Managing groundwater in a mining region: an opportunity to compare best‐worst and referendum data

Abstract

In nonmarket valuation, practitioners must choose a format for the valuation questions. A common approach in discrete choice experiments is the 'pick-one' format, often with two alternative policy proposals and a status quo from which the respondent selects. Other proposed formats, include best-worst elicitation, where respondents are asked to indicate their most and least favoured alternative from a set. Although best-worst formats can offer efficiency in data collection, they can also lead to responses that are difficult to reconcile with neoclassical welfare estimation. The current article explores methodological issues surrounding the use of pick-one versus best-worst data for nonmarket valuation, focusing on framing and status quo effects that may occur within three-alternative discrete choice experiments. We illustrate these issues using a case study of surplus groundwater use from Western Australian mining. Results identify concerns that may render best-worst data unsuitable for welfare estimation, including a prevalence of serial choices in which the status quo is universally chosen as the worst alternative, rendering part of the choice process deterministic. Asymmetry of preferences and serial choices can be obscured when models are estimated using 'naively' pooled best-worst data. Results suggest that caution is warranted when using best-worst data for valuation, even when pooled results appear satisfactory.

A nonparametric random effects model for the valuation of forest recreation services: An application to forest sites in Tuscany, Italy

“…Within the environmental and resource economics literature, many studies have valued the recreational benefits related to environmental nonmarket goods and services by analysing stated preference (SP) survey data collected via discrete choice experiments (DCEs; see, e.g., Adamowicz, 2004;Boxall & Adamowicz, 2002;Doherty et al, 2013;Juutinen et al, 2017;Legg et al, 2020;Pelletier et al, 2022;Yao et al, 2019). In contrast to revealed preference (RP) data, SP data consist of hypothetical consumption decisions made by economic agents within experimentally designed scenarios that are constructed by the analyst, involving multiple alternatives (Adamowicz et al, 1994(Adamowicz et al, , 1998Bazzani et al, 2018;Boxall et al, 2009;Boxall & Macnab, 2000;Hanley et al, 1998;Juutinen et al, 2011Juutinen et al, , 2012Louviere et al, 2000;MacDonald et al, 2019;Morrison & Bennett, 2004;Pellegrini et al, 2022;Scheufele & Bennett, 2013).…”