Comments on Papers on Life Cycle Analysis

Robert Ferber, University of Illinois
[ to cite ]:
Robert Ferber (1979) ,"Comments on Papers on Life Cycle Analysis", in NA - Advances in Consumer Research Volume 06, eds. William L. Wilkie, Ann Abor, MI : Association for Consumer Research, Pages: 146-148.

Advances in Consumer Research Volume 6, 1979      Pages 146-148

COMMENTS ON PAPERS ON LIFE CYCLE ANALYSIS

Robert Ferber, University of Illinois

As is evident from the papers, all of them deal with the life cycle concept in one way or another. Two of the papers attempt to emphasize the value of a life cycle concept for explaining consumer behavior, while the third includes a life-cycle variable among others in an attempt to assess the relative importance of these different variables in explaining one type of consumer behavior--recreation.

My approach in these comments will be to first make some remarks on each of the papers individually. Based on these evaluations, some comments will then be offered on the value of a life-cycle type of variable, and particularly on what needs to be done if the value of a lifecycle variable for analysis of consumer behavior is to be established. Also, given my awareness of the problems that exist with survey data of the type reported here, I have been unable to resist making a few final comments on how such data might be reported and interpreted when used for analytical purposes.

The Paper by Arndt

This paper tends to reemphasize the importance of the life cycle concept in influencing three types of activities--receipt of income, composition of consumption expenditures and the proportion spent on gifts (which are treated separately from other consumption expenditures). This is done by recourse to data from a national survey of consumer expenditures conducted in Norway in 1973, presenting breakdowns of these expenditures by stages of the family life cycle.

In line with these objectives, the paper presents a cross tabulation of the distribution of the 3,360 families interviewed in this survey by major expenditure categories by six stages in the life cycle, plus a seventh stage designated "other households". As in previous studies in this country and elsewhere, the results show that income (represented by total consumption expenditures) first rises over the life cycle and then falls; that the allocation of the family budget to different expenditures fluctuates in some predictable (but also some unpredictable) fashions; and that gifts as a percent of total expenditures tend to follow a U-shaped pattern.

Accordingly, do these results establish the value of the family life cycle concept in explaining spending patterns? I am afraid not. There are two types of problems with this analysis, one type associated with the analysis of the data and the other associated with the general approach.

The analysis of these data raises a number of questions, as follows:

1. The six stages of the life cycle used in this analysis account for only two-thirds of the families for which data were available. It seems to me that when over one-third of the observations in a national study have to be discarded because they do not fit a particular classification, something must be wrong with the classification. While the author points out one reason for this large loss, namely, that the upper age limit for dependent children was set at 16 by the Norwegian Central Bureau of Statistics, the fact remains that no attempt was made to deal with this problem. As the author points out in his concluding paragraph, living arrangements are changing, particularly in view of the increasing number of divorces and remarriages. Hence, a life-cycle classification to be meaningful under present circumstances has to take these changes into account if it is to be useful for explaining consumer behavior.

2. The tendency to equate income with household expenditures is questionable, unless one wants to consider such expenditures as a measure of permanent income, something that is not discussed. Total expenditures and family income are undoubtedly highly correlated, as the author says, but there is no question that appreciable differences would result in expenditures, since saving is hardly likely to be constant over the life cycle.

3. From a survey point of view, the reported response rate of 71% in this survey raises questions about sampling biases, and also about the reliability of the data for life cycle analysis, neither of which is discussed. Especially noticeable, for example, is that the base for each of the two first stages of the life cycle is only about 100 families, or about 3% of the total sample. Both absolutely and percentage-wise, these are very small bases to use for analysis, especially in a nationwide study.

4. Also not receiving any consideration is the very likely possibility of response errors in the expenditure data. Expenditure studies of the type reported here are notorious for having such errors, with substantial proportions of the total known expenditures for a particular category being unaccounted for. Thus, an evaluation of the 1972-73 survey of consumer expenditures in this country by Robert Pearl of our staff, done for the U.S. Bureau of the Census, indicates that 30-40% of expenditures for meals away from home were not reported in the survey. These errors are by no means equally distributed by different types of families, and their consideration might alter substantially the nature of the inferences drawn from these data.

5. On a more detailed level, I am not convinced that the category "other goods and services" represents only restaurant meals and entertainment; such an "other" category very likely includes also a wide variety of other services (such as financial) and other types of expenditures that could hardly be classified as meals or entertainment. Their relative importance in this category is not mentioned, nor the very real possibility that they will vary in importance by stage of the life cycle.

More generally, even if none of the foregoing problems existed, I do not see how this type of analysis can demonstrate the value of a life-cycle variable. As is well known, income and spending patterns are influenced by many different variables, including education, occupation, age and family composition; plus the fact that a major determinant of expenditure allocation is the level of income. Only if these variables are taken into account in some multivariate framework is it possible to evaluate whether life cycle is a meaningful variable for the present purposes. In particular:

1. To establish the value of life cycle as an indicator of the allocation of spending patterns, one would have to show that a life-cycle variable is relevant after allowance has been made for income, education and occupation (or some combination of these in the form of socio-economic status).

2. To establish the value of a life cycle classification for explaining fluctuations in family income, one would have to show that this variable is relevant after taking into account education and occupation.

In both instances, one would also have to show that a life-cycle variable is more meaningful than, say, age or household composition considered separately.

The Paper by Landon and Locander

This paper is conceptually similar to that of the previous one. It seeks to establish the value of a life cycle variable for explaining variations in leisure behavior and attitudes toward such facilities. The approach is essentially the same as before, except that in this case the survey was carried out in Pima County, Arizona; the focus was on leisure time behavior and attitudes; and there is no information on response rates or on how the data were obtained. There is also no conceptual framework nor is there much review of the past literature on this topic, which I believe is extensive.

In view of the similarity of these two studies, the same sorts of questions can be leveled at this one as at the previous one. In particular:

1. There is no indication of the reliability of the data or of the survey.

2. It is not clear whether the particular life-cycle classification used is the most meaningful for the purposes.

From a more general point of view, the univariate analysis used in this study, with no attention to other relevant variables, provides no basis for inferring that this particular type of life-cycle classification is of any particular value. It could be, for example, that income or age, or a combination of the two, is a much better discriminator of behavior and attitudes toward recreation than a life-cycle variable. Until such alternatives are explored, the authors' contention of the value of a life-cycle variable remains unproven.

The Paper by Settle, Alreck and Belch

Methodologically, this study differs from the preceding two in that it includes five socio-economic and five demographic variables (one of which is life cycle) to investigate their relative effect in explaining participation in 100 different leisure time activities. Based on 1,000 chi-square tests that relate each of these variables to each of the activities, the authors conclude that the demographic variables are better predictors than the socio-economic variables, because education is the best single socio-economic determinant of leisure time behavior.

Unfortunately, as with the previous two papers, the validity of the differences is not supported by the analysis. About all that can be said from these data is that the demographic variables are more frequently associated with leisure time than the socio-economic variables, and that education is the variable that is most frequently significant. Such a result gives no indication of the relative importance of different variables. This is especially so since, here again, as in the preceding papers, no attempt was made to carry out any sort of multivariate analysis. Hence, we have no information about the relative importance of the different variables in explaining leisure time behavior, and there is no basis for doing so on the basis of the frequency of significance of explanatory variables to simple chi-square tests.

From a survey point of view, one could also raise a number of questions on the procedures. Leaving the selection of a quota sample to be carried out by students in the field raises all sorts of possibilities of bias, both from a sampling and a response point of view, and none of these is given any consideration. Moreover, to have students offer to administer a personality test to respondents and provide them with interpretations is highly questionable from an ethical point of view; it is a procedure that should be strongly discouraged. (Incidentally, the assertion that mailing the interpretations of such tests to the respondents' homes provides validation of the data collection procedure has no basis, unless the respondents later were actually contacted.)

Finally, I do not understand what the authors mean by social status. Sometimes they seem to equate it with education; other times they seem to want to leave it to the judgment of the respondents; while still other times they suggest using analytical methods to obtain "candidate variables in determining social status, selecting those that perform the best". The concept itself is a highly ambiguous one and needs a great deal of additional work.

General Comments

As one may suspect from these comments, I am by no means convinced that the case for the value of life-cycle analysis has been established by these studies or, for that matter, by the studies carried out over twenty years ago in the conference volume edited by Lincoln Clark. This is not to say that I do not think that life-cycle analysis may not be a useful variable, but that the case for it has by no means been established one way or another.

To investigate if the life-cycle approach is useful in explaining consumer behavior would seem to involve two major steps, and preferably a third as well.

1. One step is to investigate alternative definitions of the family life cycle in relation to current living conditions. This means not only devising alternative definitions to take into account different forms of living, but also testing these definitions in terms of their ability to discriminate among different forms of consumer behavior. In this sense, it may well happen, for example, that a life cycle classification explaining one type of consumer behavior may not be the best for explaining some other type of consumer behavior.

2. Proceeding from these univariate tests, the best such classifications should be incorporated in a model of consumer behavior that includes other demographic and socio-economic variables. In a sense, this is the acid test, since if such a classification does not hold up when considered with other variables, it is of little analytical value. Such tests have rarely, if ever, been made, and it is only until such tests are carried out that we will have definitive information on the value of a life-cycle classification.

A third sort of test, one which would lend more generality to the preceding two steps, is to ascertain whether such results are valid across different cultures. Such cross-cultural comparisons are not easy, partly because comparable variables are not always available, and partly because errors in the data (to which practically nobody in the consumer behavior field pays any attention) may invalidate much of the results.

The latter point leads me to make the final comment that one thing badly needed in future papers of the type reported here is better reporting of the data collection process, and some evaluation of these data, an evaluation that is integrated with the substantive results. The research on survey methodology has by now documented fully the substantial errors that can result from sloppy sampling procedures, failure to consider bias due to non-response, and failure to consider the effects of response errors, and of response variations. The least that consumer researchers can do is to be aware of these dangers, and to evaluate results they obtain in the light of the errors that are known to exist in the data.

----------------------------------------