Work-family conflict: a model of linkages between work and family domain variables and turnover intentions.

Researchers have studied many outcome variables of work-family conflict (WFC) and family-work conflict (FWC), such as depression (Frone et al., 1992a), family satisfaction (Beutell and Wittig-Berman, 1999), heavy alcohol use (Frone et al., 1996), and job satisfaction (Netemeyer et al., 1996). However, relatively few have specifically examined withdrawal. While intention to quit (Burke, 1988; Netemeyer et al., 1996) and absenteeism (Goff et al., 1990) have been linked to WFC, there is some question about the generalizability of current findings. For instance, Burke (1988) used a global measure of work-family conflict and Netemeyer et al. (1996) only considered correlations. The purpose of this study is to address these issues by testing a model of work and family variables leading to conflict and, ultimately, turnover intentions.

Gaps in the Research

Work-family conflict is a form of interrole conflict that occurs when pressures associated with membership in one role interferes with membership in another (Kahn et al., 1964). It is defined as "a form of interrole conflict in which the role pressures from the work and family domains are mutually incompatible in some respect" (Greenhaus and Beutell, 1985: 77). Research in the area of work-family conflict, while informative, still has shortcomings that have yet to be addressed. In order to advance this stream of research, more consistency in the literature is needed. Comparisons between studies are still limited because some researchers continue to use a global measure of work-family conflict rather than two separate variables. The following section details important gaps in the literature that will be addressed in the current study.

First, researchers have shown that WFC and FWC are distinct constructs with discriminant validity (e.g., Gutek et al., 1991; Kossek and Ozeki, 1998; Netemeyer et al., 1996). While some researchers have adopted the use of two independent measures to capture work interfering with family conflict (WFC) and family interfering with work conflict (FWC) (e.g., Carlson et al., 2000; Frone et al., 1992a; Frone et al., 1996; Gutek et al., 1991; Netemeyer et al., 1996), recently published research continues to use a global measure of work-family conflict (e.g., Carlson and Perrewe, 1999; Greenhaus et al., 1997; Parasuraman and Simmers, 2001; Yang et al., 2000). By measuring WFC and FWC separately, we have the opportunity to see how work domain variables influence WFC and how family domain variables influence FWC (Frone et al., 1996; Gutek et al., 1991; Kossek and Ozeki, 1998).

Second, few studies examine full measurement models. Work-family conflict studies using structural equation modeling often consider a structural model and use summated scales (see Carlson and Kacmar, 2000). These methods only estimate error; they do not model all of the theorized relationships (i.e., observed and latent). By creating an average of the latent construct, they are creating a single manifest indicator. Using a full measurement model and structural model is more rigorous and accounts for measurement error above and beyond a structural model and is the recommended approach (Anderson and Gerbing, 1988). Further, simultaneously assessing the measurement and structural models provides a more thorough assessment of construct validity (Bentler, 1978). It also allows for the opportunity to use the preferred two-step modeling approach (Anderson and Gerbing, 1988). By first confirming the measurement model in evaluating a priori relationships, theory can be tested and confirmed in the second step (Anderson and Gerbing, 1988). We estimate and fix the measurement and test the structural model in the interest of using this two-step approach.

Third, a large number of studies have followed the suggestions of researchers (i.e., Frone et al., 1992a; Kopelman et al., 1983) to consider only a subgroup (e.g., those married or having children) of the population of workers (Greenhaus et al., 2001). In some cases, researchers only considered those employees with children under the age of 18 (Frone, 2000) or those with children living in the household (Netemeyer et al., 1996) or where both the subject and their spouse were professionals (Higgins et al., 1992), rather than the entire population. No rationale exists for using such groupings, except for that it was the selection criteria chosen. We believe these subgroupings may have limited generalizability to the participating organization and fail to account for the population of employees as a whole. Further, all employees have the potential to experience work-family conflict and should not be excluded because they are unmarried, do not have children living at home, or their spouse does not work. In fact, single parent households are increasing (U.S. Bureau of Labor Statistics: USDL 00-172) and may be overlooked by researchers. In order to increase the generalizability of our research findings, we do not limit our sample.

Fourth, researchers have considered various unique antecedents to FWC (e.g., Frone et al., 1992a); however, only a few researchers have assessed the relationship between family responsibility and FWC. For example, Goff et al. (1990) asked about primary child-care responsibility, yet used a global measure of work-family conflict. Other researchers have assessed family responsibility using only the number of children (Frone et al., 1997b), but neglected other dependent relationships, such as having an aging parent or a sibling's children living within the household (Kinnunen and Mauno, 1998). Measurement of family responsibility has been deficient (Rothausen, 1999). Capturing the number of individuals living within a household may provide additional insight into one's level of family responsibility. We more fully assess the construct of family responsibility and its relationship with FWC.

Fifth, while many studies assess the relationship between WFC and FWC using correlation analysis, few consider causal paths between the two. Carlson and Kacmar (2000) found a reciprocal and positive relationship between WFC and FWC. It could be argued that a spillover effect exists from WFC to FWC and vice versa. However, they did not utilize a full measurement model nor did their model consider withdrawal behaviors. Therefore, we assess the causal link between WFC and FWC.

Finally, there are problems with current research examining the relationship between WFC and FWC on turnover intentions that may limit generalizability. Burke (1988) and Good et al. (1996) showed that WFC has a positive relationship with turnover intentions, yet used a global measure of work-family conflict. Netemeyer et al. (1996) examined both WFC and FWC and found they were each positively correlated with turnover intentions in 2 out of 3 samples. However, they only considered bi-variate relationships without controlling for the effects of the other type of conflict. Barling et al. (1994) found a significant relationship between WFC and psychological withdrawal from work. Greenhaus et al. (2001) considered withdrawal from a profession with both WFC and FWC and found a significant relationship for WFC and not for FWC. However, their method does not allow for the analysis of turnover behavior because employees might have quit the organization, yet planned to stay in the profession.

In summary, this study will address current gaps in work-family conflict research. Specifically, work-family conflict is captured using two separate measures, WFC and FWC, in a full measurement and structural model where we assess paths between work stress and WFC and family responsibility and FWC. Also, we clarify and develop a measure of family responsibility. The causal paths between WFC and FWC will also be considered. Additionally, we consider relationships with turnover intentions for both WFC and FWC.

Proposed Model

Work Stress and WFC

Work stress is usually conceptualized as work-role conflict, work-role overload, and work-role ambiguity (Cooke and Rousseau, 1984; House et al., 1983; Rizzo et al., 1970). Each has the potential to affect WFC (Bacharach et al., 1991). With respect to work roles conflict, the more conflict among work roles the greater the chances that stress will spill over and cause negative behaviors that interfere with fulfilling family roles (Greenhaus et al., 1987). Role overload is the result of having too many things to do in a given time period (Bacharach et al., 1991). As time is constrained by having too many tasks to accomplish at work, the employee may need to use time allocated to the family role. This could cause work family conflict (Greenhaus and Beutell, 1985). Work-role ambiguity occurs when workers are unsure of what is expected of them in a work role. As uncertainty concerning work-roles increase, employees use more mental energy to decipher it. This requirement may drain mental energy and attention needed for their family roles. Carlson and Kacmar (2000) found that role overload and role conflict were predictors of WFC, yet did not find significant results for role ambiguity. They did not address reasons for this null finding. This result could be sample-specific. Thus, we offer the following hypothesis:

H1. Work-role conflict, work-role overload,

and work-role ambiguity will be positively

related to WFC.

Family Responsibility and FWC