Making an Informed Vehicle Scrappage Decision

Abstract

According to the US Federal Highway Administration (FHWA), the number of publicly owned vehicles in the USA reached 3 913 999 in 2003. In order to maintain a stable vehicle fleet, government agencies must repeatedly make vehicle scrappage decisions because older vehicles must first retire to make room for newer vehicles. Typically, these decisions are made based on a deterministic ranking evaluation model to select candidate vehicles for replacement. The paper applied an objective and probabilistic method to a vehicle dataset collected by the DuPage County Forest Preserve District ((DCFPD), in the state of Illinois). A Weibull‐form survival model with time‐varying covariate and unobserved heterogeneity was estimated on the dataset. The results suggest that in addition to the fact that vehicle age is negatively related to the vehicle’s survival probability, there are other variables that also appear influential. The survival probabilities of alternative fuel vehicles are similar to those of reformulated unleaded gasoline vehicles. The results suggest that a probabilistic and objective model can benefit government agencies in their vehicle scrappage decisions.

Acknowledgement

The authors thank the DuPage County Forest Preserve District, Illinois, for its generous support in providing the dataset and answering numerous questions. They also thank three anonymous reviewers and the editor whose comments have significantly improved the paper. The authors are responsible for all remaining errors.

Notes

1. Areas designated by the EPA not to have met the National Air Quality Standard.

2. Survival probability is defined as the probability that a vehicle will survive for at least a certain period.

3. DuPage County of the State of Illinois is within the EPA designated 8‐hour ozone and PM_2.5 non‐attainment area. All gasoline sold in the region is reformulated gasoline.

4. The survival model is designed to answer the following questions. (1) What is the probability that an object will survive for at least T period (termed as the survival probability) given that it has survived until now. (2) How will various factors affect the object’s survival probability?

5. If α > 1, it suggests that the hazard is positively related to the survival time. If 0 < α < 1, it suggests that the hazard is inversely related to the survival time.

6. Complete cells mean that within the defined study period, the occurrence of an event was observed. The term ‘event’ in this paper is defined as ‘vehicle scrappage’. On the other hand, incomplete cells mean that within the defined study period, the occurrence of an event was not observed. In other words, the vehicle is still active at the end of the study period.

7. The usual way to model the unobserved heterogeneity via parametric modelling is to specify a mixing distribution for the hazard rate function λ(t). However, the results of a mixing distribution are highly dependent on the type of distribution used (Heckman and Singer, 1984; Blossfeld et al., 1989, p. 100).

8. In addition to modelling vehicles owned by households, Chen and Niemeier (2005) also applied a discrete approach to model vehicle age.

9. The dataset does not distinguish between vehicles that are disposed and auctioned.

10. Alternative fuel can include the following types: compressed natural gas (CNGAS), biodiesel, ethanol E‐85 (E‐85), electric, and liquid propane gas (LPGAS).

11. These numbers are average values of the dataset.

12. ln(odometer reading in miles) is used in the current study.

13. Given the advantages of a panel dataset (Kitamura, 1990).