Inflation Then And Now | Bianco Research

Comment

In the last 30 years, there have been a number of changes in the calculation of CPI:

1981

Began outlet and item sample rotation
Began systematic replacement of outlets and their item samples between major revisions
Implemented new Point-of-Purchase Survey (POPS)
Selected retail outlets with probability proportional to consumer spending therein
Eliminated reliance on outdated secondary-source sampling frames
Began rotating outlet and item samples every 5 years
Began rotating one-fifth of the CPI pricing areas each year

1983

Introduced rental equivalence concept (January 1983 for the CPI-U; January 1985 for the CPI-W):
Introduced the flow-of-services method, which removes the investment component from homeowner indexes
Discontinued the asset-price approach, which treated the purchase of a home as a consumer good

1987

Used weights from the 1982-84 Consumer Expenditure Survey and the 1980 census
Updated samples of items, outlets, and areas
Redesigned the CPI housing survey
Improved sampling, data collection, data processing, and statistical estimation methods
Initiated more efficient sample design and sample allocation
Introduced techniques to make CPI production and calculation more efficient

1990s

Improved the housing estimator to account for the aging of the sample housing units
Improved the handling of new models of vehicles and other goods
Implemented new sample procedures to prevent overweighting items whose prices are likely to rise
Improved seasonal adjustment methods
Initiated a single hospital services item stratum with a treatment-oriented item definition:
Discontinued pricing of the inputs to the index for hospital services

1998

Weights from the 1993-95 Consumer Expenditure Survey and the 1990 census
Updated geographic and housing samples
Extensively revised item classification system
Implemented new housing index estimation system
Used computer-assisted data collection
Added the Telephone Point-of-Purchase Survey (TPOPS)

1999

Initiated a new housing survey based on the 1990 census
Estimated price change for owners’ equivalent rent directly from rents
Began using a geometric mean formula for most basic indexes
Extended the use of hedonic regression to estimate the value of items changing in quality
Directed replacement of sample items in the personal computer and other categories, to keep samples current
Implemented 4-year outlet rotation to replace the 5-year scheme
Began within-outlet item rotation for prescription drugs and other item categories

2002

Implemented biennial weight updates
Separated weight updates from major revisions to keep weights as current as possible
Increased sample size of the Consumer Expenditure Survey, so that CPI weights can be based on just 2 years of data
Added the Chained Consumer Price Index for All Urban Consumers (C-CPI-U) (August 2002)

2003

Began computer-assisted data collection for the Commodities and Services Survey
Expanded collection of price data to all business days of the month

2007

Began publishing indexes to three decimal places (Jan. 2007)

For those interested, the complete list of changes can be found on pages 8-11 of this PDF.

Today’s Methodology Through Time

Because this index has morphed so many times throughout the years, one complaint often lodged against the BLS is that inflation rates today cannot be compared to inflation rates from, say, 1980. In an attempt to offer a more apples-to-apples comparison of historical inflation, the BLS maintains a series known as the CPI-U Research Series. It reverse engineers the CPI based on today’s index weights and methodology all the way back to 1978.

The chart below shows the CPI-U Research Series in blue and the normal measure of headline CPI in red. Because the last major revision to CPI occurred in 2007, these two measures only materially differ prior to that. If the BLS would have used the current index weights and calculations back in 1980, CPI would have peaked at 11.8% as opposed to 14.8%. So, according to this measure, the current construction of CPI favors lower inflation rates.

Below the same methodology is used to create a similar series for Core CPI. Again, if the current weights and methodologies used today were used in 1980, inflation would have been much lower. The Core CPI-U Research Series peaked at 9.9% in December 1980 as opposed to the regular Core CPI peak of 13.6% in June 1980.

1980’s Methodology Brought Forward

But what if we wanted to know what the inflation rate would be today based on the index rules used in the past? To get this answer, we must turn to Shadow Government Statistics. This website, run by economist John Williams, calculates CPI using the same methodology used in 1980.

As Williams explained in his October 2005 newsletter:

By stripping current CPI reporting of all the methodological shenanigans of the last three decades that were aimed at artificially reducing reported inflation, one might expect that the resulting series would match the proposed SGS Standard CPI fairly closely. While the patterns of annual growth tend to match reasonably well, annual inflation rates appears to trend somewhat higher for the SGS Standard series than the cleansed CPI series, as had been suggested by some subscribers. The reasons for this will be discussed next month, but the differences appear to be tied largely to BLS misestimates of the impact of some of its methodological changes over time.

Based on detailed BLS estimates of the impact of its methodological changes over time, published in the Monthly Labor Review, June 1999 (“Consumer Price Index research series using current methods, 1978-98,” by Kenneth J. Stewart and Stephen B. Reed), SGS has reworked the CPI-U series to remove the impact of the various methodological changes. The following two graphs show the results of that analysis.

<http://www.shadowstats.com/alternate_data/inflation-charts>

A growing divide exists between the level of inflation based on the 1980’s methodology and that of the currently reported inflation measure. The adjustments and revisions made to CPI throughout the years have largely resulted in lower rates of inflation being reported. As a result, a 2% to 3% year-over-year change in inflation, which sounded tame in 1980, could be problematic today.