The Reliability of the Paired Comparison Method in Contingent Valuation

David C. Kingsley



Does the preference learning provided in paired comparison experiments positively affect estimated welfare measures from traditional stated preference nonmarket valuation methods?


The goal of the stated preference nonmarket valuation literature is to develop methods which accurately reveal individual preferences for public goods and resources which reside outside of markets. The most commonly used method of valuing mean willingness to pay is the single bounded contingent valuation format in which a respondent makes a single, take it or leave it, choice between the provision of a public good or resource and a dollar bid amount.


Recent research has focused the attention of economists on the role of learning in such valuation methods. Specifically, research has shown that participants in a paired comparison experiment are able to display preferences consistent with economists’ expectations in a willingness to pay – willingness to accept disparity experiment when a private good is involved. Although an important contribution it remains to be shown whether preferences for public goods measured through PCEs are more reliable than alternative stated preference nonmarket valuation methods.


Building on my work in “Contingent Valuation and the Method of Paired Comparisons” this paper will determine the effectiveness of PCEs in the valuation of a locally relevant public good. Possibilities include the water quality or recreational opportunity along the Westfield River (perhaps through the removal of existing Dams) or the avoidance of a biomass electricity plant nearby.


Two treatments are needed to measure the effect of the PCE

- Valuation estimated based on a PCE with a standard SBCV afterwards

- Valuation estimate based on a SBCV with PCE afterward


- Ideally a baseline value will be determined either using a revealed preference method or a contingent donation mechanism. If unavailable a cheap talk survey instrument may be employed as survey instruments using a cheap talk script have been shown to reflect market preferences (Champ et al. 1997).


If the type of preference learning provided in the PCE is of consequence to public good valuation than the SBCV estimated from the first treatment (after the PCE) should more closely reflect the baseline measure. The success of an easily implemented SP method able to provide accurate preference measurement for public goods would be a significant contribution to the NMV literature.


Format would be a mail survey with an internet portion. The PCE portion requires an interactive computer program that can be delivered though the internet. Funding for this project is necessary and appropriate grant funding is being pursued.