g WWll"ll‘lIWI‘UIHHWHINHUI!)WWW 295 M T TE U NERSITY LIBRARIES l1: E lmuulumrm m uwnulm mm 3 01389 7719 ll HI 1 i This is to certify that the thesis entitled HUSBANDS' PERFORMANCE OF DAILY HOUSEWORK: A RE-TEST OF THE RESOURCE HYPOTHESIS presented by MERIDETH R. TRAHAN has been accepted towards fulfillment of the requirements for MA degree in Sociology 014/4 WWW Major professor Date Mo“) 1" \‘Q‘ 0-7639 MS U is an Affirmative Action/Equal Opportunity Institution LIBRARY Michigan State Universlty PLACE It RETURN BOX to romovo this chookom hum your record. TO AVOID FINES Mum on or baton doto duo. DATE/Q“ DATE DUE DATE DUE MSU lo An Al'firmotivo Action/Equal Opportunity lnothton ans-m ,, 7w _,7 , , A , 7 , , a. _ ._ __ _ . .h _._—E v- HUSBENDS' PERFORMANCE OF DAILY BOUSEWORR: A.RR-TRST OF THE RESOURCE HYPOTEESIS By thridath a. Trnhan A.TERSIS Submitted to Michigan State Uhivorsity in partial fulfillnont of the requirements for the degree of Mister of Arts quartnont of Sociology 1996 ABSTRACT HUSBANDS' PERFORMANCE OF DAILY HOUSEWORK: A RE-TEST OF THE RESOURCE HYPOTHESIS BY Merideth R. Trahan This study tests the ability of the resource hypothesis to predict married men's performance of household tasks. I propose that the variable results from past studies have been due to the inconsistent operationalization of the relative income measure. My findings indicate that younger and non-white men are more likely to do housework. Consistent with my prediction, the different formulas used to measure relative income did alter the outcomes. These results imply that socioeconomic resources only partially explain the relationship between gender and domestic work patterns. ACKNOWLEDGMENTS This paper would not have been possible without the guidance of my major advisor, Clifford Broman, the role modeling of my parents, Carole and Edward Trahan, and the constant support of the Sociology Graduate Students. TABLE OF CONTENTS LIST OF TABLES .......................................... V INTRODUCTION ............................................ 1 THE RESOURCE HYPOTHESIS ................................ 4 Theoretical Origins ....... ..............I .......... 4 Previous Research on Housework ..... ..... ........... 7 Methodological Limitations of Past Research ....... 11 METHODOLOGY ....................................... ....115 Hypotheses ........................................ 15 Subjects .......................................... 15 Independent Variables ............................. 16 Dependent Variable. ............................... 18 Analysis .......................................... 20 FINDINGS ............................................ ...21 Correlation Matrix ................................ 21 Regression Models. ................................ 21 SUMMARY OF FINDINGS .................................... 24 DISCUSSION ............................................. 27 CONCLUSIONS ............................................ 29 APPENDICES ............................................... Table 1: Descriptive Statistics ................. ..32 Table 2: Correlation Matrix Results .............. .33 Table 3: Regression Models 1 and 2 ................ 34 Table 4: Regression Models 3,4 and 5 ............ .35 Table 5: Regression Models 6 and 7.... ........... .36 BIBLIOGRAPHY ........................................... 37 iv Table Table Table Table Table wNH LIST OF TABLES Descriptive Statistics ..................... 32 Correlation Matrix Results ................. 33 Regression Models 1 and 2... ............... 34 Regression Models 3, 4,and 5 ............... 35 Regression Models 6 and 7 ........... . ...... 36 INTRODUCTION Regardless of increases in women's labor market participation and a growing acceptance of less traditional gender roles by both women and men, women continue to perform more housework than their male partners. Many family scholars have used the resource hypothesis as a conceptual basis for understanding this unequal division of domestic labor (Geerken and Gove, 1983; Coverman, 1985; Berardo et al., 1987). The hypothesis predicts that domestic power, which is defined as the ability to influence ‘ household decisions, is related to the market resources (education, employment status and income) that each spouse brings to the marriage. Such that, more resources can be exchanged for greater decision making power in the family. Although the evidence demonstrating the influence of market resources on family power is inconclusive, it continues to be tested with and against alternative theoretical models such as time availability (Ross, 1987; Presser, 1994), demand/capabilities (Coverman, 1985), gender-role ideology (Bird, Bird and Scruggs, 1984; Ross, 1987), family life course (Coltrane and Ishii-Kuntz, 1992; Rexroat and Shehan, 1987), and more recently perceptions of fairness (Thompson, 1991; Blair and Johnson, 1992; Mederer, 1993). All of these models have contributed to our understanding of the symbolic meaning of housework and the effect of practical demands on both partners' available time for housework. However, the resource hypothesis remains important for sociologists because it tests the relationship between macro level structures - education, employment and equitable earnings - and the micro level experience of daily family life. In other words, it tests the strength of the relationship between structural indicators of gender equality and the actual level of gender equality that occurs within the marital dyad as demonstrated by the division of daily housework. Additionally, the resource hypothesis is based on an economic theory which assumes that the market influences family power. In the United States, this framework is logically seductive because our egalitarian ideology supports the premise that rising resources should result in increased gender equality. Although the resource hypothesis is widely used to understand inequalities in domestic power, the methods employed to operationalize the variables tested by the hypothesis have been inconsistent. Specifically, a number of different formulas have been used to create the variable "relative income" resulting in varied conclusions. This study explicitly tests the predictive value of three different methods that scholars have used to measure the impact of the resource "relative income" on husband's contributions to household work. In particular, I will compare the income ratio method used by Coverman (1985), the income snhttaction method tested by Ross (1987), and the income oropottion method demonstrated by Spitze (1986). These analyses also extend previous work in three other areas. First, small convenience samples of white, middle- class families have been overly relied upon in past studies (Spitze, 1988; for exceptions, see Coltrane and Ishi—Kuntz, 1992; Blair and Lichter, 1991; and Brayfield, 1992). This study is based on a large, nationally representative survey, Which over-sampled African Americans and includes data on all income levels. Using a national sample will reduce the potential for regional biases and will increase the generalizability of my findings. Second, data for this study were collected in 1986 while previous studies which critically looked at the operationalization of relative income were based upon 1970's data (Coverman 1985; Spitze, 1986; Ross, 1987). (For an exception based on a Canadian sample see Brayfield, 1992). Because the distribution of socio-economic resources is not static, I propose that re-testing the resource hypothesis over time is critical. For example, the U.S. Census Bureau report (1993) indicates that between the mid 1970's and 1980's women's absolute resources have improved. The percentage of women in the workforce increased from 44 percent in 1975 to 54 percent in 1985. In addition, the number of women who completed four years of college or more also increased from 11 percent to 16 percent during that time frame (U.S. Bureau of the Census, 1993). Since the resource hypothesis is based on these structural predictors which are aggregated in demographic reports, it provides an excellent tool to pose questions about how these macro-level trends have trickled down to shifts in gender roles at home. Finally, because the analyses are based on the first wave of the American's Changing Lives Survey, it creates a baseline for a future longitudinal study to investigate the changes in men's performance of household work over time. TEE RESOURCE HYPOTBESIS T] l' J D . . The resource hypothesis (also called the resource theory of family power) predicts that in the marital dyad, the power between spouses is a result of their labor market resources as individual contributors to the family (Blood and Wolfe, 1960; Scanzoni, 1972; Thibaut and Kelley, 1959). This framework of power was developed in Blood and Wolfe's (1960) general study of marital power which stated that "power may be defined as the potential ability of one partner to influence the other's behavior. Power is manifested in the ability to make decisions affecting the life of the family" (p. 11). Blood and Wolfe argue that because couples decision making processes vary within a culture, tradition or a patriarchal system only partially explain the source of male power. Instead, the sources of power are the comparative resources that the husband and wife bring to a marriage. According to their conceptualization of power, "a resource may be defined as anything that one partner may make available to the other, helping the latter satisfy his needs or attain his goals (p. 12). Therefore, social resources that determine socioeconomic status such as education, earnings, and occupational status are the external resources which at least partly determine the power structure in the marital relationship (Blood and Wolfe, 1960). This framework of resource-based family power has played a significant role in the study of household labor. Domestic work has been "characterized as a domain, a sphere of influence, or a locale for the exercise of power, decision-making, and conflict" (Berk,1985, p.12). In the Blood and Wolfe study, "resources" and the "resource theory" were empirically tested and the division of household labor was one example of family power processes being executed. They stated that in regard to the division of labor, the resources included 1) the time available to devote to household tasks and 2) the relative skill of the person to accomplish them (1960, p.74). ‘ However in the majority of current research applying the resource hypothesis to household labor, socioeconomic variables have become the most frequently utilized direct measures of power (Berk, 1985). In recent studies, the division of household labor is generally analyzed as either the number of hours worked or the types of task assigned to each spouse. The absolute resource hypothesis predicts that the spouse with greater earnings, more education and a higher status occupation would contribute fewer hours or be less apt to perform undesirable tasks because of the greater value of his or her resources. According to the relatita resource hypothesis, the greater the wife's resources are relative to her husbands, the more likely her husband is to perform household work. Since men have been in the privileged situation of having higher-paying, more prestigious jobs resulting in more status and material resources than women, husbands have exchanged these resources for the greater power to limit their performance of undesirable activities such as housework (Scanzoni, 1972; Perrucci et al. 1978). Therefore, the theory concludes that if a wife could achieve a position of comparable status and income to her husband, the unequal sharing of household responsibilities and valuing of careers would be eliminated. Berk (1985) argues that the resource hypothesis is flawed because people do engage in housework willingly and receive some satisfaction from their work. This contradicts the assumption that housework is both undesirable and requires force. My conceptualization of housework as an indicator of power is more closely aligned with Heidi Hartmann (1981) who contends that housework is a source of conflict, and therefore, an effective measure of power relations (p.368). In her seminal article on household labor, Hartmann (1981) critiques the notion of the unified family. Instead, she conceptualizes the family as a "locus of struggle" (p. 368) and a place where men exercise patriarchal control over women's labor. She cites evidence from empirical studies which conclude that women spend more time on housework and childcare regardless of their employment status or class position. Based on Hartmann's conception of housework as undesirable and men's continued ability to resist performing these tasks, I argue that the allocation of daily domestic work is a reflection of family power. 2rnIiono.3nanarch_on_nonsenork Household labor studies which have tested the impact of socioeconomic resources have provided highly variable, and thus, inconclusive results. Spousal Income: Two measures of domestic power frequently tested are absolute and relative income. Absolute income refers to the individual amount each spouse contributes to the family, while relative income is the relationship or gap between the husband and wife's income. Several researchers have demonstrated that the wife's absolute income has a positive effect on the amount that her husband shares in housework (Peterson and Maynard, 1981; Ross, 1987; Spitz, 1988). Blair and Lichter (1991) found that relative earnings (wife's income relative to her husband's income) were related to men's overall contributions and the types of tasks allocated. Ross (1987)also reports that the smaller the gap between a husband and wife's earnings, the greater his involvement in household work. Contradicting these findings, Brayfield (1992) found that a relative income advantage for women did not meaningfully reduce the feminine-typed chores that they perform regardless of the income level. However, men's income advantage did result in a reduction of their contribution to female-type tasks and to a greater extent for men with lower rather than higher absolute incomes. She concluded that the relative financial advantage may be more meaningful for low-income couples. Huber and Spitze (1983) also reported that relative income was not related to the husband's household work. This was reconfirmed by Spitze (1988), who again demonstrated that only absolute resources such as earnings or occupational status measured alone, have been linked to more equal contributions. Ferree (1988) concluded that wives who define themselves as family providers (a traditionally male role) often have husbands who do more housework because they are more likely to feel entitled to help. Employment and Occupational Status: Most studies have found that the status of the wife's occupation is less significant than the fact that she is employed. Huber and Spitze (1981) stated that both spouses' perceptions of the domestic division of labor are influenced more by the wife's present employment status than by her work attachment or income. Husband with employed wives, regardless of their wife's occupational status perform more housework. . Even studies not entirely supporting the resource hypothesis have found that behavioral changes may occur because of pragmatic necessity such as employment rather than relative resources or ideological shifts. Coverman (1985)reported that neither absolute nor relative resources of husbands and wives increased men's contributions. Instead, it was practical demand which increased men's performance of housework - namely the number of children, number of hours spent at their work, and spouses' employment status. Blair and Lichter(1991)found that the wife's employment status was also a significant predictor of husband's work. However, the significance of employment status as a predictor is not uncontested. Even for couples with children at home, the husband's total housework time did not vary by wife's employment status, and it had little effect on his performance of specific tasks. That is, he did not spend more time cleaning dishes or doing tasks when his wife was working (Shelton, 1990; Rexroat and Shehan, 1987). 10 The wife's occupational status has also been an inconsistent predictor of husband's household contributions. No significant differences in domestic labor patterns between dual-career and dual-earner couples were found by Berardo, Shehan, and Leslie (1987), which indicates that the status level of the wife's occupation was not an important determinant. Presser (1994) reported that variations in employment schedules such as the availability of shift and weekend work, resulting from the growth of the service economy, are significant determinants of the husband doing traditionally female tasks. These patterns parallel Beer's (1983) conclusion that "a man's class and class background are less important than the flexibility of time provided by his immediate occupational circumstances" (p. 43). Education: Unlike Coverman (1985) who reported that education was not a significant predictor of the overall hours men spent performing household chores, Presser's study found education was a determinant of housework but in the opposite direction as predicted. The more education the husband had resulted in greater hours spent on housework instead of fewer. Presser interpreted this finding as evidence that using education as a direct measure of power may not be accurate since educational attainment also reflects values which could include a less traditional gender role ideology. Gender Role Ideology: .5. 11 Although other studies have operationalized gender ideology as a resource, it is usually in combination with the wife's employment status and income. When evaluating the effect of sex-role orientation, role salience, income and family type, Bird et al. (1984) demonstrated that the husband's sex-role ideology plus the income and job status of his wife were greater predictors of men preparing meals and cleaning than the wife's sex—role orientation or role salience. The tangible resources of income and employment influenced wife's behavior and expectations, while for husbands it was a mix between their spouse's employment and their own sex-role orientation. On the other hand, Coltrane and Ishii-Kuntz (1992) reported that it was the wife's gender role ideology and her contribution to the household income and not her husband's attitudes about appropriate roles which influenced the division of labor in couples who have early birth children (defined as having children before age 28). As the evidence above suggests, the effectiveness of the resource hypothesis to explain husband's performance of household labor is varied and inconsistent. In part, the variance may be due to the unstandardized operationalization of the resources which are being included in the model. Initially, I will discuss some overall methodological issues which may confound the findings. And finally, I will explain in more detail the three methods of operationalizing 12 the relative income variable which will be tested in this study. Regression models which include both absolute and relative resource variables may be biased due to the multicollinearity of the measures. More clearly, because the relative income measure is a mathematical formula which uses the husband and wife's absolute or individual income as its components, the measures are intrinsically related. Or, the relative income measure is dependent upon the absolute income data. On the other hand, models which test absolute income measures alone only compare men to other men of the same income level men instead of comparing them to their wives. Given this comparison, a man's household behavior may be more reflective of his gender role ideology resulting from the relationship between education and income instead of concluding that his behavior is a result of a spousal resource advantage. Or, as Coverman (1985) states "the answer does not lie in expectations concerning power or economic relations within the marital dyad but rather in - expectations regarding ideological differences between professional and nonprofessional men" (p. 83). The absolute income variable measures the independent effects of each person's income, and the relative income variable tests the gap between the husband and wife's incomes. Models which include only the relative income variable may produce results that are confused by the 13 effects of absolute indicators when they are not controlled for in the equation. Given these inconsistencies, it is not evident if the appropriate model of the resource hypothesis should include absolute, relative or both absolute and relative measures of spousal resources. It also remains unclear how meaningful absolute or relative income measures are to the explanatory value of the model. Below are the three methods commonly used to test the relative income measure. The first formula was the ratio method used by Coverman (1985) who defined relative income as the wife's income divided by her husband's income. Her study concluded that husband's time spent on domestic work was not related to his relative financial advantage. Because Coverman controlled for the husband and wife's absolute income level, the effect of the relative income measure was not confused with the effects of the absolute incomes. Second, Ross (1987) used the subtraction method which was operationalized as the husband's income less his wife's income. She found that the greater the income gap was in' the husband's favor, the less his relative participation in housework. Since this model incorporates the power perspective which assumes that the wife's earnings and the husband's earnings have opposite effects on the distribution of household work, Ross argued that this method was more accurate. Unlike the ratio technique, the subtraction method uses the same units of measurement as the absolute l4 indicators (Brayfield, 1992). Ross, however, did not control for husband's or wife's absolute income, therefore, the effects she found for relative income may be confounded with the effects of absolute earnings. A third analytic strategy, which I label the proportion method was used by Spitze (1986) who re-defined relative income as the wife's proportion of the couple's total income. She did not find any evidence that higher relative earnings were related to men's household labor. Given the number of alternative methods used to test the relative income portion of the resource hypothesis, the relationship between the influencing power of absolute income and relative income on husband's contributions to housework remains unclear. Also, it is unspecified if the different operationalizations of the relative income measure would significantly alter the findings. As I previously discussed, re-testing the resource hypothesis is also crucial because it, unlike other theories, specifically tests the relationship between structural or socioeconomic factors and "doing gender" (West and Zimmerman, 1987) at home. In addition, studies based on nationally representative, longitudinal surveys increase the generalizability of the findings and provide a baseline to explore men's household labor behavior over time. 15 METHODOLOGY Hypotheses As the empirical studies discussed evidence, the ability of the resource hypothesis to predict men's contribution to household labor remains inconclusive. Although I recognize that resources such as income, education and employment status do not capture the complex symbolic meaning of housework or the negotiation process which occurs within the marital dyad, they do represent the traditional structural barriers to gender equity. This study will test the resource hypothesis both theoretically and methodologically. Theoretically, it will investigate the overall effectiveness of the resource hypothesis to predict husbands' performance of traditionally female tasks by testing the absolute resources education, income and wife's employment status and the relative income resource. Methodologically, it will compare the differential effects of three methods of measuring the relative income variable ratio, subtraction, and proportion). Each of the relative income measures will be tested both alone and controlling for the absolute income measures. 31111132111 The data used in this analysis are from Wave I of the Americans' Changing Lives Survey (ACL). Under the direction of the Survey Research Center at the University of Michigan, 16 the data were collected in 1986 using a face-to—face survey of a national probability sample. There were 3,617 respondents in total, of which approximately 25 percent were married men. Respondents had to be over the age of 25 to participate and the overall response rate for the survey was 76 percent. The data are drawn from a multi-stage probability sample of people living in noninstitutionalized housing in the continental United States. Persons over the age of 60 and African Americans were sampled at twice the rate of whites and people aged 40-59. Please see House (1986) for more detailed information. For this paper, I created a subsample of married men which consisted of 887 respondents. Because of missing data, the final sample size was N= 656. All data and analyses presented in this paper are based upon this subsample. W Independent variables operationalized for this analysis include absolute income, relative income, education and wife's employment status. Because the spouse's occupational status was not found to be a significant predictor (Berardo, Shehan, and Leslie, 1987), it will not be tested as resource. The relative income measures were developed by first creating the variable wife's income. Wife's income was computed by subtracting the husband's reported income from the total family income. Both the respondent income and total family income are ten category variables. For 17 example, 1= less than $5000, 5= $20,000 to $24,999, and 10 = $80,000 and above. This new variable (wife's income) was then used to develop the three relative income measures. The first measure is the ratio method (Coverman, 1985) which is the wife's income divided by the husband's income. Secondly, the subtraction method (Ross, 1987) is tested which is the difference between the husband's reported income and his wife's income. Finally, the wife's income as a proportion of the total family income was tested as the proportion method of relative income operationalization (Spitze, 1986). Age, race, education, and employment status are also used in the analysis. Age and education are measured in years. Wife's education, as reported by the husband was coded similarly. The variable race is dummy coded for the respondent such that 1 = non-white. Employment status for both the respondent and his wife is also dummy coded where = currently employed. On average, the subjects were older ( due to the over sampling of older adults) ~~mean age was 51. Both the husbands and their wives had at least a 12th grade education. Mean income was between $15,000-19,999 for the respondents. Sixty-nine percent of the husbands reported being currently employed, while 58 percent of their wives were employed. Approximately 30 percent of the sample were non-white. Please see Table 1 for a listing of the descriptive statistics. 18 bqpandnntIYIriahle Contrary to recent popular beliefs that families are shifting from complementary household work arrangements to more parallel responsibilities, domestic tasks are still highly segregated by gender. This occurs even in families where the husband contributes many hours at home, has non- traditional gender role attitudes, and both are highly educated (Blair and Lichter, 1991; Ferree, 1991). According to Berk (1985), gender segregation in and of itself means qualitatively different contributions for men and women. The family work traditionally performed by women is often repetitive, routine, at specific times and inside such as cooking, cleaning, or laundry. All of these daily tasks are necessary to the reproduction of family life. On the other hand, the traditional family work men generally do is infrequent, irregular, non-routine and often outside such as yard work and household repairs. Because the traditionally female-type chores are the time consuming daily tasks which must be accomplished to reproduce family life, the dependent variable, "household tasks" was measured by asking the husbands if they perform any of the following chores: 1) prepare food or wash dishes, 2) grocery shop, 3) clean or vacuum, and 4) do laundry. These were asked as dichotomous 'yes-no' questions, and the value of '1' was assigned each time the respondent indicated that 'yes' they performed that household task. The responses were then summed to form a l9 household task index which ranges from 0 to 4 with a mean of 2.65. The household task index reduces the potential bias effects of using single indicators of the husband's contributions to traditionally female household work. It provides a continuum of his overall contributions to reproductive labor instead of too narrowly focusing on one task which may have been negotiated based on the husband's preference. It also provides a wider range of variation on the household tasks measure and increases the measures explanatory power (Babbie, 1990). The Alpha Coefficient test score was .69 which indicates that the scale is reliable. Although I recognize that data limited to dichotomized response categories do not take into account the frequency of performance, the quality of the work, or if his contribution was self-initiated, the questions do capture the husband's overall response. In its simplicity, and given the limitations of asking respondents to recall details about potentially unsalient behaviors (Foddy, 1993), the closed category may provide a more accurate picture of typical behavior patterns. Sudman and Bradburn (1982) discuss the potential for "social desirability" bias which may result in husband's over-reporting that they perform the task if they believe they "should" be responsible for housework. However, there was significant variance in the number reporting that they 20 do housework such that: 44 percent reported doing laundry, 74 percent prepare meals or wash dishes, 72 percent do grocery shopping, and 69 percent clean or vacuum which indicates that the items captured differences in behavior (Babbie, 1990). Analysis First, a Pearson Correlation Matrix was performed to evaluate the strength of the relationships between variables. A series of three Ordinary Least Squares (OLS) regressions were performed to test the hypotheses. Regression model 1 regressed household tasks on the control variables (husband's race, education, age and income). Model 2 included the wife's absolute resource predictors (income, wife's education and wife's employment status). The relative income measures (ratio, subtraction and proportion) were included consecutively in models 3, 4 and 5 without the absolute income variables. Finally, controlling for the absolute income of the husband and the wife, regression models 6 and 7 included both the absolute and relative measures of income. Because of multicollinearity, the income subtraction measure could not be regressed with the absolute income measures. This may explain why Ross (1987) did not control for absolute income in her study. 21 FINDINGS Correlationcuatrix Table 2 presents the findings from the Pearson Correlation Matrix computed to assess the relationship between the household tasks and the independent variables. As shown, the strongest relationships were between household tasks and wives' employment status (r=.23, p<.001); the husbands' age (r= -.25, p<.001); husbands' education (r=.31, p<.001); and the wife's education (r=.28, p<.001). The absolute income measures wife's income (r=.l4, p<.001) and husband's income (r=.11, p<.01) were significant unlike the relative income measures. Although not shown in this Table, the correlation between husband and wife's education was strong (r=.63, p<.001). Regressioncubdals Table 3 presents the first and second regression models of household tasks on the control variables. The findings in the first model do not support the resource hypothesis. The husband's education was significant but in the opposite direction than the hypothesis predicts; more educated husbands are more likely to report performing traditionally female household tasks. Age and race were also significant predictors. Younger men report greater contributions as do non-white men. Racial differences in men's contributions to domestic work have been documented by earlier studies (Shelton and John, 1993; Ericksen et al., 1979). Contrary 22 also to the hypothesis, the husband's absolute income was not a significant indicator. (Results which are not presented show that neither the husband's employment status nor the number of children in the household were significant predictors, therefore they were not included in the models). In the second model which added the wife's resources as predictors, the only significant variable for the wife was her employment status; husband's with employed wives perform more work. Neither the wife's educational level nor her absolute income contributed to the explanatory value of the model. Therefore, the hypothesis was only partially supported. The fact that the wife was employed was a better indicator of egalitarian arrangements regardless of her income or educational resources. The amount of variance in husband's contributions was better explained by the second model which included the wife's resources as evidenced by the increase in the R square from .124 to .144. Table 4 demonstrates the results of regressing the relative income measures (model 3 - ratio_mothod; model 4 — proportiommethod; and model 5 — the subtractimethod) without controlling for the direct effects of the husband or wife's absolute income. None of the relative income measures were significant indicators contrary to the resource hypothesis. All of the previously significant predictors (husband's age, race education, and wife's employment status) maintained their status, although the magnitude of significance was slightly reduced. In 23 addition, the inclusion of the relative income variables also revealed for the first time a significant relationship between the wife's education and the husband's contribution in all three models; the greater the wife's education the more likely that her husband reported performing household work. This change can be attributed to a suppressor effect which means that the correlation between the wife's education and the relative income measures is such that when both are in the equation, the shared variance allows the relationship between household tasks and wife's education to be observed. I conducted further analyses to compare the prediction value of the relative income measures controlling for the absolute income of the husband and the wife. The sixth regression model which included the income ratio variable, was not a significant predictor, nor did it affect the significance of the previous indicators. Because the income subtraction variable was developed from the absolute income measures, multicollinearity prevented me from being able to test it in this model. Unlike the first two regressions including the relative income variables, the income proportion measure in model 7 proved to be a significant predictor of household work - but in the opposite direction as the hypothesis would expect; the lower the wife's proportion of the total family income, the more likely her husband was to contribute to household tasks. 24 The inclusion of the income proportion variable also affected the significance of the absolute income measures but in the expected direction; the greater the wife's absolute income the more likely the husband reported performing the tasks. And the greater the husband's income the less likely he was to report doing traditionally female chores. Again, this change can be attributed to a suppressor effect which means that the correlation between wife's absolute income and the proportion measure is such that when both are in the equation, the shared variance allows the relationship between household tasks and wife's absolute income to be observed. The overall variance explained by this model increased as the R square was .154. SUMMARI OF FINDINGS Consistent with the previous studies, the evidence presented in this paper has shown that the ability of the resource hypothesis to explain the division of household labor is inconclusive. First, I examined only the husband's socioeconomic characteristics as predictors which revealed that younger and non-white husbands were more likely to perform the traditionally female tasks (laundry, dishes/meal preparation, grocery shopping, cleaning/vacuuming). This may reflect a generational shift in gender-role ideology (Bird, Bird and Scruggs, 1984), or 25 support the life course perspective that when men are younger they have more time to contribute because of their position in the workforce (Coltrane and Ishii-Kuntz, 1992; Rexroat and Shehan, 1987). Education was a significant indicator but in the opposite direction which indicates that defining education as a resource may not be an effective operationalization of marital power. Rather, educational level may more accurately reflect the values of an occupational status (Presser, 1994). The analyses then revealed that the wife's only resource that explained husband's performance of housework was her employment status. Similar to Coverman's findings (1985), the practical necessity of both partners doing housework in a dual-employment household may play a greater role than the resources either spouse contributes. Studies which incorporate time availability may better address the impact of employment status (Ross, 1987; Presser, 1994). Even when the wife's resources were included, the husband's education, age and race continued to be significant indicators. I Finally, consistent with my prediction, the way in which the relative income variable is operationalized did alter the findings. When I examined the independent effects of the three relative income measures (ratio, subtraction, proportion), the analyses did not find any of the predictors significant. This supports earlier arguments that both relative and absolute measures of income are necessary to 26 uncover the signifiCance of the inCome resource. The absolute income compares all men/women within an income category while the relative income measures test the significance of the income gap within the marital dyad (Spitze, 1986; Ross, 1987). However, when controlling for the absolute income effects, the only relative income measure that was a significant predictor was the proportion method; and that relationship was in the opposite direction than predicted. This finding completely contradicts the relative resource hypothesis. In regression model seven , controlling for the income proportion also revealed a significant effect of the wife's absolute income in the predicted direction. But the effect was reverse for the husband's absolute income. Further research is necessary to fully understand these results. A limitation of the study which may contribute to the findings is the dichotomous definition of employment status which does not take into consideration the number of hours or the potential effects of shift-work on household labor. Another potential shortcoming is that the analyses is based solely on responses from the husband. Since each partner's perceptions of how daily work is allocated may differ, responses from both spouses may have enhanced my ability to understand the complexities of the negotiation process. 27 DISCUSSION These results do not imply that improved structural resources do not influence the potential for more egalitarian domestic work patterns. Rather, they force us to reconsider the role that gender plays in how household labor continues to be distributed. In other words, a gender neutral or resource based approach alone only contributes partially to our understanding. Feminist and family scholar's critiques of the resource hypothesis address these limitations. First, they argue that it is reductionistic. In other words, to reduce the explanation of who performs housework to a rational exchange of valued resources leaves the symbolic meaning of the work out of the equation. It is too simplistic to assume that the differing levels of men's involvement in domestic labor could be entirely explained by operationalizing socioeconomic variables as resources to be exchanged (Ferree, 1990). Since women as wives and mothers have been socialized to attach a symbolic meaning to housework, women themselves have bought into domestic labor as part of their self-image, which may contribute to the asymmetrical relationships at home. They too see themselves as "good" wives by doing good housework (DeVault, 1990). A second limitation with the theory is that it assumes human behavior is based on a completely rational or cognitive model of decision making. But instead, people 28 often make behavioral choices based on non—rational emotions or because it is the traditional behavior. Therefore, they do not always engage in an active thought process. In the example of housework, a highly gendered division of labor is the traditional behavior (Berch, 1983). To support the theory's assumption of a completely rationalized process, one would have to prove that women are more efficient at performing traditional female tasks. Or, by not having to do housework, men's productivity in the public labor force is greater. Finally, the resource hypothesis is based on the premise that each actor's behavior is dependent upon assumptions of equality, fairness and rationality (Turner et a1, 1989). Although the resource hypothesis recognizes that there is a power differential between husbands and wives, it generally views resource-based power as operating in gender-neutral ways. Therefore, it is inaccurate to assume that the norms of reciprocity and exchange between husbands and wives are based on egalitarian principles when the broader social structures give women less power. Instead, the use of gender role ideology as a predictor of household labor contradicts the resource hypothesis. It assumes that those men who are more educated and in higher- status occupations more likely to perform more household work instead of less. Thompson and Walker (1989) state that "there is no simple trade-off of wage and family work hours between 29 husbands and wives, nor do partners allocate family work based on time availability" (p. 856). Because it cannot account for the meaning men and women attach to gender roles as a relative resource in the exchange equation, the explanatory value of the resource hypothesis is limited. Both the continued formal and informal structural support in the workplace for housework to remain women's work and the symbolic association of housework as women's work need to be problematized (Ferree, 1990). Only by viewing domestic labor as integral to society's definition of what it means to be a women can we understand why women, even when contributing equal resources to the family, are not excused from the expectation they are responsible for the daily, reproductive work at home (Spitze, 1988). CONCEUSIONS This study's contribution to our understanding of domestic labor is first methodological. As my findings show, the methods used to test the resource hypothesis may be flawed. Since the three operationalizations of the relative income measure resulted in different outcomes, further research needs to be done to re-test these measures and develop more consistent indicators. Also, because of the suppressor effects revealed in this study, it is important that both the absolute and relative measures of income are included in future models. These methodological 3O improvements would allow scholars to more accurately compare their findings when they test the resource hypothesis; methodological consistency may potentially reduce the variability of conclusions. Secondly, re-testing the theoretical strength of the resource hypothesis to explain the distribution of domestic tasks did not result in significantly different findings from previous research. Neither the absolute nor relative measure of the husband or wife's income consistently translated into family power. However, husbands who reported performing greater household work in this study were found to be younger, non-white, more educated and their wives were employed. In light of these findings, future studies using the resource hypothesis should reconsider the significance of income and focus on the influence of women's employment status as a source of domestic power. Although the resource hypothesis only partially illustrates the connections between external resources and family power, I maintain that it remains a valuable tool for family scholars. The strength of the theory lies in its - ability to test the relationship between macro-structural changes and daily family life. On one level I am not surprised at the results of my study. The resources tested - income, education and employment - are critical indicators of power. And until social institutions ensure equal access to these resources, we cannot expect to find egalitarian relationships in the family. However with the majority of 31 wives now in the paid labor force, husbands' ability to avoid performing housework is a daily reminder that gender remains a significant source of family power. APPENDICES TABLE 1. Descriptive Statistics Husband's Education Wife's Education Husband's Age Husband's Race (1=non-white) Husband's Income Wife's Income Total Family Income Income Subtraction Income Proportion Income Ratio Wife Employed (1=employed) Husband Employed (1=employed) Household Tasks Mean 12.04 12.14 50.73 .29 4.58 1.28 5.86 3.30 .23 .42 .58 .69 650 33 TABLE 2. Correlation Matrix Results HOUSEHOLD TASKS Income Ratio .04 Income Proportion .05 Income Subtraction. .03 Husband's Age -.23** Wife's Income .14** Husband's Education .31** Wife's Education .28** Husband's Income .11* Husband's Race .05** Wife's Employment .23** **p<.001 *p<.01 34 TABLE 3. Regression Models 1 and 2 Model 1 Husband's Education .113*** Husband's Age -.011*** Husband's Race (1=non-white) .251* Husband's Income -.029 Wife's Education Wife's Income Wife Employed (1=employed) Constant 1.934 N 655 R square .125 * p< .05 **p < .01 ***p < Model 2 .083*** -.008* .248* -.027 .046 .046 .244* 1.359 649 .144 .001 35 TABLE 4. Regression Models 3, 4 and 5 Model 3 Model 4 Husband's Education .076*** .077*** Husband's Age -.008* —.008* Husband's Race (1=non-white) .259* .259* Wife's Education .047* .047* Wife Employed (1=employed) .264* .260* Income Ratio .041 Income Eroportion .127 Income Subtraction. Constant 1.30 1.297 N 649 649 R square .140 .140 * p< .05 **p < .01 ***p < .001 Model 5 .084*** .245* .048* .255* -.032 1.356 649 .144 36 TABLE 5. Regression Models 6 and 7 Model 6 Husband's Education .087*** Husband's Age -.008* Husband's Race (1=non-white) .256* Husband's Income -.051 Wife's Education .046 Wife's Income .120 Wife Employed (1=employed) .251* Income Ratio -.250 Income Proportion Constant 1.425 N= 649 R square .148 * p< .05 **p < .01 ***p < .001 Model 7 .089*** -.008* .260* -.082* .047 .248** .268* -1.61** 1.624 649 .154 BIBLIOGRAPHY 37 BIBLIOGRAPHY Babbie,Earl. 1990. Survey Research methods. CA: Wadsworth Publishing Company. Beer, William R. 1983. HOusehusbands:.Men and Hbusework in American Families. New York: Praeger. Berardo, Donna Hodgkins, Constance L.Shehan, and Gerald R. Leslie. 1987. "A Residue of Tradition: Jobs,Careers, and Spouses' Time in Housework." JOurnal of'Marriage and the Family 49:855-66. Berch, Bettina. 1983. The Endless Day: The Political Economy of Human and Wbrk. New York: Hardcourt Brace Jovanovich. Berk, Sarah Fenstermaker. 1985. The Gender Factory. New York: Plenum Press. Berk, Richard A. and Sarah Fenstermaker Berk. 1979. Labor and Leisure at Heme. Beverly Hill, CA: Sage. Bird, Gloria W., Gerald A. Bird, and Marguerie Scruggs. 1984.”Determinants of Family Task Sharing: A Study of Husbands and Wives." JOurnal of .Marriage and the Family 46:345-55. Blair, Sampson Lee and Michael P. Johnson. 1992. "Wives Perception of the Fairness of the Division of Household Labor: The Intersection of Housework and Ideology." JOurnal of Marriage and the Family 54: 570-81. Blair, Sampson Lee and Daniel T. Lichter. 1991. "Measuring the Division of Household Labor: Gender Segregation of Housework Among American Couples." Jburnal of Family Issues 12:91-113. Blood, Robert 0. Jr. and Donald M. Wolfe. 1960. Husbands and Wives: The Dynamics of.Married Living. Glencoe, IL: The Free Press. 38 Brayfield, April A. 1992. "Employment Resources and Housework in Canada." Journal of'Marriage and the Family 54:19-30. Coltrane, Scott and Masako Ishii-Kuntz. 1992. "Men's Housework: A Life Course Perspective." JOurnal of Marriage and the Family 54:43-57. Coverman, Shelley. 1985. "Explaining Husbands' Participation in Domestic Labor." Sociological Quarterly 26:81-97. DeVault, Marjorie L. 1990."Conflict Over Housework: The Problem That(Still)Has No Name." In Research in Social .Mbvements, Conflict and Change, edited by L. Kriesberg. Greenwood, CT: JAI. Ferree, Myra Marx. 1988. "Negotiating Household Roles and Responsibility: Resistance, Conflict, and Change." Paper presented at the annual meeting of the National Council on Family Relations, Philadelphia. Ferree, Myra Marx. 1990. "Beyond Separate Spheres: Feminism and Family Studies." Journal of'Mbrriage and the Family 52: 866—84. Ferree, Myra Marx. 1991. "The Gender Division of Labor in Two-earner Marriages." Jburnal of Family Issues 122:158-80. Foddy, William H. 1993. Constructing Questions for Interviews and Questionnaires: Theory and Practice in Social Research. Boston: Cambridge University Press. Geerken, Michael and Walter R. Gove. 1983. At Heme And At WOrk: The Family'Allocation of Labor. Beverly Hills,CA: Sage. Hartmann, Heidi I. 1981. "The Family as the Locus of Gender, Class, and Political Struggle: The Example of Housework." Signs 6:366-94. House, James S. 1986. Americans' Changing Lives Survey: wave I(Computer File). Ann Arbor, MI: Inter University Consortium for Political and Social Research. Huber, Joan and Glenna Spitz. 1981. "Wives' Employment, Household Behaviors, and Sex-Role Attitudes." Social Forces 60:151-69. 39 Huber, Joan and Glenna Spitz. 1983. Sex Stratification: Children, HOusework, and Jobs. New York: Academic Press. Mederer, Helen J. 1993. "Division of Labor in Two-Earner Homes: Task Accomplishment Versus Household Management as Critical Variables in Perceptions About Family Work." Jburnal of'Marriage and the Family 55:133-45. Perrucci, Carolyn C., Harry R. Pottes, and Deborah L.Rhoads. 1978. "Determinants of Male Family—role Performance." Psychology of WOmen Quarterly 3(Fall): 53—66. Peterson, Larry R. and Judy L. Maynard. 1981. "Income, Equity, and Wives' Housekeeping Role Expectations." Pacific Sociological Review 24:87-105. Presser, Harriet B. 1994. "Employment Schedules Among Dual- Earner Spouses and the Division of Household Labor by Gender." American Sociological Review 59:348-64. Rexroat, Cynthia and Constance Shehan. 1987. "The Family Life Cycle and Spouses' Time in Housework." Journal of .Marriage and the Family 49:737-50. Ross, Catherine E. 1987. "The Division of Labor at Home." Social Forces 65:816-33. Scanzoni, John H. 1972. Sexual Bargaining: Power Politics in the American.Marriage. Englewood Cliffs, New Jersey:Prentice-Hall. Shelton, Beth Anne. 1990. "The Distribution of Household Tasks: Does Wife's Employment Status Make a Difference?" JOurnal of Family Issues 11:115-35. Shelton, Beth Anne and Daphne John. 1993. "Ethnicity, Race, and Difference: A Comparison of White, Black, and Hispanic Men's Household Labor Time." Pp. 131-50 in Men, work, and Family, edited by J.C. Hood. Beverly Hills, CA: Sage. Spitze, Glenna. 1986. "The Division of Task Responsibility in U.S. Households: Longitudinal Adjustments to Change." Social Forces 64:689-701. Spitze, Glenna. 1988. "Women's Employment and Family Relations: A Review." JOurnal of'Marriage and the Family 50: 595-618. 40 Sudman, Seymour and Norman M. Bradburn. 1982. Asking Questions: A Practical Guide to Questionnaire Design. San Francisco: Jossey-Bass. Thibaut, John W. and Harold H. Kelley. 1959. The Social Psychology of Groups. New York:Wiley. Thoits, Peggy.1987. "Negotiating Roles," In Spouse, Parent, worker, edited by Faye Crosby. Thompson, Linda. 1991. "Family Work: Women's Sense of Fairness." Journal of'Family Issues 12: 181-96. Thompson, Linda. 1993. "Conceptualizing Gender in Marriages." Journal of .Marriage and the Family 55: 557-59. Thompson, Linda and Alexis J. Walker. 1989. "Gender in Families: Women and Men in Marriage, Work and Parenthood." Jburnal of Marriage and the Family 51:845- 71. Turner, Jonathan H., Leonard Beeghley, and Charles H. Powers.1989. The Emergence of Sociological Theory. Belmont, CA: Wadsworth Publishing Company. U.S. Bureau of the Census. 1993. Statistical Abstract of the united States Census: 1993 (113th edition). Washington, DC. West, Candice and Donald Zimmerman. 1987. "Doing Gender." Gender & Society 1:125-51.