Reliability of Meta-analytic Benefit Transfers of International Value of Statistical Life Estimates: Tests and Illustrations

Abstract

If there are no applicable domestic studies, there are many ways to utilize the international literature to conduct benefit transfer (BT). In the health economics literature simple unit transfer methods, rather than function-based methods, are the most commonly used. In this chapter we utilize a large database of Value of a Statistical Life (VSL) estimates, derived from stated preference studies worldwide, to investigate the reliability of meta-analytic BT (MA-BT) and compare this method with simple unit transfers in a case study illustration. Meta-regression analysis is a way to estimate how different policy-relevant factors affect VSL and is thought to improve accuracy in BT. We discuss in particular how different quality criteria to screen available studies and VSL estimates may influence BT accuracy. Results show that quality screened MA-BT models give lower transfer errors, and in the case study example MA-BT methods achieve accuracy gains over the use of unit transfer methods. However, the unscreened MA-BT method achieved around the same accuracy as the best unit transfers based on quality screened data. Hence, transfer accuracy may in some contexts depend as much on the quality of the underlying data as on the BT method itself. Many countries have a relatively small body of national valuation literature to use as a basis for benefit transfer (BT). A solution to this problem is to expand the information base to include relevant international valuation literature of acceptable quality, and then use a BT method to derive and transfer a suitable welfare estimate to the national context in question. For environmental goods, such as water quality, wetlands, coral reefs, forest conservation benefits etc., meta-analytic benefit transfer (MA-BT) has become an increasingly common method, at least for academic investigations of reliability (e.g., Brander et al. 2012; Johnston and Thomassin 2010; Lindhjem and Navrud 2008; Londoño and Johnston 2012; Stapler and Johnston 2009). Although the evidence is mixed, for the international context at least, the emerging consensus seems to be that such function-based transfers out-perform unit value transfers (Kaul et al. 2012; Rosenberger and Stanley 2006). For some reason, and as pointed out by Johnston and Rosenberger (2010), the health economics literature deriving value of statistical life (VSL) estimates from stated or revealed preference studies seems to be more doubtful of the potential reliability gains from function-based benefit transfer. Both the academic literature and many international agencies, such as the World Bank, still emphasize simple unit value transfer, typically adjusted only by gross domestic product (GDP) dif-ferences between countries. To our knowledge, there are very few examples of function-based transfers in the health economics literature. Brouwer and Bateman (2005) is a first application of a standard function-based transfer of willingness to pay (WTP) for health risk reductions (rather than VSL directly), while Dekker et al. (2011) uses a Bayesian meta-model of VSL for BT. Preceding Dekker et al., OECD initiated a project compiling a large database of VSL estimates from stated pref-erence studies worldwide, resulting in several preliminary reports and two final publications using meta-analysis (see Lindhjem et al. 2011; OECD 2012).