Printer Friendly

Individual consumption within the household: a study of expenditures on clothing.

Individual Consumption Within the Household: A Study of Expenditures on Clothing

While most studies of consumption expenditures focus on the household as the purchasing unit, it would be useful for many purposes to know how households spread purchases among family members. Studies of intergenerational or sex equity within the household require data of this sort. Information on intrahousehold patterns may be helpful in formulating policy prescriptions, for example, in setting fair levels of child support payments. Examination of demands using the individual as the unit of analysis could reveal factors that aggregation at the household level hides, improving projection of future demand patterns.

This study focuses on the distribution of expenditures on clothing among household members for two reasons. First, on a conceptual level, clothing purchases are "allocatable." That is, because households generally purchase clothing items with a specific individual beneficiary in mind, household purchases can be seen as the sum of purchases for the individual members. This property does not extend to certain other household purchases, such as expenditures on housing. Second, on a practical level, clothing is one of only two items (educational expenses being the other) for which respondents to the U.S. Consumer Expenditure (CE) Survey give information about allocation. While the clothing data from this survey are currently released for public use only in grouped form (e.g., clothing expenditures for girls ages 2 to 15), the individual-level data internal to the Bureau of Labor Statistics provide a unique opportunity for direct observation of intrahousehold allocation.

Most studies of intrahousehold allocation, for lack of just such data or out of ambition to "allocate" all expenditures (including those of a "public" or "family" nature), have been forced to impute expenditures on individuals. Usually this is done by making comparisons across households of various sizes and compositions, looking at the effect of the "marginal" person or persons on household expenditures. Recent works in this vein include Deaton and Muellbauer (1986), Gronau (1987), and Deaton (1987). Espenshade (1973) and Edwards (1981) were among those interested in the allocation of particular goods, including clothing. The theoretical justification and empirical plausibility of the results of these imputation methodologies are far from established (Deaton and Muellbauer 1986; Nelson 1986a and 1986b). This study of clothing expenditures may provide a means of "benchmarking" the empirical results of such studies.

While many studies of expenditures on clothing at the household level exist (Norton and Park 1987), there have been few direct studies at the individual level. An early study of U.S. Bureau of Labor Statistics data on expenditures by wage and clerical earners in 1934-1936 (Williams and Hanson 1941) compiled statistics on average clothing expenditures by sex and sixteen age classes. Erickson conducted the most recent study of expenditures on individuals using national survey data from the 1960-1961 CE Survey (Erickson 1968; United States Department of Labor, Bureau of Labor Statistics 1967). Very small-scale studies of clothing consumption by individuals, such as that conducted by the Iowa Agriculture and Home Economics Experiment Station in 1965 (Britton 1969) provide interesting detail but may not be generalizable to other populations.

The purpose of the present study is to examine the influences on annual clothing expenditures for individual children and adults in two-parent households in the United States. Two-parent-only households were chosen for analysis to provide some degree of homogeneity in the sample and because they have been the focus of much previous empirical work and many policy questions. Unlike the earlier Williams and Hanson (1941) and Erickson (1968) studies of consumption by individuals that used tabular analysis, this study uses multivariate regression analysis to identify the important factors.

Two caveats should be kept in mind while reviewing this study. First, the term "consumption" in the title of this paper is used as suggested in Winakor's (1969) first definition of the term, i.e., as synonymous with "money expenditures." Money expenditures do not necessarily correspond closely to value in use (Winkor's second definition) nor do they encompass the the full "process of acquiring, storing, using, maintaining, and discarding clothing" (Winakor's third definition). The data presented here include purchases of new and used clothing by household members during the survey year, but use by the members of clothing obtained from other sources (such as from stocks purchased in an earlier year, gifts from outside the household, hand-downs within the household, or homemade clothing) is not captured. Britton's (1969) work suggests that nonpurchase sources may be very important: in her sample, an average of 28 percent of individual clothing acquisition was derived from sources other than direct purchase. The percentage derived from these other sources also varied by age and sex, with young children being the heaviest recipients of nonpurchased clothing. The second caveat is that the distributional pattern of clothing expenditures should not be extrapolated to other (much less total) household expenditures. Distributional patterns could be quite different for other goods.

The next section describes the data used in this study. Subsequent sections describe hypotheses about the effects of the independent variables and the method of analysis. Discussion of the results follows. The last section contains some conclusions and suggestions for further research.


The interview component of the CE records data on characteristics of households and their members as well as on household expenditures on many large or recurring expenditures. In the case of clothing, respondents are asked to describe the type of clothing purchased,1 to identify for whom each item was bought, and to give the number of items, the month of the purchase, and the purchase price. For infant's clothing, the same pattern is followed, except that instead of identifying the recipient, the respondent is only asked whether the recipient is or is not a household member. This study does not examine purchases of clothing as gifts for persons outside the household; the survey does not collect data on gifts of clothing received. The CE sample is a national probability sample of households designed to be representative of the civilian population. Households enter the survey on a rotating basis; each quarter one-fifth of the units interviewed are new to the survey. After being interviewed for five consecutive quarters (the initial interview being used for bounding purposes only), the unit is then dropped (U.S. Department of Labor, Bureau of Labor Statistics 1986). 1 For descriptions of the items included in the clothing category see United States Department of Labor 1986.

Households that met the following criteria were selected for inclusion in this study. The household must have reported expenditures for 12 months (four quarterly interviews) between January 1984 and December 1985, inclusive. The household must have consisted of only two parents and their children under age 18, with no more than one child under age three (as of the last interview). (The lack of specificity in coding of infant clothing expenditures necessitates this last condition.) The household composition must have been stable over the full 12 months, with the exception of the birth of infants during the first interview quarter. The household must have been classified as a "complete" income reporter and consist of only one "consumer unit."2

After the original selection was made, some data cleaning was necessary. Sixty-two households that provided insufficient detail on expenditures were dropped. One additional household was deleted from the sample because of extremely high outlying values of income and total consumption expenditure, and four households were dropped because one or both of the parents were retired. The final sample contains data on expenditures by 612 two-parent households, containing 574 girls and 626 boys.3 Appendix A contains descriptive statistics for the sample. 2 For definitions of terms used in the CE, see U.S. Department of Labor 1986. 3 This sample size is considerably smaller than that used in studies of data collected in the CE of 1960-1961 (Erickson 1968) and 1972-1973 (Dardis, Derrick, and Lehfeld 1981; Wagner and Hanna 1983) because of changes in the survey design. While this survey has been conducted on a continuing basis since 1980, only data from the years 1984 and 1985 could be used for this study because changes in the sample design in 1984 (the addition of rural households) and in 1986 (redefinition of the primary sampling units) prevent the compilation of a consistent data set of annual expenditures for the full currently available survey period.

Clothing expenditures may be somewhat underreported, in general, in the CE Interview Survey. Estimated national aggregate expenditures on apparel derived from the CE are only a little over one-half as high as the personal consumption expenditure estimates from the National Income and Product Accounts (Gieseman 1987). The mean expenditures by member in this study may, therefore, be underestimated. Also, underreporting may be more of a problem for some members than for others. The respondent (i.e., the household member or members that actually provide the data to the interviewer) in husband/wife households with children is likely to be the mother (Silberstein 1987). Therefore, underreporting may be expected to be more of a problem for fathers and older children, because mothers may not be well informed about their clothing purchases. If the reasons for underreporting are correlated with the independent variables in the analysis, biases in the regression coefficients may result.


The independent variables chosen to aid in explaining individual clothing consumption are in many cases the same as those used in past research to explain expenditures at the household level. Use of the individual as the unit of analysis, however, allows both for a finer investigation of the effects of personal characteristics and for an investigation of how responses to household-level variables may vary among individuals of different types.

Personal Characteristics

The age, sex, and race of an individual are factors that are expected to influence clothing expenditure. Physiological needs would seem to dictate that growing children need more frequent replacement of their wardrobes and hence have higher expenditures than their parents. This effect, however, might be moderated or reversed if children are less sensitive to style changes than their parents, if children's clothes are less expensive (as might be expected by use of smaller fabric quantities or by desired durability), if children derive more clothing from nonpurchase sources (as suggested by Britton 1969), or if parents simply have greater preferences for clothing or indulge children's needs and preferences less than they do their own (because parents control the family purse strings). Past studies of expenditures by age class have found average expenditures to be lower for children than for nonelderly adults (Simon 1958), especially if persons ages 16-17 are not included in the "child" category (Erickson 1968). Generally, females have been found to have significantly higher expenditures in all age classes (Simon 1958; Erickson 1968).

In the current sample, the age distribution is bimodal. Households are chosen on the basis of having children under age 18. Over 80 percent of the parents of these children are aged 25 to 44. Preliminary analysis (described in the results section below) reveals that expenditures do vary significantly depending on whether the individuals studied are children or parents, female or male. The main analysis of this paper consists of multivariate regressions computed separately for four member types: girls, boys, mothers, and fathers.

All the possible effects of age mentioned above also apply to comparisons of younger and older children and younger and older parents. Because children change so much during the first 17 years of life, dummy variables for each year of age (excluding age nine, chosen as the base by virtue of being a median value) are included in the equations for girls and boys. Preliminary analysis (not reported) showed that such fine age classifications added little to explanation of adult expenditures, so adults are grouped into four categories; age less than 25, 25-34, 35-44, and over 44. The modal age classes (25-34 for mothers, 35-44 for fathers) are used as the bases, and three dummy variables representing the parent's own age are entered into each parental equation.

Tastes for clothing may vary by race. A dummy variable equal to one if the member is nonwhite is entered into each equation. Dardis, Derrick, and Lehfeld (1981) found household clothing expenditures to be higher for blacks than whites, after controlling for many other variables.


The importance of household income as a determinant of household expenditure has been proven many times over in consumer expenditure research (Deaton and Muellbauer 1980).4 Following Dardis, Derrick, and Lehfeld (1981) and Prais and Houthakker (1971), household total consumption expenditures are used as a proxy for household permanent income. Elasticities of household clothing expenditure with respect to total expenditure (i.e., the percentage increase in clothing expenditures expected with a one percent increase in "income") estimated from cross-section data usually have been greater than one (Norton and Park 1987). 4 Prices, as in most other cross-section research, are assumed constant.

Individual clothing consumption may depend on the source of income as well as on the level of income (Norton and Park 1987). Differential effects by parental earning status are captured in the equations by parental occupation variables (discussed below). For boys and girls, the equations also include a dummy variable (Personal Income), which is equal to one if the child has some personal source of income (usually earnings from a part-time job). A source of personal income is expected to increase expenditures on a child's clothing by allowing the child some freedom from parental spending constraints and/or by increasing the child's need for work-related clothing.

Household Composition

Household income per se is not likely to be the best predictor of individual clothing consumption, but rather the level of household income relative to household "needs." The number of siblings present in ten age/sex categories (ages 0-2, 3-5, 6-11, 12-15, 16-17) is entered into the children's equations. The total number of children in the same categories enters into the parental equations. The hypothesis is that these variables should have negative coefficients, reflecting increasing competition for household resources. Because all households in the sample have two parents, only the age class dummies for each parent are entered into the equations for the children and spouse. The direction of parental age effects is difficult to predict a priori.

Parental Education and Occupation

Children's education and occupation are probably adequately reflected in the age and personal income dummy variables discussed above. The education level of a parent may affect not only the parent's tastes for his or her own clothing, but also his or her preferences concerning how children or the spouse are clothed. Dardis, Derrick, and Lehfeld (1981) found that household clothing expenditures increased with the education level of the household "head." Dummies representing educational achievement of less than high school, some college, and college graduate for each parent are added to all equations.

Due to variation in clothing standards for different kinds of jobs, it is hypothesized that white-collar workers may spend more on clothing for themselves than do blue-collar workers, and both may spend more on themselves than persons not working for pay. The occupation of some workers is recorded as "self-employed" with no indication of the type of work done. Occupation may affect tastes for children's and spouses' clothing as well as one's own clothing. While most of the employed fathers in the sample work full-time, mothers' hours of work are more variable. The number of hours worked by the mother may influence her own need for clothing and also give some indication of her contribution to household income and, perhaps, her degree of influence in household spending decisions. Dummy variables are entered into each equation for blue-collar, self-employed, and not working for pay occupational status of the father. Dummy variables for mothers include white-collar part-time, blue-collar part-time, blue-collar full-time, self-employed, and not working for pay.


Variation in clothing expenditures by location may reflect climatic differences and/or variation in tastes. Dardis, Derrick, and Lehfeld (1981) found household expenditures to be higher in urban than in rural areas and to be lowest in the West. Locational dummy variables representing the degree of urbanization and the region of the United States in which the household resides are included in the regressions for all member types.


The effects of the independent variables on individual clothing consumption are investigated using the Tobit model: log C sub i = a + b log Y sub i + summation sub k C sub k X sub ki + e = 0 if the right hand side greater than 0 otherwise, where C sub i is the clothing expenditure for person i (set equal to $1.00 if no expenditure was recorded), Y sub i is total expenditure for the household to which person i belongs, X sub ki are the other independent variables, and a, b, and the c sub k are the parameters to be estimated.5 The error terms e sub i are assumed to have mean zero and be independently, identically, and normally distributed. 5 The conversion of zero expenditure values to $1.00 values is a standard method of getting around the problem that the natural log of zero does not exist (e.g., Dardis, Derrick, and Lehfeld 1981; Wagner and Hanna 1983; Maddala 1983).

Tobit analysis is appropriate when the dependent variable is censored; that is the dependent variable has a number of its values clustered at a limiting value (Maddala 1983). In the present sample, reported clothing expenditure values are zero for 1.1 percent of girls, 2.7 percent of boys, 4.7 percent of mothers, and 8.2 percent of fathers. In the Tobit case, the right hand side of the regression equation is considered to be a stochastic index of a latent variable--the propensity to purchase clothing--that is only observed when it is positive. Maximum likelihood techniques can be used to derive estimates of the parameters and their standard errors. Some adjustment is necessary to make the Tobit parameters analogous to OLS parameters. As shown by McDonald and Moffitt (1980), the total effect of any independent variable on the value of the dependent variable (evaluated at the means of the independent variables) can be found by multiplying the parameter estimate by the probability of the dependent variable taking on a value greater than zero (evaluated at the means of the independent variables). All following discussion of coefficients and presentation of their estimates should be taken as refering to coefficients after such adjustment.

The double-logarithmic functional form is used for the basic Engel curve portion of the model (i.e., the relationship of clothing consumption to the total expenditure) because it has wide use in and desirable statistical properties for the study of clothing expenditures (Dardis, Derrick, and Lehfeld 1981; Wagner and Hanna 1983). Possible drawbacks to this formulation relating to theoretical plausibility (Deaton and Muellbauer 1980) are of secondary importance when only one class of goods is being considered. In the double-log form, the adjusted coefficient on the transformed total expenditure variable has a direct interpretation as the elasticity of demand.

Adjusted coefficients of the other independent variables have an interpretation as the percentage change in clothing expenditures expected to result from a one unit change in the value of the independent variable. In most of the tables that follow, the adjusted coefficients on the dummy variables are transformed to reflect the percentage difference expected for households or individuals in the dummy variable category relative to the base category, by taking antilogs of the adjusted coefficients (Dardis, Derrick, and Lehfeld 1981).

Asymptotic t-tests are used to test the significance of individual variables. Likelihood ratio tests are used to test the significance of the joint influence of groups of variables. If L(H sub 0) is the value of the likelihood function under the null hypothesis that all the variables in the group have coefficients equal to zero, L(H sub 1) is the value of the full model, and J is the number of restricted coefficients, then the following holds asymptotically (Judge et al. 1980): -2 log (L(H sub 0) / L(H sub 1)) - X raise to 2 sub j.


Preliminary analysis (reported in Appendix B) reveals that expenditures differ significantly among girls, boys, mothers, and fathers. Controlling only for household total expenditures and family size, boys are estimated to receive 81 percent of the amount spent on clothing for girls, and fathers 62 percent of expenditures on mothers. In contrast with earlier studies, parents are estimated to receive less than children of their respective sexes: fathers' expenditures are predicted to be 57 percent of boys' expenditures, and mothers' expenditures 73 percent of girls'. While such results are indicative of substantial differences in expenditures by member type, they are not adequate to describe the variations in the allocation patterns when, for example, expenditures on children of a specific age are compared with those for their parents, or children are compared with parents at various income levels. The results reported in the tables in this paper come from a single set of Tobit regressions: one regression each for girls, boys, mothers, and fathers.6 The results are presented in a series of tables rather than in one table because of the transformations made on dummy variable coefficients. 6 As might be expected given the relatively small incidence of zero expenditures, the Tobit parameter estimates are generally fairly close to results derivable using Ordinary Least Squares on the same data. In addition, the adjustment procedure suggested by McDonald and Moffitt has no practical consequence (at the number of significant digits reported inthe tables) because the probability of the dependent variable equaling zero when the independent variables are at their means is at least .9989 in all reported regressions. The estimated elasticities of clothing expenditure with respect to total consumption expenditure, shown in Table 1, are estimated to be close to unity for girls and boys, somewhat more than half again as high for mothers, and highest (close to 2.0) for fathers. That is, a doubling of income (all else constant) could be expected to double expenditures on children's clothing, more than double expenditures on mother's clothing, and triple expenditures on father's clothing. The elasticity estimate is significantly different from zero at the one percent level in each case, and significantly different from one at the one percent level for mothers and fathers.

The variation in income elasticities among member types leads to a complicated picture of intrahousehold allocation as income rises. Figure 1 shows predicted mean clothing expenditures at five total expenditure levels for a "base" household whose characteristics are those of the modal or median class in all independent variables.7 At the $10,000 and $20,000 total expenditure levels, predicted expenditures in order of highest to lowest are, as suggested by the preliminary analysis, girl-boy-mother-father. When income is not abundant, parents may be able to get along with only minor changes in their clothing stocks, while children's constantly changing sizes makes periodic complete replacements of their wardrobes a necessity. By the addition of another $10,000, the mother's higher income elasticity pushes her expenditures above those for boys. With an additional $20,000, at $50,000, the father's high income elasticity pushes his expenditure above boys'. 7 The base household consists of a nine-year-old girl and a nine-year-old boy who have no personal sources of income, a mother age 25-34 who is a high school graduate and works full-time at a white-collar job, and a father age 35-44 who is also a high school graduate and a white-collar worker; they are white and live in an urban area in the South.

The values used in this chart are predicted mean expenditures. That is, they show the mean value of expenditure one might best predict, given information on only the variables used in the model (and dependent on the chosen functional form). Levels of expenditure, and the resulting intrahousehold allocation of expenditure, may (and does, in the sample) vary quite widely across households.

Table : 1 Elasticities of Clothing Expenditure with Respect to Household Total Expenditures(a) (a) Asymptotic standard errors in parentheses.

Table 2 presents the results of likelihood ratio tests for most of the remaining independent variables treated as groups. (Dummy variables for race and children's personal income are not included in any group). Children's ages are important in the determination of their clothing consumption. Geographic location would seem to affect household clothing expenditures largely through variation in expenditures on children; the hypothesis that the coefficients on location variables are all equal to zero cannot be rejected in the equations for mothers and fathers. Parental education appears important only for mothers, parental occupation only for boys and mothers, and parental ages only (and then marginally) for boys and fathers. The household composition variables are significant as a group only for girls.

An examination of the individual coefficients gives more insight into how these variables affect clothing expenditures. The transformed coefficients for the children's own age dummies (expressing expenditure on a child of a given age as a percentage of that for a child age nine) are given in Table 3. While the conventional wisdom in the "cost of children" literature is that expenditures rise with the age of the child, these results suggest that, though ages are important, a different pattern may prevail. Girls ages zero through one and boys ages zero through two do not appear to receive significantly less than their nine-year-old counterparts, and girls and boys age 17 do not appear to receive significantly more than the nine-year-olds. In the intermediate ages, something more like the expected pattern can be discerned. Girls ages two through four and six through eight and boys age three apparently receive less expenditure, and boys ages 12 through 16 more expenditure, than the nine-year-olds. Figure 2 displays these results graphically, showing predicted mean expenditures by age for a girl and a boy in the "base" household, at the median household total expenditure of $24,779.

The high expenditure for infants relative to other young children is especially surprising, given evidence that children in this age group may also be the heaviest recipients of gifts and hand-downs (Britton 1969) and because Erickson's study on 1960-1961 data (1968) showed infants to have the lowest level of clothing expenditure among the age classes. Further analysis shows that the growing popularity of disposable diapers since the 1960s may be the source of these newly observed high expenditures because they require sizable recurring expenditure. The equations for girls and boys were re-estimated using expenditure on outergarments (defined as total clothing less undergarments) as the dependent variable. The results are plotted in Figure 3. The omission of undergarments totally eliminates the initial spike for boys and greatly lowers predicted expenditures for infant girls. While one may want to exclude disposable diapers from analysis of clothing for some purposes (e.g., for projection of demand for textiles), it would be a mistake to exclude them from estimates of the "cost of children" unless they are included in some other category.8

The apparent drop in clothing expenditures for the oldest teens relative to younger teens could be due to a slowing in children's physical growth rates and/or to underreporting as children may do more of their own shopping and not necessarily make purchases known to the household respondent for recording during the interview. Unfortunately, little information exists within the CE for investigating this latter possibility.

The transformed coefficients on locational variables reported in Table 4 suggest that expenditures on girls are lower in nonurban than in urban areas and in the North Central and West than in the South. Expenditures on boys also appear to be lower in the West. These are similar to Dardis, Derrick, and Lehfeld's results for locational variables, but they suggest that variation in household expenditures by location may be due primarily to variation in children's, and especially girls', expenditures.

The case of the mother having less than a high school education seems to reduce expenditures on girls' clothing, and some college education appears to increase expenditures on herself and the father, while college graduation by the mother is associated with an increase in expenditures on the father. None of the coefficients for father's education is statistically significant. Correlation between the parental education variables probably explains the difference from Dardis, Derrick, and Lehfeld's results, which found increased household expenditure to be related to education of the household "head." 8 The elimination of undergarment expenditures reduces but does not remove the child/parent differentials in level of expenditure found in the preliminary analysis, evenwhen child outergarment expenditures are compared with total parental clothing consumption.

Mothers not working for pay appear to spend significantly less on their own clothing than mothers in white-collar, full-time occupations, as may be expected if these jobs require a distinct office wardrobe. Women in blue-collar occupations spend significantly less on themselves than do white-collar workers (and less than mothers not working for pay), with the difference increasing in size and significance with the hours of work. Having a mother in a blue-collar, part-time occupation seems to increase clothing expenditures on boys.

A household in which the father is not working for pay appears to spend significantly less on clothing for boys and mothers as well as on the father. This may be expected if households respond to a lack of paid employment by the father with decreased discretionary expenditures. While no significant white-collar/blue-collar differences are observed for fathers' occupations, households with self-employed fathers tend to spend less on fathers and more on boys than households with white-collar fathers.

While the mothers' age variables appear to have no significant effects on mothers' expenditures, nor fathers' age variables on fathers' expenditures, there are some significant cross-effects and effects on children. Young mothers are associated with increased expenditures on girls and older mothers with increased expenditures on boys. Young fathers are associated with increased expenditures on mothers and boys and older fathers with decreased expenditures on boys.

The effects of race are opposite for parents and boys and insignificant for girls. Nonwhite parents appear to spend less on themselves but more on boys than do whites.

Girls who have a source of personal income tend to spend significantly more on clothing than those who do not, as hypothesized. The effect is in the expected direction for boys, but it is not statistically significant.

Table 5 displays the estimated coefficients on the household composition variables. All of the statistically significant coefficients are of the expected sign, while the rest are mostly negative or very small in magnitude. Expenditures on girls seem to be reduced most by the presence of other girls ages 12 through 15, while boys' expenditures are reduced most by the presence of other boys ages 6 through 11. The number of children in three sex/age classes appears to reduce expenditures on mothers, while only one sex/age class has a (marginally) significant effect on fathers' purchases. The relative weakness of these effects, as compared with the strength of the family size effect in the preliminary regressions (Appendix B), may be due in part to collinearity with other included variables, such as parental age.


Clothing expenditures on individual girl, boy, mother, and father members of two-parent households are studied using data from the CE for the years 1984-1985. The results suggest that clothing expenditures on girls are generally higher than on boys and that expenditures on mother's clothing are higher than for fathers. Expenditures on children's clothing tend to be higher than for parents, though because parents have higher elasticities of clothing expenditures with respect to total household expenditures, this pattern is less pronounced at higher permanent income levels. Girls' and boys' predicted mean expenditures exhibit similar patterns by age, with expenditures highest in infancy and in the mid-teen years and lowest at early preschool ages. The effects of other variables, such as race, location, education, and occupation of the parents, vary in size, statistical significance, and often even in direction across member types.

The results challenge some of the popular theories and stylized "facts" about the distribution of expenditures within households. Since at least the time of Engel (1895), one common assumption has been that individuals within households could be seen as consuming goods in fixed proportions to each other, i.e., that a set of scales existed that related the consumption of each member to some standard member type. For example, Simon (1958) estimated that a female under age five is equivalent in clothing consumption to .29 males age 25 through 29. The idea of "unit consumer scales" has received wide use in the formulation of theories of demand relating to household composition (e.g., Prais and Houthakker 1971; discussion of the "Barten" model in Deaton and Muellbauer 1986). The present analysis suggests that the actual ratios of child to adult consumption of clothing within households are far from fixed, due to the variation in income elasticities (as well as responses to other variables) among member types. The results of this paper also contradict the usual assumptions that these scales are increasing functions of children's ages and are small for children relative to their parents (Engel 1895; Simon 1958).

While it is impossible to generalize from the case of clothing to the case of expenditures in general, the results of this study may make a small contribution to policy discussions in the area of child support. Although the question of the appropriate level of child support is sometimes couched in terms of minimum, prescriptive child "budgets," the more predominant approach has been to try to estimate appropriate levels of support based on observed actual expenditures, paying attention, for example, to how expenditures on children vary with age and with total household consumption (Institute for Research on Poverty 1987). The age pattern observed for clothing expenditures in this study and the result of estimated proportionality of child clothing expenditure to household total expenditure contribute at least a small piece of this puzzle, without the use of imputation techniques of dubious validity.

Further investigation using these data should shed more light on (properly limited) questions of intrahousehold distribution. Implicit in the model used in this paper is the rather naive assumption that errors in the regression equations are independent across children, even when the children are siblings. An analysis of the correlation in regression residuals between members of the same household may reveal whether unusually high (low) clothing expenditures for a member seem to be the result of high (low) household clothing expenditures in general (if positive correlation) or the result of unusually unbalanced intrahousehold distribution patterns (if negative correlation).(9) The analysis may be extended to include single-parent and childless families, in order to compare distribution patterns. Innumerable variations in equation specification could be tried, including allowances for nonconstant income elasticities. As an application of the results, predicted mean expenditures may be used to test the accuracy of various attempts by other researchers to impute individual expenditure from household data. The direct observation of purchases of clothing for individual household members in the CE opens up a wealth of opportunities for research on consumption within the household. 9 Moulton and Nelson (1988) investigate the patterns of error covariances among household members.

Table : 2 Likelihood Ratio Tests for Groups of Variables

Table : 3 Transformed Regression Coefficients for Childs' Own Dummy Variables (Base Age = 9)

Table : 4 Transformed Regression Coefficients for Other Dummy Variables(a) (a)Base category in parentheses.

Table : 5 Regression Coefficients for Other Household Composition Variables(a),(b) (a)For girls and boys the variables measure the number of the child's siblings in each sex/age class; for mothers and fathers they measure the total number of children in the sex/age class. (b)Asymptotic standard errors in parentheses.
COPYRIGHT 1989 American Council on Consumer Interests
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1989 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Nelson, Julie A.
Publication:Journal of Consumer Affairs
Article Type:bibliography
Date:Jun 22, 1989
Previous Article:Incorporating credit in demand analysis.
Next Article:The effects of consumer education on consumer search.

Related Articles
Individual consumption within the household: a study of expenditures on clothing.
An indirect test of children's influence on efficiencies in parental consumer behavior.
Consumption theories and consumers' assessments of subjective well-being.
An alternative model of U.S. clothing expenditures: application of cointegration techniques.
The determinants of consumer installment credit.
Trends in out-of-pocket spending on health care, 1980-92.
Elderly and nonelderly expenditures on necessities in the 1980s.
Are apparel trade restrictions regressive?
Are Two Incomes Needed to Prosper Today? Evidence from the 1960s to the 1990s.

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