Citation

Dubin R, Greene DL, Begovich C. Transp. Res. Rec. 1979; 726: 29-37.

Copyright

(Copyright © 1979, Transportation Research Board, National Research Council, National Academy of Sciences USA, Publisher SAGE Publishing)

DOI

unavailable

PMID

unavailable

Abstract

Interest in forecasting the fuel efficiency of the automobile population has led to the development of automobile market-shares demand models. The validity of these models depends on the automobile clasification used, yet little rigorous attention has been given to the problem of classifying automobiles for demand analysis. All existing models use classifications that are heavily subjective and rely on only one or two vehicle characteristics for classification. A cluster analysis of 125 models of 1975 automobiles was conducted in order to aggregate the vehicles into homogeneous groups suitable for modeling the demand for automobiles by vehicle type. Eight variables extracted from an automobile characteristics data base developed by the U.S. Department of Transportation were employed: curb weight, wheelbase, engine displacement, roominess, passenger capacity, fuel economy, list price, and power-to-weight ratio. Several weighting schemes, two-distance metrics, and hierarchical as well as nonhierarchical clustering techniques were used. The analysis strongly indicated that two- and six-group configurations were important. Within the six-group clustering, the three groups that had the highest average seat kilometers per liter and seats per initial cost comprised more than 80 percent of sales in 1975. A comparison of the cluster-analysis grouping with another classification used in a recent econometric automobile-demand model showed that the multivariate clustering did a consistently better job of accounting for the variability of vehicle characteristics.