The variables used to predict household trip-generation rates have long been an area of concern for transportation planners; these variables included household size, number of vehicles owned, and income. However, a recent NCHRP study that used linear regression analysis has proposed that a household-structure variable would correlate more strongly with trip rates than almost any other variable, except vehicle ownership. In particular, this should improve the model significantly where it is combined with vehicle ownership and used as a substitute for household size. The results of a trip-generation analysis performed on data from the Midwest by using multiple classification analysis (MCA) in contrast to linear regression are described. The household-structure variable was tested by using both analysis of variance and MCA to determine how well the variable performs in various model structures when compared with other variables. The other variables tested were number of cars or vehicles available to the household, household size, housing type, total number of employed persons, household income, and total number of licensed drivers. It was concluded that the household-structure variable did not perform significantly better than the other variables tested. (A)
Abstract