Last month I called attention to an analysis by BLS researchers John Greenlees and Robert McClelland of some of the claims by John Williams of Shadowstats about the consequences for reported inflation of assorted technical decisions made by the BLS. Williams asked me to update with a link to his response to the BLS study. I am happy to do so, along with offering some further observations of my own.
You can follow the link to Shadowstats’ response to Greenlees and McClelland and judge for yourself, but my impression is that the response is more philosophical than quantitative. In a separate phone conversation, Williams further clarified the Shadowstats methodology. Here’s what John said to me:
I’m not going back and recalculating the CPI. All I’m doing is going back to the government’s estimates of what the effect would be and using that as an ad factor to the reported statistics.
I had formed the mistaken impression that Williams was indeed trying to go back and recalculate measures such as the CPI based on a retrospective application of the historical BLS methodology. I found the specific quantitative results provided by Greenlees and McClelland to be convincing demonstrations that this could not be the case. I take further comfort in the understanding that Williams agrees that his numbers indeed do not represent the outcome of such a procedure.
Technorati Tags: CPI,
the response is more philosophical than quantitative.
This is such a comical statement! The BLS states currently that:
the CPI’s objective is to calculate the change in the amount consumers need to spend to maintain a constant level of satisfaction.
Which is not just “philosophical” instead of “quantitative”, but outright delusionary.
How can the BLS measure “quantitative”ly a “constant level of satisfaction” across time and across a population? Even philosophy does not succour here.
So the BLS approach is not actually to measure that level of satisfaction, but even more delusionarily to *assume* that given changes in specific goods correspond to specific changes in the level of satisfaction that they provide (and of course, even more laughably, “coeteris paribus”).
The BLS used to use a less delusionary methodology:
“The best measure of inflation for a given application depends on the intended use of the data. The CPI is generally the best measure for adjusting payments to consumers when the intent is to allow them to purchase, at today’s prices, the same market basket of consumer goods and services that they could purchase in an earlier reference period. It is also the best measure to use to translate retail sales and hourly or weekly earnings into real or inflation-free dollars.”
From: Understanding the Consumer Price Index:Answers to Some Questions US Department of Labor/ Bureau of Labor Statistics, July 1996 (Revised)
But then more on this in previous comments here:
Is it possible to get the raw data used for cpi, employment numbers?
Then it would be possible for anyone to test varying
The issue is, fundamentally, what the price index is supposed to measure.
Historically, the CPI was a fixed-weight price index. That is, the “market basket” of goods to be priced (and from which the index was calculated) was established, and the “weights”–the importance–of each of the goods in that beasket remained the same until the market basket was adjusted (which occurred roughly once every 10 years or so).
The problem with a fixed-weight index is that the weights can, and do change. With the CPI, briefly, two problems:
1) As incomes change, people’s purchases of different goods do not change in proportion. So, during periods of rising incomes, the class of goods on which spending rises faster-than-income actually becomes underweighted as time goes by–these goods are more important in actual spending than they are in the index.
2) As relative prices change (e.g., the price of chicken rises relative to the price of beef), people will tend to substitute toward the good whose relative price has declined. This means, again, that weights change in people’s consumption decisions, but not in the index. To be sure, such substitution can be shown to leave people worse off than they were at the old prices. But suppose the price of chicken rises relative to the price of beef, and people adjust their consumption of chicken (less) and beef (more). And now suppose that the price of chicken falls relative to the price of beef–consumers will adjust again. But, compared to how well off they were after adjusting to the new prices, this change will also make consumers worse off. (Odd, but true.)
So I don’t know how one wishes to evaluate the sorts of changes than accompany changing weights in the price index. When things change that affect people’s consumption decisions, the importance of various goods in their consumption really does change. Should we ignore that in calculating a price index, or try to incorporate it?
I don’t think the motivation is to try to minimize the measured rate of inflation, frankly. And I don’t think there’s an easy, or necessarily correct answer.
I should have made 3 comments, the third being the “new product” issue. As new products are (successfully) introduced, the importance of “old” goods in consumers’ decisions must fall. Using a fixed weight index, that will be incorporated into the price iddex only at fairly long intervals (my recollection is that pocket calculators didn’t make it into the CPI until the 1987 revision). If the prices of successful new products generally fall (as they did for color TVs, calculators, VCRs, personal stereos, personal computers, CDs and CD players, DVDs and DVD players, flat-panel TVs…), then failure to get these goods into the market basket quickly also provides a bias to measures of the general level of prices–and of inflation.
Also the calculation of the Owner`s Equivalent Rent CPI, which is a subcomponent of the HOUSING CPI, is seriously questionable. Remind that this measure is competely imputed and goes up if natural gas decreases…
So, let me get this straight. This guy takes the CPI inflation stats, adds 3.5% to all of them and sells them to the public for $175 per year?
Be fair, James. This is a preliminary response.
I have questioned Williams’s index, too. By his own standard of “common experience and traditional expectations that the CPI measures the cost of maintaining a constant standard of living,” his index implies a decline of living standards during the 1990s not consistent with what is observed. On the other hand, the BLS series cannot explain why Americans have stopped saving and gone so deeply in debt. A slight error in BLS methodology, smaller than Williams claims, could explain that reality. Williams also makes a fair point that there’s a lot of money riding on the calculation of inflation, and government policy makers have a motive to pressure BLS… if perhaps not the means.
The BLS makes a good case for their methodology, and people like AL should at least read the paper before commenting on it. And, no, Joseph, that is not an accurate representation of what Williams does.
Williams adds 7%, not 3.5%. Otherwise, Joseph’s comment is correct. The BLS estimated the Shadowstats numbers by adding 7% to their own numbers. Williams in his response agreed that they were very close to his index.
OK, so maybe 7% or whatever is not a reasonable estimate of the difference between the results of the original and current methodology, but it may be significant that the sum of the effects of the changes on introduction is clearly positive.
Anonymous, that is NOT a fair characterization of what Williams does. He states: “I publish two estimates of alternate CPI growth, based on methodologies in place as of 1980 and as of 1990. The alternate estimates are based on adjusting the published CPI-U for cumulative annual differences in CPI as estimated by the BLS for the impact of its various methodological changes since the early 1980s (or 1990s).”
Taking and applying previous methodology to show how its prediction contrasts with present methodology is NOT the same as robotically adding 3.5% to each year’s estimate. The BLS disputes Williams’s calculation of the magnitude of the effect, but they are NOT talking about just adding 3.5% to the value.
John Williams’ work has been very beneficial in encouraging businesspeople and investors to dig deeper into and better understand government statistics. There should be public debates about the methodology behind these statistics, conducted not just among the wonks who do the calculations for a living, but among those who depend on these numbers for business, investment, and policy decisions.
I took John’s reply to simply indicate that he is now aware of the BLS paper, and will prepare a detailed response. Shame on the BLS authors for not providing John with advance notice of the paper and perhaps a chance to serve as a reviewer.
If the CPI were used solely to estimate how high government checks need to be to allow, for example, Social Security recipients, to maintain “constant level of satisfaction”, or perhaps a level of satisfaction in line with the rest of society (as standards of living fall) then the current CPI methodology may be fine.
However, the Federal Reserve uses CPI in guiding monetary policy. Clearly, the Laspeyres index is the best measure for the effect of monetary policy on prices. Substitution and other forms of evasive action taken by consumers does not make a loose monetary policy less loose.
Also, the GDP deflator also is not a Laspeyres index, and allows for substitutions based on consumer behavior. GDP should be deflated based on the lost purchasing power of the dollar, as reflected in changes in actual prices. The understating of inflation in turn causes real GDP to be overstated and misleads policy makers by, for example, masking the detection of recessions. Using correctly calculated inflation indices, policy makers would have known the US has been in recession for some time, and could have been more prepared for the current economic turmoil.
Calculating inflation indexes is too important to be left an insular group of government funded macroeconomists.