RESUMEN
The estimation of economic benefits of health effects through direct methods, such as contingent valuation, presents the problem of preference imprecision. This paper deals with this problem by proposing an elicitation method that allows the subject to state an interval for willingness to pay, without inducing any specific amount as a response. The paper also analyzes the effects of the context where changes in health occur on the associated imprecision level and the estimates, by comparing a situation without context with another in which effects are due to atmospheric pollution. The econometric modelling develops a Bayesian estimation method for censored intervals, which models the existing uncertainty between the lower and upper limits derived from the elicitation process. Results prove that data dispersion is significantly higher for the non-contextual scenario, and increases for the most severe symptoms.