The Consumer Expenditure Survey is a large, federal data collection effort and contributes to the calculation of the U. S. inflation rate. For years, researchers have worried that the repetitive structure of the questionnaire leads respondents to underreport purchases. However, evidence for this underreporting is weak. I use a unique approach to estimate underreporting in the Consumer Expenditure Survey. Using a parallel web survey and multiple imputation, I estimate the size of the misreporting effect without experimentally manipulating questions in the survey itself. Results suggest that household purchases are underreported by 5 percentage points in the first wave of the survey. The approach used here could be applied in other large surveys where budget or logistical constraints prevent experimentation.