Re: [PEN-L:9478] geometric-mean CPI

[Peter Dorman <dorman@xxxxxxxxxxxxx>]

One point should be made concerning the geometric-mean CPI.  Press
reports of the BLS briefing made it seem as though a new, more
technically accurate method was being used which, on being applied,
revealed a quarter-point lower inflation rate.  I will withhold judgment
(for now) on the merits of the geometric mean approach, but anyone who
has worked with alternative statistics knows that the geometric mean
*automatically* gives you a lower number than the arithmetic mean.  It
comes right out of the algebra.  So the issue is not, what does this
approach "show" compared to what old approach showed, but what is the
justification for using a geometric mean approach in the first place?
If I understand the matter correctly, the case for a geometric mean
(either at the detailed level, as with the experimental CPI, or at
higher levels as the Boskin posse wanted) is ultimately axiomatic.  It
is based on hypotheses derived from utility theory, not on empirical
research.

Peter Dorman