Combining averting behavior and contingent valuation data:: an application to drinking water treatment in Brazil

This paper estimates WTP for drinking water quality in Brazil by combining averting behavior with contingent valuation data. Using bivariate probit models, alternative structures allowing for heteroscedasticity between and within data sources are incorporated by taking advantage of the different information content that characterizes each data source. We look at two covariates not yet examined in the literature when combining stated and revealed preferred data to explain the variance in the models: income and the bid in the contingent valuation questionnaire. Tests for parameter equality across data sets are performed. The results suggest that the specification of heteroscedasticity has a significant impact in WTP estimates and is crucial to legitimate the combination of data sets from different origins. The significant differences found in WTP between the two sources are discussed.