Gap between Willingness‐to‐Pay (WTP) and Willingness‐to‐Accept (WTA) Measures of Value of Travel Time: Evidence from Norway and Sweden

Abstract

A main purpose of this paper is to stress the existence of the gap between willingness‐to‐pay (WTP) and willingness‐to‐accept (WTA) measures of value of time that is larger than could be explained in a Hicksian framework. The explanation of the gap is of great importance in welfare analysis. This is a subject that has not been paid much attention to in value‐of‐time studies and there is great need for further research on the subject. This paper relies on the evidence from Norwegian and Swedish value‐of‐time studies to examine the gap between WTP and WTA measures of value of time. These studies use different techniques for data collection. The evidence from both studies suggests that the observed WTA measure of value of time is about 1.5–2.0 times larger than WTP measure. Alternative model formulations in the Swedish study to address ‘inertia factors’ for WTP and WTA, or the exclusion of zero values for WTP and WTA measures in the Norwegian study, reduce the gaps. The reductions of the observed gaps in the WTP and WTA measures are analysed in the framework suggested by Zhao and Kling in 2001. They put forward ‘commitment cost’ as an explanation for WTP and WTA gap. An agent forms a commitment cost, in a similar manner to a ‘real option’, due to a lack of information, knowledge and uncertainty about his own valuations and also because a decision about a transaction has to be made rather swiftly. The theoretical framework suggested by Zhao and Kling is in line with additive utility functions underpinning the neo‐classical economics. If indeed ‘commitment costs’ is the explanation for the observed WTP and WTA gaps, the observed small values of small time savings should be interpreted as short‐term valuations. The long‐term values of small time savings should be quite higher than the short‐term values.

Acknowledgements

The authors like to a thank Staffan Algers, Mogens Fosgerau, Knut Veisten, and Lars Hultkrantz for suggestions and comments on an earlier version of this paper.

Notes

1. The standard deviations for WTP and WTA in Table 2 are about 11 and 8, respectively. The standard deviation for the value of time in Table 3 based on the pooled sample is about 6. The standard deviations of these estimates are underestimated by about 50% when a jack‐knife procedure is used. This procedure is used to account for the fact that the choices of an individual are not independent (Lindqvist Dillén and Algers, 1999).

2. The mixed‐logit procedure described by Revelt and Train (1998) and Train (1998) has been adopted in the present study.

3. These models were the best among alternatives according to the log‐likelihood values.

4. Lindqvist Dillén and Algers (1999) estimate a non‐linear model based on the third‐order approximation to the utility function. They show that the observed distribution of value of time is closer to the estimated distribution based on the third‐order approximation than the estimated distribution based on the second‐order approximation.

5. This study addresses both urban and long‐distance (interurban) travel. It focuses on the evidence related to long‐distance travel in order to make it comparable with the Swedish evidence.

6. For mode car there is not a good record of how many telephone calls were placed to locate the targets (those who had undertaken a long‐distance travel journey in the previous week). The response rate was about 75% among the targets.

7. This technique was actually closer to the open‐ended format of contingent valuation than to transfer pricing, since the respondents were offered changed attribute levels.

8. Or it should be reduced to the EG and EL gaps.