One of the most consistent messages I’ve heard throughout my career is that the market is inexpensive or at least “fairly valued” based on next years earnings. We hear it at heights of euphoria and the depths of despair. I don’t recall ever hearing the consensus or even a vocal minority calling the market overvalued based on forward earnings estimates. In fact, we rarely even hear perma-bears cite high P/Es based on forward estimates as the primary cause for concern. Over a decade with low and often significantly negative returns, how can that be?
A Flawed Exercise
What makes market calls based on forward P/Es so dangerous is that the concept makes such good intuitive sense. We know the market’s primary function is to look forward and discount the future back to the present. We know that valuation, as James Montier puts it, is “the closest thing we have to a law of gravity” in finance. We know that earnings, while flawed in many ways (another discussion for another day), are essentially what investors want to pay for. To confuse matters worse, when analyzing an individual company’s stock, skilled investors absolutely would want to forecast forward earnings, make pro-forma adjustments from GAAP and then estimate a fair value based on a P/E multiple of that estimate.
Valuing the market on forward earnings estimates is a flawed exercise which often leads to incorrect assumptions about future market returns. It just doesn’t work very well.
1. The earnings estimates are usually inflated by improper adjustments and blind optimism
In another post criticizing a typical Wall Street approach to valuing the market, I challenged the conventional wisdom that operating earnings represented a clearer picture:
I’m no accounting expert, but I know one thing: pro-forma operating EPS is what company management uses to make their results look as good as possible without being accused of wrongdoing. Stock option expense? It’s not real money. Business restructuring? One-time event. Overpaying for an acquisition? Just a non-cash write down.
It’s not that on a company-by-company basis there aren’t legitimate adjustments that should be made, however most sell-side analysts simply take management guidance and run it through management blessed models making their forward EPS estimates questionable. Aggregating those estimates gets you to an even worse place when analyzing the market as a whole. Ed Easterling of Crestmont quantifies these adjustments for us
“As Reported” earnings reflects the past and projected (by S&P analysts) net income from the five hundred large companies in the S&P 500 Index. This measure is based upon Generally Accepted Accounting Principles (GAAP) and is the measure that historical averages are based upon. “Operating” earnings reflects a subjective measure of earnings (by other S&P analysts) that adds back certain costs and charges. It attempts to reduce the impact of the business cycle and one-time charges, yet it is generally considered to be an optimistic view of earnings. This measure of earnings per share (EPS) is NOT comparable to the long-term average P/E, since operating earnings excludes a variety of costs and charges that reduce the funds available for dividends. On average, ‘Operating EPS’ is almost 20% more than ‘As Reported EPS’.
In a recent update from Crestmont, this number was established at roughly 16% over the past 20+ years. This is clearly NOT a function of extreme events or unusual occurrences that should be omitted from our analysis, but is instead a systematic overestimation of earnings. Of course, “as reported” earnings also have their limitations like undue influence from large losses in a small number of companies. As chronicled by Denis Ouellet at his excellent blog, AIG alone shaved over $7 from S&P 500 reported earnings in the 4th quarter of 2008, even though the stock at the time had little effect on the index’s price. And both operating and reported earnings on the index level are subject to issues of non comparability when the folks at S&P simply remove failing companies (stocks) and replace them with successful ones. Yet another reason why a normalized earnings measure like Shiller’s, Ben Graham’s, or Ed Easterling’s is so useful.
2. The resulting P/E multiple is usually improperly compared to historical averages or notions of fair value
The typical analysis goes something like this: at 13x next year’s earnings, the S&P 500 is attractively valued compared to it’s long-term historical average of 15x.
That is comparing Apples to Oranges.
14x is the long-term average P/E based on reported earnings since 1900. Many cite the arithmetic mean of 15x which gets a bit of an unfair lift from the bubble years without a comparable low-end offset according to Ed Easterling. 14x is consistent with both the median and trimmed-mean and serves as a better guide for “average” in this case.
So what is the long-term average P/E based on forward operating earnings? What is the number that we can compare the oft-cited valuation measure to? While there is no good really long-term data on forward estimates, we can get a pretty good idea by going through this exercise:
Hypothetical $100 reported EPS
Operating EPS is about 16% more than reported EPS= $116 operating EPS
Forward EPS is about 6% more than current EPS historically= $123 forward operating EPS
An average P/E of 14x on $100 is only 11.4x the SAME level of earnings, using forward operating earnings
This is the approximate historical average that we should be comparing to the market P/E based on consensus estimates. While the above exercise may lack scientific rigor, the eminently quotable John Maynard Keynes would certainly approve as he too “would rather be vaguely right, than precisely wrong.“
A glance at this chart from Goldman Sachs (I added the bars at 10x and 15x) shows the more recent past of the P/E based on “Next 12 months Operating Earnings Estimates.” The P/E spent much of the first 15 years below 10x and was only pushed above 15x in 1997 before falling once again below 15x for much of the last 4 years.
source: Goldman Sachs, Compustat
Average Valuation ≠ Fair Valuation
However, our analysis on market valuation should not stop there. While the average P/E on forward earnings is approximately 11.4x, it would be unwise to assume that every instance of above average P/Es means overvaluation and every instance of below average P/Es mean undervaluation. There are many drivers of intrinsic value or fair value that have significant impact on measure such as P/E multiples. The obvious wild card is inflation. During periods of higher inflation investors rightly demand higher nominal returns to compensate them for the loss of purchasing power. This pushes P/E multiples down as investors pay less for future earnings which also tend to be of lower quality. Note, this is NOT an endorsement of the so called Fed Model. Please see Cliff Asness’ classic work Fight the Fed Model: The Relationship Between Stock Market Yields, Bond Market Yields, and Future Returns. If you are too busy to read the paper that disproves one of the most widely used asset allocation forecast tools, I’ll sum it up for you– forget bond yields when forecasting future stock market returns– the only thing that matters is starting and ending valuation.
Theoretical Concept of the Fed Model Leads to False Conclusions About The Market's Fair Value
In fact the theoretical construct of the Fed Model which says that at low nominal interest rates, P/E multiples should be higher combined with the elevated nature of future earnings allowed for some erroneous justification of wildly expensive markets in 1998-2002 and again from 2004-2007. One can conduct a good smell test by looking at the Goldman chart above and see that on forward earnings the market appeared to be “equally cheap” in 2007 as it was in 2009 (about 15x) when intuitively we know (and the 10 year normalized measures confirm) that the market was much cheaper in 2009 than it was in 2007.
A Better Way?
Obviously I think there is a better way to try to ascertain the market’s value. Valuation ratios based on Normalized Earnings is one place to start the analysis. How we determine “fair value” based on these earnings is another matter. I covered a few versions used by GMO, Hussman, and others. I recently discovered Denis Oulette’s valuation approach– “The Rule of 20″ in the Valuation section at the excellent www.news-to-use.com. I intend to spend more time applying some of his framework to a normalized earnings measure and will report back in a future post.
In the meantime, the next time you hear a pundit tell you the market is “cheap” or below the historical average based on next year’s earnings, be sure to run, not walk away to find a more suitable comparison of value.
I am open to being persuaded by the rule of 20, but I will need some persuading. That such an approach to determine fair value is better than using 15 for a trading signal may be true, though I cannot determine whether it is a statistical artifact without more research from myself and others. Certainly the raw P/E is unreliable.
You mention applying it to a normalized P/E, and I would be interested in that as well, as a trading signal.
For cheapness it makes little sense for the same reason the fed model doesn’t. While low interest rates (or inflation) may affect current P/E levels, they do not lead to higher returns in the future. If the rule was correct and 2009 was an generational period of cheapness then we would expect the following 10 years to result in some of the highest returns of the post war era. I doubt it.
Let us break down (as Easterling does, as GMO does, and I do) the compnents of returns. Low inflation (as Easterling points out) is generally accompanied by higher P/E’s. However, low inflation does not result in higher earnings growth, nor higher starting yield. Thus, if the components of returns are inflation, plus real earnings, plus yield, plus change in P/E then returns will be lower at any given P/E when they are higher than lower. So how does todays inflation rate alter future returns against that? It doesn’t.
So possibly that means such a measure could tell us what todays level is likely not to fall too far below, or to go down as a trader. What it has to do with cheapness or longer term returns is questionable at best.
Great post!
Thanks a lot! Quality or credibility of any forward estimation is always questionable.