Академический Документы
Профессиональный Документы
Культура Документы
durables, rates of depreciation of existing find that durable purchase intentions are lin-
units, and current and future expectations of early related to purchase behavior.
product characteristics. Dynamic and feed- We can specify the functional relationship
back effects are also considered by allowing for replacement intentions as
unanticipated events (e.g., product failure,
R(X) =g(X) +e, (1)
special price/ promotion, financial windfall,
unavailability of desired product) to affect where g(X) is the deterministic component
the demand for a particular durable. of replacement intentions and is dependent
As conceptualized in Pickering (1981), re- on the set of explanatory variables X. Here,
placement demand involves an implicit com- E is the error term (stochastic component) of
parison of the utility likely to be derived intentions. The set of variables in X include
from the purchase of a new item as com- characteristics of the product currently
pared with the utility obtained from the ex- owned and household characteristics. Fol-
isting unit. He suggests that this comparison lowing the conceptual model proposed by
will depend on the age of the existing unit, Pickering (1981), possible effects due to the
its reliability, an assessment of whether it is stock of durables owned on replacement in-
likely to break down or require replacement tentions for a particular product are not con-
in the near future, and to a lesser extent, the sidered (i.e., we assume the indirect utilities
perceived attractiveness of new, more up to of product replacement are separable). Re-
date units available in the marketplace. laxing this assumption is left for future re-
However, no detailed operationalization of search.
variables which influence replacement pur- Assuming that the error term and is iid
chases is given, nor are we aware of any according to a Type I extreme value distribu-
study which has empirically estimated the tion, (1) can be transformed into the familiar
influence of these variables on replacement logistic function. Letting P denote the prob-
demand. ability of replacing a durable, the model is
P=l/.[l+exp(-CY-X/3)], (2)
where X is the vector of explanatory vari-
ables, j3 is the coefficient vector, and a is an
3. A model of durable replacement inten- intercept term. We note that the marginal
tions impact on the replacement probability due to
changes in an explanatory variable (say xi) is
pjP(l - PI. Because P is a function of sev-
In this section, a descriptive model of re-
eral explanatory variables, the marginal im-
placement intentions for a consumer durable
pact of a single variable thus depends on the
is developed. Variables associated with prod-
values of the other variables. In other words,
uct and household characteristics are consid-
the interactive effects of several explanatory
ered as predictors. Intentions are expected
variables are implicitly included in (2). Since
to be an important indicator of replacements
the dependent variable we will analyze is
since durable purchases are considered a
binary (likely or not likely to replace),
planned purchase (Pickering, 1984). Empiri-
cal studies by Morrison (1979), Kalwani and As part of the empirical study described later in this paper,
Silk (19821, and Jamieson and Bass (1989) replacement purchase data for refrigerators and coffee
also demonstrate a strong relationship be- makers were also collected along with intentions data. Re-
sults not reported in this paper indicate that the same set of
tween stated intentions and actual purchases explanatory variables for replacement intentions and pur-
of durables. Kalwani and Silk (1982) further chase were significant for both these products.
260 B.L. Bayus, S. Gupta / Consumer durable replacement intentions
Table 2
Sample profile
cleaner, and coffee maker. This set of appli- owning these eight home appliances was gen-
ances was selected to represent a range of erally over 70%. Demographic profiles of the
product categories. Prior research (Pickering entire panel and the samples of respondents
et al., 1973; Bayus and Carlstrom, 1990) indi- and nonrespondents are given in Table 2.
cates that these appliances can be separated Generally speaking, the sample of respon-
into three groups based on perceptual mea- dents is a little older and has a smaller
sures: major home appliances (stove, re- percentage of households with a working
frigerator, washer, vacuum cleaner), house- spouse than the nonrespondents. No statisti-
wares (coffee maker), and entertainment cally significant differences exist in terms of
items (color TV, VCR). household income, education of the house-
Information on the age (in years) and con- hold head, and the percentage of married
dition (three-point scale: good, fair, poor) of households.
currently owned units was also collected. De- Most of the variables in Table 1 have
mographic data such as age of household natural definitions based on the survey ques-
head (in years), income (eleven-point or- tions. Replacement intentions were di-
dered scale), and whether a spouse was em- chotomized into positive (1 = definitely or
ployed (if married) were also available. Fi- likely) and negative (0 = not likely or defi-
nally, information on the length of residence nitely not) intentions by combining response
at the current address and expected future categories. 3 Since unit condition was re-
household financial situation was collected ported using a three-point scale, two dummy
(three-point scale: better, same, worse than variables are used to represent perceived unit
now). * condition (GOOD = 1 if unit condition is
Completed questionnaires were received good, 0 otherwise; POOR = 1 if unit condi-
from 407 households owning their home or tion is poor, 0 otherwise). Similarly, expected
condominium, representing a response rate future household financial situation is repre-
of over 70%. The percentage of households sented by two dummy variables (BETTER =
2
We note that Van Raaij and Gianotten (1990) use a five- Since replacement intentions were collected using a four-
point scale (much better to much worse than now) to point scale, estimation procedures such as multinomial logit
measure future household financial situation. In their analy- or probit, or ordered logit could also be used. However,
ses this scale is assumed to have equal interval properties. other analyses (using an ordered logit model) not reported
Since the true nature of such a scale has not been estab- in this paper indicate that the main conclusions remain
lished, we use a three-point ordered scale, but only assume unchanged. Thus, for ease of interpretation we report re-
ordinal properties. sults based on a binary dependent variable.
B.L. Bayus, S. Gupta / Consumer durable replacement intentions 263
1 if expected financial situation is better than tion for each of the eight products consid-
now, 0 otherwise; WORSE = 1 if expected ered individually; and (2) the product charac-
financial situation is worse than now, 0 oth- teristics summarized in Table 3 suggest that
erwise). Relating these dummy variables to the expected relationship between unit age,
Table 1, POOR and BETTER are hypothe- unit condition, and intentions are similar for
sized to have positive coefficient signs, and each separate product. Although the model
GOOD and WORSE are hypothesized to be developed in Section 3 does not consider
negatively related to replacement intentions differing effects by product category, we al-
and purchase. MOVED is defined as 1 if the low for intrinsic product specific effects in
household has lived at the present address the logistic regression model (i.e., product
for less than one year, 0 otherwise. The specific constants; see Chintagunta, 1992) by
HAZARD rate for product i is calculated as defining six appropriate dummy variables:
28,tij, where tij is the age of product i for STOKE (1 if stove, 0 otherwise), FRlDGE (1
household j and the parameter Bi is found if refrigerator, 0 otherwise), WASHER (1 if
using the overall mean replacement age of clothes washer, 0 otherwise), CW (1 if color
product i (CL= i(rr/0)/). Estimates of TV, 0 otherwise), VCR (1 if VCR, 0 other-
mean product replacement ages are available wise), and VACUUM (1 if vacuum cleaner, 0
in trade publications and are given in Table otherwise). Here, coffee maker is assumed to
3. be the base product (i.e., if all product
dummy variables equal zero). Everything else
4.2 Analysis method being equal, if a particular appliance has
some inherent replacement importance or
In order to estimate the replacement in- priority in the household, then we expect
tention model, we pooled the data for the the dummy variable for that product to be
407 respondent households across the seven statistically significant.
products. Aftering deleting observations with
missing values, the total sample size available
for analysis is 2132. Supporting this decision 5. Resdts
are two factors: (1) due to the nature of 5.1 Model fit
consumer durables and the replacement de-
cision, a relatively small number of house- The overall fit of the logistic regression
holds indicated a positive replacement inten- model is very good. Classification results for
Table 3
Product characteristics for home appliances studied
The estimated coefficients for the logistic dummy variables are significant, indicating
model are reported in Table 5. The model no intrinsic product specific effects.
chi-square is highly significant. The coeffi- Consistent with Kim (19891, we find that
cients of the four significant variables are in wifes employment status is an important
the hypothesized direction. The product variable related to replacement intentions.
characteristics of perceived unit condition Supporting the results of Van Raaij and Gi-
and hazard rate are significant, and show anotten (19901, we find that expected house-
negative and positive effects, respectively. hold financial situation is significantly associ-
Whether a spouse is working and expected ated with replacement intentions. Contrary
future household financial situation are posi- to the results of Strober and Weinberg (19771,
tive and significant. None of the product Weinberg and Winer (19831, and Winer
(1985b) for durable ownership and probabil-
ity of acquisition, we find no significant ef-
Table 4
fects for age of household head on replace-
Overall model fit
ment intentions. Unlike other studies which
Actual Predicted replace- Total
find that income is a significant variable for
replacement ment intentions
intentions durable ownership (Nickels and Fox, 19831,
Positive Negative
durable expenditures (Strober and Wein-
Positive 111 169 280
Negative 44 1808 1852
berg, 1977; Weinberg and Winer, 1983; Van
Raaij and Gianotten, 19901, and purchase
Total 155 1977 2132
probability (Winer, 1985b), our results do not
B.L. Bayus, S. Gupta / Consumer durable replacement intentions 265
show a significant income effect for replace- derlying measurement scales (e.g., it is not
ment intentions. However, this result may be clear that a change in a spouses working
due to the sample analyzed (i.e., respondents status is the same as a change in product
that owned their home) and the inclusion of condition), the values in Table 6 do show the
the spouses working status. Finally, contrary considerable importance of perceived prod-
to our hypothesis, whether a household has uct condition and the hazard rate (which is a
recently moved is not significantly related to function of the currently owned units age)
replacement intentions. We note that this on subsequent replacement intentions.
finding may be due to the fact that very few
households in the sample analyzed reported
a recent move. 6. Discussion and conclusions
about the tradeoffs a household might make In particular, current unit condition (infor-
between the set of products owned. Individ- mation which is not generally collected) is a
ual household data, perhaps collected in an significant explanatory variable of replace-
experimental fashion, is needed to study this ment intentions. Since replacements account
issue in greater detail. Studying replace- for the majority of observed sales of mature
ments as a choice among heterogeneous al- durable categories, including a measure of
ternatives is also a potential area for future unit condition in a model of sales may pro-
research (e.g., see Bayus and Rao, 1989). vide improved predictive power. Future re-
Although our survey questions were de- search might address this question.
signed to be valid measures of the underlying Other analyses not reported in this paper
constructs in Table 1, due to the nature of give an initial start in this direction. As dis-
the study (i.e., cross sectional and self-re- cussed earlier, unit condition is a function of
ported) it is possible that the causal direc- physical wear and tear, and maintenance and
tion of certain relationships is confounded. repair efforts. For six appliances (refrigera-
This is of particular concern for the measure tor, stove, washer, VCR, color TV, vacuum),
of unit condition, i.e., replacement intentions unit age (a proxy for physical deterioration)
or a recent purchase might have influenced is positively and significantly related to re-
the reported unit condition, meaning that ported unit condition, and for four of these
the observed significance of unit condition is (refrigerator, stove, washer, vacuum> house-
overstated. We examined this possibility and hold size (a proxy for usage) is also positive
concluded that such an effect, if operative, and significant. Only in a few cases did a
could not entirely account for our findings household in our sample report having made
concerning unit condition. First, over 85% of any expenditures on repairs/ maintenance
respondents that reported owning a refriger- for their refrigerator or coffee maker. Future
ator or a coffee maker in poor condition research needs to refine and extend the
specified the reason for making a replace- modelling of perceived unit condition across
ment purchase as old unit broken, ex- durable categories.
pected old unit to breakdown, or costly Finally, our results concerning the impor-
repairs needed. This indicates that the mea- tance of unit age suggest a need to model
sure has face validity. Second, across all the and track the age distribution of units in use.
appliances considered, approximately 70% of In addition, modelling and estimating the
respondents stating they would definitely or effects of variables such as price, advertising,
were likely to make a replacement purchase and product enhancements on this underly-
also reported owning an appliance in good or ing age distribution is a promising area for
fair condition. This indicates variability in further research.
the measures of unit condition and inten-
tions.
References
6.2 Implications Antonides. G., 1990. The lifetime of a durable good, Boston,
MA: Kluwer Academic Publishers.
Bayus. B.. 1988. Accelerating the durable replacement cycle
Researchers have suggested that the rea- with marketing mix variables. Journal of Product Innova-
son for the poor performance of purchase tion Management 5, 216-226.
intentions in forecasting subsequent sales is Bayus, B., 1991a. The consumer durable replacement buyer.
Journal of Marketing 5 (January), 42-51.
due to measurement and scaling issues (e.g.,
Bayus, B., 1991b. Applying targeted marketing principles to
Juster, 1974; Pickering, 1984). Our results consumer durables. MS1 working paper no. 91-108, Cam-
suggest that what is measured is important. bridge, MA: Marketing Science Institute.
B.L. Bayus, S. Gupta / Consumer durable replucement intentions 267
Bayus, B. and C. Carlstrom, 1990. Grouping durable goods. chases. Journal of Economic Psychology 10 (November),
Applied Economics, 22 (June), 759-773. 391-409.
Bayus, B. and V. Rao, 1989. A hierarchical utility model for McNeil. J., 1974. Federal programs to measure consumer
the dynamic acquisition of heterogeneous items. Market- purchase expectations, 1946-1973: A post mortem. Jour-
ing Letters 1. 71-80. nal of Consumer Research 1 (December), l-10.
Bayus, B., S. Hong and R. Labe, 1989. Developing and using Merchandising, 1986. 13th Annual consumer survey (Septem-
forecasting models of consumer durables: The case of ber), 2-29.
color television. Journal of Product Innovation Manage- Morrison, D., 1979. Purchase intentions and purchase behav-
ment 6, 5-19. ior. Journal of Marketing 43 (Spring), 65-74.
Beatty, S. and S. Smith, 1987. External search effort: An Nickels, S. and K. Fox, 1983 Buying time and saving time:
investigation across several product categories. Journal of Strategies for managing household production. Journal of
Consumer Research 14 (June), 83-95. Consumer Research 10, 197-208.
Chintagunta, P.K., 1992, Heterogeneity in nested logit mod- Pickering, J., 1981. A behavioral model of the demand for
els: An estimation approach and empirical results, Inter- consumer durables. Journal of Economic Psychology 1.
national Journal of Research in Marketing 9, 2, 161-175. 59-77.
Cox. A., D. Granbois and J. Summers 1983. Planning, search, Pickering. J., 1984. Purchase expectations and the demand for
certainty and satisfaction among durables buyers: A longi- consumer durables. Journal of Economic Psychology 5,
tudinal study. In: R. Bagozzi and A. Tybout teds.), Ad- 341-352.
vances in Consumer Research 10, 394-399. Pickering, J., J. Harrison, B. Isherwood and C. Cohen, 1973.
DeBelI, M. and R. Dardis, 1979. Extending product life: Are goods? Some empirical evidence. Applied Economics
Technology isnt the only issue. In: W. Wilkie fed.), Ad- 5, l-18.
vances in Consumer Research 6, 381-385. Schmittlein, D. and K. Helsen, 1990. Analyzing duration times
Dickson, P. and W. Wilkie 1978. The consumption of house- in marketing research. Wharton School working paper.
hold durables: A behavioral review, MS1 monograph no. Philadelphia, PA: University of Pennsylvania.
78-117, Cambridge, MA: Marketing Science Institute. Strober, M. and C. Weinberg, 1977. Working wives and major
Dickson, P., R. Lusch and W. Wilkie 1983. Consumer acquisi- family expenditures. Journal of Consumer Research 4
tion priorities for home appliances: A replication and (September), 141-147.
extension. Journal of Consumer Research (March), 432- Tse, D. and P. Wilton, 1988. Models of consumer satisfaction
435. formation: An extension. Journal of Marketing Research
Harrell, F., 1986. The LOGIST procedure. In: SAS Users 25 (May), 204-212.
Guide, Cary, NC: SAS Institute. Urban, G. and J. Hauser, 1987. The value priority hypothesis
Hosmer, D. and S. Lemeshow 1989. Applied logistic regres- for consumer budget plans. Journal of Consumer Re-
sion. New York: Wiley. search 12 (March), 446-462.
Jacoby, J., C. Berning and T. Dietvorst 1977. What about Van Raaij, W. and H. Gianotten, 1990. Consumer confidence,
disposition? Journal of Marketing 41 (April), 22-28. expenditure, saving, and credit. Journal of Economic Psy-
Jamieson, L. and F. Bass, 1989. Adjusting stated intention chology 11 (June), 269-290.
measures to predict trial purchase of new products: A Weinberg, C. and R. Winer, 1983. Working wives and major
comparison of models and methods. Journal of Marketing family expenditures: replication and extension. Journal of
Research 26 (August), 336-345. Consumer Research 10 (September), 259-263.
Juster, F., 1974. Commentary on McNeil: Federal program to Wilkie, W. and P. Dickson, 1985. Shopping for appliances:
measure consumer purchase expectations. Journal of Con- Consumers strategies and patterns of information search.
sumer Research 1 (December), 12-15. MSI working paper no. 85-108. Cambridge, MA: Market-
Kalwani, M. and A. Silk, 1982. On the reliability and predic- ing Science Institute.
tive validity of purchase intention measures. Marketing Winer, R., 1985a. A revised behavioral model of consumer
Science 1 (Summer), 243-286. durable demand. Journal of Economic Psychology 6. 175-
Kasulis, J., R. Lusch and E. Stafford, 1979. Consumer acquisi- 184.
tion patterns for durable goods. Journal of Consumer Winer, R., 1985b. A price vector model of demand for con-
Research (June), 47-57. sumer durables: Preliminary developments. Marketing Sci-
Kim. C., 1989. Working wives time-saving tendencies: Durable ence 4 (Winter), 74-90.
ownership, convenience food consumption and meal pur-