Contingent valuation (CV) or defensive behavior data is often used to estimate the economic value of water quality. Although combining these data (i.e., stated and revealed preferences) mitigates the potential bias from using either type of information, the costs of collecting both could overwhelm the benefits. We attempt to find a convenient estimation method by using a proxy indicator for revealed preferences in the analysis of stated preference data. Specifically, this study explores the effect of individuals’ reported defensive behavior on their stated preferences for groundwater quality. Logit models based on random utility theory were estimated using referendum CV data at household level collected in Maine, US. The results suggest that failure in accounting for defensive behavior in the valuation could result in a bias willingness to pay estimate for groundwater quality. We also found that the monetary value for groundwater quality was small, even though subsoil water constituted an important drinking water supply in the survey period. The results also revealed that respondents’ averting behavior were mainly influenced by their perception of groundwater quality. Implications of our findings for welfare analysis are discussed.