Comparison of Variance Estimation Techniques for a Price Index With Dependent Weights

Richard W. Toftness

Abstract

The International Price Program (IPP) of the Bureau of Labor Statistics (BLS) measures aggregate price changes for samples of U.S. exporters and importers of agricultural goods, industrial supplies and materials, capital equipment and machinery, consumer merchandise, and transportation services. These measures are used by the U.S. Department of Commerce to adjust the monthly international trade figures and the quarterly National Income and Product Accounts for inflation or deflation (BLS 1997). These adjustments, also known as deflators, take the form of price indexes of the Laspeyres type, which means that as prices change over time, quantities are left at their original level determined at a base period. Estimating the variances of price indexes is desirable as a measure of accuracy and stability, but is difficult due to the nonlinearity of the function and in the IPP's case due to a complex sample design.Therefore several variance estimation techniques need to be examined and evaluated. The length of time from the "previous" period to the "current" period is important because some customers of IPP data are interested in short-term changes, such as monthly or quarterly percent changes, and some customers are interested in long-term changes, potentially of several years. The length of time chosen for IPP variance estimation is twelve months, which is useful to both short-term and long-term customers.