The Dec 1996 Boskin report (May 1997 AER summary)
concluded that the change in the Consumer Price Index (CPI) overstates the change in the cost of living by about 1.1 percentage points per year (the range of plausible values is 0.8-1.6 percentage points) (1997, p. 78).
Breakdown (1997, p. 80):
My question: What, if any, are some more recent and similarly-"authoritative" estimates of CPI bias?
Redding and Weinstein in their paper "A Unified Approach to Estimating Demand and Welfare" develop a unified price index that nests all existing major price indexes. Existing price indexes can be thought of as arising from the imposition of parameter restrictions on their unified index.
Their examination with barcode data suggests that standard methods of measuring welfare overstate cost of living increases by ignoring new products and demand shifts. However, contrary to the Boskin report, they do not put an overall number on the bias.
There are two ways to see the CPI: (1) A measure of price changes for a weighted basket of consumer goods (2) A measure of the change in cost of obtaining a fixed level of utility (cost-of-living)
If you interpret it as the second one, then there are a number of crucial structural assumptions you need to make, mostly concerning the importance of substitution effects. Also, you may still mismeasure cost-of-living for different demographic groups. Here is an interesting recent paper on this topic. http://scholar.harvard.edu/xavier/publications/unequal-gains-product-innovations-evidence-us-retail-sector
My preferred interpretation of the CPI is the first one. The CPI is a very simple aggregate statistic and this is why it is useful. Of course it does not correctly capture the evolution of cost-of-living for every demographic and under every assumption, but this is an unreasonable requirement for a single index.
