Last night, in response to my latest update on the valuation of the U.S. market, I received a thoughtful email from Theodore Keeler, Professor Emeritus of Economics at UC Berkeley. Professor Keeler writes:

When you calculate overall stock price trends, you revise the post-1982 inflation figures with data from Shadowstats.com. Though the old method of calculating inflation may be inferior to the current one, I agree that it makes the most sense to compare stock prices based on a consistently-measured inflation figure.

My question is, why not also correct the P/E10 figures for inflation using the Shadowstats.com CPI numbers? As I understand it, the Shadowstats numbers yield generally higher inflation figures after 1982 than the official ones. That means that, all other things equal, as inflation goes up, today's price is worth less relative to last year's earnings. Therefore, use of a higher inflation correction will lower the real value of the P/E10, relative to a lower inflation correction.

Professor Keeler's email raises an important question. In my regression analysis of the S&P Composite, I posted a pair of charts showing the impact of the changes to the method of calculating the Consumer Price Index (CPI) introduced by the Bureau of Labor Statistics (BLS) in 1982. On his Shadowstats.com website, Economist John Williams calculates an "Alternate CPI" using the original BLS methodology. I used the Alternate CPI for the second chart in my regression commentary. My "Bullish" and "Bearish" labels characterize the striking difference between the two.

So let's take another look at the P/E10 ratio, an inflation adjusted measure, this time with an Alternate CPI comparison. The first chart uses the official CPI for the inflation adjustment. The second chart shows a radically different contour to the post-1982 S&P Composite and a significantly different market valuation. This alternate version shows a much cheaper valuation, one that hit single digits in March.

So the question is, which is more reliable — the Bureau of Labor Statistics or www.ShadowStats.com?

As I stated in the regression commentary, my opinion is that the optimum method for calculating consumer prices is somewhere between the revised BLS method and the historic method preserved by Williams. But for a long-term analysis of inflation-adjusted P/E ratios, consistency is essential, which may lend some credibility to the alternate CPI version of P/E10. In that version, the 2000 Tech Bubble peaked at about the same level of over-valuation as the 1929 peak. And our March valuation low was closer to the secular lows of the past, although 9.7 is still above the lows in 1921, 1932 and 1982.

Thanks, Professor Keeler, for calling attention to this pivotal issue for cyclical market valuation.