CAR COMMUTERS' MODE CHANGE IN RESPONSE TO TDM MEASURES: EXPERIMENTAL DESIGN APPROACH CONSIDERING TWO-WAY INTERACTIONS

Abstract- Many studies have shown that individuals' responses to urban traffic congestion, as usually assumed by policymakers, are significantly different from their respected actual behavior. This paper adopts a behavioral approach to examine this difference, using the design of experiment principles and binary logit models. In this approach, five transportation demand management (TDM) measures including three push and two pull measures were investigated. Then, effects and contributions of the measures in diverting car commuters to seven existing non-car modes were taken into account. This study uses the stated preferences of 288 individuals who regularly use their private cars to access their job locations in the central Tehran area, to calibrate seven non-car mode models. The results show that when considering each mode separately, pull measures are necessary to regulate the market share of each non-car mode. Analysis of the effects of the measures in considering non-car modes shows that although their contributions are about 14% for transit accessed by walking and 7% for taxi, they have never contributed more than 5% to other modes. Table 5 presents a salient representation of the final models of the seven main non-car modes, including Walk & Ride, Drive & Ride, Taxi & Ride, Motorcycle, Drive & Taxi, Taxi and Tel-taxi. Because the focus of this stage is on the general tendency of the measures' effects on consideration of noncar modes, the salient representation of such effects are presented. A positive sign of a variable in these models indicates that the condition represented by that variable increases the relative probability of considering the associated non-car mode. A negative sign also shows that the condition represented by that variable increases the relative probability of considering other non-car modes of the study. The goodness of fit value of models, ρ2, varied in a range of 0.19 for Walk & Ride to 0.91 for Drive & Taxi, which seems suitable in individual-based models. This result is confirmed by assessing each model's lack of fit using Hosmer-Lemeshow (H-L) test, which shows that the lack of fit hypothesis is rejected for each of the models at 5% level of significance. Interactions were all insignificant except when both parking cost and fuel cost measures were implemented simultaneously, which motivated less than 1% of people to choose motorcycles. According to Table 7, a decrease in transit time and improvement in the transit access hav