Valuing Unfamiliar and Complex Ecosystem Services: The influence of survey mode, knowledge and dishonesty

Abstract

Increasing demand for valuation of ecosystem services has led stated preference methods to be applied to public goods that are increasingly complex and unfamiliar. Traditionally, stated preference surveys were conducted via mail or face-to-face interviews, but over the past two decades internet panels have been used to a larger extent. As we move away from traditional methods of survey administration it is apparent that we need a better understanding of how “new” survey administration modes influence elicited preferences, particularly when the environmental good under valuation is complex and unfamiliar. Furthermore, it is unlikely that people have well defined preferences over goods for which they have no experience consuming, and evidence suggests that preferences for such goods are constructed during the survey itself. This highlights the importance of information and, by extension, familiarity and knowledge, for people to accurately state their preferences. When analyzing discrete choice data, one of the underlying assumptions is that people are rational utility maximizers, however, mounting evidence show that respondents in discrete choice experiments use simplifying strategies and decision heuristics to reduce the cognitive burden of the choice task. This type of boundedly rational behavior is likely to increase when the environmental good is complex and unfamiliar. This thesis addresses some of the challenges practitioners face when valuing complex and unfamiliar public goods.

In the first paper I compare two identical discrete choice experiments (DCEs) aimed at eliciting the Norwegian population’s preferences for increased cold-water coral protection, an environmental good that is considered both complex and unfamiliar. This is the reason why the first DCE was implemented in a series of valuation workshops and why we created videos to secure identical information and provide the same visual impact when conducting the DCE using a probability based internet panel. Our results show that it is possible to use internet panels when the environmental good is complex and unfamiliar, but that practitioners should pay close attention to information provision, emphasize consequentiality and implement procedures to reduce speeding behavior. In the second paper I explore the link between knowledge (familiarity) about the environmental good measured by a quiz on cold-water coral, and the probability that a respondent ignores one or more attributes on the choice card. We find that respondents scoring above the average on the quiz, a measure of high knowledge, is associated with a higher probability of attending to the non-cost attributes (although only significant for one) and a significantly lower probability of attending to the cost attribute, irrespective of whether they knew how well or how badly they did on the quiz. These results show that understanding what type of information affects the degree to which respondents ignore attributes, and in which direction, is crucial to reduce attribute non- attendance behavior and obtain more precise estimates. In the third paper, I identify a group of dishonest respondents who have lied on a follow-up question. I hypothesize that these respondents have spent less effort on the choice tasks and as such have a less deterministic choice process (from a practitioners point of view) and are more likely to ignore attributes. The results show that dishonest respondents are more likely to be in a scale class characterized by a relatively higher error variance and more likely to ignore the non-cost attributes (significant for two out of four). Furthermore, the results suggest that observed difference in error variance between honest and dishonest respondents can partly be explained by different propensities to ignore attributes. As such, this thesis addresses some of the challenges associated with using DCEs to value complex and unfamiliar goods.