The Consumer Expenditure Interview Survey collects data about household purchases using a series of filter questions. Many researchers have worried that respondents underreport purchases to decrease the length of the survey, but none have tested this hypothesis experimetnally, because the cost and concerns about disruption to the data trend. Using a parallel web survey and multiple imputation, I estimate the size of the misreporting effect in the first wave of the Consumer Expenditure Survey without experimentally manipulating questions in the survey itself. Results suggest that respondents underreport purchases in the first wave of the survey, but the effect does not grow over waves. The approach used here could be applied to other large on-going surveys that lack resources for methodological experiments.