A key input in cost-benefit analysis is represented by the marginal rate of substitution which expresses the willingness to pay, or its counterpart willingness to accept, for both market and non-market goods. The consistent discrepancy between these two measures observed in the literature suggests the need to estimate reference dependent models able to capturing loss aversion by distinguishing the value attached to a gain from the value attached to a loss according to reference dependent theory. This paper proposes a comparison of willingness to pay and willingness to accept measures estimated from models with both symmetric and reference dependent utility specifications within two different freight transport stated choice experiments. The results show that the reference dependent specification outperforms the symmetric specification and they prove the robustness of reference dependent specification over datasets designed according different attributes levels ranges. Moreover we demonstrate the policy relevance of asymmetric specifications illustrating the strong implications for cost-benefit analysis in two case studies.