Two very similar contingent valuation surveys eliciting willingness-to-pay (WTP) to avoid eye irritation, gastroenteritis, and coughing episodes due to seawater pollution were conducted on visitors to beaches in Portugal and Costa Rica. Various forms of the hypothesis regarding the transfer of mean WTP between the two countries were rejected, as was the hypothesis that model parameters were drawn from the same pooled sample across countries for three different illness episodes. When compared to on-site studies in Costa Rica, benefit transfer from Portugal leads to errors typically of the order of 100 per cent. Adjusting WTP for declared income or other easily accessible socio-demographic variables does not reduce transfer error. This study shows that transfer of health benefit estimates can be potentially quite unhealthy for policy analysis, questioning whether the time and resource savings are justified in this particular transfer context.