What Determines Prediction Errors In "Benefits Transfer" Models?

Abstract

The aim of BT techniques is to provide decision makers with a monetary valuation of environmental goods and service in a cost-effective and timely manner, since original valuation studies are both expensive and time-consuming. Demands for environmental valuation estimates are rising in the policy community in both Europe and the US. In the UK, widespread use of benefits transfer has already occurred within policy making and regulatory bodies, for instance in the setting of water quality targets for private water companies and in the design of agrienvironmental policy. An important question is how big the errors are resulting from this practice, and how sensitive transfer errors are to how the benefits transfer is conducted. In this study we employ a choice experiment study focusing on the value of landscape attributes in four upland farming regions of England to investigate the sensitivity of transfer error to procedures. This is done using an experimental design with the same set of attributes and levels applied in four different regions of England. The main findings to emerge are that transfer errors depend on the choice of study site at which original valuations are sought, on whether a single site or pooled model is used, and on whether a mean value or benefit function transfer is used. Large variations in transfer errors are found to be related to these choices.