The BLS has just published its response to its critics, especially John Williams at Shadow Government Statistics. The BLS writes:
One widely cited alternative index is based on an estimate that changes to the CPI since 1983 have lowered its growth rate by at least 7 percentage points per year. The use of the geometric mean alone is stated to have lowered the CPI growth rate by 3 percentage points, and other BLS changes, such as the use of hedonic models and OER, supposedlyhave lowered the growth rate by an additional 4 percentage points.
Each of these estimates of the impact of BLS changes is inconsistent with the empirical evidence. As noted earlier, the BLS has computed indexes showing that the use of the geometric mean formula has reduced the growth rate of the geometric mean of the CPI by only -0.28 percentage point per year, not 3 percentage points. Also discussed earlier, BLS analyses have shown that if the implementation of hedonic adjustment models since 1999 has had any net downward effect, it is very small.
I looked at the Williams site, and I honestly have no idea where he gets his numbers. Can anyone enlighten me?