A failure in the measurement of inflation: results from a hedonic and matched experiment using scanner data

Statistical offices use the matched models method to compile consumer price indices (CPIs) to measure inflation. The prices of a sample of models are recorded, and then price collectors visit the same stores each subsequent month to record the prices of the same matched sample of models. The matched models method is designed to control for quality changes. But new, unmatched models launched in subsequent months have their prices ignored as do old unmatched models no longer available. The paper uses retailer's bar-code scanner data on several consumer durables to show that serious sample degradation can take place and that the quality-adjusted prices of unmatched items differ from those of matched ones, leading to substantial underestimates of inflation. Hedonic indices use the whole sample. They are argued to be more useful to price measurement in markets with a rapid turnover of models in order to avoid the demonstrated bias.