Income effects reported in contingent valuation (CV) studies tend to be much smaller than those found in the literature on collective choice. This disparity has received surprisingly little attention by environmental economists. The present study uses meta-analysis to explore determinants of the presence/absence of a significant income effect in a sample of CV surveys. The probability of significant income effects–controlling for the statistical power of the reported tests–was higher when ‘progressive’ payment vehicles were used and tended to be lower when cost distribution and institutions were well defined, when the choice was formulated as a policy referendum, or when ‘passive-use’ goods were involved. A simple explanation of this pattern in terms of respondent behaviour suggests that the low income effects in contingent valuation surveys may be an artefact of the survey method. Since empirical estimates of the income effect may have important policy implications, this issue deserves attention in future research.