Men's and women's eldercare-based work-family conflict: antecedents and work-related outcomes.

Role theory was used to develop a model examining the antecedents and work-related outcomes of eldercare-based work-family conflict. This model was then tested with a path analysis framework. Proposed antecedents to Family Interference with Work (FIW) include time spent providing eldercare, marital status, and gender. Proposed antecedents to Work Interference with Family (WIF) include gender, hours worked, flexible work arrangements, family-related supervisor support, and supportive workplace culture. The work-related outcomes explored were partial absence and intention to seek new employment. For this combined model, nine hypotheses were advanced, and four were supported. Next, the model was tested separately for men and women to begin to address the understudied question of how men's responsibilities as sons relate to their work-family conflict. For example, separate model findings suggest that supervisor support is unrelated to WIF, for men, and that work-family conflict is related to partial absence for men but to intention to seek new employment for women.

Keywords: work-family conflict, men in families, eldercare

**********

Role theory (Katz & Kahn, 1978) was used in the present study to examine two different forms of interrole conflict. Interrole conflict occurs when "role pressures associated with membership in one organization are in conflict with pressures stemming from membership in other groups" (Kahn, Wolfe, Quinn, Snoek, & Rosenthal, 1964, p.20). Work-family conflict is a specific type of interrole conflict. Work-family conflict refers to the perceived incompatibility of role pressures between work and family so that participation in one role makes participation in the other role more difficult (Greenhaus & Beutell, 1985). Implicit within the definition are two types of work-family conflict, work interference with family (WIF) and family interference with work (FIW).

Our goal in conducting this research was to use role theory to introduce a conceptual model of the antecedents and work-related outcomes associated with the work-family conflict experienced by employed men and women who also provide care to the elderly. We recognize that family-related outcomes are also important and worthy of investigation; however, such outcomes were not included in the present study. Through this work, we sought to enhance the professional work-family conflict literature by specifically examining those employed adults who also provide eldercare. Additionally, we sought to enhance the literature on men's family responsibilities and work-family conflict by comparing elder-care providing men and women.

CHANGING POPULATION DEMOGRAPHICS

Several key changes in society are increasing the need to understand how employees balance eldercare responsibilities with work demands. First, the United States is undergoing a change in population demographics, with the elderly (specifically, those 80 years old and older) representing the fastest growing segment of the population (Federal Interagency Forum on Aging Related Statistics, 2000). Unfortunately for many older Americans, "... with increasing age there is greater likelihood of disability and dependence" (Halpern & Deck, 1989, p. 10).

Second, women have traditionally taken on the role of primary caregiver within the family infrastructure, as evidenced by the statistic that 75% of all informal caregivers are women (Ettner, 1995). Yet the number of unemployed women available to provide care to an elderly friend, relative, or parent is on the decline. As women continue to have a dominant presence in the workforce and as the number of older Americans increases, the number of female caregivers balancing work and family responsibilities will inevitably increase. It is unclear at this point what ramifications these demographic changes might have for men. However, it is possible that men's eldercare provision will increase much as their child care provision has over the past two decades (see Coltrane, 1998).

As the demand for eldercare provision increases and potentially requires more men to provide this care, it will be important to evaluate the impact on active fathering among these men who are "sandwiched" between the needs of their elderly parents and their children. Nichols and Junk (1997) indicate that in a random sample of individuals between the ages of 40 and 65 in Idaho, Oregon, Utah, and Michigan, !L5 percent reported that they provided care for aging parents and financially dependent children. Furthermore, the type of eldercare generally needed includes hands-on tasks such as providing transportation and going shopping rather than assisting with financial needs that might be more readily handled remotely. Thus, ironically, at just the point when active, involved fathering including engagement and accessibility is being promoted (Lamb, Pleck, Chamov, & Levine, 1985), fathers may now be proximally needed elsewhere as well.

Although there is very little research to date on eldercare providing fathers, research suggests that people filling multiple caregiving roles report having little time for themselves (Stern & Wolford, 2001). Zal (1992) suggests that these parents and caregivers experience difficulty maintaining a separate identity amidst their roles of spouse, child, and parent. Additionally, evidence from outside of the eldercare literature suggests a troubling mechanism through which youth might be negatively affected by increased eldercare provision by fathers. Crouter, Bumpus, Maguire, and McHale (1999) reported that youth psychological adjustment was predicted (negatively) by parental role overload, but that fathers' work pressure was related to both parents' reporting parental-role overload, while mother's work pressure was related to only her own parental-role overload. In other words, it could be that as fathers take on multiple care-providing roles their resulting work-family conflict affects both their fathering and their significant others' mothering.

These social trends fuel the need to provide a more in-depth examination of the work/elder caregiving relationship, especially as it impacts men. Gorey, Rice, and Brice (1992) estimate that between 7.4% and 11.8% of eldercare providers are employed and the number of employed eldercare providers is projected to continue to increase as our population ages (Kossek, DeMarr, Backman, & Kollar, 1993). Thus, it is imperative that social scientists better understand how eldercare responsibility and the role conflict that often accompanies it affect work and vice versa.

To that end, a conceptual model based on role theory is presented, and hypotheses are advanced regarding how this relationship influences work-related outcomes. Subsequently, the model is tested separately for men and women to elucidate and explore linkages within the model that are unique to men. Future studies will go beyond the work-related outcomes considered here to consider how women's and men's eldercare-based work-family conflict relates to family-related outcomes.

THE CONCEPTUAL MODEL

The conceptual model as presented in Figure 1 will be explained from left to right, with the components broken down into the following subsections: (A) Antecedents to Work-Family Conflict, (B) Work-Family Conflict, and (C) Work-Related Outcomes of Work-Family Conflict: Withdrawal Behaviors. Since eldercare is a relatively new area of research in the work and family domain, literature from the general caregiving, general work-family conflict (i.e., no focus on caregivers), and childcare based work-family conflict domains will be used when eldercare literature on the topic is not available.

FAMILY VARIABLE ANTECEDENTS TO WORK-FAMILY CONFLICT

The magnitude of work-family conflict perceived by a person providing care to an elderly parent, relative, or friend does not necessarily fall equally on all caregivers. Family variables or characteristics of the caregiver, anticipated to influence work-family conflict include time spent providing eldercare per week, marital status, and gender. Each characteristic, along with its hypothesized relationship as an antecedent to work-family conflict, is developed in the following subsections.

Time spent providing eldercare. Hepburn and Barling (1996) found both the number of hours "providing care for my parent" and "interacting with my parent" to have significant path coefficients to both parent versus work interrole conflict and work versus parent interrole conflict (p. 313). Thus, the following hypothesis is proposed.

Hypothesis 1: The total time spent providing eldercare per week will be positively related to family interference with work.

Marital status. Greenhaus and Beutell (1985) suggest that the increased time demands at home experienced by married individuals may lead them to experience more work-family conflict than unmarried individuals. Maraist (1999) notes that research to date has produced mixed results. Maraist's (1999) meta-analytic investigation did not find married individuals to experience more work-family conflict than unmarried individuals. Maraist's (1999) meta-analysis, however, did not specifically focus on caregivers. Thus, further investigation is necessary with specific focus on eldercare providers. As a result, the following exploratory hypothesis is proposed.

Hypothesis 2: Married eldercare providers will experience more family interference with work than unmarried eldercare providers.

Gender. A good portion of the caregiving literature has examined whether gender differences exist in the provision of care. It is widely accepted that women generally provide more overall caregiving assistance than men (Ettner, 1995; Stoller, 1983). Stone, Cafferata, and Sangl (1987) add support to previous researchers through their finding that 72% of caregivers to the frail elderly are women.

Stoller (1983) found that the impact of employment on the amount of time spent caring for an elderly parent was different for sons and daughters. Employed sons provided 22.9 fewer hours of assistance than unemployed sons per month, while employment did not affect the amount of care provided by daughters. In addition, many researchers have proposed that women will perceive greater family interference with work than men (Gignac, Kelloway, & Gottlieb, 1996; Gutek, Searle, & Klepa, 1991; Hochschild, 1989). Thus, the following hypothesis is proposed.

Hypothesis 3: Women will report more work-family conflict (WIF and FIW) than men.

Thus, with regard to differential patterns of work-family conflict for men and women providing eldercare, a review of existing literature supports only the hypothesis above, that women will report higher WIF and FIW than men. However, in an effort to more fully investigate how work interacts with men's family responsibilities in this understudied population of male eldercare providers, we provide a post-hoc exploratory investigation of male-female differences with regard to all hypothesized paths that do not have gender as their central focus.

WORK VARIABLE ANTECEDENTS TO WORK-FAMILY CONFLICT

Hours worked. Greenhaus and Buetell (1985) suggest that as the number of hours worked per week increases, the amount of work-family conflict perceived increases. This result has been further supported in a recent meta-analysis (Maraist, 1999) and work completed by Scharlaeh and Boyd (1989) as well as Neal and Hammer (2000). Thus, the following hypothesis is proposed.

Hypothesis 4: The number of hours worked per week will be positively related to work interference with family.

Flexible work arrangements. Maraist (1999) as well as Staines and Pleck (1984) found schedule flexibility to be negatively related to work-family conflict. Maraist's (1999) work was based on Greenhaus and Beutell's (1985) proposition that in addition to total hours worked, inflexible work schedules can produce work-family conflict. In addition, Smith, Buffardi, and Holt (1999) found that increased job flexibility is directly related to decreased work interference with eldercare. Thus, the following hypothesis is proposed.

Hypothesis 5: Flexible work arrangements will be negatively related to work interference with family.

Supervisor support. Wagner and Neal (1994) point out the need for more research examining the role of support and work-family conflict for eldercare providers. Thomas and Ganster (1995) found supervisor support to be negatively related to work-family conflict. Frone, Yardley, and Markel (1997) found supervisor support and work-family conflict to share an indirect relationship (through supervisor support's effect on work distress and work overload). Goff, Mount, and Jamison (1990) found that supervisor support predicted less overall work-family conflict in a sample of 253 parents of children five years old and younger. Thus, the following hypothesis is proposed.

Hypothesis 6: Family-related supervisor support will be negatively related to work interference with family.

Supportive workplace culture. Workplace culture provides a broader index of organizational support than supervisor support by tapping the "deep structure of organizations" (Denison, 1996, p. 624). Thompson, Beauvais, and Lyness (1999) found perceptions of a supportive work-family culture to be negatively related to work interference with family. Work-family culture explained an additional 18% of the variance in work to family conflict after controlling for demographic variables and benefit availability. In another study, Parker (1999) found that perceived army (i.e., organizational) support for family predicted unique variance in work-family confliict above and beyond demographic variables.

More specific to eldercare, Smith et al. (1999) found organizational support to be a direct predictor of work interference with eldercare (WIE). In summary, "a supportive culture should make it easier for employees to balance work and family demands and, as a result, experience less work-family conflict" (Thompson et al., 1999, pp. 397-398). Thus, the following hypothesis is proposed.

Hypothesis 7: Supportive workplace culture will be negatively related to work interference with family.

WORK-RELATED OUTCOMES OF WORK-FAMILY CONFLICT

A number of outcomes have been examined in relation to work-family conflict, including both nonwork and work-related factors. Examples of non-work-related outcomes include depression, stress, life dissatisfaction, marital dissatisfaction, and quality and quantity of family care provided (Netemeyer et al., 1996; O'Drisco]l, Ilgen, & Hildreth, 1992; Thomas & Ganster, 1995). Examples of work-related out comes studied include job dissatisfaction, job burnout, job performance, organizational commitment, turnover/absence, and intention to seek new employment (Ayree, 1992; Burke, 1988; Goff et al., 1990; Netemeyer et al., 1996; Shultz & Silverstein, 1999).

Again, we acknowledge the importance of non-work-related outcomes. However, in the present study, we focused on intention to seek new employment and partial absence. As Allen, Herst, Bruck, and Sutton (1999) note, "more research examining work-related behavioral outcomes is needed. Only a handful of studies have examined the effect of WFC on variables such as absenteeism ... and turnover" (p. 5).

Intention to seek new employment. Shultz and Silverstein (1999) point out "... little research is available that looks specifically at those providing eldercare and the potential predictors associated with their intentions to change jobs" (p. 4). Even less research is available examining the relationship between work-family conflict and intention to seek new employment. According to Rickard, Neal, Hammer, and Caubet (2000), "only a few studies in the work and family literature have included intention to quit one's job as a variable (e.g., Boles, Johnston, & Hair, 1997; Grandey & Cropanzano, 1999; Smith et al., 1999)" (p. 8).

Netemeyer et al. (1996) found intention to leave an organization to be correlated with both WFC and FWC. Boles et al. (1997) found a positive relationship between work-family conflict and emotional exhaustion as well as a positive relationship between emotional exhaustion and propensity to leave. Further, Boles et al. (! 997) found a negative relationship between work-family conflict and job satisfaction in addition to a negative relationship between job satisfaction and propensity to leave. Smith et al. (1999) found that interrole conflict, specifically work interference with eldercare, was positively related to turnover intentions, utilizing a sample of employees with eldercare responsibilities from the same Fortune 500 company. Thus, the following hypothesis is proposed.

Hypothesis 8: Work interference with family will be positively related to intention to seek new employment.

Partial absenteeism. Although caregiving duties are related to both full and partial absences (e.g., arriving late and leaving early), Hepburn and Barling (1996) believe partial absence to be "a more sensitive indicator of how work is affected by eldercare responsibilities" (p. 311). Full and partial absence may be separate phenomena, affected by eldercare responsibilities in different ways. Barling et al. (1994), for example, measured a correlation of only 0.18 between full absence (days missed) and partial absence (leaving work early, arriving late, and telephone use). Hepburn and Barling (1996) found interrole conflict to moderate the relationship between number of hours providing eldercare and partial absence. In particular, parent versus work interrole conflict (i.e., FIW) was positively related to partial absence. As a result, the following hypothesis is proposed.

Hypothesis 9: Family interference with work will be positively related to partial absence.

METHOD

PARTICIPANTS AND PROCEDURES

Data were analyzed from the 1997 National Study of the Changing Workforce (NSCW) conducted by the Families and Work Institute (see Bond, Galinsky, & Swanberg, 1998). The NSCW represents a national cross-section of U.S. workers (18 years or older). Between March 14 and July 27, 1997, 3,551 telephone interviews were completed. The present analyses draw participants from the 2,877 wage and salaried workers. Although the data set included information on self-employed persons, they were not included in the present study since self-employed persons can often set their own schedule and do not have direct supervisors. From the 2,877 respondents, the eldercare group included those individuals who answered yes to the question "Do you CURRENTLY provide special attention to someone 65 years old or older?" Thus, the final data set consisted of 329 respondents (11.44% of the original 2,877) with current eldercare responsibilities (see Bond et al., 1998, for detailed procedures). One hundred thirty-two (40.1%) of these eldercare providers are men while 197 (59.9%) are women. Comparative demographics concerning these male and female eldercare providers can be found in Table 1.

MEASURES

Means, standard deviations, intercorrelations, and alpha reliability values (as appropriate) for all variables in the combined and separate (by sex) models are presented in Tables 2 and 3, respectively.

Time spent providing eldercare. Time spent providing "personal care" was assessed by a one-item question asking participants to estimate the amount of time spent per week providing care in person, such as meal preparation, household work, physical care, transportation to medical services, and so on. Time spent providing "other care" was also assessed by a one-item question asking participants to estimate the amount of time spent per week doing other types of things, such as calling on the phone, arranging services, and handling finances. Total time spent providing eldercare represents the summation of time spent providing personal care and time spent providing other care.

Marital status. Marital status was assessed by one item, asking participants whether they are presently married, living with someone as a couple, single and never married, divorced, widowed, or separated. For the purpose of hypothesis testing, married and living with someone as a couple was coded 2 for married, with the remaining options being coded 1 for unmarried. Living with someone as a couple was considered the same as married, because these individuals are likely experiencing the same role demands as a legally married couple.

Gender. Gender was assessed and recorded by the telephone interviewer. The data were recorded as male, female, or not sure. Those coded "not sure" were excluded from the present analyses.

Hours Worked. The total number of hours worked per week was constructed from two items: number of hours worked at main job and number of hours worked at jobs other than main job.

Flexible work arrangements. This measure was assessed using five items, with mixed response scales. One item was measured with yes or no: "Are you allowed to choose your own starting and quitting times within some range of hours?" One item was answered with a four-point Likert type scale ranging from strongly agree to strongly disagree, and one item was measured with a four-option response ranging from not at all hard to very hard. The final item ("How much control would you say you have in scheduling your work hours?") had a five-option response ranging from complete control to none.

Supervisor support. Family-related supervisor support was assessed by five items utilizing a four-point Likert type scale ranging from strongly agree to strongly disagree. An example of an item from this scale is "My supervisor really cares about the effects that work demands have on my personal and family life."

Workplace culture. Family-supportive workplace culture was assessed by four items with a four-point Likert type scale ranging from strongly agree to strongly disagree. An example of an item from this scale is "There is an unwritten rule at my place of employment that you can't take care of family needs on company time."

Work-family conflict. The measure of work-family conflict consists of five items measuring family interference with work (FIW) and five items measuring work interference with family (WIF). All items asked respondents to frame their response in terms of the last three months. The responses for each item utilized a five-point frequency scale, where 1 = never and 5 = very often. For example, participants were asked to assess their response to questions such as "In the past three months, how often have you not had enough time for yourself because of your job?"

Intention to seek new employment. Intention to seek new employment was assessed using one item, which asked participants to rate their level of likelihood on a three-point scale ranging from very likely to not at all likely. Respondents were asked: "How likely is it you will make a genuine effort to find a new job with another employer in the next year?"

Partial absence. Partial absence was assessed with one item, which asked: "During the past three months, how many days did you start work late or stop early?"

DATA ANALYSES

OVERVIEW

We used a path analysis to test our conceptual model and all of the hypotheses simultaneously with the combined male/female sample and then separately for men and women. All analyses were run by analyzing the covariance matrix using the maximum likelihood method in LISREL 8.30 (Joreskog & Sorbom, 1993). The data used for the combined and separate analyses can be seen in Tables 2 and 3, respectively. Where possible (i.e., when a scale score was used as an indicator) reliability information was used for each construct to set measurement parameters. In setting the measurement parameters, the factor loadings were set equal to the square root of the reliabilities, whereas the error variances were set equal to the variance of the measure multiplied by the value (1--reliability). Reliability was assumed to be 1.00 for single item indicators.

ASSESSING AND COMPARING MODEL FITS

We assessed model fit using multiple fit statistics ([chi square], RMSEA, CFI, and GFI). A nonsignificant [chi square] indicates that the model fits the data well. However, because [chi square] is a direct function of sample size and will indicate a poor fit with models on large sample sizes, three additional fit indices were also used: the Comparative Fit Index (CFI; Bentler, 1990), the Root Mean Square Error (RMSEA; Steiger & Lind, 1980), and the Goodness of Fit Index (GFI; Joreskog & Sorborm, 1984). Although there are no objective rules for adequate model fit, an index of .90 or greater for the CFI and GFI is thought to represent a good fit (Hoyle, 1995; Mulaik et al., 1989), whereas an RMSEA of .08 to. 10 indicates mediocre fit, and a value less than .05 indicates good fit (MacCallum et al., 1996).

RESULTS

PRESCREENING OF DATA

Evaluation of the partial absence variable's distribution indicated that 247 respondents (75.08% of all respondents) reported starting work late or stopping early zero times in the prior three months. Due to the lack of variability and extreme skewness within this continuous variable, the item was dichotomized. For the dichotomized variable, respondents citing zero days partially missed were coded as zero, and respondents citing one or more days partially missed were coded as one.

DESCRIPTIVE STATISTICS

Descriptive statistics were utilized to compare the overall sample (N = 2,877) to the eldercare subsample (N = 329). Although some differences exist, overall the samples are comparable. For example, comparing the mean age of the two samples, the average age for the overall sample is 40.18 years (SD = 11.90), and the eldercare sample is similar at 41.4-8 years (SD = 10.73). In addition, a larger percentage of respondents in the eldercare sample have a four-year college degree, with 36.8% versus 30.1% in the overall sample. In terms of occupations, more respondents in the eldercare sample work in the service industry compared to the overall sample, 15.46% versus 10.67%, respectively. Again, male and female comparative demographic data are available in Table 1.

EVALUATION OF COMBINED MODEL HYPOTHESES

The model depicted in Figure 1 was run as a path analysis on the combined sample, and the fit indices indicated that the model fit the data adequately ([chi square] = 72.70, p < .05, RMSEA = .058, CFI = .85, GFI = .96). That is, the RMSEA and GFI indicated good fit while the CFI did not reach the .90 level. Thus, as a whole the fit indices indicated adequate fit of the model to the data. Since the fit indices suggested adequate fit, we proceeded to examine the individual parameter estimates of the model. The standardized path estimates between the constructs and their significance levels can be seen in Figure 2.

[FIGURES 1-2 OMITTED]

As can be seen from the model, for the overall sample, the pathways from time spent providing eldercare, marriage status, and gender to FIW (and with respect to gender also WIF) are all not significant. Thus, hypotheses 1-3 were not supported. Since the pathway from hours worked to WIF conflict was significant and positive, hypothesis 4 was supported. The pathway from flexible work arrangements to WIF conflict was not significant; thus hypothesis 5 was not supported. The pathways from supervisor support and workplace culture were both significant and negative. The negative relationship indicates that more supervisor support and "better" workplace culture is associated with less WIF conflict. Thus, hypotheses 6 and 7 were supported. The correlational pathway between WIF and FIW conflict was significant and positive. Examining the relationships between WIF, FIW, and the two outcome variables, one sees that the pathway from FIW to partial absences was not significant; thus hypothesis 9 was not supported. On the other hand, the pathway from WIF to intention to quit was significant; thus hypothesis 8 was supported.

EXPLORATORY INVESTIGATION OF SEPARATE MALE AND FEMALE MODELS

The model depicted in Figure 1 (absent the gender variable) was run as a path analysis for the male and female subsamples separately.

MALE SAMPLE

The fit indices indicated that the model fit the data adequately ([chi square] = 43.14, p = ns, RMSEA = .058, CFI = .84, GFI = .94). That is, the chi-square, RMSEA, and GFI indicated good fit while the CFI did not reach the .90 level. Thus, as a whole the fit indices indicated adequate but not superb fit of the model to the data. Since the fit indices suggested adequate fit, we proceeded to examine the individual parameter estimates of the model (see comparative results below). The standardized path estimates between the constructs and their significance levels can be seen in Figure 3.

[FIGURE 3 OMITTED]

FEMALE SAMPLE

The fit indices indicated that the model fit the data adequately ([chi square] = 47.86. p < .05, RMSEA = .059, CFI = .86, GFI = .95). That is, the RMSEA and GFI indicated good fit while the CFI did not reach the .90 level. Thus, as a whole the fit indices indicated adequate but not superb fit of the model to the data. Since the fit indices suggested adequate fit, we proceeded to examine the individual parameter estimates of the model (see comparative results below). The standardized path estimates between the constructs and their significance levels can be seen in Figure 4.

[FIGURE 4 OMITTED]

COMPARATIVE RESULTS

To examine whether any gender differences existed, we employed two strategies. The first strategy was to examine whether pathways were significant for one gender group versus another. Using this framework the results of the separate male and female models indicate that the pathway from marital status to FIW (hypothesis 2) was significant for men but not women, and the pathways from supervisor support and workplace culture to WIF (hypotheses 6 and 7) were significant for women but not men. With regard to the work-related outcomes of work-family conflict, the path from FIW to partial absence (hypothesis 9) was supported for men but not women, and the path from WIF to intention to seek new employment (hypothesis 8) was supported for women but not men.

However, inferring gender differences using this framework may be too liberal in that a pathway could be significant in one group and not significant in another, but the two pathways may not necessarily be significantly different from one another. Thus, the second strategy we used was to conduct t-tests comparing all unstandardized pathways between the male and female samples. One-tailed t-tests were used since the directionality of each comparison was known. Furthermore, we report significant differences at both the .05 and .10 level. The pathways from marital status to FIW [t(299) = 1.76, p < .05, one-tailed] and from WIF to seeking new employment [t(299) = 1.66, p < .05, one-tailed] were significantly different from one another at the .05 level. Specifically, the pathway from marital status to FIW (hypothesis 2) was significant for men but not women, and the pathway from WIF to intention to seek new employment (hypothesis 8) was supported for women but not men. At the .10 level two additional pathways become significantly different from one another: supervisor support to WIF [t(299) = 1.55, p < .10, one-tailed] and FIW to partial absence [t(299) = 1.56, p < .10, one-tailed]. Specifically, the pathway from supervisor support to WIF (hypotheses 6) was significant for women but not men and the pathway from FIW to partial absence (hypothesis 9) was supported for men but not women. However, the pathways between workplace culture and WIF (hypothesis 7) were not significantly different from one another in the two samples. Aside from this one differential finding (concerning hypothesis 7), the gender differences ascertained from the two strategies seem to overlap.

DISCUSSION

ANTECEDENTS TO WORK-FAMILY CONFLICT: COMBINED MODEL FAMILY VARIABLES

Total time spent providing eldercare per week was not related to FIW (hypothesis 1). This result was contrary to Hepburn and Barling's (1996) finding that both time spent providing care and time spent interacting with a parent had significant path coefficients to interrole conflict. One interpretation of these results is that FIW is not influenced by time constraints as hypothesized. It is evident that FIW is more than simply an outcome to a behavioral conflict (i.e., conflicting time demands). Since the number of hours providing eldercare per week was not related to FIW, there may be another factor that, along with the presence of eldercare responsibilities, results in FIW.

For example, Frone, Russell, and Cooper (1992) identified a relationship between family involvement and family-to-work conflict where family involvement was conceptualized as the degree to which the respondent's family was central to the respondent's self-concept. Identifying family responsibilities (of which eldercare responsibilities are a part) as central to one's self-concept involves much more than the gross number of hours invested in care per week. Such an interpretation would be in line with role theory (Katz & Kahn, 1978), discussed at the outset of this paper.

A second interpretation of the lack of significant relationships between time spent providing care and FIW involves adaptation. For example, Franklin, Ames, and King (1994) hypothesized that eldercare providers adapt their employment behaviors to provide care. Franklin et al. (1994) assessed employment behaviors', at two time periods, three months apart, and found that short-term work adjustments (defined as arriving late or leaving early, missing work without pay, taking sick or personal days, changing work hours, refusing a job or promotion, being kept from a job hunt or a better job, and other work effects) and leaves of absence were significantly less at Time 2. Perhaps this outcome represents a return to the balance of work and family obligations that existed before the caregiver assumed the eldercare role. It may mean the caregiver is adjusting so successfully to the eldercare role that she finds it less necessary to adapt employment (Franklin et al., 1994).

Results of hypothesis 2 indicate that married eldercare providers did not experience greater FIW than non-married eldercare providers. Married eldercare providers were expected to report greater FIW than unmarried eldercare providers because of the increased time demands associated with being married in addition to the time demands associated with providing care. The current result, however, adds evidence to Maraist's (1999) meta-analytic finding that married individuals do not experience more work-family conflict than unmarried individuals. Since Maraist's (1999) meta-analysis did not focus on caregivers, the present research aimed to examine the role of marital status and FIW in an elder caregiving population. It appears, however, that marital status was not significantly related to time demands or FIW as evidenced by the nonsignificant path coefficient in the combined model. It could be that this additional role stress is countered by couples' role sharing.

Contrary to previous literature (e.g., Gignac et al., 1996; Gutek et al., 1991) reporting greater levels of work-family conflict for women, women did not report more work-family conflict (WIF and FIW) than men (hypothesis 3). Similarly, Eagle, Miles, and Icenogle (1997) did not find gender differences in regard to the permeability of work and family boundaries. In other words, men and women reported near identical levels of work-to-family conflict and family-to-work conflict. As more female caregivers are members of the workforce than ever before, new roles as well as role exchanges have evolved for men and women. As men and women embody a variety of social roles (both family and work), the potential for interrole conflict increases for both genders, and both men and women must learn to manage their interrole conflict (Frone & Rice, 1987).

WORK VARIABLES

As anticipated, the total number of hours worked per week was positively related to WIF for the combined sample, supporting hypothesis 4. In 1983, Staines and Pleck suggested that since people usually have much less control over their work lives than their family lives, their family lives are often adjusted to accommodate their work lives. Following the logic of Staines and Pleck, the present study found that as the number of hours worked increased, the amount of WIF also increased. However, contrary to previous literature (e.g., Greenhaus & Beutell, 1985; Maraist, 1999; Smith et al., 1999; Staines & Pleck, 1984), flexible work arrangements were not related to WIF in the combined sample (Hypothesis 5). This may be due in part to the need to "see" and "be seen" at work. That is, regardless of whether workers have flextime available, individuals still may perceive the need to be present at the workplace at certain times, which may lead to WIF.

Supporting hypothesis 6, family-related supervisor support was negatively related to WIF. In other words, as supervisor support increased, WIF decreased. This finding extends the literature in the area of supervisor support and work-family conflict by supporting the proposed relationship in an eldercare provider population (Neale, Chapman, Ingersoll-Dayton, & Emlen, 1993).

As proposed, supportive workplace culture was negatively related to WIF for the combined sample (hypothesis 7). Eldercare providers who felt that their workplace was supportive of their commitment to their family responsibilities reported decreased WIF. The support for hypotheses 6 and 7 provides evidence of the impact that a worker's employment situation may have on his or her family life. In particular, the strongest association existed between supervisor support and WIF, with a standardized pathway of -.25. By being supportive of the eldercare provider's situation through supervisor support and workplace culture, the amount of WIF experienced by employed adults who also provide eldercare can be reduced.

WORK-FAMILY CONFLICT: COMBINED MODEL

Work-family conflict has typically been investigated as a unidimensional construct in much of the early work-family research. However, the nature of the construct lends itself to a bidirectional measure: work interference with family and family interference with work (Greenhaus & Beutell, 1985). "Failing to examine both types of WFC (work-family conflict) may limit our understanding of the work-family interface to the extent that each is associated with different antecedents and consequences" (Frone et al., 1992, p. 66).

Netemeyer et al. (1996) note that most research has only assessed general work-family conflict, failing to recognize the conceptual distinction. With respect to measurement issues, Kossek and Ozeki (1998) found that work-family conflict measures specifying direction (work to family or family to work) performed better than general or mixed measures. Similar results were obtained by Frone et al. (1992). A significant positive relationship was found between WIF and FIW in the present study. However, while increases in WIF were associated with increases in FIW, they did represent distinct constructs based on factor analytic results.

WORK-RELATED OUTCOMES OF WORK-FAMILY CONFLICT: COMBINED MODEL

As predicted, WIF was positively related to intention to seek new employment in the combined sample (hypothesis 8). That is, as the level of WIF increased for employed adults who provide eldercare, intention to seek new employment also increased. Employees may seek alternate work settings that provide them with better work schedules or arrangements to reduce the nonwork demands and the potential for work-nonwork conflict (Cohen, 1997).

Contrary to expectation, FIW was not related to partial absence (Hypothesis 9). Past research has documented the relationship between FIW and partial absence, causing the present result to be suspect. Because of an abundance of zeros reported in the partial absence variable (i.e., 247 respondents, or 75% of total respondents, did not report starting late or stopping early in the prior three-month period), the variable was dichotomized as no days partially missed and one or more days partially missed. Thus, the lack of variability in this population may be masking differences within another elder caregiving population (e.g., a population where eldercare responsibilities are new to the care providers; thus they have not adapted yet).

In addition, the lack of a relationship between FIW and partial absence may have been the result of the way in which the variable was operationalized in the current research. Past researchers have included arriving late, leaving early, time spent on the phone, and expanded lunch hours as measures of partial absence, yet only starting late or stopping early could be assessed in the present archival investigation. Unfortunately, the secondary dataset used in this analysis did not allow for the inclusion of other partial-absence measures. Future researchers may wish to include all the partial-absence components in an attempt to more fully explore their relationship with FIW.

SEPARATE MALE AND FEMALE MODELS

Results from the separate male and female models should be interpreted with caution due to the exploratory nature of the investigation. However, several findings warrant discussion as they shed additional light on the results of the tested hypotheses and suggest future directions for better understanding men's family responsibilities.

First, it is noteworthy that the link from marital status to FIW is significant for men but not women and that in subsequent analyses these were found to be significantly different (p < .05) This suggests that married or cohabiting men experience more FIW than single men, thus supporting hypothesis 2, but only for men. Perhaps, however, this finding is driven not by the nature of marriage, but rather by the fact that married men are more likely to be providing care to children in addition to their eldercare responsibilities than are single men. We would not expect married and single women to exhibit the same pattern, however, due to the large number of single mothers raising children. Thus, this presumed "marriage effect" that supports hypothesis 2 for men is perhaps more of a "sandwich generation" (Schlesinger, 1989) effect experienced by parents who must provide care to both children and elders.

Next, it appears that having a supportive supervisor in the workplace is perhaps less important for eldercare-providing men than for women in terms of their reported WIF. It is particularly noteworthy that we previously discussed a strong significant pathway from supervisor support to WIF, but the male and female analyses suggest that the finding was driven by the female portion of the sample. Thus, supervisor support is particularly salient for women (with a standardized pathway to WIF of--.30) but is not significant for men. Perhaps women have a stronger need for emotional support from a supervisor or a greater fear that they will be negatively judged by their supervisor for any interference with work due to their eldercare responsibilities. In terms of better understanding men's roles in families, an interesting direction for future work is to investigate whether men simply have less of a need to feel understood and supported with regard to their family responsibilities, or whether men have similar support requirements as women but simply do not seek or require this support from a supervisor.

Additionally, it is interesting to note the combination of FIW being significantly related to partial absence for men but not women and WIF being significantly related to intention to seek new employment for women but not men. These results suggest differing response patterns of men and women to eldercare-based work-family conflict. It appears that many women experiencing work-family conflict intend to leave their current job, perhaps with the belief that a different job might better support their caregiving efforts. The results suggest that men, however, take portions of time off, perhaps to meet their caregiving obligations, but are less likely to indicate that they will seek new employment. Potentially, this reflects women's less stable attachment to the workforce and a work-based identity. Cinamon and Rich (2002) indicated that men were more likely than women to ascribe high importance to their work role and low importance to their family role, while women were more likely than men to do just the opposite (ascribe high importance to their family role and low importance to their work role). It is also possible that men, who are more likely to be their family's primary wage earner, view themselves as having fewer options with regard to new employment. We suggest that future qualitative investigations follow up with this interesting, hut at this point exploratory, finding, to better understand the experiences and perspectives of these care-providing men and women.

Although the present study did not specifically investigate the impact or experience of simultaneously fathering and providing eldercare, the findings of this study allow limited speculation regarding the kinds of issues that may arise for men in this situation. Although our data did not permit a direct test of the question (i.e., we do not have data regarding which participants are fathers), perhaps men in this situation are already noticing the pile-up effect of these multiple roles, as evidenced by the married/cohabiting male group reporting more FIW than the single male group. In terms of their practical approach to meeting the needs of these caregiving roles, limited evidence here suggests that perhaps they will continue to meet their family responsibilities through partial absences from work when the need arises. Interestingly, it is possible that one reason why men in our sample did not benefit from having a supportive supervisor (in terms of decreased WIF) is that perhaps they receive the emotional support they need to juggle their multiple roles from their spouse/significant other. However, if this were demonstrated to be the case, it is possible that this mechanism explains the previously mentioned connection between men's work pressure and women's parental role overload, which is, in turn, associated with decreased youth functioning (Crouter et al., 1999). In other words, if dual care-providing men are relying on spouses/significant others to share the stress of their multiple roles rather than turning to their supervisors for support, it might be that this indirectly negatively impacts their children.

LIMITATIONS AND FUTURE DIRECTIONS

The present research contributes to the family caregiving, work-family conflict, and men in families literatures in a variety of ways. Overall, by using a large national sample of employed eldercare providers across many job categories, the present research adds strength and generalizability to the topics explored.

Inevitably, there are many factors that employers have control over in the workplace, such as work hours, schedule flexibility, supervision, and workplace culture. By acknowledging that workplace factors can interfere with their employees' family (nonwork) lives and vice versa, employers will be taking a proactive step in the right direction. Interestingly, the present investigation suggests that these organizational "tools" are perhaps less effective with family-care-providing men than women. However, hours worked was a significant factor for both women and men with regard to their experience of WIF.

Overall, there is evidence here that married men experience more work-family conflict than single men, that the family-friendly workplace policies investigated do not buffer their level of reported conflict, and that they attempt to handle their dual roles of worker and care provider through partial absences. Thus, the present study suggests that men in families might benefit more from a tangible offer of flex-time than from an abstract feeling that they are understood and supported by their coworkers and supervisors.

This study is limited in several ways. First, with regard to the data set, the present study has the advantage of being based on a large sample of U.S. workers over the age of 18. However, the dependent variables used in these analyses (intention to seek new employment and partial absences) were based on single-item measures. As such, the reliability of these measures is unknown. Additionally, partial absence was measured via self-report rather than employment records. Thus, this measure is potentially subject to reporting biases.

Second, as indicated previously, the scope of the present work does not include the relationships of FIW and WIF to family-related outcomes. Other research (Frone et al., 1992, 1997) suggests that FIW primarily impacts work-related outcomes, while WIF tends to impact family-related outcomes. The present study is not an effort to directly compare the proposed conceptual framework with that advanced by Frone; however, such a study would make a substantial additional contribution to the field. In other words, it would be interesting to assess whether WIF is a stronger predictor of specific family-related outcomes than it is of the intention to seek new employment, and this question would lend itself nicely to a Dominance Analysis (Azen & Budescu, 2003; Budescu, 1993).

In the present study, flexible work arrangements were conceptualized as being an antecedent to WIF. Although prior research is cited supporting this framework, other studies suggest that family-friendly policies moderate the relationship between work-family conflict and outcomes. Thus, future research should further evaluate this issue, perhaps with a comparison of a direct-effects model with a moderation model.

In addition, although this study did not focus on other organizational outcomes to work-family conflict, previous researchers have documented relationships between work-family conflict and job dissatisfaction, job burnout, job performance, organizational commitment, and turnover (Ayree, 1992; Burke, 1988; Goff et al., 1990; Netemeyer et al., 1996), all of which are salient to organizational as well as individual and family well-being. Thus, it will be important for future research to expand the conceptual framework of the present study to include these important work-related outcomes as well.

Table 1

Comparative Demographics for Male and Female Eldercare Providers

   Variable                                   Male           Female

1. Mean age (Standard Deviation in
   parentheses)                          42.30 (11.15)   40.92 (10.93)
2. Percent with high school degree or
   more                                      92.4%           95.9%
3. Percent with college degree or more       37.1%           36.5%
4. Percent currently married                 61.4%           51.5%
5. Percent currently divorced
   (not remarried)                           17.4%           15.2%
6. Percent never married                     16.6%           21.3%
7. Percent White                             78.0%           70.6%
8. Percent Black                             11.4%           15.2%

Table 2

Intercorrelations Used for Path Analysis Model of Combined Sample

Variable                Means       SD       1        2         3

1. Eldercare             8.04      6.43    (1.00)
2. Marital status        1.34      0.48     0.02    (1.00)
3. Gender                1.56      0.50     0.07     0.09     (1.00)
4. Hours worked         44.71     12.49    -0.05     0.01     -0.19 *
5. Flexibility          10.23      2.96     0.01    -0.05     -0.03
6. Supervisor           16.19      4.13    -0.05    -0.04     -0.01
7. Culture              11.59      3.24    -0.06    -0.06      0.02
8. FIW                   9.58      3.31    -0.01     0.06      0.09
9. WIF                  11.58      4.21     0.04    -0.06      0.01
10. Partial absence      0.34      0.48    -0.09     0.13 *   -0.02
11. New job              1.46      0.70     0.01     0.11     -0.07

Variable                  4         5        6        7

1. Eldercare
2. Marital status
3. Gender
4. Hours worked         (1.00)
5. Flexibility           0.06     (0.60)
6. Supervisor           -0.12 *    0.30 *    (0.88)
7. Culture              -0.01      0.29 *     0.31 *   (0.76)
8. FIW                   0.08     -0.10       0.01     -0.05
9. WIF                   0.28 *   -0.21 *    -0.32 *   -0.24 *
10. Partial absence      0.02      0.17 *    -0.04      0.11
11. New job              0.05     -0.10      -0.27     -0.13 *

Variable                   8         9         10        11

1. Eldercare
2. Marital status
3. Gender
4. Hours worked
5. Flexibility
6. Supervisor
7. Culture
8. FIW                  -0.83
9. WIF                   0.40 *   -0.85
10. Partial absence      0.10      0.04      (1.00)
11. New job              0.04      0.21 *     0.09     (1.00)

Note. N = 301. Reliabilities are in parentheses.

Table 3

Intercorrelations Used for Males' and Females' Path Analysis Models

Variable               Male     Male    Female   Female
                      Means      SD     Means      SD

1. Eldercare           7.32     6.24     8.26     6.23
2. Marital status      1.32     0.46     1.42      .50
3. Hours              48.00    11.40    42.19    12.30
4. Flexibility        10.38     2.87    10.05     3.03
5. Supervisor         16.28     3.83    16.11     4.31
6. Culture            11.29     3.31    11.74     3.30
7. FIW                 9.27     3.70     9.64     3.29
8. WIF                11.43     1.15    11.56     4.32
9. Partial absence     0.34     0.47     0.35      .48
10. New job            1.50     0.73     1.44      .71

Variable                1        2        3        4        5

1. Eldercare          (1.00)   -0.01    -0.13    -0.02    -0.14
2. Marital status      0.02    (1.00)   -0.10    -0.08    -0.12
3. Hours               0.03     0.12    (1.00)    0.03    -0.18
4. Flexibility         0.03    -0.02     0.07    (0.60)    0.22
5. Supervisor          0.01     0.01    -0.08     0.35    (0.88)
6. Culture            -0.07    -0.01    -0.01     0.29     0.33
7. FIW                 0.02    -0.06     0.12    -0.11     0.06
8. WIF                 0.05    -0.18     0.31    -0.22    -0.36
9. Partial absence    -0.20     0.05     0.03     0.14     0.03
10. New job           -0.02     0.12     0.03    -0.14    -0.29

Variable                6        7        8        9        10

1. Eldercare          -0.06    -0.05     0.03     0.05     0.02
2. Marital status     -0.14     0.18     0.10     0.25     0.11
3. Hours              -0.01     0.07     0.25    -0.01     0.05
4. Flexibility         0.28    -0.08    -0.19     0.20    -0.05
5. Supervisor          0.27    -0.05    -0.28    -0.14    -0.24
6. Culture            (0.76)   -0.11    -0.23     0.12    -0.13
7. FIW                -0.01    (0.83)    0.45     0.16     0.03
8. WIF                -0.24     0.37    (0.85)    0.04     0.12
9. Partial absence     0.09     0.05     0.04    (1.00)    0.06
10. New job           -0.14     0.05     0.28     0.11    (1.00)

Note. Male subgroup intercorrelations above the diagonal, N = 133,
Female subgroup intercorrelations below the diagonal, N = 168.
Reliabilities are in parentheses.

REFERENCES

Allen, T.D., Herst, D., Bruck, C., & Sutton, M. (1999). Work to family conflict: A review of consequences and an agenda for research. Paper presented at the 14th annual meeting of the Society for Industrial and Organizational Psychology, Atlanta, GA.

Ayree, S. (1992). Antecedents and outcomes of work-family conflict among married professional women: Evidence from Singapore. Human Relations, 45, 813-817.

Azen, R., & Budescu, D. V. (2003). The dominance analysis approach for comparing predictors in multiple regression. Psychological Methods, 8, 129-148.

Barling, J., MacEwen, K.E., Kelloway, E.K., & Higginbottom, S.F. (1994). Predictors and outcomes of eldercare-based interrole conflict. Psychology and Aging, 9, 391-397.

Bentler, P.M. (1990). Comparative fit indices in structural models. Psychological Bulletin, 107, 238-246.

Boles, J.S., Johnston, M.W., & Hair, J.F., Jr. (1997). Roles stress, work-family conflict, and emotional exhaustion: Interrelationships and effects on some work-related consequences. Journal of Personal Selling and Sales Management, 17, 17-28.

Bond, J.T., Galinsky, E., & Swanberg, J.E. (1998). The 1997 national study of the changing workforce. New York, NY: Families and Work Institute.

Budescu, D.V. (1993). Dominance analysis: A new approach to the problem of relative importance of predictors in multiple regression. Psychological Bulletin, 114, 542-551.

Burke, R.J. (1988). Some antecedents and consequences of work-family conflict. Journal of Social Behavior and Personality, 3, 287-302.

Cinamon, R.G., & Rich, Y. (2002). Gender differences in the importance of work and family roles: Implications for work-family conflict. Sex Roles, 47, 531-541.

Cohen, A. (1997). Nonwork influences on withdrawal cognitions: An empirical examination of an overlooked issue. Human Relations, 50, 1511-1536.

Coltrane, S. (1998). Gender in families. Thousand Oaks: Pine Forge Press.

Crouter, A.C., Bumpus, M.F., Maguire, M.C., & McHale, S.M. (1999). Linking parents' work pressure and adolescents' well being: Insights into dynamics in dual earner families. Developmental Psychology, 35, 1453-1461.

Denison, D.R. (1996). What is the difference between organizational culture and organizational climate? A native's point of view on a decade of paradigm wars. Academy of Management Review, 21. 619-654.

Eagle, B.W., Miles, E.W., & Icenogle, M.L. (1997). Interrole conflicts and the permeability of work and family domains: Are there gender differences? Journal of Vocational Behavior, 50, 168-184.

Ettner, S.L. (1995). The impact of "parent care" on female labor supply decisions. Demography, 32, 63-80.

Federal Interagency Forum on Aging Related Statistics (2000). Older Americans 2000: Key indicators of well-being. Retrieved 08/23/00, from http://www .agingstats.gov/chartbook2000/population.html

Frankiln, S.T., Ames, B.D., & King, S. (1994). Acquiring the family eldercare role: Influence on female employment adaptation. Research on Aging, 16, 27-42.

Frone, M.R., & Rice, R.W. (1987). Work-family conflict: The effect of job and family involvement. Journal of Occupational Behaviour, 8, 45-53.

Frone, M.R., Russell, M., & Cooper, M.L. (1992). Antecedents and outcomes of work-family conflict: Testing a model of the work-family interface. Journal of Applied Psychology, 77, 65-78.

Frone, M.R., Yardley, J.K., & Markel, K.S. (1997). Developing and testing an integrative model of work-family interface. Journal of Vocational Behavior, 50, 145-167.

Gignac, M.A.M., Kelloway, E.K., & Gottlieb, B.H. (1996). The impact of caregiving on employment: A mediational model of work-family conflict. Canadian Journal on Aging, 15(4), 525-542.

Goff, S.J., Mount, M.K., & Jamison, R.L. (1990). Employer supported child care, work/family conflict, and absenteeism: A field study. Personnel Psychology, 43, 793-809.

Gorey, K.M., Rice, R.W., & Brice, G.C. (1992). The prevalence of elder care responsibilities among the work force population. Research on Aging, 14(3), 399-418.

Grandey, A.A., & Cropanzano, R. (1999). The conservation of resources model applied to work-family conflict and strain. Journal of Vocational Behavior, 54, 350-370.

Greenhaus, J.H., & Beutell, N.J. (1985). Sources of conflict between work and family role. Academy of Management Review, 10, 76-88.

Gutek, B.A., Searle, S., & Klepa, L. (1991). Rational versus gender role expectations for work-family conflict. Journal of Applied Psychology, 76, 560-568.

Halpern, P.L., & Deck, F.W., III. (1989). Building responsive systems for public employee caregivers: Eldercare program and policy options. Washington, DC: National Association of State Units on Aging: International Personnel Management Association.

Hepburn, C.G., & Barling, J. (1996). Eldercare responsibilities, interrole conflict, and employee absence: A daily study. Journal of Occupational Health Psychology, 1, 311-318.

Hochschild, A. (1989). The second shift. New York, New York: Viking Penguin.

Hoyle, R.H. (1995). Structural equation modeling: Concepts, issues, and applications. Thousand Oaks, CA: Sage.

Joreskog, K.G., & Sorborm, D. (1984). LISREL VI user's guide (3rd ed.). Mooresville, IN: Scientific Software.

Joreskog K.G., & Sorbom, D. (1993). LISREL 8: User's reference guide. Chicago: Scientific Software.

Kahn, R.L., Wolfe, D.M., Quinn, R., Snoek, J.D., & Rosenthal, R.A. (1964). Organizational stress. New York: Wiley.

Katz, D., & Kahn, R.L. (1978). The social psychology of organizations (2nd ed.) New York: Wiley.

Kossek, E.E., DeMarr, B.J., Backman, K., & Koller, M. (1993). Assessing employees' emerging elder care needs and reactions to dependent care benefits. Public Personnel Management, 22, 617-638.

Kossek, E.E., & Ozeki, C. (1998). Work-family conflict, policies, and the job-life satisfaction relationship: A review and directions for organizational behavior-human resources research. Journal of Applied Psychology, 83, 139-149.

Lamb, M.E., Pleck, J.H., Charnov, E.L., & Levine, J. A. (1985). Paternal behavior in humans. American Zoologist, 25, 883-894.

Maraist, C.C. (1999). Will the "true" predictors of work-family conflict please stand up? Paper presented at the 14th Annual Conference of the Society for Industrial and Organizational Psychology, Atlanta, GA.

MacCallum, R.C, Browne, M.W., & Sugawars, H.M. (1996). Power analysis and determination of sample size for covariance structure modeling. Psychological Methods, 1, 130-149.

Mulaik, S.A., James, L.R., Van Alstine, J. Bennett, N., Lind, S., & Stillwell, C.D. (1989). An evaluation of goodness of fit indices for structural equation models. Psychological Bulletin, 105, 430-445.

Neal, M.B., & Hammer, L.B. (2000). Transitions in work and family roles: How they affect work-family fit among sandwiched generation couples. Paper presented at the 53rd Annual Scientific Meeting of the Gerontological Society of America, Washington, DC.

Neale, M., Chapman, N., Ingersoll-Dayton, B., & Emlen, A. (1993). Balancing work and caregiving for children, adults, and elders. Newbury Park, CA: Sage.

Netemeyer, R.G., Boles, J.S., & McMurrian, R. (1996). Development and validation of work-family conflict and family-work conflict scales. Journal of Applied Psychology, 81,400-410.

Nichols, L.S., & Junk, V.W. (1997). The sandwich generation: Dependency, proximity, and task assistance needs of parents. Journal of Family and Economic Issues, 18, 299-326.

O'Driscoll, M.P., Ilgen, D.R., & Hidreth, K. (1992). Time devoted to job and off-job activities, interrole conflict, and affective experiences. Journal of Applied Psychology, 77, 272-279.

Parker, C.W. (1999). An examination of the relationship between perceived army support for family and job attitudes of first-term enlisted soldiers. Paper presented at the 14th Annual Conference of the Society for Industrial and Organizational Psychology, Atlanta, GA.

Rickard, A.K., Neal, M.B., Hammer, L.B., & Caubet, S.L. (2000). Work-family conflict as a mediator between care role rewards and stressors and work outcomes among dial-earner couples in the sandwiched generation. Paper presented at the 53rd Annual Scientific Meeting of the Gerontological Society of America, Washington, DC.

Rosin, H.M., & Korabik, K. (2002). Do family-friendly policies fulfill their promise? An investigation of their impact on work-family conflict and work and personal outcomes. In D.L. Nelson & R.J. Burke (Eds.), Gender, work stress, and health (pp. 211-226). Washington, DC: American Psychological Association.

Scharlach, A., & Boyd, S.L. (1989). Caregiving and employment: Results of an employee survey. The Gerontologist, 29, 382-387.

Schlesinger, B. (1989). The "sandwich generation": Middle-aged families under stress. Canada's Mental Health, 37, 11-14.

Shultz, K.S., & Silverstein, M. (1999). Gender, eldercare responsibilities, and intention to change jobs. Paper presented at the 52nd Annual Scientific Meeting of the Gerontological Society of America, San Francisco, CA.

Smith, J., Buffardi, L., & Holt, R. (1999). Antecedents and consequences of elder care responsibility and the role of organizational and supervisor support. Paper presented at the 14th Annual Conference of the Society for Industrial and Organizational Psychology, Atlanta, GA.

Staines, G.L., & Pleck, J.H. (1983). The impact of work schedules on the family. Survey Research Center, Institute for Social Research, The University of Michigan.

Staines, G.L., & Pleck, J.H. (1984). Nonstandard work schedules and family life. Journal of Applied Psychology, 69(3), 515-523.

Steiger, J.H., & Lind, J.C. (1980). Statistically based tests for the number of common factors. Paper presented at the Psychometric Society Annual Meeting, Iowa City, IA.

Stern, S.H., & Wolford, B. (2001). Putting the "elder care" in elder law. Journal of Gerontological Social Work, 34, 55-59.

Stoller, E.P. (1983). Parental caregiving by adult children. Journal of Marriage and the Family, 45, 851-858.

Stone, R., Cafferata, G.L., & Sangl, J. (1987). Caregivers of the frail elderly: A national profile. The Gerontologist, 27, 616-626.

Thomas, L.T., & Ganster, D.C. (1995). Impact of family supportive work variables in work-family conflict and strain: A control perspective. Journal of Applied Psychology, 80, 6-15.

Thompson, C.A., Beauvais, L.L., & Lyness, K.S. (1999). When work-family benefits are not enough: The influence of work-family culture on benefit utilization, organizational attachment, and work-family conflict. Journal of Vocational Behavior, 54, 392-415.

Wagner, D.L., & Neal, M.B. (1994). Caregiving and work: Consequences, correlates, and workplace responses. Educational Gerontology, 20, 645-663.

Zal, H.M. (1992). The sandwich generation: Caught between growing children and aging parents. New York: Plenum Press.

An earlier version of this paper was presented as part of the symposium, K.S. Shultz and S.E. Rix (Co-Chairs), The Aging Workforce: Eldercare, Training and Pension Issues, at the annual meeting of the Gerontological Society of America, Chicago, IL, November 2001. The authors thank Mark Agars and Joanna Worthley for their comments on earlier drafts of this paper.

Correspondence concerning this article should be addressed to Kenneth S. Shultz, Department of Psychology, California State University, 5500 University Parkway, San Bernardino, CA 92407. Electronic mail: [e-mail].

Jaime L. Barrah

Kenneth S. Shultz

California State University, San Bernardino

Boris Baltes

Wayne State University

Heide E. Stolz

University of Tennessee, Knoxville