This morning I received a fascinating email from Robert "Bob" Bronson (Principal, Bronson Capital Markets Research) questioning of the value of inflation adjustment in calculating the cyclical 10-year P/E ratio. See my recent articles on this metric using the official CPI for inflation adjustment and a variant using the Alternate CPI. Over the past few years I've focused exclusively on the inflation-adjusted cyclical P/E, which dates from the early 1930s (devised by Benjamin Graham and David Dodd) and popularized in recent years by Yale professor Robert Shiller.
Bob Bronson writes:
Since the S&P Composite has inflation intrinsically included in it, dividing by earnings, averaged of course, which also has inflation intrinsically in it, cancels inflation out of the P/E result. This is what the most thoughtful economists do, largely because it eliminates the need to find/argue about the "right" inflation index since what investors and companies priced it at the time is what is important, which ends the debate.
So let's take another look at the P/E10 ratio — this time using nominal values for both the price and earnings:
Click the links above the chart to compare the three P/E10 versions. Instead of a quintile analysis as in the two real charts, for the nominal chart I've followed Bob's example and drawn a "best-fit" channel using the first two peaks for the upper boundary and a parallel lower boundary at the 1932 low. The dotted line shows the channel bottom if the 1932 low is excluded. The blue line is an exponential regression through the P/E10 data. The slope is equivalent to a 0.47% annualized rate of growth.
With the nominal P/E10 analysis, the current value is just above the regression trend line. The nominal P/E10 at the time of the 2009 low was well above the low in 1982 and dramatically higher than the 1932 low.
In his own P/E10 analysis, Bob Bronson uses intraday highs and lows for calculating the peaks and troughs. I've used monthly averages of daily closes in my nominal version to be consistent with my two inflation-adjusted versions. To the naked eye (at least to mine), the differences between Bob's version and the one above are not apparent. However, as you might expect, the actual ratios at the peaks and troughs are slightly different. I came up with 48.9 for the nominal peak in 2000. Bob's number is 49.9 using the intraday high on March 24, 2000. The March 2009 low of 15 in the chart above would slip to 14.5 using the intraday low on March 6 instead of the March monthly average of daily closes.
As I mentioned earlier, I've been accustomed to thinking about cyclical P/E10 strictly in real terms. But I find this nominal version quite intriguing. I plan to include it as a part of my future monthly valuation updates.
Meanwhile, I highly recommend a close reading of Bob Bronson's Revealing Supercycles: BAAC and Economic. Bob periodically updates this report, and I'll alert readers here when those updates are available.