A Model of U.S. Automobile Demand

V. Kumar, University of Texas
Robert P. Leone, University of Texas
Rajendra K. Srivastava, University of Texas
ABSTRACT - The general theory of psychological economics proposed by George Katona [1951, 1979] specifies that in order to better understand consumer expenditures, two factors must be considered. One is an objective factor Katona labelled "ability to buy" and the other a subjective factor he labelled consumers' "willingness to buy." This paper will explore the explaining and forecasting abilities of this framework in investigating U.S. automobile sales.
[ to cite ]:
V. Kumar, Robert P. Leone, and Rajendra K. Srivastava (1984) ,"A Model of U.S. Automobile Demand", in NA - Advances in Consumer Research Volume 11, eds. Thomas C. Kinnear, Provo, UT : Association for Consumer Research, Pages: 387-390.

Advances in Consumer Research Volume 11, 1984      Pages 387-390


V. Kumar, University of Texas

Robert P. Leone, University of Texas

Rajendra K. Srivastava, University of Texas

[Doctoral Student and Assistant Professors, respectively, Department of Marketing, The University of Texas at Austin, Texas 78712. The authors would like to thank the Institute for Constructive Capitalism for support during the formation stages of this research.]


The general theory of psychological economics proposed by George Katona [1951, 1979] specifies that in order to better understand consumer expenditures, two factors must be considered. One is an objective factor Katona labelled "ability to buy" and the other a subjective factor he labelled consumers' "willingness to buy." This paper will explore the explaining and forecasting abilities of this framework in investigating U.S. automobile sales.


The early tests of Katona's theory were performed on a cross-sectional basis, across all households in a sample done at a given period. The results from these early studies (Tobin 1951, Klein and Lansing 1955) consistently indicated that consumers' "willingness to buy," as measured by the Survey Research Center's index of consumer sentiment, significantly helped in explaining actual purchase behavior on an individual basis. However, when Mueller (1957) compared cross-sectional results for several different periods, she concluded that consumer attitude information was effective in explaining purchase behavior only when there was a marked divergence between changes in the levels of income and sentiment.

Katona and Schmiedeskamp (1967) and Katona (1979) argued that the objective in using sentiment data is not to find out which individual will make a purchase, but is anticipatory in nature and is intended to predict trends and turning points in expenditures. Therefore, any evaluation of the usefulness of sentiment data for understanding expenditures or forecasting must be tone on a time series basis.

Since that time, many time series studies of aggregate spending have been conducted and have shown inconclusive results (Adams 1964, 1965, Dunkelberg 1973). On the other hand, Curtin (1982) found the index of consumer sentiment (lagged) to be a more useful predictor of discretionary consumer expenditures than traditional market measures such as price and interest rates. Linden (1982) found the consumer confidence index to be a good indicator of major turning points in the business cycles. Finally, Wetzel and Hoffer (1982) found the index of (high income) consumer sentiments to be significantly related to the demand for standard and intermediate automobiles, but not for compacts and imports. All three studies used quarterly lata.

There are some plausible reasons for why consumer sentiment/confidence indices sometimes lead to inconclusive results. Juster and Watchel (1972, 1973) argued that this inconclusiveness is due to the amount of uncertainty perceived by consumers with regard to current economic conditions. In other words, consumer attitudes would reflect actual plans only for periods when they perceive economic conditions as stable. Consumers perceive that they cannot derive the maximum benefit from the decision making process during the period of uncertainty. In order to achieve the fullest satisfaction they tend to contemplate on their decisions. Therefore, for periods where a high degree of uncertainty exists, consumer attitudes are poor predictors of future plans. Tobin (1951) argued that another problem was that data were collapsed over large time periods (usually semi-annual or quarterly) to avoid auto-correlation problems and given that attitudes are often short-lived, this use of large time intervals tends to obscure any relationship that might exist. Consequently, monthly intervals may lead to further insights.

This paper will explore the usefulness of Katona's framework in investigating the monthly sales of automobiles in the U.S. Models based on wholly objective variables such as income, price, etc. will be contrasted with models that include only sentiment, as well as joint models that incorporate both types of variables. While a number of studies have explored Katona's framework, few (if any) have done so with monthly data and with a systematic, unbiased process.


The structures of the various models for U.S. auto sales which will be tested are based on various combinations of the components shown in Table 1. The two sets of variables represent the basis structure of Katona's two-factor scheme.

These variables will be operationalized by data which have been collected on a monthly basis from September 1977 through March 1982. The dependent variable is United States automobile sales in units as reported in the Survey of Current Business (UAUTO). The willingness to buy dimension will be represented by the Survey Research Center's Index of Consumer Sentiment (SRC). The components of the ability to buy dimension primarily economic and industry related factors, will be operationalized as follows:

COST (PAYMENTS) - Rather than use price, a variable was computed called purchasing COST which is the consumer price index for new cars x the interest rate since the "real" cost for purchase of durables depends on the interest rates as well (Graber, 1982). This measure hence reflects the size of monthly payments for those financing the purchase and an opportunity cost for those paying in cash. Obviously, one would anticipate a negative relationship between COST and UAUTO. Since no explicit measure of discounts were available, the manufacturers' suggested retail price index was used. The manufacturers' intent to discount should be reflected by the inventory measure (next paragraph). Further, based on the construction of our COST variable, as cost increases, monthly payments also increase leading to a decline in automobile sales. Also, as variable cost increases due to the interest rate increase, people may have a tendency to invest their money in savings deposits, which would earn them higher interest payments.

INDUSTRY OUTLOOK/CONDITIONS - This factor was broken down into two components representing the effect of industry supply and the effect the industry hopes to have on demand. industry supply is represented by INVENTORY and the industry's effect on demand is represented by ADVERTISING (Newspaper and Magazine expenditures). ADVERTISING was expected to have a positive relationship with UAUTO. INVENTORY, reflecting industry supply, was also expected to have a relationship with UAUTO. In most industries INVENTORY may reflect product availability and may thus facilitate sales. In the automobile industry, the higher than desired (by management) inventories during the time period covered in this study would lead to discounting (REBATES; not incorporated in the price index) and therefore sales. Consequently, the relationship between inventory (particularly lagged values) and sales can be expected to be stronger than in other industries that are not plagued with over-supply/excess capacity.

Table 1 summarizes the expected (hypothesized) relationships between the predictor variables and automobile demand (UAUTO).



All independent variables (other than SRC) were obtained from the Survey of Current Business.

While these variables appear in this paper, several other variables were initially considered as potential independent variables. For example, it was initially hypothesized that retail sales on nondurables would be a viable choice, having a negative relationship with automobile sales since expenditures on non-durables would reduce the monies available for durables such as automobiles. However, it was eliminated since it was highly correlated with our COST variable. Also, variables such as the personal disposable income, personal savings, the cost of living index, service expenditures, and used car prices were considered but eliminated due to similar correlation problems.

The following table illustrates the intercorrelations among the potential independent variables and the dependent variable.


One of the most interesting relationships exist between cost and automobile sales where it is clear that the constant trend upward in cost is being followed by a corresponding automobile sales series which has a downward (negative) trend. Another interesting observation is that the relationship between automobile sales and SRC exhibit a similar behavior to that just observed. This relationship does not appear to be nearly as similar as previous researchers have illustrated through analysis of quarterly data. This is an interesting finding since there are some clear implications about the existence of variations in the independent variables as well as the dependent variable. This may be attributed to the smoothing effect due to the aggregation of events within each quarter in previous studies. The smoothing effect is more limited in this study as we use monthly intervals.


Given the purpose of the study is to investigate the relationships between our specified independent variables and automobile sales, several models were estimated. One factor considered to be important was whether the independent variables had a current period effect or possibly a lagged effect on automobile sales. Also, the value of including the sentiment data was of interest, and therefore SRC was excluded from some of the models estimated.

Table 3 shows the results from the four models estimated. All coefficients are significant at an alpha of 0.05, and the values reported are the standardized regression coefficients (beta weights).



The independent variable that is most likely to have lagged effects is inventory. It may be that automobile manufacturers, upon observing high inventory levels in one month, may have a greater emphasis on price promotions (rebates, etc.) during the next. As such, lagged inventory should have a positive coefficient and should provide more explanatory power than current period inventory. This is easily observed by comparing Models 1 (current period inventory) and 2 (lagged inventory).

The increase in the Beta weights for inventory for Model 2 compared to Model 1 suggests that manufacturers may implement price promotions (not captured in price, and consequently COST, which is based upon un-discounted prices) upon observing past period accumulations in inventories, rather than current period inventories.

The effect of current and lagged consumer sentiment indices can be examined by comparing Models 3 and 4 with Model 2. Both model comparisons (3 with 2 and 4 with 2) illustrate that the consumer sentiment index is a useful predictor of automobile sales. The larger beta weight for lagged consumer sentiment index compared to the current one suggests that the purchase decision cycle for automobiles may extend well beyond a month.


All the coefficients have expected signs, and each model produced R-square values that were reasonable in a comparative sense (discussed above). However, the estimates indicate that lagged variables (Models 2,3 and 4) seem to have a stronger relationship than the current period variables (Model 1). When the plots for all four models were compared to chose for actual sales, Model 4 which included lagged terms for both SRC and Inventory seemed to capture the trends in the automobile sales with the greatest accuracy. [Note that Model 2 also has lagged inventories. Hence the ability of Model 4 to better predict turning points can be attributed, at least in part, to lagged SRC.] This finding is in agreement with that by Linden (1982) who found that the consumer confidence index to be a good predictor for turning points in business cycles. The plot for Model 4 and actual sales is provided in Figure 1.



This finding is supportive of Katona's argument for the value of using sentiment data for explanation of trends and as a leading indicator and not for pure "fitting" purposes. This finding is also similar to past research findings which showed the consumer sentiment indices' ability to act as leading indicators in recessionary periods. These models provide some explanation for the trends in automobile sales and point out the need to use such models in tandem with the best "fit" models, since the two models provide different information.

The Durbin-Watson statistic for all four models were quite close to the lower bound of the inconclusive range. Future research could be carried out with the use of multivariate time series analysis similar to that of the univariate analysis done by Leone and Kamakura (1982).

Also, the high intercorrelations observed among the set of independent variables considered initially could well be taken care of by using Ridge Regression.

Finally, the R-squares for the models reported in this study are somewhat lower (about .6, rather than .7) than those reported in the past studies. However, the reader should note that these models are based on: (1) monthly, rather than quarterly, data (consequently, greater variability in the dependent variable as well as in the independent variables is present); and (2) data over a shorter duration (1977-1982), rather than data from the 60's to the 80's, which included relatively smooth sales in the 60's and earls 70's.


Adams, F. G. (1964), "Consumer Attitudes, Buying Plans and Purchases of Durable Goods: A Principal Components Time Series Approach," The Review of Economics and Statistics 46 (November) 347-55.

Adams, F. G. (1965), "Prediction with Consumer Attitudes: The Time Series-Cross Section Paradox," The Review of Economics and Statistics 47 (November) 367-78.

Curtin, Richard T. (1982), "Indicators of Consumer Behavior: The University of Michigan Surveys of Consumers," Public Opinion Quarterly 46: 340-352.

Dunkenberg, W. C. (1973), "The Impact of Consumer Attitudes on Behavior: A Cross-Section Study," In Human Behavior in Economic Affairs. Stormpel, B. et al. (eds.) San Francisco: Jossey-Bass, 347-71.

Graber, Doris A. (1982), "Reading Between the Lines of Consumer Confidence Measures," Public Opinion Quarterly 46: 336-339.

Juster, F. T. & Wachtel, P. (1972). "Uncertainty, Expectations and Durable Goods Demand Models," in B. Strumpel et al., eds., Human Behavior in Economic Affairs: Essays in Honor of George Katona. Amsterdam.

Juster, F. T. and Wachtel, P. (1973). "Anticipatory and Objective Models of Durable Goods Demand." The American Economic Review 564-579.

Katona, G. (1979), "Consumer Expectations as a Guide to the Economy, " Public Opinion (Jan./Feb.) 15-19.

Katona, G. & Schmiedeskamp, J. W. (1967), "An Appraisal of Consumer Anticipations. Approaches to Forecasting: Discussion," Proceedings of the Business & Economics Statistical Section of the American Statistical Association, 126-7.

Klein, L. R. & Lansing, J. B. (1955), "Decisions to Purchase Consumer Durable Goods," Journal of Marketing 20 (October). 109-132.

Leone, R. P. & Kamakura, W. (1982). "The Usefulness of Indices of Consumer Sentiment in Predicting Expenditures." Proceedings for the 1982 Association for Consumer Research 10 (October), 195-199.

Linden, Fabian (1982). "The Consumer as Forecaster." Public Opinion Quarterly 46: 353-360.

Mueller, E. (1957). "Effects of Consumer Attitudes on Purchases." The American Economic Review, 946-66.

Tobin, J. (1951). "On the Predictive Value of Consumer Intention and Attitudes." The Review of Economics and Statistics 41 (February) 1-11.

Wetzel, James & Hoffer, George (1982). "Consumer Demand for Automobiles: A disaggregated market approach." Journal of Consumer Research 9 (September) 195-199.