Analysis of Divergence between Chained CPI‐U and Regular CPI‐U for the All_US‐All_Items Indexes (2000‐2002

Owen J. Shoemaker


In February 2004, the BLS calculated and published its third annual set of C-CPI-U indexes for the 12 months of 2002. The C-CPI-U (Chained Consumer Price Index - Urban) is calculated and published every year, with a one-year lag, using a Tornqvist formula, and its set of weights are updated yearly so that a unique set of monthly weights are available for both time t as well as for time t-n. The C-CPI-U can thus be labeled a "superlative" index. By contrast the Regular CPI-U uses weights that are, at a minimum, at least two years old, and uses a combination (hybrid) of Geomeans and Laspeyres formulas as its final estimator. The set of All_US-All_Items Chained C-CPI-U index results continue to diverge (lower) from Regular CPI-U index results. We investigate the nature of this divergence. We also analyze the two different weight structures, possible response biases, and the standard errors that we calculate for these indexes.