John Greenlees of the BLS was kind enough to reply by email to my recent post on the CPI, and has agreed to allow me to post his comments:
Thank you for the kind words on your The Fly Bottle blog about my article with Rob McClelland, “Addressing Misconceptions about the Consumer Price Index.” You also say there that you would be interested in our response to the concern that the CPI is too conservative in accounting for gains to consumers.
In calculating the CPI, the BLS faces a set of well-known but difficult challenges, including dealing with consumer substitution behavior, accounting for product quality change, and handling the introduction of new products like Ipods and new distribution channels like the Internet. With respect to each of these issues, we attempt to employ the most advanced methods available, subject to the requirement that those methods also be objective and reproducible. There is no perfect way, however, to measure the welfare gains and losses that consumers experience when prices change and products appear and disappear. Thus, we know that the CPI is not perfect, but we do not have an estimate of any statistical “bias” in the index; if there was an accurate and reliable method to estimate such a bias, we would use that method to improve the CPI by eliminating the bias.
The most recent summary of the BLS approach to CPI measurement problems is probably the article “Working to Improve the Consumer Price Index” by Commissioner Katharine Abraham and others in the Winter 1998 Journal of Economic Perspectives (at http://www.jstor.org/stable/pdfplus/2646935.pdf). More recently, at an American Economics Association session in 2006, I reviewed the BLS considerations underlying changes it made during the years surrounding the 1996 Boskin Commission report on the CPI. That review was subsequently published in the International Productivity Monitor, available at http://www.csls.ca/ipm/12/IPM-12-Greenlees-e.pdf.
I hope these citations are helpful, and thank you again for mentioning our article.
John S. Greenlees, Ph.D.
Division of Price and Index Number Research
US Bureau of Labor Statistics
Thanks, John (and Rob)!
Here's a few thoughts… I don't think it's surprising that there is no statistical estimate of conservative bias, given current modes of measurement. As John says, if there was, they'd use it to correct the measure. But I wonder what kind of measure would be convincing in establishing an overall conservative bias. In particular, I wonder about the possibility of objectively measuring average subjective consumer welfare gains.
I don't think we can do this with current life satisfaction survey methods. But suppose some future science of the measurement of affective quality could track changes in the levels of the various hormones and the activation of the various neural pathways underpinning the various positive and negative feelings. Then suppose we were able to use statistical analysis to isolate the portion of these changes attributable to changes in consumption. We should then be able to say something more precise about the real hedonic (in the psychological sense) value of quality changes, new products, new outlets, etc. Shouldn't we?
Of course, subjective states aren't the only things we care about and don't exhaust well-being, on plausible accounts of well-being. Longevity, health, and various capabilities are plausible constituents of well-being, too. And it strikes me that we can measure some of these things quite directly. So why not estimate the effects of new products and quality changes on all these aspects of well-being?
Let's say a “theory of well-being” is a list of plausible candidate constituents of well-being together, where each element is assigned a weight corresponding to its relative importance. We could poll people and do other experiments to reveal the most popular theories. Then we could calculate changes in well-being, relative to the most widely accepted theories of well-being, as typical consumption baskets change in composition and quality over time. Your real income might go up according to an index built on one theory of well-being and down on another. Then, instead of arguing over suspected biases in the CPI, pundits could argue directly about the constituents of welfare. That would be better. That's what the argument is really about anyway. Could pluralistic indexing be the future?
One last thing that you should skip unless you care: I find the Austrian subjectivist argument against economic measurement based in the heterogeneity of preferences uninteresting. It's still useful to know the average effect of a change in the set of consumption options, and preference profiles don't randomly differ. They cluster in rough types. So heterogeneity is no argument against discovering usefully general truths. Anti-“scientism” is often little more than obscurantism or methodological laziness.
Once again, thanks to the guys from the Division of Price and Index Number Research for so kindly responding!