Marshall B. Reinsdorf

Using Scanner Data to Construct CPI Basic Component Indexes

This article considers how scanner data could be used in constructing component indexes for the U.S. Consumer Price Index. One product, coffee, in two cities generates over 1.8 million observations in just over two years, so coping with the sheer volume of data would be a challenge. Some other findings are (1) some aggregation of prices into unit-value averages is necessary for practical reasons and to avoid bias, (2) chained Laspeyres indexes are very high, (3) modified Laspeyres indexes have some upward bias but much less than a true Laspeyres index, (4) Fisher ideal or modified Edgeworth indexes perform well, and (5) aggregating prices across outlets to form city-level unit values reduces the discrepancies between index-number formulas.

An exact price index for the almost ideal demand system

Abstract We show that an exact price index for the Almost Ideal Demand System can be evaluated using data on expenditure shares and prices at two comparison points, and at their geometric mean.