Public Choice for Flood Defence

Abstract

Why do we want to value the environment? Environmental assets provide a flow of goods and services over time which benefit mankind. Valuing these services contributes towards their protection and enhancement, however many of these benefits cannot be valued in traditional markets and as such rely on non-market valuation techniques. One of these is contingent valuation (CV) which directly asks respondents whether they are willing to pay for an improvement in the good or service. This thesis seeks to explore methodological issues associated with this method by undertaking a CV survey to elicit willingness to pay (WTP) for a new type of flood defence (managed realignment) on the Tay Estuary, Scotland. One challenge for survey designers is to provide high quality, readily understandable information to mitigate bias in WTP estimates. This thesis contributes to the information provision literature by examining whether prior knowledge or new information has a greater effect on the WTP estimate when controlling for respondent experience and familiarity with the good. A field experiment was designed to test for respondent’s prior knowledge; allow for varying levels of information to be presented to respondents and identify information acquisition for each respondent. Specifically tested was the notion that respondents who learn the most about the good during the survey process will have a more robust WTP estimate. Results were mixed: a causal relationship between information provision and learning was established with respondents in the higher treatment groups scoring higher in the second quiz. However, there was no relationship identified between prior knowledge, information provision and WTP. Personal motivations were the strongest predictors of WTP: those who were most concerned about flood risk and who lived closest to the proposed flood defence were willing to pay the most. A second issue in CV is consequentiality. Carson and Groves (2007) argue that for a survey to produce meaningful information about respondent’s preferences the respondent must view their responses as potentially influencing the supply of the public good. This thesis seeks add to this relatively new literature by exploring the observable factors which may influence respondents perceived consequentiality; specifically the effects of familiarity and information. Respondents were asked to state how confident they were that the results of the survey would be used by policy makers on a Likert scale ranging from “very unconfident” through to “very confident”. Results conformed to the Carson and Groves knife edge result: consequential respondents had significantly different WTP distributions compared to inconsequential and unsure respondents and were willing to pay significantly more towards the scheme. Consequential respondents also conformed the theoretical considerations of construct validity whilst inconsequential respondents did not. Respondents with more prior knowledge also appeared to be more likely to perceive the survey as consequential, although this was not consistent across all treatment groups. There is a concern that WTP and consequentiality are endogenous: respondents who want the policy to go ahead may be more likely to state the survey is consequential and state a high WTP in the hope these responses combined contribute to the policy maker’s decision. From a policy perspective the high level of support for the new scheme was encouraging and in contrast to previous findings on preferences for managed realignment. From a flood risk management perspective a “miss-match” between actual and perceived flood risk was highlighted, with many respondents stating they were not at risk from flooding when they in fact were. This is potentially concerning as respondents may not be taking adequate steps to protect their home from future flood risks. Overall it is recognised that values derived from the CV survey form one small part of the planning process and while informative, the decision for a scheme to take place should not be based on these values alone.