Using the Retail Distribution of Sellers to Impute Expenditure Shares

Many price indices must be constructed without quantity data at the elementary level. We show that for some consumer goods in the United States and other countries, one can approximate expenditure shares using weights derived from the retail distribution of sellers. These weights are based on the share of outlets selling an item, or the share of outlets adjusted by the total number of items sold in each. Relative to using no weights, we find that using such imputed weights substantially reduces bias in the frequency of price changes, in annual inflation, and in price comparisons across countries