The scope test reconsidered: Theory and experimental evidence

• Main Author: Murphy, James Havord
• Format: Dissertation
• Language: English
• Published: ProQuest Dissertations & Theses 01.01.2000
• Subjects: Agricultural economics; Alliances; Binomial distribution; Cognitive ability; Communication; Consumption; Contingency tables; Copyright; Data analysis; Dependent variables; Diamonds; Economic theory; Economics; Experiments; Hypotheses; Laboratories; Polls & surveys; Preferences; Probability; Research methodology; Taxonomy; Training; Utility functions; Validation studies; Validity; Valuation; Variables; Violations; Willingness to pay
• ISBN: 0599818999; 9780599818996

This study investigates a key issue in the use of surveys employing hypothetical scenarios to value non-marketed public goods: how to validate survey-based estimates. Specifically, we analyze the relationship between the sensitivity of survey responses to changes in the survey scenario, a commonly-used yardstick recommended by the NOAA Panel as a “burden of proof” requirement, and the construct survey-derived values are purported to measure, responses consistent with regular individual preferences.1 We define these constructs in a manner amenable to theoretical analysis and show that satisfying sensitivity to scenario is not theoretically equivalent to consistency with regular preferences. Revealed preference theory, developed by Samuelson and reformulated by Houthakker, Richter, Varian and others, yields conditions, the Strong Axiom of Revealed Preference (SARP) and the Generalized Axiom of Revealed Preference (GARP), that are theoretically equivalent to the consistency of responses with regular preferences when demand is single-valued (SARP) or multi-valued (GARP). Assuming single-valued demand, binomial probit analysis of data collected using controlled, experimental methods suggests weak support for an empirical relationship between responses satisfying the SARP and sensitivity-to-scenario (H1), stronger support for a relationship with the hypothetical nature of payment and provision in the scenario (H2) and subjects' cognitive abilities (H5), and no support for a relationship with whether subjects are trained (H3) or the time subjects employed formulating their responses (H4). Assuming multi-valued demand, these hypotheses are not supported. Assuming single-valued demand, count data analysis yields limited support for a relationship between the number of SARP violations and sensitivity-to-scenario (H1) and robust support for a relationship with training (H3). Assuming multi-valued demand, count data analysis reveals only a less than robust relationship between the number of GARP violations and sensitivity-to-scenario (H1). 1 A preference relation, R, is regular if it is reflexive, total and transitive. A preference relation is reflexive if for all possible consumption bundles, x, xRx; total if for any two bundles x and y, either xRy, yRx, or both; and transitive if for any three bundles, x, y and z, xRy and yRz implies xRz.