“It is difficult to get a man to understand something when his salary depends on him not understanding it.” Upton Sinclair
Recently, we listened to Stephan Auth, an uber-bullish Chief Investment Officer from Federated Investors, waltz around a bearish argument. Specifically, the interviewer asked how he reconciled his optimism with Shiller’s CAPE10 price to earnings calculation, which currently portends a high valuation, ergo a significant drawdown.
His argument against using Shiller’s CAPE, which compares the current price to the average of earnings from the last ten years, is that the current ten year data set includes poor earnings from the Financial Crisis. His preference is to compare prices to expected future earnings.
While we would like to share in Auth’s optimism, the truth is that prior earnings are a very good predictor of the future earnings, and forward earnings expectations are quite often abused by Wall Street to make stocks appear cheaper. The incentive for this bias is obvious in that people who work on Wall Street make money by selling financial instruments and their analysis is directed toward advancing that purpose.
In this piece, we provide both qualitative and quantitative analysis to take the basis of his arguments to task — namely, his rationale for using unknown earnings expectations versus known historical earnings. Our opinion, as you may surmise by the opening quote, is that such a perspective on valuations is not only wrong but likely the “spin” of a spokesman that makes a much better living when markets are rising.
Before presenting quantitative analysis showing that excluding earnings data from the crisis period does not make stocks cheaper, let us first explain why using long term earnings is appropriate and why recessionary data should be included.
First and foremost, despite what any salesman says, recessions and stretches of declining earnings are commonplace and occur with regularity. Per Wikipedia and the NBER:
The National Bureau of Economic Research dates recessions on a monthly basis back to 1854; according to their chronology, from 1854 to 1919, there were 16 cycles. The average recession lasted 22 months, and the average expansion 27. From 1919 to 1945, there were six cycles; recessions lasted an average 18 months and expansions for 35. From 1945 to 2001, and 10 cycles, recessions lasted an average 10 months and expansions an average of 57 months.
As tracked by the National Bureau of Economic Research (NBER), the average period of economic expansion has increased since 1945 to 57 months or 4.75 years. In other words, in any ten-year period chosen at random, there are likely to be two recessionary periods. Recessions are a natural part of the economic cycle and should be expected. Given this fact, is it really analytically irresponsible to factor in recessions and periods of weaker earnings into equity analysis? On the contrary, it is irresponsible to diminish or ignore entirely such periods.
CAPE10 uses ten years of earnings and, in our opinion, provides a long enough period, containing full economic cycles, through which to evaluate future earnings and the premium paid for those earnings. The flaw of CAPE, as pointed out correctly by Auth, is that it is not forward-looking. Unfortunately, forward looking analysis that is not tethered to the past is likely the result of using crystal balls, tarot cards and a good measure of hope. Just ask yourself, how many professional investors and esteemed economists saw the 2008 financial crisis coming in advance? Indeed, former Fed Chairman Bernanke denied the likelihood of a recession four months after it had already begun!
If one thinks the pattern of earnings will change notably in the future, then the value of CAPE analysis is compromised. In Auth’s defense, if one thinks, for example, that earnings will grow at twice the rate seen historically, then valuations may be cheap.
The graph below shows the evolution of earnings growth over the past 100+ years. The upward trend in earnings growth peaked decades ago. Stephen Auth may have some special insight, but the economic and earnings trends of the past 50 years leave us little reason to suspect that tomorrow will be vastly different than yesterday. Taking it a step further, earnings growth, a record ten years into an economic expansion, is only peaking at the point where prior earnings growth troughed over the last 50 years. More concerning, this is occurring despite a decade of extraordinary monetary and fiscal stimulus.
Data Courtesy Robert Shiller/Yale
If you agree that earnings are likely to take their cue from the past, then we should explore the price-to-earnings ratio using various time frames and not just the “misleading” ten-year CAPE.
In the analysis below, we compare price to earnings from the trailing twelve months (TTM) as denoted in the charts below as well as for terms of three years, five years, seven years, ten years (CAPE), 15 years, and 20 years. The four shortest time frames are void of data from the 2008 crisis or any recessions. The three longest time frames include the crisis, and in the case of the 20-year period also include the recession of 2001.
The graph below shows the minimum and maximum range of the seven P/Es for every given month over the past 100+ years.
Data Courtesy Shiller/Yale
As shown by the horizontal dotted lines, the current lowest (TTM) and highest valuation (20-year) of the seven time-frames are more expensive than all prior valuations except those of the late ‘90s tech bubble. To emphasize, even the “friendliest” of the seven time frame P/Es deem the markets historically expensive.
The graphs below detail current P/E valuations for the seven periods using average and standard deviation from the norm.
Data Courtesy Robert Shiller/Yale
Investors are paying a high premium for earnings based on a wide spectrum of valuations. Eliminating data from the financial crisis does not better justify current valuations as we have shown. Rather, it argues for more caution especially given the record-length of the current economic expansion and statistical likelihood that it will end sooner rather than later.
As noted earlier, forward valuations are optically cheap only because hope, expectations, and sales tactics are behind those highly optimistic earnings forecasts, which diverge widely from expected economic growth.
Stocks can certainly go higher, but we must apply analytical rigor to assess the data. History tells us stocks are currently very expensive regardless of which time frame we use to make the comparison. Saying otherwise is shoddy analysis at best and intellectually dishonest at worst.
Michael Lebowitz, CFA is an Investment Analyst and Portfolio Manager for RIA Advisors. specializing in macroeconomic research, valuations, asset allocation, and risk management. RIA Contributing Editor and Research Director. CFA is an Investment Analyst and Portfolio Manager; Co-founder of 720 Global Research.