USE OF THE TOBIT-MODEL IN CONTINGENT VALUATION - EXPERIMENTAL-EVIDENCE FROM THE PEMIGEWASSET WILDERNESS AREA

Two statistical techniques commonly used in contingent valuation studies-ordinary least squares estimation and Tobit analysis-are compared using the data obtained from a 1989 survey of users of the Pemigewasset Wilderness Area in New Hampshire. Parameter estimates are contrasted for differences insign, size and significance. An examination of zero bids is also conducted, and inferences are drawn as to which analytical technique is preferred. Results indicate that coefficient values estimated using Tobit analysis may differ substantially from those estimated using OLS. In addition, the direction of this difference in coefficient values is not consistent. Since Tobit analysis is the more theoretically correct method for willingness-to-pay (WTP) data sets with large numbers of zero bids, past research conduced using least squares analysis may have provided incorrect estimates of coefficients for explanatory variables. Tobit also allows decomposition of the data set to examine more closely the effects of the independent variables on current non-zero WTP bidders, and estimates the probability of zero bidders changing to positive bids based on changes in the independent varaibles. However, extra care must be taken in culling protest bids from the sample. Results of this study indicate that users of the Pemigewasset Wilderness Area are willing to pay between $4·82 and $5·65 per trip to the area. Variables which tend to affect these bids most strongly are the probability of returning to the area and the importance of a primitive camping experience. © 1991 Academic Press Limited.