September 10, 2008

__________ Response to BLS Article on CPI Misconceptions

The Bureau of Labor Statistics (BLS) published "Addressing misconceptions about the Consumer Price Index" in its August 2008 Monthly Labor Review, an article by John S. Greenlees and Robert B. McClelland that attempted to debunk "a number of longstanding myths regarding the Consumer Price Index [CPI]."

Where there are many of critics of the BLS’s CPI methodologies and reporting, the government has a natural and significant interest in protecting the credibility of its CPI series.  In fairness, much of what has become distorted in the CPI series has resulted from pressures outside of the Bureau itself, ranging from the perceived political needs of overseeing administrations, to Congress and the Federal Reserve. For some of the issues raised in the article, I have no argument with the BLS. For example, as to core inflation [pages 11 and 12], the concept of "core" inflation — CPI net of food and energy inflation — indeed is not used in current Social Security cost of living adjustments. The BLS went on to explain that it "makes no claims about the predictive or analytical value of that index."

The article, however, also addressed issues directly tied to my writing and research, and where I otherwise have added fuel to the public controversy. Those areas are at least touched upon here, and most have been discussed previously in the SGS Primer Series on CPI, to which this text will be added as a supplement, as well as regularly in SGS newsletters. I stand by and am extremely comfortable with my previously expressed positions and published numbers. The following comments have been prepared quickly, in response the just-published article, of which I had no advance notice. More detailed discussion will be covered in a later, revamped Primer Report on the CPI.

As to the SGS-Alternate Consumer Inflation series I publish (the article’s authors miscalculated the ten-year inflation rate), I shall make available to the public in the not-too-distant future the detailed series and calculations of same in collaboration with someone in academia (a search is in progress), who will have the opportunity to review and replicate or to challenge the alternate CPI data I have been publishing, and who will be free to publish those findings with peer review.

As always, questions and comments are invited:

The Traditional CPI Concept Has Been Politically Mauled. At the heart of the differences over CPI reporting is the way CPI is viewed or defined. My basic approach to looking at CPI inflation is from the standpoint of common experience and traditional expectations that the CPI measures the cost of maintaining a constant standard of living, that reflects costs out of pocket to get a products or services in hand, not some nebulous benefits estimated by the BLS of having to pay for an expensive new gasoline additive when filling a gas tank.

The reason for the preceding is that inflation measures commonly are used as an indication of how much income has to increase, in order for living standards to be maintained, or of how much return is needed on an investment in order to stay ahead of inflation. Changing BLS methodologies have caused CPI inflation reporting to stray sharply from those needs

I contend that most people view inflation as being much higher than currently reported by the BLS, due largely to those methodological changes over the decades that have moved CPI inflation away from basic, traditional reporting. Changes tied particularly to quality, weighting and definitional issues have moved reported inflation ever further from broad, common experience, with resulting reporting biases in the CPI that usually are to the downside. Consumers have a pretty good sense of where basic inflation stands, and whether or not they are able to make ends meet, let alone maintain a constant standard of living. 

Here are the issues raised in the BLS article that I feel need to be touched upon:

Hamburger versus Steak. From page 5 of the article: "Some critics have incorrectly claimed, for example, that the BLS assumes that consumers are no worse off when they substitute hamburgers for steak."

A later clarification: "To begin, it must be stated unequivocally that the BLS does not assume that consumers substitute hamburgers for steak."

I have never claimed that the BLS "assumes that consumers are no worse off when they substitute hamburgers for steak." Quite to the contrary, I have always argued that those giving such a rationale as to why the CPI purportedly overstated inflation knew very well that consumers were worse off with hamburger, and that the pitchmen were looking to change the concept of the CPI so that it no longer measured the cost of maintaining a constant standard of living. Such clearly was the case in the early-to-mid 1990s and was reflected in official arguments and press of the time. Much of what happened here was forced upon the BLS by the political system, but such is the BLS’s primary client.

I did not invent the example used by the early CPI detractors, who argued that the Index was overstated because it did not reflect people substituting hamburger for steak, when steak got too expensive, or their arguments that reducing a so-overstated CPI would help to cut cost of living adjustments on Social Security and relieve federal budget pressures. I am still searching through boxes of papers to find the earliest reference to same, though I believe I first heard that pitch from then Federal Reserve Chairman Alan Greenspan.

Consider from the New York Times, "Panel Sees a Corrected Price Index as Deficit-Cutter," September 15, 1995, by Robert D. Hershey, Jr.:

"Speaker Newt Gingrich, Republican of Georgia, suggested this week that fixing the [CPI] index, with its implications for lower spending [Social Security, etc.] and higher revenue [tax bracket adjustments], would provide maneuvering room for budget negotiators …"

"Alan Greenspan, chairman of the Federal Reserve, is among the other Government officials who have spoken optimistically about financial benefits of a more accurate [CPI] index …"

"[E]conomists believe one of the most important [CPI upside biases] is when consumers shift their buying patterns in response to changing prices, substituting one product for another. The [CPI] index is based on a fixed market basket of goods and services. But for, for example, if the price on an item like steak gets too expensive, consumers may switch to hamburger."

The Boskin Commission Report, December 4, 1996, actually used steak and chicken for its substitution example. The examples being used to argue for changing the CPI clearly were tied to prices rising and resulting consumer demand shifting to a lower-quality product. Simply put, that is a cost-of-maintaining-a-constant-standard-of-living issue and was a primary consideration of those seeking to change the CPI, although other issues would come into play.

From the San Francisco Chronicle, "Government’s economic data misleading, he says," May 25, 2008, by Sam Zuckerman:

"In the 1990s, for example, Republicans wanted to make changes in calculating inflation along the lines recommended by a special commission, including more use of quality adjustments. By lowering the official inflation rate, such changes promised to reduce the annual cost-of-living adjustments for Social Security and other federal programs.

"[Katherine] Abraham, the Clinton bureau [of Labor Statistics] commissioner, remembers sitting in Republican House Speaker Newt Gingrich’s office:

"’He said to me, If you could see your way clear to doing these things, we might have more money for BLS programs.’"

Related changes to methodologies were made beginning in the mid-1990s. As noted in the 1999 Economic Report of the President, page 93:

"A final reason for the slowing of reported price indexes has been methodological changes to both the CPI and the indexes used in the national income accounts …"

What happened was that geometric weighting (replacing arithmetic weighting) was introduced for narrow product categories in the CPI. The categories were narrow enough to allow weight shifts between different types of steak, but the "mimicking" of the steak versus hamburger or steak versus chicken substitution was not possible based on geometric weightings, since steak, hamburger and chicken all were in different categories.

Nonetheless, if steak prices were to rise, strapped consumers indeed likely would shift to cheaper meats, and such would be reflected in the broader weighting categories. The problem from the BLS standpoint, though, was that those weightings traditionally were recast only every ten years. So, the broad-category reweighting process was accelerated, with a reweighting in 1998, and, thereafter, reweightings were structured formally for every two years starting in 2002. This process moved the CPI closer to a fully substitution-based index.

These approaches, in conjunction with other methodological changes ranging from increased use of discount-store surveying to shifting quality and hedonic adjustments, resulted in meaningful downside adjustments to reported annual CPI inflation of roughly 300 basis points (3%), of which 28 basis points currently is estimated by the BLS as the effect of geometric weighting. The period involved here, from the early-1990s to date, was dominated by efforts to address the Greenspan/Boskin contention that the CPI "overstated" inflation. 

Nonetheless, the resulting, current CPI was and still is not a fully-substitution-based inflation index. An experimental substitution-based index, the Chain Weighted CPI-U (C-CPI-U), is published monthly by the BLS along with the CPI-U and CPI-W. At present, the C-CPI-U is showing the annual inflation rate running about 80 basis points (0.8%) below the official CPI-U. It is plotted as one of the three CPI measures in the inflation graph on the home page.

Geometric Weighting is a Mathematical Adjustment, Not a Model of Consumer Behavior. The BLS touts the use of geometric weighting in the narrow CPI categories as a way of measuring shifting consumer preferences based on changes in prices in related items. The weights that shift based upon price changes (relatively higher price changes end up with relatively lower weightings) do so by straight mathematical adjustment that the BLS once described as "mimicking" substitution effects. The shifts are not calculated based on any consumer surveying done, for example, as to how candy bar consumption would vary given relative price changes.

The BLS claims support for using geometric weightings in the CPI, because everyone else does it. One also could argue that other sovereign statistical agencies, by their nature, have a tendency to want to reduce reported inflation as much as possible, as did Messrs. Greenspan and Boskin.

On page 6 of the article, the authors argue that the shifting of weights within geometrically weighted categories does not affect a consumer’s standard of living, since the earlier arithmetic weighting always overstated cost of living, based on common academic thinking.

"There is also no dispute among economists [except for John Williams as at least one] that the price index formula used in all of the basic CPIs prior to 1999 (called the Laspeyres formula) tends to overstate changes in the cost of living; specifically, the change in a Laspeyres is an ‘upper bound’ on the change in the cost of maintaining a [note: no "constant"] standard of living. …"

"The Laspeyres answer is correct, however, only if the consumer is completely unconcerned with changes in price …"

I would argue, to the contrary, that it is the so-called "overstatement" in the cost of living that enables the maintenance of a constant standard of living, where the consumer does not have to be concerned with changes in price. The BLS claims that with the geometric weighting, weighting shifts are measuring a "constant level of satisfaction," that there is no "declining standard of living" in the numbers, because geometric weighting is not applied to broad enough categories to allow hamburger substitution for steak.

Nonetheless, the geometric weighting shifts have impact on a constant standard of living basis, as discussed above. Further, as mentioned earlier, the increased frequency of the reweighting of the broader categories impacts the standard of living on the steak to hamburger issue.

Hedonic and Quality Adjustments. Quality adjustments that are directly quantifiable and adjustable in terms of price impact are a necessary part of the CPI process. For example, a package that contained 12 ounces of crackers that now has been reduced to 10 ounces needs to be adjusted proportionately in pricing, in order for period-to-period price comparisons to have proper meaning.

When the quality change is not easily quantifiable (which usually includes hedonic adjustments), price adjustments for quality change are not so obvious, particularly from the standpoint of how individuals would assess them against perceived common experience, a constant standard of living and variance with official CPI reporting. Here are two examples of somewhat varying nature.

Consider the following from the February 15, 1995 CPI release: "A quality adjustment has been made to gasoline prices in the January CPI to account for the effects of the mandated introduction of reformulated gasoline in selected areas of the United States. The gasoline index rose 0.4 percent in January, following seasonal adjustment. Without the quality adjustment, it is estimated that this index would have increased 1.1 percent."

As I recall, that additive later was found to be harmful to the environment and eventually was removed from gasoline. Where consumers generally will look at the cost of filling a gas tank in terms of dollars laid out, irrespective of any theoretical improvement to the environment, counting the cost of the additive as a price increase, as opposed to ignoring it as a quality improvement, would seem reasonable. 

Quality adjustments of the hedonic, more-theoretical kind, however, have tended to reduce reported inflation meaningfully.  In the article the BLS indicated that hedonic adjustments had increased prices in certain products. Those "increases," though, often were based on comparisons against prices that already had been previously adjusted and reduced based on simpler quality assessments.

Getting more into the hedonics area, I’ll get personal. I use two personal computer systems purchased about 10 years apart for roughly the same price in nominal terms, about $800. While the most recent computer has greater memory and is faster than my old system, both systems generally perform the same tasks for me. Based on the BLS’s adjustments to computer prices, in terms of quality/hedonics, my old system should have been replaceable for about $85.00 in current dollars, which was not doable. I do have a nicer picture screen, with the new system, but I also unexpectedly had to buy a new printer, because the new system was not able to function with my antique work-horse printer. The new computer also was not able to use certain key programs that had not been rewritten to the standards of the new system. How does one compare and value such systems in the CPI? While some quality adjustment in the case of computers seems appropriate, I argue it has been heavily overdone from the practical standpoint of the average consumer.

SGS-Alternate Consumer Price Inflation vs. CPI-U. The BLS argued that our alternate CPI measure showed a 155% increase in the ten-year period of April 1998 to April 2008, and that such was much too high given average prices the BLS reported in underlying CPI categories. 

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). Some of the 1990s and later estimated changes have not been published by the BLS. I do not recalculate the CPI, only adjust for the reported, aggregate biases, generally using the BLS numbers. Since 1980, the aggregate change in annual CPI inflation reporting due to methodological shifts has been a reduction of roughly 700 basis points (7%). Again this is based primarily on published BLS estimates. 

The 1980-base SGS index calculates out to a 147% increase for the period (the BLS had access to the SGS data), not 155%, although the variance there does not affect the overstatement claim significantly. In contrast, the official CPI-U was reported up by 32% over the same period. While 1980-base SGS inflation overstates the inflation suggested by most of the proffered data, the government’s CPI-U understates inflation in most of the same data. For the same period, the 1990-base SGS inflation would have been up about 77%.

These issues will be fully explored and discussed in the upcoming academic study, including a careful analysis of inflation reflected in raw price data for various periods, versus the various CPI measures.