Printer Friendly

Impact of married women's employment on individual household member expenditures for clothing.

The dramatic increase in married women's labor force participation has resulted in two-earner, husband-wife families outnumbering families in which the wife is not employed (Hayghe and Haugen 1987). In the past, wife's employment was viewed as secondary or intermittent but married women are now a significant and permanent part of the paid work force.

Wives' employment frequently leads to more dependence on the marketplace for goods and services as wives' time for home work and leisure declines (Stroker 1977; Vickery 1979). Employed-wife families appear to substitute one-use goods and purchased services, e.g., meals away from ome and child care, for wives' home time (Bellante and Foster 1984), but appear to spend less on at least some durables when wives devote more time to paid labor due to complementarity with wives' home time (Bryant 1988). In addition, the number of family earners may affect child-adult income apportionments, as well as service flows from market purchases and real income (Lazear and Michael 1980; 1986).

Clothing expenditures have been studied extensively, (1) but questions still exist about the impact of married women's employment on family members' clothing. Such information can be useful to producers and retailers in the textile and apparel industry. Heightened domestic and international competition has led the industry toward a strong marketing orientation, more dedicated to understanding and addressing consumer markets (Harding 1985) and thus to a need for more information about factors affecting expenditures. With rising labor force participation by married women, the industry has tended to focus on employed women themselves, often equating "working women" with career professionals, and giving limited attention to implications of their employment for other family members (McCall 1984; Retail Week 1978). Research on women's employment effects also can be useful to individuals who assist families with management and budgeting and to policy-makers interested in influencing household expenditures, as through incentives and transfer payments.

The purpose of the present study is to examine the effects of wife's employment, unearned income, and implicit wage (imputed value of market time) on individual family members' and household per capita clothing expenditures. Data from the 1980-1981 Bureau of Labor Statistics (BLS) Consumer Expenditure Survey (CEX) are analyzed via tobit.

PREVIOUS RESEARCH

It is well established that wife's employment increases aggregate household clothing expenditures when controlling for income and other variables (Dardis, Derrick, and Lehfeld 1981; Frisbee 1985; Hafstrom and Dunsing 1965; Hager and Bryant 1977; Norum 1989; Vickery 1979). Dardis, Derrick, and Lehfeld (1981) and Hafstrom and Dunsing (1965) examined how family clothing expenditures vary by wife's employment status (working/not working). Frisbee (1985) and Vickery (1979) focused on wife's employment hours (full- or part-time), Norum (1989) her occupation (white- or blue-collar), and Hager and Bryant (1977) and Vickery (1979) her earnings. Frisbee (1985) found that two-earner families, with the spouse employed full- part-time, spend more on clothing than do one-earner families. Norum (1989) found higher clothing expenditures with white-collar wives than with nonemployed wives (in spring and winter quarters), but lower expenditures with blue-collar wives than with nonemployed wives in all quarters except winter. Hager and Bryant (1977) and Vickery (1979) found a higher marginal propensity to consume clothing out of wife's income than out of other income sources.

Because clothing is one of the work-related expenses of women in the paid labor force, increased household expenditures for clothing sometimes are attributed to the wife's job-related wardrobe (Dardis, Derrick, and Lehfeld 1981; Vickery 1979). Drake, Ruffin, and Kocher (1983) and Nelson (1989) analyzed wife's employment effects on expenditures for the wife's clothing, work-related or not. They indicated that white-collar or professional wives spend the most on their clothing and blue-collar wives spend less than, or as much as, nonemployed wives. Nelson also analyzed effects on other family members' expenditures. Nelson (1989) found that boys' expenditures increase with mothers in blue-collar, part-time jobs; otherwise, the mother's occupation significantly affects only her own expenditures. Hafstrom and Dunsing (1965) found average clothing expenditures to be equal for wives in one- and two-earner families, but high for husbands and children in two-earner families.

A strength of Nelson's (1989) study was the use of individual-level 1984-1985 CEX expenditure data that are internal to BLS. These data identify expenditures for specific family members, unlike the public-use CEX. However, Nelson used full- and part-time, blue- and white-collar occupation categories, which may not represent realistic patterns of employment and associated clothing needs and may bias the regression coefficients. (2)

The present research makes use of the 109 occupation categories identified in the 1980-1981 public-use CEX data to develop new occupation classifications considered to reflect differing work-related clothing needs as well as family lifestyles. Cassill and Drake (1987) found that women's employment orientation has a significant effect on their lifestyles and clothing selection criteria. McCall (1977) and Schaninger and Allen (1981) indicated that wife's paid employment, in addition to influencing her own clothing expenditure behavior, affects the lifestyle and consumption of the entire family.

Analyses of wife's employment effects on clothing expenditures typically lead from utility functions on market goods, with wife's earnings and other income at times split in the budget constraint. The research reported here differs in that the clothing expenditure model derives from a utility function on the services of goods and wife's home time, subject to income and wife's time constraints.

MODEL

The model relies on a basic tenet of household production theory (Becker 1965; Michael and Becker 1973) in assuming that the household gains satisfaction from the services of clothing (c), the wife's household time (n), (3) and a composite good (g) consisting of all goods and services. Utility is maximized subject to: [P.sub.c.C] + [P.sub.g.g] = W(T -- n) + V, where [P.sub.c] and [P.sub.g] equal prices of c and g, respectively; W equals wife's wage; T -- n equals wife's total available time (T) less n, the difference being m which is her market time; and V equals household income minus wife's earnings, henceforth referred to as wife's unearned income. Utility maximization leads to a set of demand equations of the following form (Bryant 1988):

[Mathematical Expression Omitted]

where [q.sub.it] equals quantity demanded of good i in period t; [Z.sub.it] equals vector of observed family characteristics acting as preference shifters; (4) and [u.sub.it] equals error term. The [Z.sub.it] are fixed and not subject to choice in period t. Multiplying the clothing demand equation (1) by [P.sub.c] yields the conditional clothing expenditure equation, upon which the analysis is based.

A key feature of the model is the recognition that wife's household time is endogenous; the household simultaneously makes choices about the wife's market and nonmarket time allotments and about purchases of clothing and of other goods and services (Killingsworth 1983). Wife's time allocations affect total household income because W(T -- n) = Wm. Because n is endogenous, so are m and Wm (wife's earnings). To sum Wm and V would counter the theory and might produce biased regression coefficients due to correlation between m and [u.sub.it] (Johnston 1984).

The presence of W and V in the expenditure function allows evaluation of the uncompensated cross-price effect of W on clothing expenditures, as well as the income effect. This approach may enhance the discrimination between families with and without employed wives, because such families tend to differ in ways beyond wives' work activities, e.g., lower husbands' earnings and fewer members in employed-wife families (Vickery 1979). Separating W and V, in addition to including "Z" variables, helps to account for these differences. Lazear ad Michael (1980) made the added point that families' clothing and other purchases respond to the prices of service flows obtain from the goods, which in turn depend on the family environments in which the goods are used.

Because the data are cross-sectional, [P.sub.c] and [P.sub.g] goods' prices do not vary systematically and are suppressed. Urban/rural and location variables are included to capture taste and weather differences.

SAMPLE

A sample of 2,161 households was selected from the public-use tapes of the interview component of the 1980-1981 CEX. Sample households meet the following criteria: husband-wife families with at least one child less than 16 years of age; husband employed full-time and wife nonemployed or employed in one of three occupation categories (to be described) for the 12 months preceding the interview quarter; no additional earners; and pretax annual household income greater than zero in the year before the interview. The age ceiling of 15 for children is imposed in order to separately estimate clothing expenditures for children and for parent. The CEX tapes identify clothing expenditures for men and for women 16 and older, for girls and for boys 2-15, and for infants less than two. Without the child age limit, expenditures for parents and for older children would be indistinguishable.

Data are from the fourth quarter of 1980 through the third quarter of 1981. Data in each quarter are considered independent (Bureau of Labor Statistics 1985), thus the potential for repeated measures bias (caused by data from some households appearing in multiple quarters) is ignored.

VARIABLE DEFINITIONS AND HYPOTHESES

The analysis consists of six separate estimations of the clothing expenditure model. The dependent variables differ in order to examine expenditures for the household and for types of family members. Independent variables are the same in each estimation. Sample characteristics appear in Table 1.

Dependent Variables

The dependent variables in the six estimations are quarterly clothing expenditures for the household (per capita), the woman/wife, the man/husband, and an average girl, boy, and infant in the family. The age range for girls and boys is 2-15 years and for infants it is less than two years. Families without at least one child in a respective age/gender expenditure category are dropped from that estimation. While the main interest is in expenditures for types of family members,

[TABULAR DATA OMITTED]

examining household per capita expenditures provides comparative results. Average expenditures for each of the three child categories are used because sample families vary in numbers of children in these groups. Wives' expenditures cannot be differentiated as work-related or otherwise. Clothing services and gifts of clothing that transfer out of the household are excluded.

Independenct Variables

Implicit wage

Each wife's implicit wage is imputed by a three-step procedure. This instrumental variable approach is taken to avoid bias of two types: simultaneous equations bias due to the endogeneity of wives' earnings and thus of observed wages, and sample selection bias resulting from wives who opt for nonemployment, consequently displaying zero wages though their time is not valueless (Kinsey 1984; Maddala 1983).

The issue of obtaining estimates of the value of nonemployed women's time has been the focus of much research (e.g., Gronau 1973; Heckman 1974; Kinsey 1984; Zick and Bryant 1983). In the present research, wage data on employed wives are used to impute implicit market wages for all wives in the sample, both employed and not employed. Although corrected for sample selection bias, this method may underestimate market wage rates for nonemployed wives because theory indicates that nonemployed women value their time at a higher rate than the offered wage. (5)

In the first step, an employment probability model is estimated over all wives in the sample (both employed and not employed) via probit to obtain the inverse of the Mill's ratio ([Lamdba]). Lamdba is an instrumental variable subsequently used as an independent variable in a market wage equation to correct for selectivity bias. The wage equation is estimated only for employed women and the coeficients are then used to predict each wife's implicit wage. Appendix A presents the results of the probit and wage equation estimations. Table 2 presents comparison data on observed and implicit wage rates. (6)

A positive relationship between wife's implicit wage and clothing expenditures was hypothesized. Clothing expenditures may rise with wife's implicit wage for one or more reasons: higher prices paid per unit with quality either constant or nonconstant or increased quantities of clothing purchased. The CEX data do not allow differentiation

[TABULAR DATA OMITTED]

of these effects, which is also the case with the other explanatory variables in the model.

Wife's unearned income

Pretax household income minus that contributed by the wife for the year previous to the interview quarter is referred to as wife's unearned income. A positive relationship with clothing expenditures is expected, as has been found repeatedly with various income measures (e.g., Dardis, Derrick, and Lehfeld 1981; Hager and Bryant 1977; Horton and Hafstrom 1985; Lee and Phillips 1971; Nelson 1989).

Wife's employment

Four categories for wife's occupation are developed to reflect different types of work clothing of wives and different family lifestyles linked to occupation. Occupation is measured with dummy variables; the omitted category is nonemployed.

Researchers have often used husband's (or head's) occupation to represent household preferences for clothing (e.g., Dardis, Derrick, and Lehfeld 1981). With wives' expanded labor force participation, across a range of occupations, wives' occupations may be a better indicator of their employment-related clothing needs and preferences and consumption patterns of their families (Schaninger and Allen 1981). This research classifies employment according to expectations or standards of dress assumed to vary by type of occupation.

The occupation categories are labeled professional, traditional, uniformed, and nonemployed. (7) Professionals are managers, administrators, and others, such as professors and engineers, who often hold high-status positions. The traditional group includes jobs traditionally held by women, e.g., school teaching, retail sales, and clerical. Uniformed occupations include service jobs--cleaning, cooking and serving food, providing health care, delivering mail, etc. Across professional, traditional, and uniformed occupations, there is a general decline in the employee's level of authority and responsibility for the status of the place of employment and in the dress or personal appearance standards which prevail. Although wives' expenditures are for all their clothing, they are expected to be strongly affected by occupation (Cassill and Drake 1987). It is hypothesized that employed wives' expenditures would exceed those of the nonemployed who, it is assumed, lack work-specific wardrobe needs. Uniforms, employer-provided or not, are prescribed work attire. Professional and traditional workers have more differentiated work wardrobes which may require more search to obtain. When a person is employed, time to search for lower prices may decline.

It is also hypothesized that higher expenditures for other family members occur when wives are employed. A family lifestyle with more activities outside the home often coincides with the wife working away from home (Cassill and Drake 1987; McCall 1977). Higher expenditures could result if larger, more diverse wardrobes and perhaps more fashionable, higher priced clothing are purchased for these activities. The wife's search-time limitations also may increase prices paid for other family members' clothing.

Wife's other characteristics

Wife's age squared is an explanatory variable in addition to age because a nonlinear relationship between age and clothing expenditures is expected, i.e., expenditures increase with age but at a decreasing rate. Research provides evidence of such a relationship (Dardis, Derrick, and Lehfeld 1981; Frisbee 1985). Wife's education level is expected to be positively related to expenditures. This relationship, perhaps reflecting different social activities or lifestyles as well as representing consumption efficiency gains that positively affect real income and thus the ability to purchase goods and services (Bellante and Foster 1984), is supported by findings on the education level of either spouse (Abdel-Ghany and Foster 1982; Dardis, Derrick, and Lehfeld 1981; Frisbee 1985; Hager and Bryant 1977; Norum 1989). A dummy variable for race (white versus nonwhite) is used as a taste shifter. No hypothesis is formed because the impact of race in prior clothing expenditures studies is mixed. Dardis, Derrick, and Lehfeld (1981) and Horton and Hafstrom (1985) found higher expenditures by nonwhites than by whites. Nonwhite parents in Nelson's (1989) sample spent less on their own clothing, but more on clothing for their male children than did comparable whites. Hager and Bryant (1977) and Norum (1989) found race is not significant.

Family composition and location

Family composition is represented by the numbers of children in seven categories: girls and boys in the age groups of 2-5, 6-12, and 13-15 years; and infants less than two years. Coefficients on family composition variables are expected to be negative, because, all else constant, the more children present the less can be spent per member. Moreover, family clothing expenditures overall and for young children in particular can be offset by the common practice of "handing down" within and among households, mainly to young children, and of giving clothing gifts to celebrate births (Britton 1969). Interfamily transfer may lower expenditures for children's clothing in general, while intrafamily handing down may decrease average expenditure per child when more same-gender children are present.

Location is represented by dummy variables for urban/rural and region. The CEX does not identify regional location of individual rural households, but inclusion of both urban/rural and region variables allows aggregate comparison of urban northeast, urban south, and urban west households with rural north central families, while urban north central families are compared to all rural households in the sample. Past work generally has shown higher urban than rural expenditures (Dardis, Derrick, and Lehfeld 1981; Fisbee 1985; Lee and Phillips 1971) but Nelson (1989) found this urban/rural difference only for girls. Regional differences have been shown, but no general pattern is evident across past studies. It is anticipated that urban households in all regions would exhibit higher expenditures than rural households.

Season

Dummy variables are used to denote season of purchase. Summer is the base category because higher expenditures in other seasons are expected. Clothing consumption varies seasonally with activities and climate (Miron 1986; Winakor 1969).

ANALYSIS

A linear expenditure model is estimated via tobit. Tobit is a maximum likelihood estimation technique appropriate for analyzing censored response models with truncated data on the dependent variable (Kinsey 1984; Maddala 1983). (8) A portion of the sample has zero values for the dependent variable in each estimation. Seven percent of sample households have zero quarterly expenditures for total family clothing, 25 percent for women's 44 percent for men's , and 23 percent, 28 percent, and 44 percent for girls', boys', and infants' clothing, respectively.

Zero expenditures occur because clothing is a durable that provides a flow of services over time. Expenditures may be "lumpy," that is, households may purchase some types of clothing in one period, but not in the next (Jones 1960; Winakor 1969). This is often evident in quarterly clothing expenditure data (Norum 1989).

Clothing Expenditure Model

[Mathematical Expressions Ommitted]

where:

[E.sub.1] = per capita quarterly expenditures for clothing; [E.sub.2] = quarterly expenditures for women's clothing; [E.sub.3] = quarterly expenditures for men's clothing; [E.sub.4] = quarterly expenditures for girls' clothing; [E.sub.5] = quarterly expenditures for boys' clothing; and [E.sub.6] = quarterly expenditures for infants' clothing.

For each estimation of the clothing expenditure model the explanatory variables are

[X.sub.1] = number of income; [X.sub.2] = number of infants < 2 years old; [X.sub.3] = number of girls 2-5 years old; [X.sub.4] = number of girls 6-12 years old; [X.sub.5] = number of girls 13-15 years old; [X.sub.6] = number of boys 2-5 years old; [X.sub.7] = number of boys 6-12 years old; [X.sub.8] = number of boys 13-15 years old; [X.sub.9] = wife's age squared; [X.sub.10] = wife's occupation (1 = professional, 0 = nonemployment); [D.sub.1] = wife's occupation (1 = traditional, 0 = nonemployed); [D.sub.3] = wife's occupation (1 = uniformed, 0 = nonemployed); [D.sub.4] = wife's occupation (1 = some grade school, 0 = high school graduate); [D.sub.5] = wife's education (1 = some high school, 0 = high school [D.sub.6] = wife's education (1 = some college, 0 = high school graduate); [D.sub.7] = wife's education (1 = college graduate, 0 = high school graduate); [D.sub.8] = wife's race (1 = white, 0 = nonwhite); [D.sub.9] = urban/rural (1 = urban, 0 = rural); [D.sub.10] = northeast (1 = northeast, 0 = north central); [D.sub.11] = south (1 = south, 0 = north central); [D.sub.12] = west (1 = west, 0 = north central); [D.sub.13] = fall (1 = fall, 0 = summer); [D.sub.14] = winter (1 = winter, 0 = summer); [D.sub.15] = spring (1 = spring, 0 = summer); W (*) = wife's implicit wage; and [u.sub.i] = error term.

RESULTS

Table 3 contains the results of each estimation of equation (2). The impacts of independent variables on expenditures for women's, men's, and childen's clothing are discussed first and then those for the households per capita. The significant impact of an independent variable on clothing expenditures can be attributed to one or more inextricable effects: higher prices per unit, with quality either constant or nonconstant, or increased quantities of clothing purchased. The nature of these data makes it impossible to determine the source of a specific significant effect on expenditures.

Contrary to expectations, wife's implicit wage has no significant effect on expenditures for family members' clothing. (9) However, wife's unearned income has a significant positive effect on expenditures for family members, as hypothesized, except for infants. Infants' wardrobes are often supplemented by gifts and handing down which may diminish the importance of income. The income elasticities indicate that clothing is a normal good. Elasticities for women's and men's clothing expenditures are similar; a one percent increase in wife's unearned income results in a 0.43 percent increase in women's expenditures, and a 0.41 percent increase in men's. The elasticities for girls' and boys' expenditures (0.24 and 0.20 percent, respectively) are also similar. Children's clothing expenditures appear

[TABULAR DATA OMITTED]

to respond less to income differences than do adults', as Nelson (1989) found, though she measured income by household total consumption expenditures and controlled only for family size. The elasticity estimates here are considerably lower than Nelson's (1989) which is characteristic of elasticity measures obtained using income as opposed to told expenditure data. (10)

Employed-wife households spend significantly more on women's clothing than do nonemployed-wife households, as hypothesized. Compared to nonemployed-wife households' expenditures for women, families with professionally employed wives spend almost $33 more per quarter, those with wives in traditional occupations spend over $12 more, and those with wives in uniformed occupations spend $16 more. The results suggest that women in a range of occupations respond to employment-related wardrobe needs, either by purchasing more clothing or by paying higher prices.

Wife's occupation significantly affects expenditures for men's and

children's clothing, as hypothesized, though only in certain cases. The results show employment effects that previous research has not. Households with wives in professional occupations spend over $18 more on men's clothing per quarter than do the nonemployed-wife households. Otherwise, wife's occupation appears not to affect men's expenditures. (11) Nor are boys' clothing expenditures affected by mother's occupation, unlike Nelson's (1989) findings. Significant effects on children's expenditures of mother's occupation are evident in only two cases. Quarterly expenditures for infants are $18.48 higher when wives are employed as professionals than when not employed, perhaps adding credence to the popular notion that professionals purchase elaborate and extensive wardrobes for their very young children. Expenditures for girls' clothing are $9.53 more per quarter in families of wives in traditional occupations than in comparable nonemployed-wife households.

The hypothesis that all family members' clothing expenditures would be simultaneously influenced by the wife's occupation is not fully supported. The results indicate that the influence of the wife's occupation on family clothing expenditures varies by what her occupation is, when controlling for the other variables in the analysis. Differences between households with and without employed women are most apparent when employment is in a professional position, perhaps capturing lifestyle effects. Income and other variables to be discussed may then account for possible clothing consumption differences which exist between sample families with and without employed wives.

Other characteristics of the wife included in the model are race, age, and education. As in some previous studies, the effect of race is mixed. Wife's race affects only girls' expenditures which are almost $25 more when the wife is white than when she is nonwhite. Wife's age and age squared are not significant for any category of clothing expenditures estimated.

When significant, effects of wife's education are as hypothesized. Compared with wives who finished high school but did not go to college, those with only some high school education spend less on their clothing, and college graduates spend more. Husbands' clothing expenditures are significantly higher when their wives have attended or graduated from college than when their wives are high school graduates only and are lower when their wives are not high school graduates. Girls' expenditures are unaffected by mothers' education, but boys' and infants' expenditures are higher when their mothers' education extended into college rather than ending at high school.

The remaining explanatory variables control for family composition, location, and season. The impacts of the age/gender categories of children representing family composition, when significant, are as expected except for the positive impact of teenage girls on women's expenditures. An increase in the number of girls 13-15 years old increases expenditures for women's clothing by almost $64 per year. The imposed ceiling of 15 years for girls in the sample may have failed to effectively partition women's and girls' clothing expenditures, or some women may have tended to purchase more clothing for themselves if they thought their older daughters might wear them as well. All other significant coefficients on family composition variables have negative signs. This effect is anticipated because, all else constant, the more children the family has, the less resources it can apportion to each member. The negative impact on average expenditures for children could also be capturing handing down among siblings.

Because the CEX does not identify region of rural households, urban/rural and region effects must be tabulated from the coefficients on the two types of location variables. Coefficients on the urban/rural variable are added to those on the northeast, south, and west dummies to compare urban northeast, urban south, and urban west households with rural north central families. The coefficient on urban/rural indicates the difference between urban north central families and all rural households in the sample. Expenditures for women's clothing are higher in the urban south than those in rural north central families, ceteris paribus. Those for infants are higher in the urban northeast when compared to rural north central households. Winter expenditures for men's and women's clothing are higher than summer expenditures, perhaps due to higher unit costs for cold weather clothing. Fall and spring expenditures do not differ significantly from those in the summer. Expenditures for boys' clothing are higher in fall and lower in spring than in summer, possibly reflecting an annual cycle of stocking up in late summer and early fall on clothes for school. Girls' and infants' clothing expenditures are unaffected by season, implying a constant purchase pattern across seasons for these children.

Contrasting these results with those from the per capita expenditures equation points to the added insight which can be gained by disaggregating expenditures by family member type. Most clothing expenditure research has examined total household expenditures for clothing (e.g., Dardis, Derrick, and Lehfeld 1981; Frisbee 1985; Norum 1989). The studies by Nelson (1989) and Drake, Ruffin, and Kocher (1983) are exceptions.

Some of the per capita results simply mirror the ones for family members. For example, it is not surprising that wife's implicit wage and age variables are not significant for per capita expenditures and that race is not significant, given that race affects just girls' expenditures. Wife's unearned income is significant, as is the case in estimations for all family members except infants.

Per capita expenditures are higher in households with the wife employed in each occupational category, than in nonemployed-wife families. This general effect of wives' employment on household expenditures is better understood in conjunction with the information from the family member estimations. The other estimations provide evidence that increased expenditures for employed wives' clothing contribute rather strongly to the increase in household clothing expenditures. It appears especially true when the wives are in uniformed occupations because having wives in such jobs significantly affects only women's expenditures. In the remaining two occupation categories, clothing expenditures for one or two types of family members, in addition to the wives', are significantly affected.

Per capita clothing expenditures are lower when the wife has not completed high school and higher when she has gone to college, than when she is a high school graduate. Wives' completion of high school seems to be a factor which demarcates distinct clothing consumption patterns of families. Yet, as noted earlier, the effect of wife's education varies among family members; per capita results most closely follow those for men.

All family composition variables except for number of boys 13-15 years of age significantly affect per capita clothing expenditures. Significant coefficients on these variables are negative except for that on teenage girls. There is a positive relationship between per capita expenditures and residing in the urban northeast or urban south. Seasonal effect on per capita expenditures is isolated to that of winter.

SUMMARY AND CONCLUSIONS

Clothing expenditure data in the 1980-1981 public-case CEX are analyzed via tobit, primarily to assess the impact of wife's employment, unearned income, and implicit wage on expenditures for family members, when controlling for other variables. The analysis is based on a conditional expenditure model derived from a utility function on the services of goods and wife's home time, subject to income and wife's time constraints. Modeling and variable specifications are designed to avoid bias due to simultaneity of wives' earnings, observed wages, and employment hours, as well as sample selection bias from lack of wage data on nonemployed wives. Separate equations are estimated for expenditures for women, men, girls, boys, and infants, and for per capita household expenditures.

Wife's unearned income has a significant positive effect in each estimation except for infants' expenditures; income elasticities for adults' clothing exceed those for children. The imputed implicit wage variable is not significant for any clothing category.

Wife's employment, represented by occupation categories developen to capture employment-related clothing needs and family lifestyles linked to occupation, significantly affect wives' clothing expenditures and sometimes expenditures for other family members. Women in professional, traditional, and uniformed occupations spend more on clothing than do nonemployed wives. Otherwise, the effect of wives' employment is most apparent in families with professionally employed wives, in which both men's and infants' expenditures are higher when compared to nonemployed-wife families. The only additional case of a significant employment effect is for girls with mothers in traditional jobs. Yet, per capita expenditures are found to be higher when wives are employed in any of the three occupation categories than when they are not employed. Overall, these results show that the effect of wife's employment on clothing consumption varies by occupation, which becomes evident through examination of disaggregate family member types.

Some of the remaining variables in the analysis have no significant effect in any or most of the estimations. Others, particularly the wife's education and family composition, have significant impacts on per capita clothing expenditures though the impact differs among family members.

A major conclusion to be drawn from this research is that it is not possible to generalize across family members about the influence of income and sociodemographic variables on clothing expenditures. Regarding the effect of wife's employment, increased expenditures for wives' clothing appear to contribute importantly to increased household clothing expenditures when wives hold jobs, although this research indicates increased expenditures for some others in the family as well. Thus, an incomplete picture of the impact of wives' employment on clothing expenditures is obtained if the focus is solely on the women or on their families as aggregate units. It can also be said that a limited perspective forms by considering only women's employment while ignoring other influences, such as wife's unearned income, education, and family composition, on clothing expenditures.

APPENDIX A

The method used to obtain implicit wage rates for each wife in the sample is a modification of the Zick and Bryant (1983) technique. The first step is to estimate an employment probability model over all wives in the sample (both employed and not employed). The results are labeled probit in the following table. Explanatory variables represent factors that affect a wife's decision to hold a market job. Lambda is obtained from the probit estimation for use as a variable in the wage equation to correct for selectivity bias. The wage equation is estimated only for employed women (thus the need to include [Lambda]); the explanatory variables represent factors that affect the wage the wife is "offered" in the market. (In contrast, the Zick-Bryant technique (1983) imputes the wage "asked," see footnote 5.) The wage equation coefficients (see table) are used to calculate implicit wage rates for each wife in the sample, both employed and not employed.

[TABULAR DATA OMITTED]

(1) For a review, see Norton and Park (1987).

(2) BLS developed its occupational classification in the 1980-1981 CEX from approximately 1,000 occupations coded by the Bureau of the Census. The occupational classification for the present research is developed after examining the finely disaggregate Bureau of the Census occupations that BLS has aggregated into 109 categories. More recent data contain fewer, more aggregated categories. For example, BLS now includes the Census occupations of lawyer, prekindergarten teacher, and dental hygienist in "managerial and professional" (white collar). In the present research these occupations are classified as professional, traditional, and uniformed, respectively.

Since 1947 the federal government has defined part-time workers as those who work less than 35 hours a week and full-time workers as those who work 35 or more hours (Nardone 1986). Nardone asserted that the "less than 35 hours" criterion does not accurately describe part-time employment and the standard 40-hour week does not describe full-time employment, because for many workers a full-time work week is less than 40 hours. Thus, these government classifications neither realistically represent patterns for work nor clearly delineate full-time part-timers. Also, as explained in the model section, actual hours of employment, along with actual market wage, are endogenous.

(3) Only wife's time in household activities is considered because wives, employed or not, have primary responsibility for running households, including clothing acquisition and maintenance (Sanik 1981; Vanek 1974; Walker and Woods 1976).

(4) Wife's occupation is used as a preference shifter. It is conceivable that occupation is correlated with observed wage rate, however, descriptive statistics (Table 2) of imputed implicit wage across occupation categories are similar. This suggests that, to the extent that occupation is important, it is not diluting the wage effect.

(5) Overidentification of the reservation ("asking") wage equation obtained via the Zick and Bryant method (1983) produces implicit wage rates that,

although meaningful in relation to each other, are so beyond the range of realistic wages (enormous or miniscule in magnitude depending on the identifying variable) that they are unusable as measures of the impact of the value of wives' time on expenditures. The authors thank W. K. Bryant for suggesting the alternate approach taken here, which produces sample selection bias-corrected "offered" wage rates.

(6) Data in Table 2 indicate correspondence between the two wage measures within occupation categories, except between those for women who are professionals. The fact that the wage equation produces imputed values similar to observed wage rates of the women employed in traditional or uniformed occupations and underestimates those of the professionally employed women, may indicate that the model is missing variables which contribute to the determination of offered wages for professionals, e.g., intangible or not easily measured characteristics such as competitiveness and/or drive.

(7) Certain occupations, thus certain households, are not included in the sample. Exclusions are for the following reasons: jobs rarely held by women (e.g., blacksmiths, boilermakers, forgemen); CEX aggregation too encompassing as to work-related clothing needs (e.g., categories that included dental lab technicians, window dressers, stonecutters, and telephone linespersons); occupations not exclusive to a category (e.g., physicians could be either uniformed or professional); and nonemployment reasons (ill, in school) implying unusual expenditure patterns. See DeWeese (1987) for more detail on the occupation categorization.

(8) OLS estimates would have been unbiased but inconsistent. Deleting zero-value observations to allow OLS estimation would have introduced sample selection bias.

Tobit estimates a vector of normalized coefficients which, when multiplied by the standard error of the estimate, gives a vector of regression coefficients. Multiplying the regression coefficients by the cumulative normal distribution function yields transformed derivatives analogous to least squares coefficients (McDonald and Moffit 1980). A derivative for a continuous variable indicates the change in expenditures per unit change in that variable. For dummies, each derivative indicates the category versus base difference in expenditures.

(9) Collinearity diagnostics indicate some degree of collinearity with respect to the implicit wage variable in the expenditure model. A potential effect is that tests of significance are too strong, i.e., there is potential for increase in Type II errors. Although the likelihood of failing to reject may be higher, these estimates are unbiased.

(10) See Dardis, Derrick, and Lehfeld (1981) for a discussion of the use of total expenditures versus income as an explanatory variable.

(11) It could be argued that professionally employed women may be married to professionally employed men who also have high levels of wardrobe needs. Yet some unemployed women (the base of comparison) have husbands who are professionals, although the probability is higher for professional women. So the significant effect of wife's professional employment on men's expenditures may truly reflect the impact of the wife's occupation and not of the husband's, although more work is needed to test this proposition.

REFERENCES

Abdel-Ghany, Mohamed and Ann C. Foster (1982), "Impact of Income and Wife's Education on Family Consumption Expenditures," Journal of Consumer Studies and Home Economics, 6: 21-28.

Becker, Gary S. (1965), "A Theory of the Allocation of Time," Economic Journal, 75: 493-517.

Bellante, Don and Ann C. Foster (1984), "Working Wives and Expenditure on Services," Journal of Consumer Research, 11(2, September): 700-707.

Britton, Virginia (1969), "Gifts and Handed-Down Clothing Important in Family Wardrobes," Family Economics Review (September): 3-5.

Bryant, W. Keith (1988), "Durables and Wives' Employment Yet Again," Journal of Consumer Research, 15(1, June): 37-47.

Bureau of Labor Statistics (1985), 1980-1981 Interview Survey Public Use Tape Documentation, Washington, DC: U.S. Department of Labor, Bureau of Labor Statistics.

Capps, Oral, Jr. (1983), Qualitative and Censored Response Models, prepared for the American Agricultural Economics Association Refresher Course in Econometrics, Purdue University, West Lafayette, IN (August).

Cassill, Nancy and Mary Frances Drake (1987), "Employment Orientation's Influence on Lifestyle and Evaluative Criteria for Apparel," Home Economics Research Journal, 16: 23-35.

Dardis, Rachel, Frederick Derrick, and Alane Lehfeld (1981), "Clothing Demand in the United States: A Cross-Sectional Analysis," Home Economics Research Journal, 10: 212-222.

DeWeese, Gail (1987), The Impact of Married Women's Employment on Household Expenditures for Clothing, doctoral dissertation, Virginia Polytechnic Institute and State University, Blacksburg, VA.

Drake, Mary F., Marilyn Ruffin, and Leona Kocher (1983), "Clothing Expenditures of Wife in Relation to Wife's and Husband's Employment," paper presented at the National Meeting of the Association of College Professors of Textiles and Clothing, Honolulu, HI (July).

Frisbee, William R. (1985), "Economic Analysis of Household Clothing Expenditures," Canadian Home Economics Journal, 35(4): 201-206.

Gronau, Reuben (1973), "The Effect of Children on the Housewife's Value of Time," The Journal of Political Economy, 81(2, Part II): S168-S199.

Hafstrom, Jeanne L. and Marilyn M. Dunsing (1965), "A Comparison of Economic Choices of One-Earner and Two-Earner Families," Journal of Marriage and the Family, 27(3): 409-413.

Hager, Christine J. and W. Keith Bryant (1977), "Clothing Expenditures of Low Income Rural Families," The Journal of Consumer Affairs, 11(2, Winter): 127-132.

Harding, Peter W. (1985), " Is the Apparel Industry the Weak Link in the Soft Goods Chain?" Textile Chemist and Colorist, 17(2): 13-15.

Hayghe, Howard V. and Steven E. Haugen (1987), "A Profile of Husbands in Today's Labor Market," Monthly Labor Review, 111(10): 12-17.

Heckman, James (1974), "Shadow Prices, Market Wages, and Labor Supply," Econometrica, 42(4): 679-694.

Horton, Sally E. and Jeanne L. Hafstrom (1985), "Income Elasticities for Selected Consumption Categories: Comparison of Single Female-Headed and Two-Parent Families," Home Economics Research Journal, 13: 292-303.

Johnston, J. (1984), Econometric Methods, 3rd edition, New York: McGraw-Hill.

Jones, Robert C. (1960), "Transitory Income and Expenditures on Consumption Categories," American Economic Review, 50: 584-592.

Judge, George G., William E. Griffiths, R. Carter Hill, and Tsoung-Chao Lee (1980), The Theory and Practice of Econometrics, New York: John Wiley & Sons, Inc.

Killingsworth, Mark R. (1983), Labor Supply, Cambridge, MA and New York: Cambridge University Press.

Kinsey, Jean (1984), "Probit and Tobit Analysis in Consumer Research," in Proceedings of the 30th Annual Conference of the American Council on Consumer Interests, K. P. Goebel (ed.): 155-161.

Lazear, Edward P. and Robert T. Michael (1980), "Family Size and the Distribution of Real Per Capita Income," American Economic Review, 70(1): 91-107.

Lazear, Edward P. and Robert T. Michael (1986), "Estimating the Personal Distribution of Income with Adjustment for Within-Family Variation," Journal of Labor Economics, 4: S216-S239.

Lee, Feng- Yao and Keith E. Phillips (1971), "Differences in Consumption Patterns of Farm and Nonfarm Households in the United States," American Journal of Agricultural Economics, 53: 573-582.

Maddala, G. S. (1983), Limited-Dependent Variables in Economics, 2nd edition, Cambridge, MA and New York: Cambridge University Press.

McCall, Suzanne H. (1977), "Meet the 'Workwife'," Journal of Marketing, 41(3): 55-65.

McCall, Suzanne H. (1984), "The Workwife: A Powerful Change Agent of this Century," in Marketing to the Changing Household: Management and Research Perspective, M. L. Roberts and L. H. Wortzel (eds.), Cambridge, MA: Ballinger: 91-116.

McDonald, John F. and Robert A. Moffit (1980), "The Uses of Tobit Analysis," Review of Economics and Statistics, 62: 318-321.

Michael, Robert T. and Gary S. Becker (1973), "On the New Theory of Consumer Behavior," Swedish Journal of Economics, 75: 378-396.

Miron, Jeffrey A. (1986), "Seasonal Fluctuations and the Life Cycle-Permanent Income Model of Consumption," The Journal of Political Economy, 94: 1258-1279.

Nardone, Thomas J. (1986), "Part-Time Workers: Who Are They?" Monthly Labor Review, 109(2): 13-19.

Nelson, Julie A. (1989), "Individual Consumption Within the Household: A Study of Expenditures on Clothing," The Journal of Consumer Affairs, 23(1, Summer): 21-41.

Norton, Marjorie J. T., and Jae O. Park (1987), "Household Demand, Expenditures, and Budgets for Clothing," in Research Conference Proceedings Fall 1986, Information Series 87-1, Michael J. Sporakowski, Rebecca P. Lovingood, Judy A. Driskell, and Barbara E. Densmore (eds.), Blacksburg, VA: Virginia Agricultural Experiment Station.

Norum, Pamela S. (1989), "Economic Analysis of Quarterly Household Expenditures on Apparel," Home Economics Research Journal, 17(3): 228-239.

Retail Week (1978), "Chipping Away at the Myth" (November): 44-54.

Sanik, Margaret M. (1981), "Division of Household Work: A Decade of Comparison," Home Economics Research Journal, 10(2): 175-180.

Schaninger, Charles M. and Chris T. Allen (1981), "Wife's Occupational Status as a Consumer Behavior Construct," Journal of Consumer Research, 8(September): 189-196.

Strober, Myra H. (1977), "Wives' Labor Force Behavior and Family Consumption Patterns," American Economic Review, 67(1): 410-417.

Vanek, Joann (1974), "Time Spent in Housework," Scientific American, 231(5): 116-120.

Vickery, Clair (1979), "Women's Economic Contribution to the Family," in The Subtle Revolution, Ralph E. Smith (ed.), Washington, DC: The Urban Institute: 159-200.

Walker, Kathryn E. and Margaret E. Woods (1976), Time Use: A Measure of Household Production of Family Goods and Services, Washington, DC: American Home Economics Association.

Winakor, Geitel (1969), "The Process of Clothing Consumption," Journal of Home Economics, 61: 629-634.

Zick, Catherine D. and W. Keith Bryant (1983), "Alternative Strategies for Pricing Home Work Time," Home Economics Research Journal, 12(2): 133-144.
COPYRIGHT 1991 American Council on Consumer Interests
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1991 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:includes appendix
Author:DeWeese, Gail; Norton, Marjorie J.T.
Publication:Journal of Consumer Affairs
Date:Dec 22, 1991
Words:7279
Previous Article:Social issues and socially responsible investment behavior: a preliminary empirical investigation.
Next Article:The emergency food relief system: an empirical study.
Topics:


Related Articles
Employment and wage changes of families from CE Survey data.
Families of working wives spending more on services and nondurables.
Household Demographics.
Determinants of household expenditure on computer hardware and software.
Survey: Norwegian household spending.
LEAD: Japan's household spending falls 0.4% in March.
Abu Dhabi Households Spend Dh135,300 Per Year on Average.

Terms of use | Copyright © 2016 Farlex, Inc. | Feedback | For webmasters