Choice modelling (CM) is a stated preference technique which has been increasingly used to generate estimates of non-market values. In CM applications, respondents to a survey are presented with several sets of options, which contain common attributes but at different levels, and are asked to select their preferred alternative from each set. The information provided by respondents’ choices is used to estimate marginal rates of substitution and, hence, estimates of value. An advantage of using CM when estimating passive use values is the capacity to present to respondents options that explicitly consider environmental, social and developmental outcomes. Consideration of all of these outcomes has been recommended by members of the NOAA panel.

In this paper the results from three separate CM applications are presented in which estimates are made of the value of improving the quality of wetlands affected by reductions in instream flow. The first two applications focused on the Gwydir Wetlands in New South Wales, Australia. One application was conducted in Sydney, a major urban centre, and the second application was conducted in Moree, a rural centre close to the Gwydir Wetlands. The third application focused on the Macquarie Marshes, another wetland in New South Wales.

The questionnaires used in the CM applications were developed using 8 focus groups and a pretest. Five environmental and developmental attributes were selected for use in the choice sets. These were water rates, employment, wetland area, frequency of waterbird breeding and endangered and protected species present. Modifications to the questionnaire were also required to minimise bias, implausibility resulting from the experimental design and confusion.

Apart from producing estimates of non-market values, an objectives of the study was to test the suitability of using choice modelling derived estimates for benefit transfer, both across different population and across different wetlands. The results of tests conducted to determine the statistical equivalence of the three models are reported. These in turn enable an assessment of the suitability of the models for use in benefit transfer. To undertake the equivalence tests, an estimate of the scale parameter which are implicit in the results of choice models was found. While the magnitude of the scale parameter for each model does not affect the estimates of value derived, it can confound model comparisons. The paper also details attempts to derived a combined benefit transfer model based on the results of the three applications.