We investigate three sources of bias in valuation methods for ecosystem risk: failure to consider substitution possibilities between goods, failure to consider nonseparability of ecosystem services with market goods, and failure to consider substitution possibilities between ecosystems. The first two biases are known in the literature, and we offer insight on the size of the bias for a specific example. Our work on spatially transferable risk is novel. We develop the concept and show how it may undermine typical invasion prevention strategies. We find three key results. First, partial equilibrium estimates of welfare loss are significantly overestimated relative to general equilibrium estimates. If ecosystem services and market goods are substitutes the partial equilibrium bias is greater than if they are compliments. Second, well-intended policies do not necessarily reduce overall risk; risk reduction actions can transfer risk to another time or location, or both, which may increase total risk. Third, policies of quotas and inspections have to be extreme to improve welfare, with inspections having advantages over quotas.