Response anomalies in discrete choice experiments for environmental valuation

Abstract

The use of discrete choice experiments (DCEs) to value environmental goods and services has been rapidly increasing over the last years. However, many studies call into question the validity of this approach. Biases in welfare estimates derived from DCEs may arise from response behavior which is not fully consistent with one or more of the assumptions underlying stated preference approaches. Since preferences are elicited in surveys, the central question is whether anomalous response behavior is triggered by the survey design. This cumulative dissertation analyzes response anomalies in DCEs for environmental valuation. By designing and evaluating two large-scale, nation-wide online surveys each time implementing a split sample approach, the influence of choice task complexity, attribute thresholds and survey consequentiality is investigated systematically. First, the impact of choice task complexity on survey drop-outs, choice consistency, the attendance to attributes and the status quo effect is studied by systematically varying the design dimensions of the DCE following a design plan originally introduced in transportation research. It is shown that the complexity of the DCE impacts marginal as well as non-marginal welfare estimates. With respect to each individual design dimension, survey drop-outs are found to be positively related to the number of choice sets, alternatives and attributes. Choices tend to become more consistent the more choice sets are presented as well as with a narrow level range. While the attendance to attributes is negatively affected by the number of choice sets and alternatives, status quo choices are observed to decrease with the number of alternatives and increase with the number of choice sets, the level range and the similarity between alternatives. Second, stated information on attribute cut-offs is used to study threshold-based decision making. Results suggest that prior cut-off elicitation influences willingness to pay estimates. Moreover, respondents are observed to employ cut-offs, which they are, however, in many cases willing to violate when having to make trade-offs in the DCE. Third, it is investigated whether information on how survey results would be used as well as perceived survey consequentiality affect preferences. It is observed that neither the information about, nor the perception of survey consequentiality impact preferences. Overall, the results of this dissertation suggest the design of the survey, in general, and the complexity of the DCE, in particular, might influence preferences. This emphasizes the need to further identify and accommodate heterogeneous decision making in DCEs and discrete choice models.