The embedding effect represents one of the most challenging problems of the contingent valuation (CV) method. The purpose of this study is to address the issue of embedding effect in the context of museums, by estimating the size and extent of the problem. A CV study conducted in Denmark asked respondents to evaluate both the Danish museum-system (inclusive good) and the specific local museum in their municipality of residence (embedded good). Typically, an equal or very close willingness-to-pay (WTP) for the two goods would indicate the presence of an embedding effect. However, we show that this will only be the case for some respondents, while for others the similarity between the two WTP reflects an exclusive interest for the local museum. We demonstrate the existence of these two different types of respondents by using a latent class model where respondents are clustered based on answers to the CV survey. Consequently, the aggregate WTP and the bias associated with the embedding effect are estimated. Our results have implications in terms of the reliability of CV studies and its implementation in actual policy issues, demonstrating that the right questions and techniques will facilitate the separation of ‘true’ answers from biased ones.
