Comparison of Variance Estimation Techniques for a Price Index with Non‐Independent Weights

Richard W. Toftness


We present an economic approach to variance estimation of price indexes, applied to import and export trade. Instead of generating weights and price changes independently, we generate one total share value for each period for each industry, using a Gamma model of log trade dollar values, and then we calculate base period weights and quantities based on the Fisher model. To generate price changes, a spike at 0% is estimated; then positive and negative price changes are modeled separately, with the model chosen depending on the industry. The variance estimation methods explored are: the Ratio Biased, the Taylor Linearized, the Stratified Randomly Grouped, and the Stratified Jackknife.