&Quot;Spending Power&Quot; and the Subjective Discretionary Income (Sdi) Scale

ABSTRACT - New measures of "spending power" from the economics literature are reviewed as alternatives to the total household income measure typically used in consumer research. The simplest of these, the Subjective Discretionary Income measure proposed by O'Guinn and Wells, is evaluated in a cross-national replication study with financial products. SDI appears to be a valid predictor of "investments,"whereas total household income predicts "necessities." Lengthening of the SDI scale is suggested to increase its reliability.

Citation:

John R. Rossiter (1995) ,"&Quot;Spending Power&Quot; and the Subjective Discretionary Income (Sdi) Scale", in NA - Advances in Consumer Research Volume 22, eds. Frank R. Kardes and Mita Sujan, Provo, UT : Association for Consumer Research, Pages: 236-240.

Advances in Consumer Research Volume 22, 1995      Pages 236-240

"SPENDING POWER" AND THE SUBJECTIVE DISCRETIONARY INCOME (SDI) SCALE

John R. Rossiter, Australian Graduate School of Management

ABSTRACT -

New measures of "spending power" from the economics literature are reviewed as alternatives to the total household income measure typically used in consumer research. The simplest of these, the Subjective Discretionary Income measure proposed by O'Guinn and Wells, is evaluated in a cross-national replication study with financial products. SDI appears to be a valid predictor of "investments,"whereas total household income predicts "necessities." Lengthening of the SDI scale is suggested to increase its reliability.

INTRODUCTION

This paper has two purposes. The first purpose is to review the limitations of income as a measure of "spending power" in consumer behavior studies and to introduce consumer researchers to some alternative measures. The most important of these alternative measures of "spending power" come from economic research and may prove useful in consumer research. The second purpose is to provide a cross-national replication of one measure of "spending power" that has been used in consumer researchCthe Subjective Discretionary Income (SDI) measure proposed by O'Guinn and Wells (1989). To their credit, O'Guinn and Wells conducted extensive replications of the SDI measure in the U.S.A. However, with so many companies and advertising agencies now going "global," international replication of measures is of major importance. The opportunity arose to conduct a replication study of the SDI in Australia and the findings are reported in this paper.

MEASURES OF "SPENDING POWER"

Problems with Income as a Measure

Simply considered, consumer behavior is made possible by income. Yet there are severe problems with using income (usually measured as total household income) as an indicator of a household's current consumption capability or "spending power" (see also Lovell, Richardson, Travers and Wood 1990): (a) Inaccurate reporting of income. Consumer surveys may produce inaccurate reports of income. At the presumed top of the income scale, refusal rates on income questions are high, as indeed occurred in the survey reported the second part of the present study. At the presumed lower end, unreported income from "moonlighting" and from illegal businesses contributes a presumably large error, as does restricted-use income such as food stamps. (b) Failure to add the value of household members' time not spent in income-earning employment. The value of the full-time "houseperson's" time and the value of the time spent on household work by people working part-time and full-time is never figured in to household income. Nor is the value of tasks done by schoolchildren, such as washing the car or mowing the lawn. This "free" work effectively adds spending power in that the household is not having to use income to pay for the services thus provided. (e) Payments received "in kind." Similarly, value is received in many instances from employers, such as free gifts, free loans of equipment, and so forth; and, for some people, from the government, in the form of personal use of government equipment such as automobiles. Many of these "fringe benefits" are not recorded as income. (d) Services received from ownership of household durables. Products owned, such as housing itself, automobiles, vacuum cleaners, and the like, provide value which would otherwise have to be paid for each time, as in housing or apartment rental, car rental, cleaning services, and so forth, at quite a high cost. Presumably, service value from these products exceeds in the long run the price paid for them and therefore is a form of income. (e) Current income focus. Most income surveys report only current incomeCa statistic that fails to take into account household or individual actions to "smooth" consumption over the family's or person's life cycle. A common occurrence of this is "borrowed" income in the form of housing or automobile loans. A long-term debt is incurred but the "spending power" is available "now," and therefore functionally is current income. Evidently, stated current income has many limitations as a measure of a household's actual capacity to engage in consumer behavior.

Alternative Measures

There have been a number of potential solutions proposed for getting around the limitations of income as a measure of the true "spending power" available to consumers. These include: Full Income, Equivalent Income, Resources, Objective Discretionary Income, and Subjective Discretionary Income (SDI). The first four are explained briefly below. The fifth measure, SDI, is explained in the empirical study that follows.

1. Full Income is a measure first proposed by Decker (1965) and operationalized recently by Richardson (1989). Full Income is the maximum income that could be received if the household devoted all of its human and financial resources solely to maximizing income. In particular, hours which could have been spent in paid work are treated as if they were in fact spent that way, and the after-tax income is imputed. Services received from consumer durables (which include owner-occupied houses) are valued for the after-tax return which would have been obtained had the money used for their purchase been invested in financial securities, less depreciation. The value of "payments in kind" is similarly imputed. Numerically, the value of non-market time and of an owner-occupied house are the major adjustment factors.

Full Income is cumbersome to measure and compute in a standard consumer research survey. However, it is being seriously considered for government surveys in Australia as a measure for deciding social policy and in particular for defining the "poverty line" for consumer welfare payments (Harding 1994). The present measure, the "Henderson line," which is also used in the U.S., is based mainly on cash income and neglects the value of homes owned. By the Full Income measure, the emphasis shifts from retirees, who have low cash incomes but tend to be asset-rich, to blue-collar families, single parents, and unemployed youths, who are asset-poor as well as cash-poor.

2. Equivalent Income is a measure devised by Travers and Richardson (1990) which takes the after-tax, cash income of the household and divides it by an "equivalence" scale that rises proportionally with: the number of family members; children's ages, and thus their consumption costs; and the presence of a working spouse, who contributes additional household income not obtained if the spouse is present but not working. Each adult in the household is then assumed to have the same "equivalent income"; that is, income is adjusted after accounting for "need" differences between households.

Equivalent Income is quite readily computable from measures typically included in consumer surveys and would appear to be worth investigating by consumer researchers. It is also relevant for social policy in the area of taxation. For instance, France uses a "pooled dependent" system (Garran 1994) whereby family income is combined for tax purposes and the tax threshold is varied according to the number of dependents, including a non-working spouse, live-in relatives and disabled family members.

3. Resources is a measure devised by Lovell, Richardson, Travers and Wood (1990) to measure potentially available spending power rather than current spending power. It attempts to assess the extent of "resources," including income, that the household has available to draw upon if necessary. These resources, combined with current income, are thought to reflect the household's "standard of living." Standard of living is a concept that is commonly acknowledged but not well measured. When standard of living has been investigated in previous research, it has been estimated with a single overall rating which is of doubtful validity (see, for example, Andrews and Withey 1976). The Resources measure begins with Equivalent Income as measured above and adds scale points for 13 other variables such as whether the house is owned outright, the value of the residence, ownership of a second dwelling, value of automobiles, durables, shareholdings and so forth as well as the more idiosyncratic characteristics of whether the household recently fell behind in payment of bills (reverseBscored), received material help from a welfare agency (also reverse-scored), and whether the household is able to raise $5,000 in a week in an emergency. As might be evident, this is more a measure of potentially available spending power than current spending power.

Resources requires measures to be added to a consumer survey (13 single-item variables in addition to the standard measures used to compute Equivalent Income) but is quite easy to calculate. As pointed out by a reviewer, Resources would provide an interesting economically-based alternative to demographically-based measures of social class (e.g., Coleman 1983). It might be the other side of the hyphen in the term "socio-economic status" in that Resources seems to reflect economic status without the social values implications. In today's material world, standard of living is perhaps more relevant than the historical concept of social class.

In an extremely important study, conducted only in Australia at present, Lovell et al. (1990) demonstrated that the population's "standard of living" (as indexed by the Resources measure above) is much more homogeneous than the population's income (Equivalent Income). Although 95% of Australians are spread, by definition, across nine and a half deciles in terms of income, they are located within the seventh and eighth deciles in terms of standard of living, which is only a 20% spread. It is the latter statistic that confirms the popular perception that Australia is a very egalitarianCand materially comfortableCsociety. Looked at another way, it should be of great interest to consumer researchers that after-tax equivalent income and, by implication, the usual total household income measure, is correlated extremely weakly (r=.05, d.f.=1,067, p=.08 but consider the large sample size) with the very effectC"standard of living"Cthat such income is purported to produce! This suggests that "disposable" or "discretionary" income is being significantly overlooked when only total income is measured.

4. Objective Discretionary Income, or "Disposable Income," should, in theory, come closest to measuring "spending power." Discretionary Income (DI) is a widely-used concept for purposes of economic research. For example, the U.S. Bureau of the Census, working with the Consumer Research Center of the Conference Board, has developed a DI measure that begins with gross income and then subtracts the cost of housing, food, clothing, medical and other common household budget items as well as tax paid. These costs are assumed to be non-discretionary, that is, necessities.

Herein lies the major problem with "objective" measures that define spending power as what is left after money has been spent. The presumed "necessities" expenditures, for clothing, food and even the type of house that is purchased or rented, are obviously not 100% necessitated. The truth is that an unknown proportion of the "essential" expenditures is, for most individuals, discretionary.

The problem with Objective Discretionary Income is revealed in the following economic survey results. DI estimates for the U.S.A. for 1987 indicate that over two-thirds of households (71%) had no discretionary income after paying for so-called essentials (Bureau of the Census, 1989). Similarly, the present author's estimates for Australia, using data from the Australian Bureau of Statistics (1990), when cross-tabulated by gross income quintiles, produced the intuitively unacceptable result that, whereas the average DI per household in 1988-89 was approximately $3,500, this ranged from a negative $3,800 in the lowest income quintile; stayed negative in the second lowest income quintile; rose to marginally positive for the next two quintiles; and then reached a somewhat more credible $17,600 for only the uppermost income quintile. The negative DI for such an unbelievably large proportion of the population (well over 50% in the U.S.A. and 40% in Australia) is the problem: clearly, discretionary expenses are being misrepresented as necessary expenses.

5. Subjective Discretionary Income (SDI) stems from the work of Katona (1975) and his Index of Consumer Sentiment. O'Guinn and Wells (1989) trace the history of the SDI measure and it is their particular version of the measure that is examined in the present study. SDI is a measure of perceived spending power. As such, it should be a significant determinant of many types of consumer purchases. It is essentially a subjective measure of capacity to spend and in this sense is an "attitude" rather than an objective behavioral resource. (For a somewhat related approach to estimating spending power subjectively, see Wachtel and Blatt, 1990.) As in the O'Guinn and Wells (1989) study, the present study compares this subjective measure with the typical objective measure of total household income. The SDI measure is described in further detail next.

The SDI Measure

The O'Guinn and Wells (1989) measure of SDI has been employed in the advertising agency Needham Worldwide's Needham Lifestyle Survey for some years. It is composed of the sum of ratings on the following three items:

Item 1: No matter how fast our income goes up we never seem to get ahead (reverse-scored).

Item 2: We have more to spend on extras than most of our neighbors do (see modification below).

Item 3: Our family income is high enough to satisfy nearly all our important desires.

To better fit modern and more individualistic residential contexts, especially in urban areas, the wording of the second item was modified in the present study from "neighbors do" to "friends seem to."

In O'Guinn and Wells' studies, the three items were rated on 6-point scales ranging from definitely disagree (1) to definitely agree (6), with no neutral rating allowed. The present study employed a 5-point scale ranging from disagree strongly (1) to agree strongly (5), with a neutral point (3). As will be seen, this minor rating scale difference, which was insisted upon by the market research firm that collected the data, seemed to be more a matter of taste rather than substance. Total SDI scores on the O'Guinn and Wells scale thus can range from 3 to 18, whereas on the present SDI scale the scores can range from 3 to 15.

In the study reported here, the SDI was tested in an Australian context to examine its generalizability. Like O'Guinn and Wells' analysis, the present study analyzes the scale's reliability and its predictive validity.

THE STUDY

The data for the Australian study came from a sample of 1,427 households in one large and typical state, and were graciously made available for academic research by an Australian bank which wishes to remain anonymous. The door-to-door sampling method was as near random as is practically possible in commercial market research surveys. The present analysis is based on 1,187 households, constituting the 74% of the total sample in which respondents reported full financial data, including total household income. The 26% incidence of incomplete financial data is typical but apparently did not bias the final sample. Sex, age categories, and occupational categories estimated from the sample were within "3% of population values reported from the Australian Bureau of Statistics census, and with a sample size of 1,000, sampling variance of 2 to 4% could be expected at the 95% confidence level due to sample size alone.

The Australian survey happened to duplicate one of the two applications investigated by O'Guinn and Wells (1989) in their U.S. surveys: namely consumers' financial behavior. The Australian survey measured, among many other variables, consumers' usage of personal loans, common stocks, mutual funds, and money market fundsCvariables that were examined for predictive validation purposes in the U.S. study. The Australian study also measured usage of several other financial products which clarify the contribution of SDI.

Reliability of the SDI Scale

International comparisons provide an extreme test of a scale's reliability in the sense of reliability as generalizability (Cronbach 1951). In the present study, the international comparison is probably less extreme because of the likely cultural similarities between the U.S.A. and Australia. In this section, comparisons between the SDI scale's psychometric properties in the U.S.A. and Australia are presented.

The distribution of scores on the SDI scale in the U.S.A. was shown by O'Guinn and Wells (1989, their Table 1) to be approximately normal with a slight leftward skew (lower frequencies to the right). Their quadrichotomized distribution exhibited the following percentage frequencies: Low (scores of 3-6), 20.4%; Medium-Low (7-10), 35.2%; Medium-High (11-14), 32.9%; and High (15-18), 11.5%. Because the Australian version of the SDI scale had a different range of possible scores, it could not be identically subdivided and was instead trichotomized with 3-4-3 cutoffs, thus providing a slightly wider middle. The Australian data also exhibited a pronounced leftward skew while still being approximately normal. Percentage frequencies were as follows: Low (3-6), 28%; Medium (7-11), 66%; and High (12-15), 6%. The tendency for Australians' self-reports of SDI to cluster toward the middle is consistent with the homogeneity of the Resources measure of "standard of living" in Australia found by Lovell et al. (1990), as noted earlier.

The inter-item and item-total correlations in the two countries were highly consistent, as shown in Table 1. These results indicate that the SDI items "behave" in the same way in both nations.

The internal consistency of the SDI scale (Cronbach's coefficient alpha) was not calculated directly for the U.S. data by O'Guinn and Wells; instead, they stated that "the three items in the SDI scale...have always loaded on the same factor with factor loadings above .8" (1989, p. 35), implying a very high alpha. However, coefficient alpha for the Australian SDI scale was only .57, a level usually regarded as marginal if not unsatisfactory.

Despite the marginal internal-consistency reliability of the SDI scale in Australia, it was decided to proceed with it unchanged for the validity phase of the study to ensure comparability with the U.S. findings.

Validity of the SDI Scale

Two types of validity are of interest regarding the SDI scale: the discriminant validity of subjective discretionary income (SDI) and total family income (TFI) which should demonstrate that they are measuring different dimensions of spending power; and the predictive validity of SDI, especially in comparison with TFI, for consumers' financial behavior.

Discriminant Validity. In the U.S. data (for 1986), the correlation between SDI and TFI was r=.34 only. Though highly significant with the large sample size, the correlation suggests that SDI and TFI, in the U.S.A., have only 11.6% common variance. Using cross-tabulation, O'Guinn and Wells showed that this correlation allows wide variations in SDI at each level of total family income. In the Australian data, the correlation between SDI and TFI was lower still, at r=.25; even allowing for attentuation due to unreliability, the common variance of SDI and TFI, at 6.2%, is effectively zero. Most interestingly, the individual SDI scale item that was least correlated with TFI in the Australian data was the second, that "our family income is high enough to satisfy nearly all of our important desires," suggesting that high total income does not necessarily mean greater perceived spending power (see also the earlier result suggesting the homogeneity of "standard of living" in Australia). In any event, the lack of relationship between SDI and TFI in Australia makes their relative predictive capabilities all the more relevant.

Predictive Validity. O'Guinn and Wells compared the predictive validities of SDI and TFI by combining them in a multivariate equation and using logistic regression, an appropriate procedure given that the two predictor variables were somewhat correlated in the U.S. data. In the Australian data, however, SDI and TFI were effectively uncorrelated, so it is quite appropriateCand indeed practically realistic, as most market research managers use separate cross-tabs anywayCto compare their respective univariate predictions. O'Guinn and Wells reported Wald statistics, a form of maximum likelihood chi-square that indicates the strength of association of each predictor variable after adjusting for their joint effect in the regression. Therefore, the chi-square statistic was chosen for the Australian data also, though a simple ^c2, not an MLE-based adjusted ^c2. While not comparable precisely, ^c2 was a reasonably common statistic available for the two studies.

TABLE 1

SDI INTER-ITEM CORRELATIONS FOR THE U.S.A. (BELOW DIAGONAL) AND AUSTRALIA (ABOVE DIAGONAL) AND ITEM-TOTAL SCORE CORRELATIONS

TABLE 2

PREDICTIVE RESEULTS FOR SDI AND TFI FOR FINANCIAL PRODUCTS COMMON TO THE U.S. AND AUSTRALIAN SURVEYS AND FOR AUSTRALIA ONLY (U.S. RESULTS IN PARENTHESES): APPROXIMATELY COMPARABLE X² VALUES

The predictive results are shown in Table 2. As the top half of the table indicates, the Australian predictive results for SDI and TFI were consistent withCbut stronger thanCthe U.S. results for the four financial products included in both studies. Of the eight comparisons, the only inconsistencies involved two non-significant (p>.05) ^c2 values compared with significant but very small ^c2 values. Overall, SDI proved to be a much more important predictor variable in the Australian data than in the U.S. data for two of the products (mutual funds and common stocks), similarly highly predictive for a third (money market funds) and similarly unpredictive for the fourth (personal loans through a bank). Also, TFI proved to be a stronger predictor in the Australian data than in the U.S. data, directionally, for all four products and a much better predictor for two of these (common stocks and personal bank loans). Ignoring magnitude and focusing only on positive significance (p<.05), the U.S.-Australian results are consistent in six out of eight comparisons.

The results for the four other financial products measured only in the Australian survey, shown in the bottom half of the table, help to construct a fairly straightforward interpretation of the relative importance of SDI and TFI. SDI is evidently of major importance for discretionary investments (real estate, the stockmarket, and other high-return opportunities). TFI, on the other hand, is relatively more important for necessities (many would regard credit cards as necessities and TFI is clearly the single dominant predictor of personal loans and home loans). Again, it is important to observe that the "investment" types of financial products are not predicted well by total household income alone. Subjective discretionary income is therefore of considerable relevance to marketers of the more "high flying" financial products.

Future Research with SDI

The Australian findings suggest that whereas SDI is an important consumer variable, its measurement in an international or global context may suffer from marginal scale reliability if the O'Guinn and Wells SDI scale is used. The obvious solution to low reliability is to increase the number of items. However, scale development is recommended, based on theory and further research, rather than the mere extension of items "similar" to those in the 3-item scale at present.

Conceptually, items 2 and 3 seem to best capture what is meant by "subjective discretionary income," namely: "We have more to spend on extras than most of our friends seem to" and "Our family income is high enough to satisfy nearly all our important desires" (emphases added). In contrast, item 1 may have become too much of a cultural clichT, "...we never seem to get ahead," and it also makes the presumption that household income rises, "No matter how fast our income goes up...," which is not true for the majority of households in real terms. Also, item 2 seems to relate to discretionary expenditure on non-essential "luxuries" whereas item 3 allows for discretionary expenditure on better quality "necessities." These two items, which are correlated only about .3, might form the basis for two types of additional items. In any event, additional items reflecting degree of choice in how to spend income would appear to be conceptually appropriate.

SDI scale development would also be best pursued in several different countries, rather than in the two similar countries compared here, to develop a truly internationally applicable scale.

Predictive tests of SDI should also be extended to a broader range of products, as in the original O'Guinn and Wells' study, rather than being limited to financial products as investigated here.

SUMMARY

The typical practice of relying on simple reports of total household income in consumer behavior studies to represent the fundamental consumer concept of "spending power" reflects a limited perspective. A broader perspectiveCand deeper understanding Cwould be achieved in consumer research by using measures from economic research. Full Income (Becker 1965; Richardson 1989) appears to correct for the additional spending power released by household labor and products already acquired rather than having to be continually paid for and is relevant for social policy in the area of welfare payments. Equivalent Income (Travers and Richardson 1990) appears to be a useful "per capita" measure of spending power within the household and, since it corrects for the number and nature of dependents, it bears on taxation policy as well. Resources (Lovell et al. 1990) is a measure of potentially available spending power which appears to capture the important but surprisingly underinvestigated concept of "standard of living." Objective Discretionary Income (U.S. Bureau of Census) does not seem worth pursuing as it leads to an intuitively unrealistic distribution across the population by failing to acknowledge that much of the amount spent on "necessities" is in fact discretionary and, therefore, subjective. Subjective Discretionary Income, accordingly, appears to be a conceptually necessary aspect of "spending power" for consumer research.

Assuming that the traditional Total Household Income or total "family" income measure, TFI, continues to be used, then O'Guinn and Wells' (1989) Subjective Discretionary Income scale, SDI, appears to be a conceptually valid measure to use in addition. The dual use of TFI and SDI was shown in their U.S. study and in the present Australian replication to provide significant explanatory power for major financial products. TFI tends to predict "necessities" well, such as home loans and personal loans; SDI tends to predict "investments" well, such as mutual funds and money market funds; and both tend to be good predictors of credit card ownership, common stock ownership, and real estate investment beyond one's own home. However, the SDI scale may require further development for international or global use. Its internal-consistency reliability was estimated to be only .6 in the Australian context and low reliability would necessarily limit its predictive validity when correlated with dependent variables (for all types of products and services). Improvement may be obtained by adding items tapping the "choice" aspect of subjective discretionary income.

Indeed, as the alternative income measures introduced in this article imply, having a choice of how to spend income appears to be the main concept missing when consumer researchers use "income" to mean "spending power." Amount of income matters, but so too does the extent to which the consumer is free to spend it.

REFERENCES

Andrews, F.M. and S.B. Withey (1976), Social Indicators of Well-Being: Americans' Perceptions of Life Quality, New York: Plenum.

Australian Bureau of Statistics (1990), ABS 1988-89 Household Expenditure Survey, Canberra, AS: Australian Government Printing Office.

Becker, G. (1965), "A Theory of the Allocation of Time," The Economic Journal, 75 (September), 493-517.

Bureau of the Census (1989), A Marketer's Guide to Discretionary Income, Washington DC: Government Printing Office.

Coleman, R.P. (1983), "The Continuing Significance of Social Class to Marketing," Journal of Consumer Research, 10 (December), 265-280.

Cronbach, L.J. (1951), "Coefficient Alpha and the Internal Structure of Tests," Psychometrika, 16 (September), 297-334.

Garran, R. (1994), "Howard Outlines Income Split Plan," The Weekend Australian, January 8-9, p. 8.

Harding, A. (1994), "Just Where Should We Draw the Poverty Line?" The Australian, August 9, p. 15.

Katona, G. (1975), Psychological Economics, New York: Elsevier.

Lovell, C.A.K., S. Richardson, P. Travers, and L. Wood (1990), "Resources and Functionings: A New View of Inequality in Australia," Working Paper no. 90-7, Department of Economics, University of Adelaide, South Australia 5001.

O'Guinn, T.C. and W.D. Wells (1989), "Subjective Discretionary Income," Marketing Research, 1 (March), 32-41.

Richardson, S. (1989), "Cash Income and Full Income: Does the Difference Matter," Mimeo, Department of Economics, University of Adelaide, South Australia 5001.

Travers, P. and S. Richardson (1990), "Measuring the Standard of Living," Working Paper no. 90-4, Department of Economics, University of Adelaide, South Australia 5001.

Wachtel, P.L. and S.J. Blatt (1990), "Perceptions of Economic Needs and of Anticipated Future Income," Journal of Economic Psychology, 11 (September), 403-415.

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