The usual implementation of contingent valuation (CV), in the context of priorities setting for allocation of public funds in health care, is to develop as many surveys as there are programmes, i.e. to perform separate evaluations (SE). In the EuroWill project, three health programmes (for heart disease, breast cancer and a service of helicopter ambulance) were however simultaneously evaluated, i.e. a joint evaluation (JE) was performed. The paper examines the issue of the econometric techniques that should be used to estimate WTP values obtained in the context of JE by comparing the application of independent OLS regressions for each programme versus simultaneous estimations using seemingly unrelated regressions (SUR) on data of the French EuroWill survey. It shows that separate estimations may lead to misspecifications because they cannot take into account that JE exogenously provides a reference structure to the respondent which affects the estimates of WTP for each programme. Therefore, the potential advantage of JE versus SE as an elicitation technique in CV studies applied to health care (to better control the referents used by respondents for evaluating different programmes) only holds if simultaneous rather than independent techniques are used in the estimation of WTPs. Copyright © 2002 John Wiley & Sons, Ltd.