Task independence in stated preference studies: A test of order effect explanations

We present an experiment designed to investigate the presence and nature of ordering effects within repeated response stated preference studies. We formulate a general structural model of such effects and use this to isolate signature patterns for position-dependent effects (learning about preferences or institutions, and the impact of fatigue) and precedent-dependent effects (starting point effects, reference pricing and various forms of strategic behavior). This is tested within a large sample, full factorial study designed to mitigate against misspecification bias and design-induced error variance problems. Non-parametric and parametric analyses are applied, the latter adopting a novel data-driven approach to the detection of ordering patterns. While we find little evidence of position dependent effects, we do find evidence of a starting point effect and various types of strategic behavior including a reference price effect where respondents tend to reject alternatives that are priced higher than recently seen alternatives.