Sampling and Weighting of Commodity and Service Units for the Elementary Level of Computation of the U.S. Consumer Price Index

Mary Lee Fuxa


Elementary-level indexes are weighted and aggregated to produce the all-items U.S. Consumer Price Index each month. An elementary-level index is computed for each combination of an item stratum and index area. Given an item stratum and a geographic sample unit representing an index area, the probabilities used to select specific items and the outlets that sell them are not always equal to the desired probabilities that would lead to (near) unbiased estimation of targeted indexes. In this case, the sample is "imbalanced" -- the expected sample proportion for a sample unit is not equal to what it should be. In the estimator for the elementary-level index, weighting factors are used to correct the imbalance. This paper shows the construction of the weighting factors.