"Studies comparing household surveys with on-site interceptor surveys have typically accounted for over-sampling avid users in the on-site interceptor surveys (that is, endogenous stratification). However, these studies have typically not accounted for the possibility that the household sample may contain a large presence of zero observations. If a large proportion of the population does not recreate at the site for any value of the price vector, this inflation of zero observations leads to biased welfare estimates and an inadequate comparison with the on-site survey. In this paper, the authors estimate and compare three models which correct for these measurement issues in both the household and on-site surveys. Results from an application to recreation at Lake Sevan (Armenia) indicate that household consumers' surplus is not statistically different from that of the on-site survey once the authors account for zero-inflation in the household sample and endogenous stratification in the on-site sample. "--World Bank web site.