not have a significant influence on job satisfaction, organizational satisfaction, or affective commitment; however, participating in life-oriented programs is significantly and positively related to all of the outcome variables.

This study contributes to the literature on WLPs in several ways, both theoretically and methodologically. First, we expand the existing conceptualization of WLPs by suggesting two different types of WLPs. Previous studies have produced a sizable amount of empirical evidence regarding the antecedents and consequences of WLPs. In the vast majority of them, however, it is implicitly assumed WLPs have unidimensional benefits, and oftentimes, researchers have simply combined dissimilar WLPs into a single variable, preventing investigation of the distinct mechanisms of each program. They have not taken into account that these programs can be divided into multiple types and, therefore, have not paid much attention to how individuals perceive and respond to various forms of WLPs. Our study is one of the first to empirically examine the potential differential effects of WLPs based on their orientations in work and nonwork domains.

The findings of this study also provide additional insight into the social exchange framework that has been widely used to explain the outcomes of WLPs. Regardless of their orientations, WLPs aim to provide employees with certain resources—tangible or intangible—that they can utilize to balance their work and personal lives. Social exchange theory suggests that individuals who participate in WLPs react positively to such benefits and become willing to repay the organization with improved satisfaction and commitment. However, scholars have not given much attention to the possibility that different levels of social exchange may develop depending on the type of WLP. This study aims to expand the existing theoretical framework based on social exchange theory by arguing that different forms of benefits are associated with dissimilar levels of social exchange. Specifically, employees are likely to perceive greater benefits and feel stronger obligations to return the favor after participating in life-oriented programs than in work-oriented programs. More research should be conducted to investigate the mechanisms by which employees choose the level of reciprocation for work- and life-oriented programs.

In this study, we use a quasi-experimental approach to reduce selection bias in estimating the effects of WLP participation on employee satisfaction and commitment. Participation in WLPs is voluntary in most situations, and therefore, treatment assignment is not randomly determined. There may exist systematic differences between participants and nonparticipants, which introduces the potential for selection bias. Matching methods help resolve such problems by enabling researchers to choose well-matched samples based on observable characteristics and compare how outcomes differ for participants and nonparticipants (Heinrich et al., 2010). Another advantage of matching is that the program impact is estimated within the region of common support in which there is sufficient overlap between treated and untreated groups. Compared with matching, traditional regression models do not explicitly clarify the region of common support, and thus, biased results may be obtained if the treated and untreated groups are very different in terms of the means and variances of covariates (Rubin, 2001; Rubin & Thomas, 2000; Stuart, 2010).

To our knowledge, this study is one of the few studies in the field of public administration that employ matching methods (Bhatti, Gørtz, & Pedersen, 2015; Meier & O’Toole, 2013a, 2013b). Given that experimental methods are often not feasible in many social science research situations, this method can be a valuable tool for public administration and public policy research to resolve potential problems in observational data, such as selection bias. For example, matching is applicable to evaluations of effectiveness for a range of HR policies and practices. Using matching methods, employees who are affected by a specific HR policy (e.g., a new incentive plan) can be systematically compared with unaffected employees who are most similar to estimate the impact of the policy. In addition, large databases available in the public sector, such as the FEVS and the Merit Principles Survey, allow more accurate evaluation of a program’s impact as they ensure a sufficient number of observations in the region of common support to construct counterfactuals. Thus, we encourage public administration researchers to consider using matching methods to evaluate the impact of a particular program or policy.