Stated preference surveys for electric and alternative fuel vehicles: are we doing the right thing?

Abstract

This perspective paper analyzes the practice of stated preference (SP) surveys for electric vehicles (EVs) and identifies a number of limitations in the way they are currently undertaken. Specifically, it is found that SP surveys may not adequately represent several dimensions that are relevant in the context of EVs purchase such as: garage ownership, second v. first car, and refueling conditions, and they often neglect transitory technologies (plug-in hybrid) which are in fact an important element in the diffusion of EVs. This paper also provides a number of recommendations for practitioners to conceive more realistic SP surveys, which could increase the validity of policy recommendations formulated by applied economists.

Keywords:

• SP survey
• Electric vehicles
• Car purchase forecast

Notes

1 The willingness to pay (WTP) deriving from the coefficient values (gasoline car, household income) in model (IV) is €0·13 per household income, which indicates that a €1 increase of household income implies a 0·13 extra WTP for gasoline vehicles.

2 One may want to check this conjecture by looking at the results of SP surveys comparing the monetary value of the largest reasonably possible variation of each attribute. The intuition behind is that, if the value of raising fuel availability from 0 to 100% corresponds to limited number of range kilometers, it could be wiser to concentrate on the latter attribute. Stated preference surveys where both refueling stations and range were present provide a mild conclusion on this point. We can suppose that an electric vehicle (EV) would typically be in the range of 100 v. 800 km for conventional engines, while refueling stations could not represent more than a 100% increase. Based on the meta-analysis made by Massiani (2012) such an increase in range could be valued as much as €9000. This is slightly smaller than the amount provided by Batley et al. (2004) for refueling stations, but slightly more than the amount provided by Brownstone et al. (2000). So this method fails to provide an unambiguous indication on the pre-eminent importance of one rather than the other attribute.