WagnerHanna FamilyLifeCycle JCR 1983

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/24098345

The Effectiveness of Family Life Cycle Variables in Consumer Expenditure


Research

Article  in  Journal of Consumer Research · February 1983


DOI: 10.1086/208967 · Source: RePEc

CITATIONS READS

150 1,321

2 authors, including:

Sherman D. Hanna
The Ohio State University
257 PUBLICATIONS   3,509 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

risk tolerance of households in China View project

retirement age View project

All content following this page was uploaded by Sherman D. Hanna on 05 May 2018.

The user has requested enhancement of the downloaded file.


The Effectiveness of
Family Life Cycle Variables in
Consumer Expenditure Research"^
JANET WAGNER
SHERMAN HANNA**

Traditional famiiy life cycle, revised family life cycle, and family composition vari-
ables are compared in models predicting total family clothing expenditures. There
is little difference In the predictive ability of the three sets of variables. In models
controlling for socioeconomic and demographic variables, family life cycle and
family composition variables have little predictive ability, independent of income.

F amily iife cycle is a construct which has been popular


in consumer expenditure research for the past 30 years
(e.g., Amdt 1979; Lansing and Morgan 1955; Wells and
ment of Health and Human Ser\'ices 1982) have resulted
in an increase in the number of families who are excluded
from traditional family life cycle categories. As a result,
Gubar 1966). A set of composite variables, family life cycle there has been controversy in the consumer behavior liter-
combines trends in income and family composition with ature over the appropriate defmition of the family life cycle.
changes in demands placed on income. In economic theor)', Consumer researchers ha%'e proposed increasing the number
income and family composition are considered the most of family life cycle stages in order to accommodate non-
important variables influencing family expenditures (Brown traditional family forms, such as single parent families and
and Deaton 1972). If family life cycle successfully captures married couples who choose not to bear children (Gilly and
the effects of changes in income and family composition Enis 1981; Murphy and Staples 1979; Stampfl 1978). Such
on expenditures, then family life cycle should be a ver>' revisions presuppose that increasing the number of family
powerful determinant of expenditures for goods and ser- life cycle stages would be useful in predicting family ex-
vices. penditures for goods and services. If this is a worthwhile
Despite its popularity, family life cycle has been criti- exercise, then the superiority of revised defmitions should
cized. Ferber (1979) has argued that when socioeconomic be demonstrated. To date, there has been no substantiating
and demographic variables—especially income—are con- research.
trolled, family life cycle variables may have little effect on This paper reviews prior use of family life cycle variables
expenditures. Other critics (Murphy and Staples 1979; Trost in consumer expenditure research and discusses altemative
1974) have suggested that the effect of family composition definitions of the family life cycle. The predictive ability
is obscured by family life cycle variables. There is no em- of a traditional family life cycle model (Wells and Gubar
pirical evidence to support or refute these contentions. 1966) is compared to that of revised family life cycle and
The well-documented incidences of divorce, postpone- family composition models. The predictive effectiveness of
ment of marriage and childbearing. and voluntary child- each of these models is then compared to a set of socio-
lessness (U.S. Bureau of the Census 1982; U.S. Depart- economic and demographic variables. Clothing is the ex-
penditure category used for the purposes of this compari-
son.
*This article received an honorable menlion in the 1983 Robert Ferber
Award for Consumer Research competition for the best interdisciplinary
anicle based on a recent doctoral dissertation. The award is cosponsored
by the Association for Consumer Research and the Journal of Consumer
PRIOR RESEARCH
Research.
Family life cycle variables have been used to study a
**Janei Wagner is an Assistant Professor in the Department of Textiles
and Consumer Economics, University of Marjiand. College Park. MD
wide range of family financial characteristics and expen-
20742. Sherman Hanna is an Associate Professor in the Department of diture pattems. Lansing and Morgan (1955) studied income
Family Economics. Kansas State University. Manhattan. KS 66506. The and assets as well as expenditures for housing and durables,
authors would like to thank both Elizabeth McCullough and three anon- using a seven-stage family life cycle model based on the
ymous reviewers for their comments on the preliminarj' drafts of this
paper. The research was conducted at Kansas State University.
age and marital status of the household head and the age
of the youngest child. The pattem of income over the family
281
C JOURNAL OF CONSUMER RESEARCH • Vol. 10 • December 198.'
282 THE JOURNAL OF CONSUMER RESEARCH

life cycle followed an inverted U-shape. Income increased by the addition of a stage called the Mature Nest, which
over the life cycle, peaking when there were dependent included households with more than two adults, but no
children in the family. As the children left home, income children under the age of 18; the Murphy and Staples model
decreased until the household head retired. was modified by the inclusion of never-married and pre-
Wells and Gubar (1966) expanded the Lansing and Mor- maturely widowed parents. An analysis of covariance
gan family life cycle to include the employment or retire- showed that in both models, expenditures for energy varied
ment of the household head. Using data from both the 1963 by stage in the family life cycle. However, there was no
Sur\'ey of Consumer Finances and the 1960-1961 Survey comparison of the effectiveness of the two models in pre-
of Consumer Expenditures, the effectiveness of family life dicting energy expenditures. Derrick and Lehfeld (1981)
cycle variables was compared to that of age of the house- proposed a simplified version of the family life cycle, in
hold head in predicting expenditures for food, durables, which family life cycle variables were treated individually.
housing, and vacations. In each comparison, family life They compared two models in predicting expenditures for
cycle variables were more effective. Wells and Gubar also food: a model that included the family life cycle variables
used the data from the 1960-1961 Survey of Consumer used by Wells and Gubar (1966), and a model that included
Expenditures to compare the effectiveness of age of the the individual variables used to define family life cycle (age
household head to the effectiveness of stage in the family and marital status of the household head and age of the
life cycle in predicting expenditures for selected clothing youngest child). There was a slight improvement in pre-
items. They found that age of the household head was su- dictive ability when these variables were treated individ-
perior to family life cycle in predicting expenditures for ually.
luxury items such as fur coats, as well as for more obviously To date, the family life cycle stages proposed by Wells
age-related garments such as women's shorts and dunga- and Gubar (1966) and Murphy and Staples (1979) have not
rees. Much of the ensuing expenditure research using fam- been used in models designed to explain the variance in
ily life cycle variables has been based on the Wells and family clothing expenditures. Theoretically, the Murphy
Gubar model, which has been used to study the consump- and Staples model might be expected to explain more of
tion of food (Derrick and Lehfeld 1980), ser\'ices (Amdt the variance than would the Wells and Gubar model because
1979), and leisure time (Hisrich and Peters 1974; Landon it is composed of more stages and describes families more
and Locander 1979; Settle, Alreck, and Belch 1979). accurately.
Murphy and Staples (1979) argued that a great deal of
information is lost because nontraditional families are ex-
cluded from the Wells and Gubar family life cycle model.
Conceptual Limitations
Their modemized family life cycle, unique in recognizing A second criticism of family life cycle research has been
both divorce and childlessness, contained 13 stages. Al- that family life cycle has conceptual limitations. Family life
though the Murphy and Staples model was designed to be cycle variables obscure the effects of both family size and
more inclusive than the Wells and Gubar model, it still the age/sex structure of the family (Murphy and Staples
excluded cohabiting couples, never-married individuals 1979; Trost 1974), variables which are used in defining
with children, married couples who are separated, and family composition. Next to income, family composition
young and middle-aged widows and widowers. The Mur- is considered to be the most important variable influencing
phy and Staples model has been used by Frey and LaBay family expenditures for goods and services (Brown and
(1983), Fritzsche (1981), and Zimmerman (1981) in anal- Deaton 1972).
yses of household energy expenditures. As of this study, Family life cycle variables are related to family compo-
there has been no research comparing the predictive ability sition variables in that family composition changes as it
of the Murphy and Staples model to that of the Wells and progresses through the life cycle. In a family composition
Gubar model. model, each variable represents the number of individual
family members in one of a series of mutually exclusive
age and sex categories. Since clothing is usually purchased
CRITICISMS OF for and wom by individual family members, family com-
FAMILY LIFE CYCLE RESEARCH position variables may be more effective than family life
cycle variables in predicting family clothing expenditures.
Failure to Test Altemative Definitions
In a multiple regression analysis of clothing expenditures
Consumer expenditure research involving family life by low income families participating in the Rural Income
cycle variables has been critized for three reasons. First, Maintenance Experiment (RIME), Hager and Bryant (1977)
altemative definitions have not been tested over a compre- used nine age and sex categories to define family compo-
hensive range of goods and services (Ferber 1979). Recent sition. Among wage-eaming families, the number of male
studies have been designed to correct this inadequacy. dependents between the ages of three and 15 and the num-
Fritzsche (1981), for example, compared expenditures for ber of female dependents 11 years of age or older were
energy over the family life cycle using modified versions positively related to family clothing expenditures. Among
of both the Wells and Gubar (1966) and Murphy and Staples nonwage-eaming families, the number of infants, the num-
(1979) models. The Wells and Gubar model was modified ber of male dependents 11 years of age or older, and the
EFFECTIVENESS OF FAMILY LIFE CYCLE 283

number of female dependents seven years of age or older family life cycle model, a revised family life cycle model,
were positively related to clothing expenditures. The pres- and a model composed of socioeconomic and demo-
ence of a wife 55 years of age or older was negatively graphic variables that are not used in defming family life
related to family clothing expenditures. There was no at- cycle.
tempt to determine the size of the effect of the family com- Although this study was limited to total family clothing
position variables. expenditures, it was expected that the results might be gen-
eralized to other expenditure categories.
Failure to Control for Socioeconomic
and Demographic Variables METHOD
Finally, research using family life cycle variables has The Sample
been criticized for failing to control for socioeconomic and
demographic variables, especially income (Ferber 1979). The sample used in this study was taken from a public
This has been a problem of studies limited to a single ex- use tape file of expenditure data collected in the 1973 quar-
penditure category (Hisrich and Peters 1974; London and terly interview component of the 1972-1973 Consumer
Locander 1979) and of studies encompassing all of the Expenditure Survey (CES). This survey is the most recent
major consumption categories (Amdt 1979). Family life and most comprehensive source of information available on
cycle is a multivariate concept defined by the age, marital the expenditures and incomes of families classified accord-
status, and employment status of the household head and ing to socioeconomic and demographic characteristics. One
by the age of the youngest child. As those variables change limitation in using these data for a family cycle analysis is
over the life cycle, family income changes. Much of the that the CES data are cross-sectional and not longitudinal.
observed effect of family life cycle is then assumed to be In using cross-sectional data, it was assumed that the future
related to changes in income as the household head, in clothing expenditures of families could be inferred from
particular, ages. While this may be the case, there is no their expenditures at the time of the survey.
research demonstrating how much of the effect of family The unit of obser\'ation used in this study was the con-
life cycle is attributable to income and how much is attrib- suming unit. As defined by the Bureau of Labor Statistics,
utable to the variables used to define family life cycle. In a consuming unit consists of (1) a family of two or more
other words, there is no research comparing the predictive persons who pool their income and draw from a common
ability of family life cycle to that of (1) income alone or fund for major items of expenditure, or (2) a single con-
(2) income in conjunction with socioeconomic and demo- sumer who is financially independent of any family group
graphic variables that are not used in defining family life (Carlson 1974). The head of the consuming unit is defined
cycle. as the husband in husband'wife families. In non-husband/
If income and other socioeconomic and demographic wife families, the head is that individual designated as such
variables are controlled, the amount of the variance in ex- by other members of the consuming unit.
penditures directly attributable to family life cycle variables Certain consuming units were dropped from the sample,
can be isolated. In a model controlling for socioeconomic including those that reported having more than nine mem-
and demographic variables, Dardis, Derrick, and Lehfeld bers and those that reported total consumption expenditures
(1981) found a set of family life cycle variables significant of over S 100,000. To minimize the effect of those extreme
in predicting total clothing expenditures. However, they did values on the results, 72 units were eliminated from the
not report the proportion of the variance in clothing expen- original sample of 10,106 units (U.S. Department of Labor
ditures that was directly attributable to variance in the fam- 1978), leaving 10,034 consuming units to be used in this
ily life cycle variables. study.
Based on these criticisms of family life cycle studies, the
present study had the following objectives: The Independent Variables
Four sets of independent variables were used in this
• To compare the effectiveness in predicting total family study. The set of traditional family life cycle variables was
clothing expenditures of a set of traditional family life defined by the age or employment of the household head,
cycle variables (Wells and Gubar 1966) to that of a set the marital status of the household head, and the age of the
of revised family life cycle variables based on the model youngest child. The set of revised family life cycle variables
proposed by Murphy and Staples (1979). was defined by the age or employment of the household
• To propose a set of family composition variables as an head, the marital status of the household head, and the age
altemative to either set of family life cycle variables. of the oldest child. The family composition variables each
• To determine the size of the effect of family life cycle represented the number of family members in one of 13
variables in models controlling for socioeconomic and mutually exclusive age and sex categories. The set of so-
demographic variables, including total consumption ex- cioeconomic and demographic variables included variables
penditures (a proxy for income). other than those used in defining family life cycle which
• To compare the predictive ability of total consumption are known to infiuence or were hypothesized to infiuence
expenditures (a proxy for income) to that of a traditional family clothing expenditures.
284 THE JOURNAL OF CONSUMER RESEARCH

TABLE 1
REGRESSION RESULTS FOR TOTAL FAMILY CLOTHING EXPENDITURES: THE TRADITIONAL FAMILY LIFE CYCLE VARIABLES

Clothing expenditures as percentage


of expenditures by bachelors

Mean Model controlling for


N expenditures socioeconomic and
Variables and descriptions (N = 10.034) in dollars Basic model demographic variables

Bachelor (base variable) 811 450 100 100


Head less than 45 years ot age.
never married
Newly married 674 644 162* 88*
Head less than 45 years of age.
married, no children
Full nest la 1.607 760 211* 101
Head less than 45 years of age.
married, youngest child less than 6
Full nest lb 1.046 959 285* 111'
Head less than 45 years of age.
married, youngest child 6 or older
Full nest It 1.029 989 283" 109*
Head 45 years of age or older,
married, youngest child 6 or older
Empty nest I 1.621 626 145" 82*
Head 45 years of age or older,
married, employed, no dependent children

Empty nest II 622 294 64* 78*


Head 45 years of age of older,
married, retired, no dependent children
Solitary survivor 1 724 ' 339 62* 80*
Head widowed and employed
Solitary survivor II 838 217 36* 80*
Head widowed and retired
Other 1.062 691 169* 125
W (0.24) (0.54)
F value 350.55 347.67
Degrees of freedom 9/10024 34/9999

'Signilicantly ditferent Irom 1.00 at Ihe 0 01 level.


'%gnilicanlty diHerenl Irom 1 00 at Ihe COS level.

Traditional Family Life Cycle Variables. The set of tra- Revised Fatnily Life Cycle Variables. The revised set
ditional family life cycle variables was chosen for two rea- of family life cycle variables was adapted from the mod-
sons. First, it is the set of family life cycle variables de- emized family life cycle proposed by Murphy and Staples
veloped by Wells and Gubar (1966) and used most (1979; see Table 2). Like the Murphy and Staples model,
frequently in empirical work. Second, the ability of the this set of variables was designed to be flexible enough to
basic .traditional family life cycle model to effectively pre- accommodate both changing demographic trends and non-
dict family clothing expenditures was thought to be limited traditional flows through the family life cycle. Individual
by its exclusion of nontraditional family forms. It was then categories were defined narrowly in order to describe fam-
assumed that most families follow the traditional sequence ilies as accurately as possible. At the same time, categories
of stages through the family life cycle. We used nine were collapsed when necessary in order to include sufficient
dummy variables to represent the basic traditional family information for the purposes of analysis.
life cj'cle model (see Table 1). Expenditures by bachelors The Murphy and Staples family life cycle model was
was the base variable. modified in three ways. First, all single household heads—
EFFECTIVENESS OF FAMILY LIFE CYCLE 285

TABLE 2
REVISED FAMILY LIFE CYCLE VARIABLES BEFORE AND AFTER COLLAPSING STAGES

Defining variables N
N (after
Original family life Original Age of household Age of oldest (before Stage or substage collapsing)
cycle stage substage head child collapsing) used in analysis

Young single adult 18-34 654 Young single adult 654


(base variable)
1 0-2 29]
il 3-5 68 » Young single parent 1 205
Young single parent III 18-34 6-11 108]
IV 12-17 62 Young single parent II 62

Young married adult 18-34 549 Young married adult 549

1 0-2 423 Young married parent 1 423


II 3-5 382 Young married parent II 382
Young married parent III 18-34 6-11 486 Young married parent Itl 486
IV 12-17 167 Young married parent IV 167

Mature single adult 35 or older 1.220 Mature single adult 1.220


and employed
1 0-2 8]
II 3-5 6 • Mature single parent 1 50
Mature single parent III 35 or older 6-11 36 J
IV and employed 12-17 317 Mature single parent II 317

Mature married adult 35 or older 1.746 Mature married adult 1.746


and employed
1 0-2 ^^ — » Mature married parent 1 117
II 35 or older 3-5
Mature married parent III and employed 6-11 393 Mature married parent II 393
IV 12-17 1.830 Mature married parent III 1.830

Older single adult 726 Older single adult 726

Older married adult 622 Older married adult 622

Other 85 Other 85

NOTE: Brackets indicate stages that have been collapsed

whether divorced, widowed, or never married—were as- There were originally 10 major stages in our revised fatn-
signed the satne status. Neither never-married nor prema- ily life cycle. The four childbearing and childrearing stages
turely widowed household heads were included in the Mur- each included four substages, so that there were a total of
phy and Staples model. In the revised family life cycle, we 22 possible variables. Despite the size of the sample, some
assumed that the clothing expenditure pattems of all single stages with children were underpopulated and had to be
household heads are similar. This assumption was neces- collapsed (see Table 2). Consequently, the revised family
sary because the CES data did not distinguish among di- life cycle used in the regression analysis included only 18
vorced, widowed, and never-married individuals. Second, stages. There were 17 dummy variables used to represent
the Murphy and Staples model was modified by defining the basic revised family life cycle model. The base category'
childbearing and childrearing stages according to the age was young single adults.
of the oldest—not the youngest—child. Parents may invest According to Murphy and Staples (1979), one advantage
in clothing for the oldest child with the intention of handing to using their family life cycle is that it classifies more
it down to siblings who follow. Finally, Murphy and Sta- families than does the traditional family life cycle of Wells
ples recognized three categories of families with children: and Gubar (1966). While 8.4 percent of the obser\ations
families with infants aged zero to four, families with chil- were categorized as "other" in the traditional family life
dren aged five to 11, and families with children aged 12 to cycle, only 0.8 percent were categorized as "other" in the
17. In our revised family life cycle, infancy was defined as revised version of the Murphy and Staples model.
ending at age two because supplementary sources of cloth-
ing such as gifts, hand-me-downs, and borrowed garments Family Cotnposition Variables. For the comparison in
become less important after that age (Britton 1969). this study, the family composition model included 13 age
286 THE JOURNAL OF CONSUMER RESEARCH

TABLE 3 TABLE 4
REGRESSION RESULTS FOR TOTAL CLOTHING REGRESSION RESULTS FOR TOTAL FAMILY CLOTHING
EXPENDITURES: THE FAMILY COMPOSITION VARIABLES' EXPENDITURES; THE SOCIOECONOMIC MODEL*

Percentage change in clothing Variable F value d.f.


expenditures
Region 10.80 3,10030
Controlling for
socioeconomic and City Size 7.79 7,10026
Basic demographic Race
Variable Age 230.05 1,10032
models variables
Education 12.22 6,10027
Infant 0-2 104 103
Employment 19.30 7,10026
Male preschooler 3-5 114" 101
Total consumption expenditures 5813.32 1,10030
Female
preschooler 3-5 107 102 (0.53)
Male grade F value" 313.24
schooler 6-11 115" 103 Degrees of freedom 25/10008
Female grade
schooler •N - 10.0M.
6-11 114" 104 "p < 0.0001.
Male teenager 12-17 121" 108"
Female teenager 12-17 129" 117"
to be related to total family clothing expenditures (Dardis
Young adult male 18-34 147" 92" et al. 1981). Each of these variables was treated as a set of
Young adult dummy variables. Hafstrom and Dunsing (1972) suggested
female 18-34 180" 116" that since women are often the most important purchasing
Mature adult male 35-64 167" 94" agents for their families, either the effect of the wife's ed-
ucation or the combined effect of the educations of the
Mature adult
husband and wife may influence expenditures for goods and
female 35-64 152" 107"
services. Dummy variables which represented the educa-
Older adult male 65 or older 101 89" tional levels of the household head and the spouse were
Older adult then included in the set of socioeconomic and demographic
female 65 or older 83" 88" variables.
(0.25) (0.54)
The variable total consumption expenditures—a proxy
for income—was also included. The use of total consump-
F value 253.67 313.24 tion expenditures as a proxy for income is defensible for
Degrees of a number of reasons.' Previous research shows that total
freedom 13/10021 389990 consumption expenditures give a better fit than does income
• N = 10.034.
in models designed to predict total clothing expenditures
"Significani at the 0.01 level. (Dardis et al. 1981). The use of total consumption expen-
ditures can also be defended on the basis of the permanent
income hypothesis: in the short run, families have more
and sex categories (see Table 3). Age categories were es- control over their expenditures than over their incomes
tablished based on two considerations: the school placement (Houthakker and Taylor 1970; Prais and Houthakker 1955).
of children and the employment or retirement of adults. Finally, there is a long tradition of using total consumption
Although not all individuals retire at 65, it is likely that expenditures as a proxy for income in expenditure research.
many reduce their clothing expenditures in anticipation of Because one objective of this study was to test the predic-
retirement. There were 13 variables used to represent the tive ability of income, we tested total consumption expen-
family composition model. Each of these variables repre- ditures alone in a regression equation designed to predict
sented the number of family members in a given age and total clothing expenditures.
sex category. Every family was then described specifically The socioeconomic and demographic variables were in-
by its size and its age and sex structure. cluded in the analysis for two reasons. First, these variables
were necessary in order to determine the size of the effect
Socioeconomic and Demographic Variables. Socio-
economic and demographic variables included in this study 'Each model was also tested with gross income and disposable income.
were the occupation and education of the household head, The results were similar to those obtained with total consumption expen-
ditures. However, the family life cycle variables explained slightly more
the employment status of the spouse, geographic region, of the variance when either gross income or disposable personal income
and city size (see Table 4), all of which have been shown was included in the model.
EFFECTIVENESS OF FAMILY LIFE CYCLE 287

of family life cycle and family composition variables in where g, the relative effect on Y, can be shown to be (Ken-
models controlling for socioeconomic and demographic nedy 1981):
variables. Second, these variables were included in order
g = exp [C - V2V (C)] - 1
to compare the predictive ability of variables other than
those used in defining family life cycle to the predictive Consequently, the coefficients of the dummy variables for
ability of family life cycle itself. Because the focus of this the family life cycle variables represent clothing expendi-
paper is on family life cycle variables, discussion of the tures as percentages of expenditures by families in the base
socioeconomic and demographic variables will be limited category. The interpretation ofC„ in the family composition
to the predictive ability of the socioeconomic model. See model is somewhat different. In that model, C^ represents
Wagner (1982) for more discussion of the socioeconomic the percentage change in clothing expenditures that occurs
and demographic variables. with the addition to a family of an individual in a given age
and sex category.
The sets of family life cycle and family composition vari-
The Dependent Variable ables were tested in models controlling for socioeconomic
and demographic variables. Each of these three models in-
This study focused on two parts of the process of clothing cluded a set of traditional family life cycle, revised family
consumption: expenditures for the acquisition of clothing life cycle, or family composition variables, plus the set of
and expenditures for the maintenance of clothing by the socioeconomic and demographic variables. These models
family. The three sources of acquisition studied were the were necessary to determine the effect of controlling for
purchase of new and ready-to-wear clothing, the purchase socioeconomic and demographic variables on the predictive
of used clothing, and the purchase of clothing materials for ability of the family life cycle and family composition vari-
home production. Of these expenditure categories, the av- ables.
erage family acquires most of its clothing new and ready- The seventh model was called the socioeconomic model.
to-wear (Winakor 1969). Maintenance included expendi- This model included all of the variables in the set of socio-
tures for drycleaning, shoe repair, alterations, and coin- economic and demographic variables. This model was nec-
operated laundering. Expenditures for laundry equipment essary'—in conjunction with the models which controlled
were omitted because of the difficulty of placing a dollar for socioeconomic and demographic variables—to deter-
value on the flow of services from such durable goods as mine the size of the effect of the sets of family life cycle
washing machines and clothes dryers. Gifts of clothing to and family composition variables.
family members were included, while expenditures for Both the socioeconomic model and the models which
clothing gifts to individuals outside the consuming unit controlled for socioeconomic and demographic variables
were excluded. The dependent variable total family clothing were of the following functional form:
expenditures was then defined as expenditures by the family
for all clothing, sewing materials, and clothing-related ser- lny = lnPo + Pi'n^i + P:^2 + C,D, + . . . + C„D„
vices purchased in the market and intended for consumption
where p, represents the income elasticity of total clothing
by members of the consuming unit.
expenditures and X, represents the variable total consump-
tion expenditures. P^ is the coefficient of X^, which rep-
resents the difference between the education of the house-
The Models hold head and the education of the spouse, and D^
Seven models were tested. Each of the three basic models represents the dummy variables for the independent vari-
included one set of inde()endent variables: a set of tradi- ables. The interpretation of C„ is the same as in the basic
tional family life cycle variables, a set of revised family models.
life cycle variables, or a set of family composition vari- Multiple regression, a technique used frequently in the
ables. Each of these models was of the following semi- analysis of clothing and other household expenditure data,
logarithmic functional form: was used to analyze the relationships between the indepen-
dent and dependent variables. Regression analysis is based
lny = Po + C,D, + . . . + C A on the assumption that the relationships among variables
where lny is equal to the natural log of total clothing ex- are both linear and additive. On the basis of both theory
penditures and £)„ is a variable representing one of the fam- (Brown and Deaton 1972) and past research (Dardis et al.
ily life cycle stages or family composition categories. If 1981; Houthakker and Taylor 1970), it was determined that
clothing expenditures were equal to zero, then the log of the double-logarithmic functional form is the best choice
one was used, because the log of zero does not exist. The for describing the relationship between total consumption
appropriate interpretation of the coefficient C„ of the expenditures and total family clothing expenditures. Pre-
dummy variable £)„ (assuming there is only one dummy vious research has demonstrated that an increase in total
variable in the equation) is shown by the transformation consumption expenditures leads to an increase in total
(Halvorsen and Palmquist 1980): clothing expenditures which is proportionally larger than
the increase in total consumption expenditures (Dardis et
C„ = In (1 + g,) al. 1981). When this is the case, the double-logarithmic
288 THE JOURNAL OF CONSUMER RESEARCH

functional form is the appropriate choice for the regression variables were much more successful than the basic models
equation (Prais and Houthakker 1955). in predicting clothing expenditures. This finding confirms
For each model; the proportion of the variation in total the importance of the set of socioeconomic and demo-
clothing expenditures attributable to variation in the inde- graphic variables in determining family expenditures for
pendent variables was evaluated using 7?^, the adjusted clothing. As in the basic models, there were no differences
square of the multiple correlation coefficient. For the in explanatory power among the family life cycle and fam-
models in which socioeconomic and demographic variables ily composition models that controlled for socioeconomic
were controlled, the size of the effects of the sets of family and deinographic variables.
life cycle and family composition variables was measured The R^ value of the socioeconomic model was 0.53,
by a series of square part correlations, calculated using the demonstrating that the set of socioeconomic and demo-
following formula: graphic variables was much more effective than the basic
family life cycle and family composition models in explain-
Rj =
ing the variance in family clothing expenditures. The results
where: of the regression analysis of that model are summarized in
Table 4. All of the sets of dummy variables were signifi-
Rj is the proportion of the variance in clothing expenditures cantly related to total clothing expenditures, as was total
explained by the set of family life cycle or family com- consumption expenditures.^
position variables
The R^ value for total consumption expenditures alone
Rl is the proportion of the total variance explained by the was 0.50, indicating that this variable was highly effective
family life cycle or family composition model in which in explaining the variance in family clothing expenditures.
socioeconomic and demographic variables are controlled It was more effective than either the basic family life cycle
R] is the proportion of the total variance explained by the or the basic family composition model, but it was slightly
socioeconomic model (Cohen 1977). less effective than either the socioeconomic model or the
models that were controlled for socioeconomic and demo-
graphic variables. This finding suggests that independent
RESULTS of income (as represented by the proxy variable total con-
sumption expenditures), variables such as traditional family
Comparison of Family Life Cycle life cycle, revised family life cycle, and family composition
and Family Composition have very little predictive ability.
No major differences were found in the predictive ability
of the three basic models. The^" values presented in Tables The Size of the Effect of Family Life Cycle
1.3, and 5 indicate that when socioeconomic and demo- and Family Composition
graphic variables were not controlled, the traditional family
life cycle, revised family life cycle, and family composition Both the square part correlations and the F-values for the
models were all moderately successful in explaining the sets of family life cycle and family composition variables
variance^ in family clothing expenditures. are presented in Table 6. When socioeconomic and de-
The R' values of the traditional and revised family life mographic variables were controlled, the sets of traditional
cycle models were nearly identical. Apparently there is no family life cycle, revised family life cycle, and family com-
advantage in increasing the number of family life cycle position variables were all significant, but explained very
stages in order to accommodate nontraditional family iittle of the variance in clothing expenditures. This implies
forms, as proposed by Murphy and Staples. that the effects of family life cycle and family composition
The basic family composition model was only slightly variables in the basic models represent the confounding
more successful than the two family life cycle models in influence of the socioeconomic and demographic variables
explaining the variance in family clothing expenditures. with which they are correlated.
The failure of the family composition model to account for The confounding effects of the socioeconomic and de-
much more of the variance in total clothing expenditures mographic variables and the family life cycle or family
than either of the family life cycle models demonstrates that composition variables are further illustrated by the com-
there is no empirical advantage in accounting for every parisons of the percentage effects of individual variables in
family member by age and sex. The family composition Tables 1, 3, and 5. Discussion of the effects of the indi-
model was expected to be superior to both family life cycle vidual variables will be limited to comparisons between the
models because it accounts for all family members, while basic models and the models that were controlled for so-
the family life cycle models directly account for three fam- cioeconomic and demographic variables. This is done for
ily members at the most. two reasons. First, the emphasis of this article is upon the
size of the effects of family life cycle and family compo-
sition variables independent of socioeconomic and demo-
The Effect of Controlling for Socioeconomic
and Demographic Variables •While some mullicollincarity is inherent in all noncxperimenlal re.
search, examination of both the standard errors of the regression coeffi-
The R^ values reported in Tables 1, 3, and 5 show that cients and the simple correlations among the independent variables showed
the models that included socioeconomic and demographic that this was not a serious problem in this study.
EFFECTIVENESS OF FAMILY UFE CYCLE 289
TABLE 5
REGRESSION RESULTS FOR TOTAL FAMILY CLOTHING EXPENDITURES: THE REVISED FAMILY LIFE CYCLE VARIABLES*

Clothing expenditures as a percentage


of expenditures by young single adults

Mean Controlling for


expenditures socioeconomic and
Variable in dollars Basic model demographic variables

Young single adult (base variable) 433 100 100


Young single parent I 487 132" 124"
Young single parent II 760 189" 160"
Young married adult 626 *165" BT
Young married parent I 631 182" 100
Young married parent II 684 200" 94
Young married parent III 756 224" 97
Young married parent IV 880 250" 103
Mature single adult 363 69" 81"
Mature single parent I 856 160" 109
Mature single parent II 714 187" 129"
Mature married adult 634 152" 81"
Mature married parent I 770 219" 92
Mature married parent II 838 249" 102
Mature married parent III 1,028 312" 111=
Older single adult 217 37" 80"
Older married adult 294 66" 78"
Other 499 126 124'
R' (0.24) (0.54)
F value 190.62 282.35
Degrees of freedom 17/10016 42/9991

• N = 10.034.
"Signjlicantly difierenl Irom 1 00 at the 0.01 level.
'Signilicantly diflerent Irom 1.00 at the 0.05 level.

graphic variables, especially income. Second, the results ently the confounding effect of family life cycle variables
of this research demonstrate the importance of controlling with socioeconomic and demographic variables is greatest
for socioecotiomic and demographic variables in family life during those stages.
cycle and family composition models. The percentage effects for the family composition vari-
Table 1 presents the percentage effects of the traditional ables are presented in Table 3. Mean clothing expenditures
family life cycle variables from both the basic models and are not reported for these variables because of the large
the models that included socioeconomic and demographic number of possible age and sex configurations among the
variables. These figures can be interpreted as representing families in the sample. By controlling for socioeconomic
percentages of the base variable, expenditures by bachelors. and demographic variables, the number of family compo-
For example, in the equation that included socioeconomic sition categories in which expenditures were significantly
and demographic variables, expenditures for total clothing different from zero was reduced. The reduction in the num-
by newly married couples were 88 percent of expenditures ber of significant differences occurred among children un-
by bachelors. The percentage effects of the revised family der 12 years of age. When socioeconomic and demographic
life cycle variables are presented in Table 5. For those variables were included in the regressions, the addition to
variables, the figures represent expenditures as percentages a family of a child under 12 years of age had no significant
of the level of expenditures by young single adults. effect on family clothing expenditures.
For both the traditional and revised family life cycles,
the number of significant differences between the base vari- IMPLICATIONS
ables and the other dummy variables was reduced by con-
trolling for socioeconomic and demographic variables. Re- In models of consumer behavior, family life cycle is
ductions in the number of significant differences occurred considered part of the social and cultural environment
in the early stages in which children were present. Appar- which determines family purchase behavior. The results of
290 THE JOURNAL OF CONSUMER RESEARCH

TABLE 6 In terms of methodological simplicity, the use of family


SQUARED PART CORRELATIONS AND F VALUES
life cycle variables is awkward. Easy-to-use models with
FOR THE SETS OF FAMILY UFE CYCLE AND a small number of stages exclude large numbers of families.
FAMILY COMPOSITION VARIABLES' When the number of stages is increased, fewer families are
excluded from the model, but some cells are underpopu-
Model lated, as was the case in this study. A set of socioeconomic
and demographic variables is easier to use and explains
Traditional Revised
(amily (amily Family
more of the variance in expenditures. Compared to the
li(e cycle li(e cycle composition models which included both family life cycle and socio-
economic and demographic variables, the socioeconomic
Squared part 0.01 0.01 0.01 model was neariy as effective in predicting family clothing
correlation expenditures. Occam's razor suggests that, given two
models of comparable predictive ability, the simpler of the
F value" 27.44 15.86 22.17
two models should be used. While family life cycle vari-
Degrees o( (reedom 9/9999 17/9991 13/9995 ables offer a convenient conceptual framework, socioeco-
• N = 10.034. nomic and demographic variables may be better choices for
"All values were signilic.irt at the 0.001 level. expenditure research.

this study challenge the acceptance and use of family life


SUMMARY AND CONCLUSIONS
cycle variables in consumer theory and research. This study tested the effectiveness of family life cycle
Family life cycle is purported to be a composite frame- variables in predicting expenditures for goods and services.
work that combines trends in income with demands placed While family life cycle variables have been used frequently
on income. When socioeconomic and demographic vari- in consumer expenditure research, their use has been crit-
ables—esjjecially income—are not included in family cycle icized. As a result of this research, conclusions have been
models, family life cycle is moderately successful in rep- reached conceming the validity of three of those criticisms.
resenting the effect of those variables on expenditures. First, expenditure research has been criticized for failing
Models which include socioeconomic and demographic to test altemative definitions of the family life cycle concept
variables are much more successful. In fact, when socio- over a comprehensive range of goods and ser\'ices (Ferber
economic and demographic variables are controlled, the 1979). During the last five years, consumer researchers
size of the effect of family life cycle is very small. This have proposed revised definitions that include nontradi-
study confirms the importance of income, in particular, in tional family forms. However, interest in those definitions
determining expenditures for goods and ser\'ices. has been primarily at the conceptual level. Including non-
Speculation about altemative definitions of the family life traditional family forms in the family life cycle involves
cycle, while interesting conceptually, must be either sup- increasing the number of stages. In this study, a revised
ported or refuted by research. Empirical justification for the family life cycle was no more successful than the traditional
use of a given definition is essential to the building of the- family life cycle in explaining the variance in expenditures
ory. This research was limited to a comparison of two def- for clothing. The traditional family life cycle may be an
initions of family life cycle in predicting expenditures for outdated concept because it does not accommodate a variety
one consumption category. Further research might be di- of family types, but this study concludes that increasing the
rected toward testing these models with other consumption number of family life cycle stages to include nontraditional
categories, or comparing these models to those proposed families does not improve the predictive ability of family
by Stampfl (1978) or Gilly and Enis (1981). Researchers life cycle models, at least in the case of clothing expendi-
might even experiment with reducing the number of stages tures.
in the traditional family life cycle. For example, the first Family life cycle variables have also been criticized by
two stages of the full nest variable might be collapsed. Murphy and Staples (1979) and Trost (1974) for obscuring
However, the results of this study imply that the manipu- the effects of family composition on expenditures. Our re-
lation of the variables underlying family life cycle generates search demonstrates that this criticism is unwarranted. The
no increase in predictive ability, at least in the case of family composition model offered only slight improvement
clothing expenditures. in predictive ability over the family life cycle models. Once
While the size of the effect of family life cycle variables socioeconomic and demographic variables were controlled,
is small, it is still significant. In terms of predictive ability, there was no difference between the family life cycle and
the results of this study support the inclusion of either the family composition models. This study concludes that fam-
traditional or the revised family life cycle in models of ily life cycle variables successfully represent the effects of
consumer expenditure behavior. However, it should be kept family composition on expenditures.
in mind that the effects of socioeconomic and demographic Finally, family life cycle research has been criticized for
variables, particularly income, are much larger than the failing to control for socioeconomic and demographic vari-
effect of family life cycle on expenditures. ables, especially income (Ferber 1979). The results of our
EFFECTIVENESS OF FAMILY LIFE CYCLE 291

study demonstrate that this criticism is valid. In models that penditures of Low Income Families," Journal of Consumer
controlled for socioeconomic and demographic variables, Affairs, II (Winter), 127-132.
the sets of family life cycle variables were significant, but Halvorsen, Robert and Raymond Palmquist (1980). "The Inter-
pretation of Dummy Variables in Semilogarithmic Equa-
their effects were small. We conclude that clothing exf)en-
tions," American Economic Review, 70 (June), 474-475.
ditures vary over the family life cycle, but that the impor-
Hisrich, Robert D. and Michael P. Peters (1974), "Selecting the
tance of stage in the family life cycle in determining ex- Sup>erior Segmentation Correlate," Journal of Marketing, 38
penditures has been exaggerated. As compared to income (July), 6 0 - 6 3 .
or to income in conjunction with other socioeconomic and Houthakker, H. S. and Lester D. Taylor (1970), Consumer De-
demographic variables, sets of family life cycle variables mand in the United States: Analysis and Projections, Cam-
are of limited value in predicting clothing exjjenditures. bridge, MA: Harvard University Press.
Kennedy, Peter E. (1981), "Estimation With Correctly Interpreted
[Received September 1982. Revised July 1983.] Dummy Variables in Semilogarithmic Equations," The
American Economic Review, 71 (September), 801.
Landon, E. Laird, Jr. and William B. Lcicander (1979), "Family
Life Cycle and Leisure Behavior Research," in Advances in
REFERENCES Consumer Research, Vol. 6, ed. William Wilkie, Ann Arbor,
Amdt, Johan (1979), "Family Life Cycle as a Detemiinant of MI: Association for Consumer Research, 133-138.
Size and Composition of Household Expenditures," in Ad- Lansing, John B. and James M. Morgan (1955), "Consumer Fi-
vances in Consumer Research, Vol. 6, ed. William L. nances Over the Life Cycle," in Consumer Behavior, Vol.
Wilkie, Ann Arbor, MI: Association for Consumer Research, 2, ed. Lincoln H. Clark, New York: New York University
128-132. Press, 36-50.
Britton, Virginia (1969), "Gifts and Handed-Down Clothing Im- Murphy, Patrick E. and William A. Staples (1979), "A Modem-
portant in Family Wardrobes," Family Economics Review, ized Family Life Cycle," Journal of Consumer Research, 6
(September), 3 - 5 . (June), 12-22.
Brown, Alan and Angus Deaton (1972), "Surveys in Applied Prais. S. J. and H. S. Houthakker (1955), The Analysis of Family
Economics: Models of Consumer Behavior." Economic Budgets, Cambridge, MA: The University Press.
Journal, 82 (December), 1145-1236. Settle. Robert B., Pamela L. Alreck, and Michael A. Belch
Carlson, Michael D. (1974), "The 1972-73 Consumer Expendi- (1979), "Social Class Determinants of Leisure Activity," in
ture Survey," Monthly Labor Review, 97 (December), Advances in Consumer Research, Vol. 6, ed. William L.
16-23. Wilkie, Ann Arbor, Ml: Association for Consumer Research.
Cohen, Jacob (1977), Statistical Power Analysis for the Behav- 139-145.
ioral Sciences, New York: Academic Press. Stampfl, Ronald W. (1978). "The Consumer Life Cycle." The
Dardis, Rachel, Frederick Derrick, and Alane K. Lehfeld (1981). Journal of Consumer Affairs, 12 (Winter), 209-219.
"Clothing Demand in the United States: A Cross-Sectional Trost. Jan (1974), "This Family Life Cycle—An Impossible Con-
Analysis," Home Ecotiomics Research Journal, 10 (Decem- cept." International Journat of the Sociology of the Famih.
ber), 212-222. 4. 37-54.
Derrick, Frederick W. and Alane K. Lehfeld (1980). "The Family U.S. Bureau of the Census (1982), Current Population Reports.
Life Cycle: An Alternative Approach," The Journal of Con- Series P-20, No. 75, Fertility of American Women: June
sumer Research. 7 (September), 214—217. 1980, Washington, D . C : U.S. Govemment Printing Office.
Ferber, Robert (1979), "Comments on Papers on Life Cycle Anal- U.S. Department of Health and Human Services (1982). Vital
ysis," in Advances in Consumer Research, Vol. 6. ed. Wil- Statistics of the U.S., 1978, 3 , Hyattsville. MD: National
liam L. Wilkie, Ann Arbor, Ml: Association for Consumer Center for Health Statistics.
Research, 146-148. U.S. Department of Labor (1978). Consumer Expenditure Sunry:
Frey, Cynthia J. and Duncan G. LaBay (1983), "A Comparative integrated Diary and Interview Survey Data 1972-73, Bul-
Study of Energy Consumption and Conservation Across the letin 1992, Washington, D . C : U.S. Govemment Printing
Family Life Cycle," in Advances in Consumer Research, Offiee.
Vol. 10, eds. Richard P. Bagozzi and Alice M. Tybout, Ann Wagner, Janet S. (1982), "Family Clothing Consumption: A
Arbor, MI: Association for Consumer Research, 641-646. Cross-Sectional Comparison of Family Life Cycle and Fam-
Fritzsche, DavidJ. (1981), "An Analysis of Energy Consumption ily Composition Models," unpublished Ph.D. dissertation.
Pattems by Stage of Family Life Cycle," Journal of Mar- Department of Family Economics, Kansas State University.
keting Research, 28 (May), 227-232. Manhattan, KS 66502.
Gilly, Mary C. and Ben M. Enis (1982), "Recycling the Family Wells, William D. and George Gubar (1966). "Life Cycle Con-
Life Cycle: A Proposal for Redefinition," in Advances in cept in Marketing Research," Journal of Marketing Re-
Consumer Research, Vol. 9, ed. Andrew Mitchell, Ann Ar- search, 3 (November), 355-363.
bor, MI: Association for Consumer Research, 271-276. Winakor, Geitel (1969), "The Process of Clothing Consump-
Hafsu-om, Jeanne L. and Marilyn M. Dunsing (1972). "Satisfac- tion," Journal of Home Economics, 61 (October), 629-634.
tion and Education: A New Approach to Understanding Con- Zimmerman, Carol A. (1981), "Household Travel Pattems by
sumption Pattems," Home Economics Research Journal, 1 Life Cycle Stage," in Consumers and Energy Conservation:
(September), 4 - 1 2 . International Perspectives on Research and Policy Options,
Hager, Christine J. and Keith W. Bryant (1977), "Clothing Ex- eds. John D. Claxton et al. New York: Praeger, 81-95.
View publication stats

You might also like