The VTI National Transport Library Catalogue

Models for car ownership, transactions and vehicle type Ramjerdi, Farideh ; Rand, Lars ; Saetermo, Inger-Anne

By: Ramjerdi, FaridehContributor(s): Rand, Lars | Saetermo, Inger-AnnePublication details: Lunds universitet. Tekniska högskolan, 2000; Teknik och samhälle. Trafikplanering, ; Bulletin, Description: nr 187:2, s. 425-43Subject(s): Sweden | Conference | Vehicle ownership | Car | Mathematical model | Questionnaire | Household | Choice | Statistics | 11Online resources: Publikation/Publication Bibl.nr: VTI P1494:187:2Location: Abstract: This paper describes the integrated survival models for car ownership, car transaction and vehicle type. These models are estimated based on a two-wave panel study. The design of the questionnaire in the first wave (1993) allows the reconstruction of the history of car ownership and transactions and the history of the household characteristics to different extent. NMNL models for household's choice of car transaction (disposal, purchase, replacement or do nothing) and vehicle type were estimated for years 1988 until 1994 as cross-sectional data. Logsum variables from the NMNL models are used in survival models for car ownership and car transactions. This was the main reason for not using lagged variables in the NMNL models. With lagged variables, the models would be dynamic models of car transaction and car type choice. Survival models for car ownership and car transactions were estimated using a Weibull distribution as well as an exponential distribution. The log likelihood shows that the Weibull model is the better one for all the model formulations. The exponential duration model is very restrictive. It does not handle hazards that are increasing or decreasing, while the Weibull duration model does so. We failed to estimate models based on other distributions.
Item type: Reports, conferences, monographs
Holdings: VTI P1494:187:2

This paper describes the integrated survival models for car ownership, car transaction and vehicle type. These models are estimated based on a two-wave panel study. The design of the questionnaire in the first wave (1993) allows the reconstruction of the history of car ownership and transactions and the history of the household characteristics to different extent. NMNL models for household's choice of car transaction (disposal, purchase, replacement or do nothing) and vehicle type were estimated for years 1988 until 1994 as cross-sectional data. Logsum variables from the NMNL models are used in survival models for car ownership and car transactions. This was the main reason for not using lagged variables in the NMNL models. With lagged variables, the models would be dynamic models of car transaction and car type choice. Survival models for car ownership and car transactions were estimated using a Weibull distribution as well as an exponential distribution. The log likelihood shows that the Weibull model is the better one for all the model formulations. The exponential duration model is very restrictive. It does not handle hazards that are increasing or decreasing, while the Weibull duration model does so. We failed to estimate models based on other distributions.

Powered by Koha