Data limitations frequently prevent using actual consumer behavior in determining natural resource values, so stated preference methods are used. Whether value estimates show sensitivity to the scope of resource valued is a key test for their validity, which several studies fail. Developing a correlated panel mixed logit model of households' water quality valuations in California, we show that false negatives in scope tests can result when individual preference variation and correlation are ignored and split-sample comparisons are used. Monte Carlo simulations further demonstrate potentially prevalent false rejections of scope sensitivity even when within-subject comparisons of willingness to pay portray strong scope sensitivity. Copyright © 2013 John Wiley & Sons, Ltd.