Gender and Relationship Differences in Caregiving Patterns and Consequences Among Employed Caregivers 1

Transcription

1 Copyright 1997 by The Cerontological Society of America The Cerontologist Vol. 37, No. 6, Gender and relationship differences in caregiving (i.e., for a spouse, parent, parent-in-law, other relative, or friend) are investigated among a sample of employed caregivers (N = 2,174). No gender differences were found in the provision of 7 of 13 caregiving tasks, but women devoted more time and were more likely to be primary caregivers. After controlling for other key variables, relationship added modestly to the ability to predict all five measures of caregiving consequences, and gender further added nominally to the ability to predict three of the measures. The findings lend support to the gender-role socialization hypothesis (Miller & Cafasso, 1992) and suggest the targeting of services to caregivers of spouses and parents, especially female caregivers. Key Words: Absenteeism, Elder care, Informal caregivers, Work and family, Work performance Gender and Relationship Differences in Caregiving Patterns and Consequences Among Employed Caregivers 1 Margaret B. Neal, PhD, 2 Berit Ingersoll-Dayton, PhD, 3 and Marjorie E. Starrels, PhD 4 Research conducted during the past few decades has documented the importance of family and friends in the provision of long-term care to the elderly. Much of this literature has focused on who these informal caregivers are (i.e., the composition of the caregiving network) and what they do (i.e., the nature of their responsibilities and the amount of time spent caregiving) (Stone, Cafferata, & Sangl, 1987). The negative consequences of caregiving for family members also have received a great deal of attention. Much of this research, however, has tended to treat family caregivers as a homogeneous group, with the effect of clouding differences in the social context of caregiving (Dwyer & Coward, 1992a), such as those in type of relationship to the elder, gender, health, and work status of the caregiver (Cantor, 1983). This manuscript contributes to the body of caregiving research by examining both 1 The data reported in this article were collected as part of the study, "Work and Elder Care: Supporting Families in the Workplace," funded by the Meyer Memorial Trust. An earlier version of the manuscript was presented at a symposium, "Family Care of the Elderly: The Role of Caregiver Gender Re-Visited," organized by Cliff Barber, at the 57th Annual Conference of the National Council on Family Relations, in Portland, Oregon, in November Sincere thanks to David Morgan for his generous methodological consultation, and to Maria Talbott, as well as two anonymous reviewers, for their comments and suggestions on previous versions. institute on Aging and Department of Urban Studies and Planning, Portland State University, Portland, OR. Address correspondence to Margaret B. Neal, Institute on Aging, Portland State University, P.O. Box 751, Portland, OR nealm@pdx.edu 3 School of Social Work, The University of Michigan, Ann Arbor, Ml. 'Department of Human Development and Family Studies, Colorado State University, Fort Collins, CO. gender and relationship differences in caregiving tasks and outcomes among informal caregivers of the elderly, and specifically, caregivers who are employed outside the home. A growing number of studies has focused on differences between men and women in their enactment of the caregiving role (Miller & Cafasso, 1992). In brief, these studies have consistently demonstrated three key findings. First, the majority of caregivers to elders are women (Stone et al., 1987), and the bulk of care provided is given by women (Finley, 1989; Horowitz, 1985; Stoller, 1983). Second, men and women display different patterns of assistance that are consistent with a gender-based division of labor (Stoller, 1990). For example, most day-to-day, personal, and hands-on care is provided by women (Horowitz, 1985; Miller & Cafasso, 1992), whereas men are more likely to engage in arrangements for services or care management (Chang & White- Means, 1991; Montgomery & Kamo, 1989), transportation (Young & Kahana, 1989), home repair and maintenance tasks (Coward, 1987), or in intermittent assistance with occasional tasks (Stoller, 1990). Third, women experience greater negative consequences of caregiving, typically referred to as "burden" or "stress" (Horowitz, 1985; Stoller, 1983). A large and growing number of women, the traditional caregivers, are employed outside the home, with 58% of all American women aged 16 and older in the labor force in 1993, with an additional gain of 5% forecasted by the year 2005 (MetLife, 1994, cited in Doty, 1995). Concerns about the rapid increase in labor force participation by women have raised 804 The Gerontologist

2 questions about women's continued availability to provide care (Doty, 1995) and have intensified the debate over gender differences in the patterns and consequences of caregiving. For example, a recent study of employed, nonspousal caregivers revealed that female employees were more likely to help with care provision tasks than were male employees (Kramer & Kipnis, 1995). Earlier, Stoller (1983) found that although being employed decreased the hours of caregiving assistance provided by sons, it did not have a significant impact on the hours of assistance provided by daughters. In contrast, Stueve and O'Donnell (1989) found that daughters who spent more hours in employment provided less care. Brody and Schoonover (1986) found no differences in the overall amount of care provided to aging mothers by daughters who were employed and those who were not. They did discover that employed daughters assisted less with personal care and cooking than nonemployed daughters, but this difference was offset by purchased help. A recent analysis of the nationally representative 1989 National Long-Term Care Survey data revealed that employed primary caregivers provided a considerable number of weekly hours of informal help, but still significantly fewer hours of help personally than nonemployed primary caregivers (Doty, 1995). Moreover, among employed caregivers of elders, nearly all studies have revealed a more profoundly negative effect on work-family conflict and workrole strains for women than for men (Anastas, Gibeau, & Larson, 1990; Kramer & Kipnis, 1995; Stone et al., 1987; Stone & Short, 1990). An exception is Horowitz's (1985) finding of no difference in caregiving sons' and daughters' self-assessments of their job performance. Spitze, Logan, Deane, and Zerger (1994) found that men who work more hours were less distressed by caregiving than were those who worked fewer hours. This suggests that employment may serve as a justification for not providing more help (Spitze et al., 1994). Limitations in Previous Research Previous caregiving research that has compared the types and levels of task involvement by gender has been inconsistent with respect to the numbers and kinds of tasks included in the measures (Walker, Pratt, & Eddy, 1995). For example, Barusch and Spaid (1989) asked about the provision of 37 tasks, whereas Stoller (1983) included 7. Other researchers exploring this area include Young and Kahana (1989), who asked about the provision of 12 tasks; Pruchno and Resch (1989), who included 15 tasks; Horowitz (1985), who listed 10 tasks; Miller, McFall, and Montgomery (1991), who asked about 8 tasks; and Miller and Montgomery (1990), who listed 14 tasks. The provision of social/emotional support has not been included consistently as a task, and few researchers, except for Miller et al. (1991), have specified as caregiving tasks the administration of medications and injections. Furthermore, the ways in which tasks have been summed, or grouped, vary considerably from study to study. For example, Finley (1989) grouped 13 tasks into four categories: activities of daily living (ADLs), cognitive assistance, instrumental tasks (ladls), and expressive tasks. Brody and Schoonover's (1986) categorization of tasks consisted of personal care, ladls, shopping/transportation, and emotional support. In contrast, some researchers (e.g., Barusch & Spaid, 1989; Pruchno & Resch, 1989) have placed tasks into only two groups, according to whether they related to ADLs or ladls. Others have combined tasks that are typically performed by men with tasks typically performed by women, thus obscuring some potentially important gender differences. For example, in our own research (Starrels, Ingersoll-Dayton, Neal, & Yamada, 1995), a health/ care management factor included both personal care tasks, which are typically performed by women, and care management tasks, which have been considered as tasks commonly performed by males. Another limitation of previous caregiving research is its failure to consider a broad range of characteristics, in addition to gender, that are associated with the provision of care and stress-related outcomes (Barber, 1988; Dwyer & Coward, 1991, 1992a). Thus, as Dwyer and Coward (1991, p. S259) note, "it is not clear whether reported differences in the gender structure of caregiving are due to gender per se or to other unmeasured factors with which it covaries." In addition to gender, relevant caregiver characteristics are the caregiver's age, marital status, proximity to the care recipient, and employment status (Dwyer & Coward, 1991). Furthermore, important elder and care situation characteristics include gender, marital status, age, income, number of living children, location of residence, ADL limitations, and IADL limitations (Dwyer & Coward, 1991). Also critical to include in models of the effects of caregiving are available social resources of the caregiver. Such resources will likely affect the caregiver's experience of strain or stress (Pearlin, Mullan, Semple, & Skaff, 1990; Thoits, 1987). In addition, the family relationship of the caregiver to the care receiver (husband, wife, son, daughter) should be examined as a possible modifier of caregiver burden (Miller & Cafasso, 1992) and with regard to limitations on the social activities of family caregivers (Miller & Montgomery, 1990). Most caregiving research, however, has focused on caregiving to individuals of just one relationship type, such as parents (e.g., Horowitz, 1985; Stoller, 1983) or spouses (Barusch & Spaid, 1989; Fitting, Rabins, Lucas, & Eastham, 1986). Three exceptions are studies by Stoller (1990), Miller and Montgomery (1990), and Neal, Chapman, Ingersoll-Dayton, and Emlen (1993). Miller and Montgomery (1990) examined care provided by spouses and adult children, but excluded data on other informal caregivers (e.g., in-laws, other relatives, and nonrelatives). Neal et al. (1993) included these other informal caregivers, but combined nonrelatives and relatives other than parents or spouses in a single "other" category. Also, similarities and differences in the Vol. 37, No. 6,

3 nature of care provided by relationship were not examined, nor were the contributions of gender and relationship to stress over and above those of other personal, caregiving demand, and resource variables. Our recent research has refined the investigation specifically of parent care by considering care provided to step-parents as well as parents (Starrels et al., 1995) and has contrasted such caregiving relationships with the care provided to parents-in-law (Ingersoll-Dayton, Starrels, & Dowler, 1996). To our knowledge, there has been no attempt, as yet, to compare the full spectrum of caregiving relationships with respect to either caregiving outcomes or type of care of provided. Finally, previous caregiving research has generally failed to investigate gender and type of relationship to the care recipient simultaneously (Young & Kahana, 1989). As Young and Kahana note, gender and relationship must be distinguished: "Efforts to consider either without controlling for the other may be masking important information" (p. 665). In their research, Young and Kahana found pronounced differences in both the process of care and the effects of caregiving among wives, daughters, and husbands. We are aware of only one other study that has concurrently addressed gender and type of relationship. Specifically, in their survey of 262 individuals caring for victims of Alzheimer's disease, Barber and Pasley (1995) found that the impacts of caregiving varied by both caregiver gender and generational relationship between the caregiver and the care recipient. Female caregivers experienced greater strain in family relationships and greater declines in health as a result of caregiving than did male caregivers; filial caregivers reported more strain in their relationships with family members than did spousal caregivers, yet spousal caregivers experienced more negative health impacts and greater restrictions in their social activity (Barber & Pasley, 1995). The present research addresses these limitations in a number of ways. First, we simultaneously examine a spectrum of caregiving relationships (i.e., spouses, parents and step-parents, parents-in-law, other relatives, and friends). This set of relationships represents a significant expansion over the kinds of caregiving relationships that have been previously examined. Second, we perform a finegrained analysis of caregiving tasks. Each task is examined individually, rather than in the aggregate, in contrast to our own previous research and that of other researchers. This analysis includes an investigation of gender differences in the performance of each individual task. Also, we have created a new measure, number of tasks performed, in order to capture diversity of task involvement, which could have an impact on stress. Third, in this study we focus on various forms of stress (personal, family, and work-related) as outcomes of caregiving. By including a broader array of stress-related variables, previous research on the outcomes of caregiving is expanded upon and refined. Finally, we examine the contribution of gender to stress, over and above that of various other personal, caregiving demand, and resources variables, as well as that of relationship to the elder. In each of these four areas, the present research represents a unique contribution to existing studies on the nature and patterns of care provided by employed caregivers, and the consequences associated with employment and the provision of elder care. Research Questions The present study responds to Miller and Cafasso's (1992) recommendation that further studies examining gender differences in caregiving be highly focused and involve large samples. It also addresses Dwyer and Coward's (1992b) criticism, "Few studies have been designed in a way that would permit the investigators to isolate that part of the variance in caregiving outcomes that is due solely to gender" (p. 159), and to Young and Kahana's (1989) exhortation that gender and type of relationship to the care recipient must be simultaneously investigated and distinguished. Utilizing data from a large sample of informal caregivers who are in the labor force, gender differences in the patterns and outcomes of caregiving are examined. These gender differences are examined in the context of the care recipient's relationship to the caregiver (i.e., parent, parent-in-law, spouse, other relative, or friend). This is a much wider variety of caregiving relationships than has been examined previously in a given study. Specifically, three research questions are addressed. First, what gender differences exist in the type and amount of caregiving provided by employees to elders? Second, do the type and amount of caregiving by employees differ by gender and relationship to the elder? Third, do the effects of caregiving differ by relationship and gender among employed caregivers when other characteristics of the caregiver, care recipient, and caregiving situation, as well as available resources, are controlled? Hypotheses Suggested to Explain Gender Differences in Caregiving Several hypotheses have been suggested to explain the gender differences in the amount and nature of the caregiving tasks performed and in levels of caregiver stress. As noted by Miller and Cafasso (1992), two basic hypotheses are the gender-role socialization hypothesis and the social-role hypothesis. According to the gender-role socialization hypothesis, women are more likely to experience stress because of the personality orientations they have internalized. Specifically, women are hypothesized to have stronger affiliative orientations and behaviors (e.g., sensitivity to relationships, nurturing behaviors), compared with a greater emphasis on autonomy, differentiation, and instrumental behaviors in men. This internalization occurs due to socialization to gender roles during childhood, as well as continuing reinforcement of cultural norms. 806 The Gerontologist

4 According to this hypothesis, women and men would be expected to engage in different amounts and types of caregiving tasks, with women being more likely to perform more tasks overall, and to perform more personal care and household tasks than men, and men performing tasks that are more time-limited and that can be performed at the discretion of the caregiver, such as money management. Also, women would be expected to experience greater caregiving stress than men. According to the gender-role socialization hypothesis, gender differences are fundamental; they are within the person due to his or her socialization and would be expected to persist after other personal and situational characteristics were controlled. With regard to relationship to the elder, no differences in stress or in types of tasks performed would be expected. The social role hypothesis focuses on current social role and situational demands, suggesting that negative effects (e.g., higher stress levels) are due to those very roles and are determined by exposure to stressful situations. This hypothesis argues that once situational demands are taken into account, gender differences should be minimal, and women would have greater levels of distress because they have more stressful experiences and more role demands (Miller & Cafasso, 1992). Thus, the severity of the impairment will determine gender differences in task involvement. When providing care to persons with similar levels of impairment, the nature of male and female caregivers' task involvement should be similar. Also, when the pressures of competing career and family needs decline, gender differences are hypothesized to become less pronounced (Miller & Cafasso, citing Gutmann, 1980). Thus, gender differences in stress among daughter and son caregivers would be expected to be greater than those among spouse caregivers. The data from the analyses conducted will be examined to see which of these two alternative hypotheses for gender differences is best supported. Method Sample The analyses presented are based on data gathered in 1987 through a mailed survey of the employee populations of 33 businesses and agencies in the Portland, Oregon, metropolitan area. The employers chosen for participation were selected to represent large, medium, and small employers within seven major categories of the Standard Industrial Classification system (SIC). [The SIC covers all economic activities and defines industries according to the composition and structure of the economy. It is used to facilitate the comparability of establishment data describing various facets of the U.S. economy (Executive Office of the President, 1987).] Information elicited through the employee survey included social, demographic, and work-related characteristics, caregiving responsibilities and arrangements for children, adults with disabilities, and elders, and the effects of caregiving on work and family. [See Neal et al. (1993) for a detailed description of the sample and the survey instrument.] Employees were identified as providing elder or adult care if they responded "yes" to the following question: Do you have responsibilities for helping out adult relatives or friends who are elderly or disabled? This includes persons who live with you or who live somewhere else. By "helping out" we mean help with shopping, home maintenance or transportation, checking on them by phone, making arrangements for care, etc. The subsample for these analyses consists of those 2,188 persons who indicated that the adult for whom they were providing the most care was aged 60 or over. Fourteen of the 2,188 employed caregivers failed to provide data on either their gender or their relationship to the elder whom they were helping. Thus, the subsample for these analyses consists of 2,174 employees. Of this subsample, 63% were women, and 37% were men. The majority (58%) were caring for parents or step-parents. Fifteen percent were caring for parents- or step-parents-in-law. Seventeen percent were caring for relatives other than spouses or parents, and 8% were caring for friends or other nonrelatives. Only 2% were caring for spouses. Table 1 displays the descriptive statistics of the independent and dependent variables used in the multivariate analyses for the sample as a whole and for men and women separately. Measures Dependent Measures, Research Questions 1 and 2. As detailed in the Analysis Plan, three sets of analyses were conducted to address the study's three research questions. In the first two sets of analyses, the dependent measures concern the amount and nature of caregiving. In the third set of analyses, the dependent variables are various consequences of caregiving. The contributions of relationship and gender to these consequences are explored, after controlling for various other personal and caregiving situation characteristics (e.g., elder's gender, age, and degree of difficulty with activities of daily living), and resources available to the employee (e.g., flexibility in work schedule). Thus, in the third set of analyses, the total number of tasks performed, the average number of hours per week spent caregiving, and the level of responsibility for care (high = sole) are included as independent variables, while they serve as dependent variables in the first two sets of analyses. Dependent variables in the analyses pertaining to Research Questions 1 and 2 include measures of the nature and amount of caregiving performed. Respondents were asked how often they performed each of 13 tasks: house and yard maintenance, transportation, shopping, check on by phone, fix or bring meals, personal (dressing, bathing, etc.), Vol. 37, No. 6,

5 Table 1. Descriptive Statistics for Dependent and Independent Variables Total Sample (N = 2,174) Males (n = 799) Females (n = 1,375) Mean SD Range Mean SD Range Mean SD Range Independent Variables Control variables Age Race(1 = white) Has partner (1 = yes) Household income (in thousands) Number of children Occupation (1 = professional) Numer of hours worked Number of elders caring for Elder's gender (1 = female) Elder's age Distance (0 = live together, 6 = > 1,000 miles) Elder has difficulty with ADLs (5 = most/all of time) Eder exhibits problem behaviors (5 = most/all of time) Average number of hours caring Total number of tasks performed Responsibility for care (1 = secondary, 4 = sole) Ease in finding/maintaining care (6 = high) Family-friendly policies (6 = high) Flexibility in work schedule (4 = high) Ability to continue caregiving (1 = yes) Relationship (omitted category = parent) Spouse (1 = yes) Parent-in-law (1 = yes) Other relative (1 = yes) Friend (1 = yes) Gender (1 = female) ^1 1^ Dependent variables Chronic elder care stress (in past year) Recent elder care stress (in past 4 weeks) Time taken off work Reduced work effectiveness Difficulty combining work and family housekeeping, continuous supervision, nursing care, help with expenses (give money), manage legal and financial affairs (write checks, fill out insurance forms, etc.), visit, give emotional support, and arrange and manage health or social services (1 = never or seldom, 2 = several times a year, 3 = once a month, 4 = a few times a month, 5 = once a week, 6 = a few times a week, 7 = daily). In addition, three global measures of the amount of caregiving involvement are included: the average number of hours spent per week in caring for the elder (ranging from less than 1 to 96), the total number of caregiving tasks performed (ranging from 0 to 13), and the employee caregiver's level of responsibility for care. For this last measure, respondents were asked, "In the past year, when this person has needed help, who has usually been the one who has given it or seen that it was given?" (4 = "I have been the only one," 3 = "I have been the main one, with some help from others," 2 = "I have shared equally with one or more others," 1 = "others, with my help"). Since previous research has grouped these tasks inconsistently, the tasks are not aggregated here by content (e.g., through factor analysis or other data reduction/scaling techniques) to avoid obscuring any potential gender-related differences. (See Table 2 for descriptive statistics of these variables for the total sample and for men and women separately.) Dependent Measures, Research Question 3. Five dependent measures are used to represent various consequences of caregiving, including personal and family-related outcomes, work productivity-related outcomes, and work-family conflict. Personal and family-related outcomes are repre- 808 The Gerontologist

6 Table 2. t Tests of Differences in Task Performance and Involvement by Gender Total Sample M (SD) Males (n = 799) M (SD) Females (n = 1,375) M (SD) t Value Tasks" Personal/health care Nursing care Personal care (dressing, bathing, etc.) Continuous supervision Social/emotional support Transportation Shopping Check on by phone Visit, give emotional support Household chores House and yard maintenance Housekeeping Fix or bring meals Care management Manage legal and financial affairs Help with expenses Arrange and manage health or social services (1.12) (1.20) (1.45) (1.74) (1.71) (1.92) (1.69) (1.55) (1.58) (1.83) (1.55) (1) (1.47) (2.62) (8.61) (.96) (1) (1.16) (1.37) (1.65) (1.63) (1.93) (1.72) (1.49) (1.41) (1.65) (1.56) (1) (1.45) (2.64) (1.14) (1.22) (1.50) (1.77) (1.73) (1.88) (1.64) (1.58) (1.65) (1.91) (1.54) (1) (1.49) (2.60) (9.67) (.97) *** -5.37*** -7.35*** -7.36*** *** -6.46*** ^.10*** -5.69*** -7.12*** Total number of tasks performed Average weekly hours spent caregiving Responsibility for care b (6.27) (.91) "1 = Never or seldom, 2 = Several times a year, 3 = Once a month, 4 = A few times a month, 5 = Once a week, 6 = A few times a week, 7 = Daily. b 1 = Secondary, 2 = Shared equally with others, 3 = Primary, 4 = Sole. *p < ; **p < ; ***p < 1. sented by two measures: chronic elder care stress and recent elder care stress. Chronic elder care stress is the mean of two items (alpha =.69): "In general during the past year, how easy or difficult has it been for you to help out this person?" (1 = very easy, 6 = very difficult) and "In the past year, how often have your caregiving responsibilities interfered with your social and emotional needs and other family responsibilities?" (1 = never, 5 = most or all of the time). Recent elder care stress consists of responses to a single item: "In the past 4 weeks, to what extent has care for elderly or disabled adult family members been a source of stress to you?" (1 = no stress at all, 4 = a lot of stress). Thus, recent elder care stress focuses on caregiving stress of a more acute nature (over the past four weeks), whereas chronic elder care stress measures caregiving stress over the past year. Work productivity outcomes consist of two additional single-item measures: time taken off from work due to caregiving ("In the past year, how often have you taken time off from work to do something for the person?"; responses of 1 = never or seldom to 7 = daily were logged to more closely approximate a normal distribution) and reduced work effectiveness due to caregiving worries ("How often have you worked less effectively at your job because you are worried or upset about this person?," 1 = never, 5 = most or all of the time). The final dependent measure is difficulty in combining work and family ("Circumstances differ and Vol. 37, No. 6, some people find it easier than others to combine working with family responsibilities. In general, how easy or difficult is it for you?," 1 = very easy, 6 = very difficult). Independent Variables. The independent variables in all analyses include the employee's gender (1 = female, 0 = male) and the elder's relationship to the employee. For the analyses pertaining to Research Questions 1 and 2, relationship to the elder was measured as a categorical variable (1 = spouse, 2 = parent or step-parent, 3 = parent-in-law or stepparent-in-law, 4 = other relative, 5 = friend or other nonrelative). For the analyses pertaining to Research Question 3, relationship to the elder was measured as a set of dummy variables (with parent as the omitted category). Building on earlier research concerning the impact of various characteristics of the caregiver, the care recipient, and the caregiving situation, as well as available resources, the third set of analyses uses multivariate modeling to investigate the impact of gender, after controlling for these previously identified variables, on several personal, family, and workrelated consequences of caregiving among employed caregivers. These control variables, identified in earlier research as being associated with caregiver burden or stress (Dwyer & Coward, 1991; Pearlin et al., 1990), include other personal and job-related characteristics of the employee caregiver, characteristics of the elder care recipient and the care situa-

7 tion, and available resources. These variables measure the role and situational demands suggested by the social role hypothesis. Personal and job-related characteristics of the employee caregiver include: age, race (white = 1, nonwhite = 0), whether or not the caregiver has a spouse or cohabitating partner (1 = yes, 0 = no), number of children under age 18 in the household, occupation (professional, managerial or technical = 1, other = 0), number of hours worked, and household income [10 categories, ranging from "under $10,000" to "$70,000 or more," coded in thousands and to the category midpoint; for example, the lowest income level was 5 ($5,000), and the highest was 75 ($75, 000)]. Also included as controls are characteristics of the care recipient and the caregiving situation. These consist of the total number of elders or disabled adults being cared for by the employee, the elder's gender (1 = female, 0 = male), the elder's age, elder's proximity to the employee (0 = lives with, 1 = less than 5 miles, 2 = 5-24 miles, 3 = miles, 4 = miles, 5 = miles, 6 = 1,000 or more miles), how often the elder is unable to manage activities of daily living (1 = never, 5 = most or all of the time), and how often the elder exhibits problem behavior (based on a scale averaging three items, 1 = never, 5 = most or all of the time: wanders or is confused; acts inappropriately, is disruptive; and is aggressive or uncooperative) (alpha =.82). Total number of tasks performed (range = 3), average number of weekly hours spent caring for this elder, and level of responsibility for care (1 = secondary, 4 = sole) are additional control variables in these analyses (they were dependent measures in the analyses pertaining to the first two research questions). The third set of control variables consists of resources available to the caregiver. The first is ease of finding and maintaining care arrangements for the elder. This measure averages responses to two items (alpha =.81): "How easy or difficult has it been to find care arrangements for this person?" and "How easy or difficult has it been to manage or maintain these arrangements?" (1 = very difficult, 6 = very easy). Thus, a high score on this item represents the availability of community resources, whereas a low score represents the absence of resources. The second measure is family-friendly policies, based on the item, "To what extent do personnel practices in your department make it easy or difficult to provide care for this person?" (1 = very difficult, 6 = very easy). The third measure is degree of flexibility in work schedule (1 = none, 4 = a lot), and the fourth is a dichotomous measure of ability to continue providing care for as long as is necessary (1 = yes, with the same help have now; 0 = no, not sure, or yes but only with more help). Analysis Plan To address the first research question concerning the relationship between employee's gender and the type and amount of caregiving to elders, t tests were performed comparing the means of male and female employees for each of the 13 tasks and the three global measures of task involvement. To compare the type and amount of caregiving by employees across gender and relationship groups (the second research question), two-way analyses of variance were conducted. The findings from these analyses revealed the importance of examining both the main and interactive effects of gender and relationship with regard to caregiving outcomes. Thus, to explore the third research question concerning gender differences in the effects of caregiving, hierarchical multiple regression was performed. In the first step, the control variables (various personal, family, caregiving situation, and available resource characteristics) were entered. In the second step, four dummy variables representing the relationship of the elder to the employee were entered (spouse, parent-in-law, other relative, friend; parent was the omitted category). In the third step, gender was entered, and in the fourth step, interaction terms for gender by relationship were entered. This strategy made it possible to ascertain the effects of relationship, and then gender, controlling for other variables shown to influence caregivers' experience of the caregiving role, and to examine gender-relationship interaction effects. Results As shown in Table 2, male and female employees differ significantly in their frequency of performance of six of the 13 tasks, and for all three of the global task involvement measures. Female employees consistently provide transportation, help with shopping, check on the elder by phone, and visit and give emotional support more frequently than do male employees. Similarly, they do more housekeeping and fixing or bringing of meals than do male employees. However, contrary to the gender-role socialization hypothesis, there are no significant gender differences with respect to personal/health care tasks (nursing care, personal care, continuous supervision) (although neither the employed men nor women in our sample provide this sort of care often), care management tasks (including management of legal and financial affairs, help with expenses, and arrangement and management of health or social services), or house and yard maintenance. Men were not significantly more involved than women on any of the measures studied. In sum, with regard to the first research question concerning gender differences in the frequency of task performance, there are no gender differences for seven of the 13 tasks. However, employed women are clearly more involved overall in caregiving than are employed men. They perform a slightly higher number of tasks, they spend just over two hours more each week in caregiving (6.2 hours for women, compared with 4.1 for men), and more women are primary caregivers. It appears that, for employed caregivers, some of the gender differences in task performance found in previous re- 810 The Gerontologist

8 search are attenuated. However, women continue to provide more care overall in terms of their total number of hours and their level of responsibility for the elder, thus providing more support for the gender-role socialization hypothesis than the social role hypothesis. Concerning the second question regarding gender and relationship differences, relationship of the elder to the employee also is associated with performance of caregiving tasks (see Table 3). Main effects for relationship were found for 12 of the 13 tasks (all but nursing care). This indicates that the pattern of caregiving differs by relationship of the elder; that is, certain tasks are performed with greater frequency for certain relations than for others. For example, one-way analyses of variance (not shown) reveal that transportation is provided significantly more often to parent-in-laws than to other nonspouse, nonparent relatives (p < 1). Also, spouses spend significantly more hours per week caregiving than the other groups (15.75 hours compared with 5.8 for parents, 5.1 for parents-in-law, 4.4 for friends, and 3.64 for relatives other than spouses, parents, and parents-in-law) (p < 1). Table 3 also demonstrates that there are gender by relationship interactions for transportation, housekeeping, meals, helping with expenses, and arranging health and social services. Thus, the patterns, or types, of care provided to different care recipient groups (i.e., spouses, parents, parents-inlaw, other relatives, friends) differ for male and female employees. Moreover, with respect to level of involvement, as measured by average weekly hours spent in caregiving, gender differences also vary by relationship, that is, there are gender x relationship interaction effects. Specifically, males spend one-third as much time as females in spousal care (7.4 hours compared with 23.1 hours), and about two-thirds the time spent by women in the care of a relative who is not a spouse, parent, or parent-in-law (2.7 hours compared with 4.1 hours). Interestingly, men spend almost identical amounts of time caring for parents-in-law and parents (4.4 hours and 4.3 hours, respectively), whereas women spend about one hour less caring for parents-in-law than for parents (5.7 hours and 6.7 hours, respectively). These findings reveal the importance of including both main and interaction terms for relationship to the elder and gender in analyses that address gender differences in effects of caregiving. Findings for the third question, which concerns gender differences in the effects of caregiving when other key predictor variables are controlled, are summarized in Table 4. The hierarchical regression analyses conducted reveal that relationship to the elder (Step 2) adds between 1% and 4% to the explained variance for each of the five dependent measures (chronic elder care stress, recent elder care stress, time taken off work, reduced work effectiveness, and difficulty combining work and family). The most consistent findings here are that care for a friend or a relative who is not a spouse, parent, or parent-in-law creates fewer negative outcomes than does care for a parent. Similarly, care for a parent-inlaw is less likely than care for a parent to result in time taken off from work or reduced work effectiveness due to caregiving worries. Care for a spouse is the only relationship that is more burdensome than caring for a parent; it is associated with more chronic elder care stress than is care for a parent. Table 4 also shows that when gender is added to the analyses (Step 3), a small but statistically significant amount of the variance is explained for three of the measures, after controlling for other relevant variables. Specifically, gender explains an additional 1% of the variance for chronic elder care stress, reduced work effectiveness, and difficulty combining work and family. This finding is, therefore, supportive of the gender-role socialization hypotheses. None of the gender by relationship interaction terms, however, accounts for significantly more variance for any of the dependent measures. Thus, after controlling for variables shown in previous research to be associated with negative consequences of caregiving, relationship and gender show several main, but no interactive, effects. To further explore the gender-role socialization hypothesis, which would suggest that different factors will influence the stress outcomes for male and female employed caregivers, additional analyses (not shown) were conducted. Specifically, interaction terms combining each of the 20 control variables with gender were created and then added to the regression equations depicted in Table 4 for each of the five outcome measures. Thus, a total of 100 gender X control variable interaction tests (each of the 20 control variables x gender for each of five outcome measures) were conducted. Only four of these interaction terms, involving three of the outcome measures, were significant. None of these interaction effects with gender affected either recent elder care stress or time taken off work to do something for the elder. With respect to chronic elder care stress, ease in finding and managing care served as a resource more for women than for men (p = 3). Concerning reduced work effectiveness due to worry about the elder, only average number of hours spent caregiving each week interacted with gender, with spending more hours caregiving having a greater negative effect for women than for men on this outcome measure (p = 1). Stated differently, for men, spending more hours caregiving did not affect concerns about reduced work effectiveness due to worry about the elder. With respect to difficulty combining work and family, caring for an elder who frequently exhibited problem behavior created more difficulty for women than for men (p = 7). In addition, for women, being older made combining work and family easier (p = 5); age had no such effect for men. In sum, where there were interaction effects (in only four of 100 tests), there was a small but consistent pattern of effects being larger for women than for men. Vol. 37, No. 6,

9 Table 3. Analyses of Variance for Task Involvement by Gender and Relationship to Elder Spouse Parent Parent-in-Law Other Relative Friend M F M F M F M F M F ^Values" (n = 20) (n = 23) (n = 439) (n = 814) (n = 152) (n = 175) (n = 122) (n = 254) (n = 66) (n = 109) Gender Relation Tasks Mean" Mean" Mean" Mean" Mean" Interaction Personal/Health Care Nursing care Personal care Continuous supervision Social/Emotional Support Transportation Shopping Check on by phone Visit, give emotional support Household Chores House and yard maintenance Housekeeping Fix or bring meals Care Management Manage legal and financial affairs Help with expenses Arrange and manage health or social services Total Number of Tasks Performed Average Weekly Hours Spent Caregiving Responsibility for Care c * *** *** 5.55*** 3.77** 31.36*** 6.25*** *** 18.63*** *** 5.20*** *** *** 8.62*** 5.59*** 44.33*** 9.67*** 2.78* 9.58*** *** 3.54** *** 2.41* 20.22*** 13.85*** *** 18.74*** 6.29*** 48.23*** 62.79***.93 a 1 = Never or seldom, 2 = Several times a year, 3 = Once a month, 4 = A few times a month, 5 = Once a week, 6 = A few times a week, 7 = Daily. b Main effect: Gender Main effect: Relationship Gender x Relationship interaction C 1 = Secondary, 2 = Shared equally with others, 3 = Primary, 4 = Sole. *p<;**p<;***p< The Gerontologist

10 Table 4. Hierarchical Multiple Regression Analyses Predicting Consequences of Caregiving for Employed Caregivers Chronic Elder Care Stress ( n = 1,567) Recent I^Ider Care Stress i(n = 1,572) Time Taken Off Work (n=1, 554) Reduced Work Effectiveness (n = 1,573) IDifficulty Combining Work and Family (n = 1,547) Independent Variables BI SEB B SEB B SEB B SEB Bi SEB 1. Control variables Age Race Has partner Household income Number of children Occupation Number of hours worked Number of elders caring for Elder's gender Elder's age Distance from elder Elder has difficulty with ADLs Elder exhibits problem behaviors Average number of hours caring Total number of tasks performed Responsibility for care Ease in finding/maintaining care Family-friendly work policies Flexibility in work schedule Ability to continue caregiving Change in R 2 -** *** ***.20 * ***.07** -.19*** -.15*** ***.44*** *** *** *** ** * *** ** *** * -.15**.26*** - - *** - - -* * - ** *** *** *** -.20 *** * -.09*** -* *** * ** ***.11*** -*** -.17*** * -.13***.30 *** *** *** *** -.17 **.21*** 2. Relationship of elder Spouse Parent-in-law Other relative Friend Change in R 2 3. Gender Gender Change in R 2.47** **.19* *** *** -.82***.10 *** *** -** -.11** *** *** -.16** -.39*** *** *** * **.11 ** *** 4. Gender X relationship interactions Gender x spouse Gender X parent-in-law Gender x other relative Gender x friend Change in R Constant 2.45*** ** 3.87*** Overall R Overall F 43.97*** 23.50*** * *** 15.76*** Note: Coefficients for control variables are reported from the first step in the regression. Coefficients for relationship to elder variables are reported from the second step, and those for gender are reported from the third step. Two-way interaction terms (gender by relationship) are reported from the fourth step. Change in R 2 indicates additional variance explained by each subsequent step. *p <, ; **p < ; ***p < 1. Discussion Using data from a large sample of employees who are informal caregivers to elders, this study has examined gender differences in the level of performance of various caregiving tasks, the level of involvement in caregiving overall, and the consequences of caregiving for various facets of wellbeing. One important finding is that there are similarities as well as differences between male and female employees with regard to the types of caregiving tasks performed. These similarities may be more striking than the differences, given previous research that has emphasized the distinctions in task performance by male and female caregivers. In contrast to the bulk of previous caregiving research, which consistently has found patterns of assistance that are congruent with a gender-based division of labor (Stoller, 1990), the male and female employee caregivers in this study did not differ with respect to Vol. 37, No. 6,

11 the provision of personal/health care-related tasks or of care management tasks. Among this sample of employed individuals, these tasks seem to be assumed more equally by men and women. The relatively low average level of impairment of the elders receiving care, however, should be noted. Nonetheless, these findings are bolstered by Kramer and Kipnis (1995), who observed that gender differences in tasks such as providing financial assistance, arranging for services, and helping with home chores or repairs typically are not as dramatic as differences in other kinds of tasks (Finley, 1989; Stoller, 1983). Similarly, Miller and Cafasso (1992) found no gender differences in money management tasks, which are generally considered to be tasks performed more often by males. At the same time, consistent with previous research, the male employee caregivers in this study were less likely to perform social/emotional support-type tasks and to help with household chores (except for house and yard maintenance) than were the female employees. The findings concerning gender differences or similarities in the types of caregiving tasks performed, then, are mixed, not clearly lending support to either the gender-role socialization or the social role hypothesis. In considering these findings, it must be noted that the tasks examined in the present study and in most previous studies may be gender-biased in favor of those activities performed more frequently by women than by men (Coward, 1987; Dwyer & Coward, 1991,1992b; Kramer & Kipnis, 1995; Stoller, 1990). This point is particularly salient when considering gender differences in level of involvement as measured by the number of tasks performed. The bivariate analyses conducted in this study with respect to all three measures of level of involvement, however, are consistent with previous research on gender differences in level of involvement in caregiving: Female caregivers were unquestionably more involved in caregiving than their male counterparts, not only as measured by the number of tasks performed but also in average hours spent caregiving and level of responsibility for care. This finding is congruous with the gender-role socialization hypothesis. This study's findings also are consistent with previous research concerning the effect of the caregiver's relationship to the care recipient on the type and amount of care (Young & Kahana, 1989). Whereas prior research has focused on only one or two types of relationships, the present study extends this line of inquiry by examining differences in patterns and consequences of care for five different relationship groups: spouses, parents, parents-in-law, other relatives, and friends. The findings confirm the existence of different patterns of caregiving across relationship types. They also provide some evidence of gender by relationship interactions. Furthermore, relationship influences the consequences of caregiving after controlling for other factors. This study indicates that caring for a parent is more problematic than caring for a parent-in-law, other nonspouse relative, or friend. Only spousal care has more deleterious consequences for wellbeing than parent care, at least with respect to chronic elder care stress. This latter finding is inconsistent with the social role hypothesis, which would indicate greater differences associated with parent care than with spousal care. Our findings do suggest that caring for elders who are closely related by marriage or blood is associated with greater caregiver stress. Possibly this is because such relationships involve greater emotional attachment. Finally, the findings of this study demonstrate the persistent role of gender in some forms of caregiver stress, lending support to the gender-role socialization hypothesis. In these data, gender contributed to three of the five measures of caregiver stress over and above other personal characteristics of the caregiver, characteristics of the elder and the caregiving situation, and available resources. In other words, when the role and situational demands that are suggested by the social role hypothesis were controlled for, gender differences in stress remained. Thus, on balance, the results provide greater support for the gender-role socialization hypothesis than for the social role hypothesis. It is important to note, however, that although statistically significant, the differences in outcomes that can be attributed to gender alone are not large. Limitations and Implications for Future Research, Policy, and Practice The present study has several limitations that should be addressed in future research efforts. Of particular importance is the relatively low average level of impairment of the elders receiving care, thus limiting the generalizability of the findings of the study. Another limitation is the use of several single-item outcome measures, although this is compensated for by the size of the sample. Also, these data did not consider the employee's family structure. As suggested by Dwyer and Coward (1991) and Matthews (1995), future research that examines care provided to parents and parents-in-law needs to account for the family configuration of the caregiver (e.g., whether the caregiver is an only child, has male siblings only, has female siblings only) and degree of choice as to whether to provide care. Furthermore, this study did not appraise the quality of the relationship between the caregiver and the elder or, for those providing care to parents, the sense of filial obligation of the caregiver. Most previous research, as well, has suffered from this limitation, as noted by Barber (1988), Finley (1989), and Miller (1987). One exception is a study by Lechner (1992), which found that a positive relationship between the employed caregiver and the parent was the best predictor of sustained commitment to caregiving. Quality of the relationship between the caregiver and the care recipient clearly is a variable that should be included in future research on gender differences in caregiving tasks and stress. Similarly, this study did not address the meanings 814 The Gerontologist

12 associated with caregiving. Gender role explanations may affect not only the meaning assigned to the caregiving experience (Miller & Cafasso, 1992), but even the definition of caregiving (Walker, Pratt, & Eddy, 1995). Nurturance, performance of personal care, and household duties may define caregiving as a female activity, yet caregiving also involves supervisory activities and issues of authority (Miller & Cafasso, 1992). Longitudinal research is needed (Dwyer & Coward, 1992b; Horowitz, 1992), and additional work remains to be done to understand the meanings attached to caregiving (Horowitz, 1992). Moreover, future research should explore the connotations associated with care for elders of various relationships (e.g., parent, spouse, friend) to the caregiver. Both qualitative and quantitative research, cross-sectional and longitudinal, to explore the contribution of the effects of family structure, the meanings of caregiving, the quality of the relationship between the caregiver and the care recipient, and sense of filial obligation will undoubtedly inform and improve the study of caregiving. Despite its shortcomings, the present study has important implications for policy and practice. One set of implications concerns the targeting of specific populations who appear to be at increased risk of negative effects of caregiving. Employed women are more vulnerable than employed men to the stress of caregiving. Also, employed caregivers who are assisting spouses and parents are more vulnerable to stress than those who are providing care to elders of different relationships. Taken together, these two findings suggest that services to employed caregivers should be particularly focused on those caring for spouses and parents, and especially on female caregivers. Moreover, intervention should occur not only in later life, but also during the formative years of childhood. The way in which maturing boys are socialized to the provision of care must be examined. Boys and girls alike should be provided opportunities and encouraged to nurture and care for others. This study also found that caregivers who had work and community resources experience less stress than did those with less adequate resources. These robust findings have implications for both workplace-based and community services. Work organizations with supervisors who are sensitive to the needs of caregiving employees and organizations that have policies that provide flexibility in the scheduling of work hours, as well as workplaces that offer specific services for caregivers, will have employees with less caregiver stress. Community-based social service agencies can also be valuable sources of help. Since difficulties in finding and maintaining care and inability to continue caregiving without additional support are related to more stress among employed caregivers, social service agencies should focus their efforts on helping employees find and manage satisfactory care arrangements. Extending the hours during which offices are open and recognizing elders' families as well as the elders themselves as the unit of care are examples of ways to enhance the accessibility and utility of community-based services to individuals who are employed. In conclusion, this study expands understanding of the importance of gender and relationship in caring for elders. In particular, it explicates the patterns and consequences of caregiving by employed women and men to spouses, parents, parents-inlaw, other relatives, and friends. Future research needs to augment this important area of investigation by exploring family configuration, relationship quality, and the connotations of caregiving. References Anastas, J., Cibeau, J., & Larson, P. (1990). Working families and eldercare: A national perspective in an aging America. Social Work, 35, Barber, C. E. (1988). Correlates of subjective burden among adult sons and daughters caring for aged parents. Journal of Aging Studies, 2, Barber, C. E., & Pasley, B. K. (1995). Family care of Alzheimer's patients: The role of gender and generational relationship on caregiver outcomes. Journal of Applied Gerontology, 14, Barusch, A. S., & Spaid, W. M. (1989). Gender differences in caregiving: Why do wives report greater burden? The Cerontologist, 29, Brody, E. M., & Schoonover, C. B. (1986). Patterns of parent-care when adult daughters work and when they do not. The Gerontologist, 26, Cantor, M. H. (1983). Strain among caregivers: A study of experience in the U.S. The Gerontologist, 23, Chang, C, & White-Means, S. (1991). The men who care: An analysis of male primary caregivers who care for frail elderly at home. The Journal of Applied Gerontology, 10, Chapman, N., Ingersoll-Dayton, B., & Neal, M. (1994). "Balancing the Multiple Roles of Work and Caregiving for Children, Adults, and Elders." In C. P. Keita & J. J. Hurrell, Jr. (Eds.), Job stress in a changing workforce: Investigating gender, diversity, and family issues (pp ). Washington, DC: American Psychological Association and the National Institute for Occupational Health and Safety. Coward, R. T. (1987). Factors associated with the configuration of the helping networks of noninstitutionalized elders. Gerontological Social Work, 10, Doty, P. (1995, August). Eldercare: The impact of family caregivers' employment on formal and informal helper hours. ASPE Research Notes, 14. Dwyer, J. W., & Coward R. T. (1991). A multivariate comparison of the involvement of adult sons versus daughters in the care of impaired parents. Journal of Gerontology: Social Sciences, 46, S259-S269. Dwyer, J. W., & Coward, R. T. (1992a). Gender, family, and long-term care of the elderly. In J. W. Dwyer & R. T. Coward (Eds.), Gender, families, and eldercare (pp. 3-17). Newbury Park, CA: Sage. Dwyer, J. W., & Coward, R. T. (1992b). Gender and family care of the elderly: Research gaps and opportunities. In J. W. Dwyer & R. T. Coward (Eds.), Gender, families, and eldercare (pp ). Newbury Park, CA: Sage. Executive Office of the President, Office of Management and the Budget (1987). Standard Industrial Classification Manual. Order No. PB For sale by the National Technical Information Service, 5285 Port Royal Road, Springfield, VA Finley, N. J. (1989). Theories of family labor as applied to gender differences in caregiving for elderly parents. Journal of Marriage and the Family, 51, Fitting, M., Rabins, P., Lucas, M. j., & Eastham, J. (1986). Caregivers for dementia patients: A comparison of husbands and wives. The Gerontologist, 26, Gutmann, D. L. (1980). Psychoanalysis and aging: A developmental view. In S. I.Greenspan & G. H. Pollock (Eds.), The course of life (Vol. 1). Bethesda, MD: National Institute of Mental Health. Horowitz, A. (1985). Sons and daughters as caregivers to older parents: Differences in role performance and consequences. The Gerontologist, 25, Horowitz, A. (1992). Methodological issues in the study of gender within family caregiving relationships. In J. W. Dwyer & R. T. Coward (Eds.), Gender, families, and eldercare (pp ). Newbury Park, CA: Sage. Ingersoll-Dayton, B., Starrels, M. E., & Dowler, D. (1996). Caregiving for parents and parents-in-law: Is gender important? The Gerontologist, 36, Kramer, B. J., & Kipnis, S. (1995). Eldercare and work-role conflict: Toward an understanding of gender differences in caregiver burden. The Gerontologist, 35, Lechner, V. M. (1992). Predicting future commitment to care for frail parents among employed caregivers. Journal of Gerontological Social Work, 18, Matthews, S. H. (1994). Gender and the division of filial responsibility be- Vol. 37, No. 6,