Abstract Sound cost-benefit analysis should acknowledge differences in the spatial distribution of cost-bearers, environmental effects and beneficiaries. Where the first two are often well-known by policymakers, identifying the area of affected beneficiaries through a common spatial distribution of values is still under debate. Using general rules for the spatial distribution of values has obvious appeal for cost-benefit analysis. With a five-country contingent valuation dataset of water quality, we study the performance of international benefit transfer at different spatial scales, making use of the EU regional statistics for NUTS 1, NUTS 2 and NUTS 3 levels. Unit value transfers yield the smallest transfer errors on average. For function transfers, spatially explicit models yield lower transfer errors. However, caution should be exercised in choosing a proxy for substitutes, as the choice of an intuitive proxy can cause unintuitive predictions. The choice between the NUTS 2 and 3 regional level statistics induces, on average, almost no difference in transfer errors when used as policy site data. However, a blind choice of transfer function form can have large effects on aggregate WTP estimates on the national and regional level when significant non-use values are present.