We elicit willingness to pay for different types of consumption goods, systematically manipulating irrelevant anchors (high vs. low) and incentives to provide true valuations (hypothetical questions vs. Becker-DeGroot-Marschak mechanism). On top of a strong hypothetical bias, we find that anchors only make a substantial, significant difference in the case of hypothetical data, the first experiments to directly document such an interaction. This finding suggests that hypothetical market research methods may deliver lower quality data. Moreover, it contributes to the discussion examining the mechanism underlying the anchoring effect, suggesting it could partly be caused by insufficient conscious effort to drift away from the anchor.