The Moderating Effect of Seasonality on Household Apparel Expenditure.
Seasonality is an environmental factor that plays a significant role in consumer expenditure decisions. While expenditures in most product categories are seasonal, fluctuating in response to changes in weather, the holidays, and the school calendar (Moehrle 1994; Scott and Buszuwski 1993), apparel is considered the classic example of a seasonal good (Winakor 1969). For consumers, apparel meets a wide range of seasonally-related utilitarian and social needs, offering protection from the elements in the winter and summer, and fostering social interaction during holiday and other calendar events (Horn and Gurel 1981). A well-developed body of research shows that apparel expenditures are affected by consumer characteristics, including income, social class, family type, and location (Dardis, Derrick, and Lehfeld 1981; Horton and Hafstrom 1985; Wagner and Hanna 1983). However, little is known about if and how seasonality moderates the effect of consumer characteristics on household apparel expenditure.
The role of seasonality in moderating the relationship between customer characteristics and apparel expenditures is an important research issue with theoretical and practical implications. Among applied economists, seasonality is known to influence demand for goods and services by shifting consumer preferences from one quarter to another. As such, seasonality may moderate the relationship of income and prices to expenditures, an issue of central theoretical importance. From a practical standpoint, the apparel expenditures of U.S. consumers are substantial. They are estimated to account for 5.7 percent of the typical household budget and to contribute $200 billion annually to retail sales (U.S. Department of Labor 1993). Information on how seasonality affects the apparel expenditure of consumers with various characteristics should enhance the understanding of environmental factors affecting household budget decisions.
This paper explores the effect of seasonality, in conjunction with income and other consumer characteristics, on apparel expenditure. The analysis is based on data from the Quarterly Interview Component of the 1990 to 1991 Continuing Consumer Expenditure (CE) Survey, the largest and most comprehensive source of data on the income and expenditures of U.S. households. The primary objective is to evaluate the moderating effect of seasonality on apparel expenditure by analyzing quarterly expenditure data. In support of this objective, a number of theoretical and methodological problems encountered in research involving the CE Survey data are addressed. Results show that the income elasticity of apparel expenditure differs by season. In addition, seasonality moderates the effect of age, African-American ethnicity, household size, region of residence, and home tenure.
RESEARCH ON SEASONALITY IN APPAREL EXPENDITURE
The Bureau of Labor Statistics' (BLS) Consumer Expenditure Survey has been the major source of data for the analysis of expenditures on goods and services (including apparel) by U.S. households for a century. In 1979, the Consumer Expenditure Survey was converted from a decennial survey to an ongoing survey. Data are now collected by quarter from a rotating panel of consumer units  and are published in quarterly, rather than aggregated form. With the availability of the quarterly data, researchers have begun to explore the effect of seasonality on apparel expenditure. The major purpose of the CE Survey is to collect data for revision of the Consumer Price Index. However, the results of research using CE Survey data are also used in evaluating public policy with respect to poverty and household well-being.
Seasonality is the tendency of consumer expenditures on a good or service to vary in some pattern over the course of a year. Expenditures for products ranging from furniture (see Moerhle 1994) to cocktail mixers (see Turcsik 1995) are affected by seasonality. However, apparel is considered by retailers and applied economists (e.g., Moehrle 1994) to be the classic example of a seasonal good.
Previous research on seasonality in expenditures has been based on both time series (to some extent) and cross-sectional micropanel (to a greater extent) data. The few time series analyses of the quarterly CE Survey data (e.g., Moehrle 1994; Scott et al. 1993) reveal strong seasonal patterns in all expenditure categories except health care and rent. Seasonal patterns are of two forms. In the first pattern (typical of food and gasoline), expenditures are lowest in the first quarter, peak in the third quarter, and then decline. In the second pattern (typical of apparel and all other categories), expenditures are lowest in the first quarter, increase through the year, and peak in the fourth quarter (Moehrle 1994). Peak expenditures in the fourth quarter can be explained by environmental factors, such as the weather and the holiday season. Time series analyses also show the effect of seasonality varies by age of the reference person and region of residence (Scott et al. 1993). The results of cross-sectional stud ies (using micropanel data) show apparel expenditures are higher in winter than in summer (DeWeese and Norton 1991; DeWeese 1993). Moreover, the effects of age, social class, and marital status appear to differ by quarter (Norum 1989).
Conceptual and Methodological Issues
A review of the literature on apparel expenditures reveals a number of conceptual and empirical issues that need to be addressed. While time series analysis is useful in identifying seasonal effects, the number of independent variables is limited. Cross-sectional analysis makes it possible to extend the number of independent variables but still presents a set of challenges with respect to specification of the model, definition of the income variable, and treatment of the data. The major stream of research involving cross-sectional analysis of apparel expenditures follows the applied economics paradigm (based on neoclassical demand theory), with a secondary stream in the consumer behavior paradigm (derived from psychology).
In working with cross-sectional data, there are two issues with respect to model specification: completeness of the model and appropriateness of the functional form. Researchers working in the consumer behavior paradigm have employed restricted models with a limited number of variables. Results indicate that apparel expenditures are affected by income, wife's age, and stage in the family life cycle (Rubin, Riney, and Molina 1990; Wilkes 1995). In the applied economics paradigm, researchers have developed models including a broader range of socioeconomic and demographic variables. While income is the central theoretical variable, other variables are included to control for the confounding effects of tastes and preferences on consumer expenditures (Deaton 1992). Benchmark studies include the work of Dardis et al. (1981), Horton and Hafstrom (1985), and Wagner and Hanna (1983). While none of these studies contains information on seasonal effects, the results have influenced research on seasonality in apparel ex penditures by setting standards for complete model specification. Across these studies, apparel expenditures have been shown to be affected by income, characteristics of the household head (age, marital status, sex, race, employment status, and education), and characteristics of the household (family size, stage in family life cycle, housing tenure, and region of residence).
The second issue with respect to model specification is functional form. In the consumer behavior paradigm, both Rubin, Riney, and Molina (1990) and Wilkes (1995) use the linear functional form. In their classic exposition on functional form, Prais and Houthakker (1971) argue that nonlinear models are more representative of expenditure data and recommend the double-log for all expenditure categories except food. Thus, the double-log has become the standard for expenditure research (e.g., Dardis et al. 1981; Horton and Hafstrom 1985; Wagner and Hanna 1983).
Definition of Income Variable
In neoclassical demand theory, the relationship of income, prices, and expenditures is of fundamental importance. In analyzing cross-sectional data, such as the CE Survey data, prices are assumed to be constant, so the relationship between income and expenditures becomes central. The appropriate definition of income is subject to debate among applied economists. In most expenditure research, household income is defined as either disposable income (DI) (e.g., Norum 1989) or total expenditure (TE). Income is defined as TE for theoretical and practical reasons.
At a theoretical level, the argument for using TE as a proxy for income is based on the permanent income hypothesis. According to Friedman's (1957) classic work, income has both a permanent and a transitory component. In modeling expenditures for a good or service, TE is more appropriate than DI because it captures the permanent component of household income. There are also practical reasons for using TE as a proxy for income. First, incomplete and inaccurate reporting of income by consumers is a well-documented problem. Rubin et al. (1990) attempted to address this problem by selecting only households classified in the CE Survey as complete income reporters. However, the CE Survey definition of complete income reporting is very liberal. To be considered a complete income reporter, one needs to report income from only one source. Thus, some complete income reporters may have reported only part of their income (Garner and Blanciforti 1993). A second reason for using TE rather than DI is that TE information is available for every quarter. In contrast, information on DI is collected only in the second and fifth quarters, with updating in the third and fourth quarters under certain circumstances.
Treatment of Quarterly Data
In analyzing apparel expenditures, researchers have annualized (Rubin et al. 1990), aggregated (Wagner and Soberon-Ferrer 1990; Wilkes 1995), and pooled (DeWeese and Norton 1991; DeWeese 1993) the quarterly CE Survey data. Each approach has implications for information on seasonal effects. Annualizing the data (multiplying one quarter's worth of data by four) is based on the assumption that apparel expenditure does not vary by quarter. This assumption is called into question by both the results of previous research and the experience of marketing practitioners. When the data are aggregated (summing data for all households participating in four consecutive quarters), seasonal differences are captured; however, the variation in expenditures by season is masked. In pooling the data (compiling information for all consumer units participating during the survey year), there are two assumptions. First, the coefficients of the parameters do not differ by quarter, an assumption that is likely to be too restrictive; s econd, all observations are independent, an assumption that is unlikely to be supported because the CE Survey involves an overlapping panel.
In the research, the model was specified as completely as possible (based on the available data), expressed in the double-logarithmic function form, with total expenditures used as a proxy for income. To capture expected seasonality in apparel expenditures, unpooled quarterly CE Survey data were analyzed.
Data were from the quarterly interview component of the 1990 to 1991 CE Survey. The subsample consisted of 955 consumer units  reporting four consecutive quarters of expenditure data in 1990. By restricting the sample in this way, a micropanel of data was created to allow the capture of seasonal fluctuations in apparel expenditures. The seasons were defined as winter (January-March), spring (April-June), summer (July-September), and fall (October-December).
Deaton (1992) argues that per capita expenditure (rather than household expenditure) should be used in demand analysis because it provides a closer approximation of the decision-making process of the individual consumer. Thus, the dependent variable, Quarterly Apparel Expenditure (QAE), was defined in per capita terms. This variable was created by dividing total apparel expenditure for a given quarter by household size. QAE was converted to a logarithm to fit the double-logarithmic model.
The independent variables were chosen on the basis of theory, the results of previous research, and the availability of data. Those variables included total expenditure (the proxy for income), characteristics of the reference person,  and characteristics of the household. Similar to apparel expenditure, total expenditure was expressed in quarterly terms (TQE), per capita, and was converted to a logarithm to fit the double-logarithmic model.
Characteristics of the reference person included age and age-squared, marital status, sex, occupation, education, and ethnicity. Age, marital status, and education are variables reported to be significant in previous research (Dardis et al. 1981). Age-squared was included to capture the life cycle effect of apparel consumption. Sex and ethnicity (African-American and Hispanic) of the reference person were included because of recent increases in the proportion of female-headed and minority-headed households in the U.S. While occupation has been found significant in other research (e.g., Norum 1989), no other research has explored the effect of retirement on consumer apparel expenditure. Variables representing the interaction of ethnicity and TQE were created to evaluate differences in income elasticity between African-Americans and Caucasians and Hispanics and Caucasians (see Dorsett, Durand, and Wagner 1998).
Characteristics of the consumer unit included household size, region of residence, and housing tenure. Next to total expenditure, household size is considered to be the most important variable influencing expenditure on goods and services (Prais and Houthakker 1971). Region of residence was included to capture seasonal effects due to weather, as well as the possibility of regional price effects. Winakor (1969) suggests that apparel expenditure may be influenced by inventory effects. That is, clothing that is out of season may have to be stored by the consumer unit. Thus, housing tenure was included to control for the possible effects of differences in storage facilities between homeowners and renters. In addition, for consumer units with equity in their homes, housing tenure may capture a wealth effect.
For each quarter, the model was specified as follows:
1nQAE = [[alpha].sub.i] + [[beta].sub.1] ln[TQE.sub.i] + [lambda][Age.sub.i] + [[lambda].sub.2] [[Age.sup.2].sub.i] + [[beta].sub.2] [Married.sub.i] + [[beta].sub.3] [Married.sub.i] *ln[TQE.sub.i] + [[alpha].sub.2][Female.sub.i] + [[alpha].sub.3] White [collar.sub.i] + [[alpha].sub.4][Retired.sub.i] + [[lambda].sub.3][Education.sub.i] + [[alpha].sub.5]African-[American.sub.i] + [[alpha].sub.6][Hispanic.sub.i] + [[beta].sub.4] African-American*ln[TQE.sub.i] + [[beta].sub.5]Hispanic*ln[TQE.sub.i] + [[lambda].sub.4]Household [size.sub.i] + [[alpha].sub.7][Northeast.sub.i] + [[alpha].sub.8][Midwest.sub.i] + [[alpha].sub.9][South.sub.i] + [[alpha].sub.10][Mortgage.sub.i] + [[alpha].sub.11][Rent.sub.i] + [[alpha].sub.12][NoRent.sub.i] + [[epsilon].sub.i],
where i = 1, 2, ... 955 identifies each observation (per capita), [alpha], [beta] and [lambda] are parameters, * shows an interaction term, and [[epsilon].sub.i] is the error term (assumed to be normally distributed with zero mean and a constant variance).
Means and standard deviations for Total Quarterly Expenditure (TQE) and Quarterly Apparel Expenditure (QAE) are reported in Table 1 in per capita terms. In addition, the ratio of QAE to TQE is reported. All three descriptors suggest a seasonal pattern in apparel expenditure. The me an value for QAE was lowest in the winter quarter and highest in the fall quarter, which includes Christmas. The standard deviation showed the least variance in the summer quarter and the most variance in the fall quarter.  In addition, the ratio of QAE to TQE was lowest in the winter and summer quarters (.061) and highest in the fall quarter (0.075). Overall, the ratio of QTE to TAE was slightly higher than the national average for the year, which was about 5.7 percent.
The mean value for total apparel expenditure was $535.62. QAE represented 21 percent of TAE in the first quarter and 32 percent in the fourth quarter, suggesting seasonality and highlighting the fact that pooling data across quarters may result in lost information.
Characteristics of the reference person included age, marital status, gender, ethnicity, occupation, and education. The sample was constrained to only those households reporting four consecutive quarters of expenditure data. The mean age of the reference person was fifty years, higher than that of the average household head in the U.S.  Thirty-four percent of the consumer units were headed by females, and 60 percent were headed by married individuals. Nine percent of the consumer units were headed by African-Americans, 6 percent by Hispanics, and 84 percent by Caucasians. The average education level was 16.2 years, which was higher than the U.S. average. Forty percent of the consumer units reported reference persons who were white collar, 31 percent reported reference persons who were blue collar, and 29 percent reported reference persons who were retired or not working.
Characteristics of the consumer unit included household size, home tenure, and geographic location. The average household size was 2.8. Forty-five percent of the consumer units reported owning their homes and having mortgages, 28 percent reported owning with no mortgage, 26 percent reported renting, and 1 percent reported living in a dwelling but paying no rent. Twenty percent of the households were located in the northeast, 23 percent in the mid-west, 26 percent in the south, and 31 percent in the west.
Differences in Expenditure Patterns by Quarter
The purpose of the research was to explore the moderating effect of seasonality on household apparel expenditures. This exploration could be accomplished either by pooling the quarterly data and using dummy variables to represent three of the four quarters or by analyzing the unpooled data on a quarter-by-quarter basis. Given the research objectives, there were two reasons to avoid pooling the data. First, the only immediate gain from pooling the data would be increased efficiency (lower variances) of the parameter estimates. Second, and more importantly, pooling the data would obscure consumer responses to changes in season, which would be reflected in coefficients on the independent variables that change size or direction from one season to another. An F-test was used to evaluate the poolability of the data;  the results (F = 7.42; d.f. = 20, 955, p [less than] 0.01) showed that the data were not poolable. Consequently, a quarter-by-quarter analysis of the unpooled apparel expenditure data was conducted .
The results of the quarter-by-quarter regression analysis, based on data from the 1990 to 1991 CE Survey are presented in Table 2. The results of the PE test (Bera and Jarque 1982), which was used to compare the linear and double-logarithmic functional forms, supported use of the double-logarithmic model. The results of the F-test showed the model to be significant in all four quarters. The adjusted R-squared values were 0.295 (winter), 0.260 (spring), 0.326 (summer), and 0.416 (fall), revealing that considerable variation in quarterly apparel expenditures was explained by the model.
The income elasticity of apparel expenditures varied by season (as discussed below). Other variables showing seasonal variation (affecting apparel expenditures in some, but not all, quarters), included age, age-squared, marital status, the interaction of marital status and TQE, African-American ethnicity, the interaction of African-American ethnicity and TQE, household size, region, and housing tenure. Variables affecting apparel expenditures in all four quarters were sex, education, white collar occupation, Hispanic ethnicity, and the interaction of Hispanic ethnicity and TQE.
In the discussion that follows, ways in which the results support the theory are noted. Similarities and differences between the results and those of previous research are reported and explained. Finally, results of the research that represent new findings concerning the effect of seasonality on apparel expenditure are highlighted.
Given the importance of the income-expenditure relationship in neoclassical demand theory, the discussion begins with the relationship of TQE (the proxy for income) to QAE for the consumer units in the sample. Of particular interest is the difference in income elasticity observed by season. The seasonal (and nonseasonal) effects of socioeconomic and demographic variables are then discussed in the order in which they are presented in the model.
Total Quarterly Expenditure
One attractive feature of the double-logarithmic functional form is that it allows reading of the elasticity of QAE directly from the coefficient of TQE. As shown in Table 2, the effect of income was positive in all four quarters. However, the income (TQE) elasticities appeared low compared to previous cross-sectional studies, ranging from 0.40 for spring to 0.62 for fall. The results of the Chow test showed the elasticity was highest for fall. The result was expected for two reasons. First, fall and winter apparel, most of which is purchased during the fall quarter (October through December), tends to be higher priced than that of spring and summer apparel. Second, fall includes the holiday season, when many retailers realize as much as 25 percent of annual sales. In previous research, apparel has been shown to be the most popular gift item (Belk 1979; Caplow 1982; Wagner and Garner 1992).
While the elasticities in this study are consistent with economic theory by showing apparel to be a normal good, they are lower than elasticities for apparel expenditure reported in previous cross-sectional research (based on TQE). The elasticities of this study are well below one, suggesting that apparel is a necessity. In previous research, elasticities across all households have ranged from 1.17 (Dardis et al. 1981) to 1.44 (Wagner and Soberon-Ferrer 1990), suggesting that apparel is a luxury. The elasticities of this study may differ from those reported previously because of decreases over time in the price of apparel, relative to other goods, leading to a decline in apparel's share of the household budget (Winakor 1989). There is also evidence that the status of apparel as a luxury good may be diminishing (Hughes 1996). However, the robustness of the results of this study are supported by the results of a sensitivity analysis, which showed elasticities lower than one for all quarters, under a variety of conditions,  and the results of a time-series analysis of apparel expenditure by Mokhtari (1992), which showed a long-run elasticity of 0.50.
Characteristics of the Reference Person
Both age and age-squared were included in the analysis to test for a possible life cycle effect, as suggested by Norum's (1989) analysis. The results confirm that for the consumer units in the sample, the relationship between age and apparel expenditure differs by quarter. In spring and summer, both age and age-squared were significant, suggesting an inverted U-shaped relationship between age and apparel expenditure, in which apparel expenditure peaks at age forty-eight in spring and age forty-three in summer. Neither age nor age-squared show any effect in winter.
The marital status of the reference person was not significant in any quarter. However, the interaction of marital status with TQE was significant in the fourth quarter, which includes the holiday season. This indicates that per capita apparel expenditure is more elastic for married couples (TQE elasticity was 0.77) than for unmarried individuals. Married couples may have more extensive social networks than single individuals, with more active social lives during the holiday season. Consequently, expenditure for apparel for personal use (as well as for gifts) may account for a higher proportion of the budget during the fourth quarter.
Consumer units with a female designated as the reference person spent more on apparel in all quarters than comparable CUs with a male reference person. This result is consistent with the results of previous research by Wagner and Soberon-Ferrer (1990), which revealed that consumer units with female reference persons were reported to spend more annually on apparel than other CUs.
For the consumer units in the sample, significant effects for education and occupational status were observed in all quarters. Households in which the reference person had more education had higher QAE, and white collar households had higher QAE than blue collar households. Apparel is considered one marker of status, so households with higher educational and occupational status might be expected to spend more per capita than others on apparel. The result for education is consistent with previous research, based on both the annualized and the quarterly data (e.g., Dardis et al. 1981; Horton and Hafstrom 1985; Norum 1989; Wagner and Soberton-Ferrer 1990), showing college graduates to spend more than others on apparel. The result for occupational status is new; this research is the first to demonstrate that households with a white collar reference person spend more per capita than comparable blue-collar households. In previous research, the analysis of occupation has been based on a working/nonworking dichotomy .
The apparel expenditure of the African Americans in the sample differed from that of Caucasians in fall quarter, when they spent less than Caucasians. Previous research on the effect of race has yielded mixed results. On the one hand, studies based on annualized data (e.g., Dardis et al. 1981; Dorsett et al. 1998; Horton and Hafstrom 1985; Wagner and Soberon-Ferrer 1990) suggest that African Americans spend more than Caucasians. On the other hand, studies based on the quarterly data (e.g., DeWeese 1993; DeWeese and Norton 1991; Norum 1989) show no effect for race. All previous research based on the quarterly data has involved some measure of income, rather than TQE; in the case of African Americans, there may be a relatively high incidence of incomplete income reporting, creating bias in the results. Unlike most other research, the data in this study include expenditure for gifts of apparel, which are less likely to be given by African Americans than Caucasians (Wagner and Garner 1992). The variable represen ting the interaction of African-American ethnicity and TQE was significant in the fall quarter, showing the TQE elasticity of apparel to be higher (0.84) for African-Americans than Caucasians
For the consumer units in the sample, Hispanics spent more per capita than Caucasians on apparel in every quarter. However, the coefficients of the interaction terms showed that the TQE elasticities of Hispanics were lower than those of Caucasians in all quarters, with values ranging from 0.13 in the summer to 0.31 in the winter. This is a new finding, the cause of which is not readily apparent.
Characteristics of the Consumer Unit
Characteristics of the consumer unit included household size, region of residence, and housing tenure. Household size affected QAE in the winter quarter only, when the relationship was negative. As household size increases, economies of scale may become operative as apparel items are shared or handed down (Winakor 1969). Indeed, DeWeese et al. (1991) demonstrate that the number of children (except teenage boys) in a household is negatively related to per capita apparel expenditure.
Consumer units in the sample from the Northeast reported higher QAE than those in the West in spring and summer. This result may reflect differences in lifestyle, in that the Northeast may be somewhat more formal than the West. It may also reflect differences in climate, in that the Northeast may be cooler than the West, on average, in the spring and the summer. Households in the South reported expenditures than those of households in the West for spring.
Housing tenure was first used by Horton et al. (1985), who showed that among households with children, renters spend more than homeowners on apparel. Wagner and Soberton-Ferrer (1990) showed that across all households, renters spend less than homeowners. The results of the research reported here, which are based on quarterly per capita data, show that compared to homeowners with no mortgage, renters in the sample spend less on apparel in winter, spring, and fall. Homeowners with mortgages spend less than homeowners with no mortgages in the fall quarter only. Consumers occupying a dwelling but paying no rent reported spending less than homeowners with mortgages in the fall.
In this research, the moderating effect of seasonality on per capita apparel expenditure was evaluated using cross-sectional, micropanel data from the Bureau of Labor Statistics' Consumer Expenditure Survey. The results contribute to development of literature on the empirical analysis of apparel demand in two ways. First, theoretical and methodological issues encountered in previous research using the CE Survey data (one of the major sources of data for analyzing apparel expenditure) have been addressed. These include specification of the model, the use of total expenditure as a proxy for income, testing for appropriateness of the double-logarithmic functional form, and testing for the poolability of the quarterly data. The second contribution has been to add to the body of knowledge about environmental factors influencing apparel expenditure. The results demonstrate that seasonality does, in fact, moderate the relationship between consumer characteristics and apparel expenditure. For the consumer units in t he sample, income elasticity varied by quarter, as did the effects of age, marital status, African-American ethnicity, household size, region of residence, and home tenure. Other results show that Hispanics spend more on apparel than Caucasians but have lower income elasticity and that there is a life-cycle effect for age.
This research is subject to certain limitations. One limitation is the use of a single equation model rather than a system of simultaneous equations. Such a system would provide insight into how consumers with various characteristics trade off seasonal expenditures between one category of good (or service) and another. A second limitation is that the sample was restricted to only those households reporting apparel expenditure in four consecutive quarters. Therefore, it is important to exercise caution in generalizing from the results.
The CE Survey is a rich source of information on consumer expenditures. Data on expenditures in numerous product categories, ranging from food to electronic goods are available. While the effect of seasonaity on expenditures in many product categories has been analyzed using time series data, there has been little methodologically sound research on the effect of seasonality using the cross-sectional, micropanel data. Thus, little is known about how seasonality moderates the effect of numerous socioeconomic and demographic characteristics of households on expenditures in most product categories. Clearly, the CE Survey is a potentially valuable resource for future research on the effect of seasonality on consumer expenditures.
Janet Wagner is Associate Professor, Department of Marketing, and Manouchehr Mokhtari is Associate Professor, Department of Family Studies, University of Maryland, College Park.
(1.) A consumer Unit (CU) is defined by the B.L.S. as two or more persons, usually living together, who pool income to share expenses. A CU may also be a consumer, living alone or sharing a household, who is financially independent (Garner, Zieschang, and Miller 1989).
(2.) In the remainder of this paper, the terms consumer unit and household will be used interchangeably.
(3.) The reference person is the first person named in the interview as owning or renting the dwelling.
(4.) The calmness (low variation) in third quarter apparel expenditure before the storminess (high variation) in the fourth quarter is also reflected in the range of values for expenditure in the third quarter ($1,576) compared to that of the fourth quarter ($3,425).
(5.) This may reflect the fact that older households tend to move less frequently than younger. Thus, older households were more likely to have four consecutive quarters of expenditure data.
(6.) Use of the F-test is based on the assumption of homoscedastic variance. We tested the null hypothesis of homoscedasticity using White's (1980) test. In every quarter, the results confirmed the null hypothesis. The results are presented in Table 2.
(7.) Elasticities were calculated for household apparel expenditure at the mean, using linear and double-log models. For the linear model, elasticities ranged from 0.26 for winter to 0.51 for fall. For the double-log model, elasticities ranged from 0.43 for spring to 0.68 for fall. Elasticities were also calculated for per capita expenditure at the mean, using the linear model. For the linear model, elasticities ranged from 0.21 to 0.38 for fall.
Belk, Russell. 1979. Gift-Giving Behavior. In Research in Marketing, edited by J. Sheth, Greenwich, CT: JAI Press.
Bera, A. K. and C. M. Jarque. 1982. Model Specification Tests: A Simultaneous Approach. Journal of Econometrics, 20:59-82.
Caplow, T. 1982. Christmas Gifts and Kin Networks. American Sociological Review, 47 (June):383-392.
Dardis, Rachel, Frederick Derrick, and Alane Lehfeld. 1981. Clothing Demand in the United States: A Cross-Sectional Analysis. Home Economics Research Journal, 10 (December):212-222.
Deaton, Angus. 1992. Understanding Consumption, Oxford: Clarendon Press.
DeWeese, Gail. 1993. Female-Headed Households: Impact of Family Structure on Quarterly Expenditures for Clothing. Home Economics Research Journal, 22 (September):39-57.
DeWeese, Gail and Marjorie J. T. Norton. 1991. Impact of Married Women's Employment on Individual Household Member Expenditure for Clothing. The Journal of Consumer Affairs, 25,2:235-257.
Dorsett, A. D., Richard Durand, and Janet Wagner. 1998. The Effects of Social Class and Ethnicity on Consumer Expenditures for Apparel. In A.M.A. Winter Educators' Conference Proceedings, edited by D. Grewal and C. Pechman, Chicago: American Marketing Association.
Friedman, Milton. 1957. A Theory of the Consumption Function, Princeton, NJ: Princeton University Press.
Garner, Thesia I. and Laura Blanciforti. 1993. Household Income Reporting: An Analysis of U.S. Consumer Expenditure Survey Data. Journal of Official Statistics, 10, 4.
Garner, Thesia I., Kimberley D. Ziesehang, and Richard D. Miller. 1989. Using the Consumer Expenditure Survey Data: Past, Present, and Future Research. Journal of Economic and Social Measurement, 15, 3-4:237-279.
Horn, Marilyn J. and Lois M. Gurel. 1981. The Second Skin, 3rd edition, Boston: Houghton Muffin.
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 (March):292-303.
Hughes, Kathleen A. 1996. Kids, Cabins, and Free Time Say Status in Understated 90's. The Wall Street Journal, (September 30):B1 and B2.
Moehrle, Thomas G. 1994. Seasonal Adjustment of Quarterly Consumer Expenditure Series. Monthly Labor Review, (December):38-46.
Mokhtari, Manouchehr. 1992. An Alternative Model of U.S. Clothing Expenditures: Application of Cointegration Theory. The Journal of Consumer Affairs, 26, 2:305-323.
Norum, Pamela. 1989. Economic Analysis of Household Expenditures on Apparel. Home Economics Research Journal, 17 (March):228-240.
Prais, S. J. and H. S. Houthakker. 1971. Analysis of Family Budgets, Cambridge: Cambridge University Press.
Rubin, Rose M., Bobye J. Riney, and David J. Molina. 1990. Expenditure Pattern Differentials Between One-Earner and Dual-Earner Households. Journal of Consumer Research, 17 (June):43-52.
Scott, Stuart and James Buszuwski. 1993. Seasonal Pattern Assessment of Consumer Expenditures. U.S. Department of Labor, Bureau of Labor Statistics Statistical Note No. 32, Washington, DC: U.S. Government Printing Office.
Turcsik, Richard. 1995. Drink Stirrers: Cocktail Mixers that Go With White Liquors are Bracing Sales in the Beverage Aisle. Supermarket News, 45 (October 9):49-50.
U.S. Department of Labor, Bureau of Labor Statistics. 1993. Consumer Expenditure Survey, 1990-91, Bulletin 3425, Washington, DC: U.S. Government Printing Office.
Wagner, Janet and Sherman Hanna. 1983. The Effectiveness of Family Life Cycle Variables in Consumer Expenditure Research. Journal of Consumer Research, 10 (December):281-291.
Wagner, Janet and Thesia Garner. 1992. Extrahousehold Giving in Popular Gift Categories: A Socioeconomic and Demographic Analysis. In Advances in Consumer Research, Vol. XX, edited by Leigh McAlister and Michael L. Rothschild, Provo, UT: Association for Consumer Research.
Wagner, Janet and Horatio Soberon-Ferrer. 1990. The Effect of Ethnicity on Selected Household Expenditures. The Social Science Journal, 27, 2:181-198.
White, Halbert. 1980. A Heteroscedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroscedasticity. Econometrica, 48, 4:817-838.
Wilkes, Robert E. 1995. Household Life-Cycle Stages, Transitions, and Product Expenditures. Journal of Consumer Research, 22 (June):27-42.
Winakor, Geitel. 1969. The Process of Clothing Consumption. Journal of Home Economics, 61, 8:629-634.
Winakor, Geitel. 1989. The Decline in Expenditures for Clothing Relative to Total Consumer Spending, 1929-1986. Home Economics Research Journal, 17:195-215.
|Printer friendly Cite/link Email Feedback|
|Author:||WAGNER, JANET; MOKHTARI, MANOUCHEHR|
|Publication:||Journal of Consumer Affairs|
|Date:||Dec 22, 2000|
|Previous Article:||Out-of-Pocket Health Care Expenditure Patterns and Financial Burden across the Life Cycle Stages.|
|Next Article:||Relationships among Information Search Activities When Shopping for a Credit Card.|