UC Berkeley

We address how using different censoring thresholds and imputation procedures affects the baseline results of Coibion, Gorodnichenko, and Hong (2015). Higher censoring thresholds introduce measurement error and outliers that generate wide variability in results across weighting schemes, but methods that explicitly control for outliers confirm the results of Coibion et al. (2015) for all censoring thresholds. We also illustrate how the BLS's approach to imputing missing prices can introduce a cyclical bias into measures of posted price inflation when store-switching is present in the data.