A New-and-Improved Shiller CAPE: Solving the Dividend Payout Ratio Problem

A common criticism of Professor Robert Shiller’s famous CAPE measure of stock market valuation is that it fails to correct for the effects of secular changes in the dividend payout ratio.  Dividend payout ratios for U.S. companies are lower now than they used to be, with a greater share of U.S. corporate profit going to reinvestment.  For this reason, earnings per share (EPS) tends to grow faster than it did in prior eras.  But faster EPS growth pushes up the value of the Shiller CAPE, all else equal.  Distortions therefore emerge in the comparison between present values of the measure and past values.

To give credit where it’s due, the first people to point out this effect–at least as far as I know–were Professor Jeremy Siegel of Wharton Business School and his former student, David Bianco of Deutsche Bank.  Siegel, in specific, wrote about the problem as far back as late 2008, during the depths of the financial crisis, when the Shiller CAPE was steering investors away from a market that he considered to be extremely cheap (see “Jeremy Siegel on Why Equities are Dirt Cheap”, November 18, 2008, link here).

In a piece from 2013, I attempted to demonstrate the effect with two tables, shown below:



The tables portray the 10 year earnings trajectories and Shiller CAPE ratios of two identical companies that generate identical profits and that sell at identical trailing-twelve-month (ttm) P/E valuations. The first company, shown in the first table, pays out 75% of its profit in dividends and reinvests the other 25% into growth (in this case, share buybacks that grow the EPS by shrinking the S). The second company, shown in the second table, pays out 25% of its profit in dividends, and reinvests the other 75% into growth.

As you can see, even though these companies are identically valued in all relevant respects, they end up with significantly different Shiller CAPEs.  The reason for the difference is that the second company reinvests a greater share of its earnings into growth than the first company.  Its earnings therefore grow faster.  Because its earnings grow faster, the act of “averaging” them over a trailing 10 year period reduces them by a greater relative amount.  Measured against that trailing 10 year average, the company’s price, appropriately set in reference to its ttm earnings, therefore ends up looking more expensive.  But, in truth, it’s not more expensive–its valuation is exactly the same as that of the first company.

The following chart illustrates the effect:


To summarize the relationship:

  • Lower Payout Ratio –> Higher Earnings Growth –> Higher CAPE, all else equal
  • Higher Payout Ratio –> Lower Earnings Growth –> Lower CAPE, all else equal

Now, how can we fix this problem?  A natural solution would be to reconstruct the CAPE on the basis of total return (which factors in dividends) rather than price (which does not). But that’s easier said than done.  How exactly does one build a CAPE ratio–or any P/E ratio–on the basis of total return?

Enter the Total Return EPS Index, explained here and here.  The Total Return EPS Index is a modified version of a normal EPS index that tells us, hypothetically, what EPS would have been, now and at all times in history, if the dividends that were paid out to shareholders had not been paid out, and had instead been diverted into share buybacks. Put differently, Total Return EPS tells us what earnings would have been if the dividend payout ratio had been 0% at all times.  In this way, it reduces all earnings data across all periods of history to the same common basis, allowing for accurate comparisons between any two points in time.

Crucially, in constructing the Total Return EPS, we assume that the buybacks are conducted at fair value prices, prices that correspond to the same valuation in all periods (equal to the historical average), rather than at market prices, which are erratic and often groundless.  To those readers who continue to e-mail in, expressing frustration with this assumption–don’t worry, you’re about see why it’s important.

The following chart shows the Total Return EPS alongside the Regular EPS from 1871 to 2015.  In this chart and in all charts presented hereafter, the index is the S&P 500 (and its pre-1957 ancestry), the values are appropriately inflation-adjusted to February 2015 dollars, and no corrections are made for the effects of questionable accounting writedowns associated with the last two economic downturns:


Now, if all S&P 500 dividends had been diverted into share buybacks, then the price of the index would have increased accordingly. We therefore need a Total Return Price index–an index that shows what prices would have been on the “dividends become buybacks” assumption.

Calculating a Total Return Price index is straightforward.  We simply assume that the market would have applied the same P/E ratio to the Total Return EPS that it applied to the Regular EPS (and why would it have applied a different P/E ratio?). Multiplying each monthly Total Return EPS number by the market’s ttm P/E multiple in that month, we get the Total Return Price index.

In the chart below, we show the Total Return Price index for the S&P 500 alongside the Regular Price, from 1871 to 2015:


Generating a CAPE from these measures is similarly straightforward.  We divide the Total Return Price by the trailing 10 year average of the Total Return EPS.  The result: The Total Return EPS CAPE.

Shiller himself proposed a different method for calculating a CAPE based on total return in a June 2014 paper entitled “Changing Times, Changing Valuations: A Historical Analysis of Sectors within the U.S. Stock Market: 1872 to 2013″ (h/t James Montier). The instructions for the method are as follows: Use price and dividend information to build a Total Return Index. Then, scale up the earnings by a factor equal to the ratio between the Total Return Index and the Price Index.  Then, divide the Total Return Index by the trailing ten year average of the scaled-up earnings.  In a piece from August of last year, I tried to build a CAPE based on Total Return using yet another method (one that involves growing share counts), and arrived at a result identical to Shiller.  The technique and charts associated with that method are presented here.

It turns out that both of these methods produce results identical to the Total Return EPS CAPE method, with one small adjustment: that we conduct the buybacks that form the Total Return EPS at market prices, rather than at fair value prices as initially stipulated. The following chart shows the three types of Total Return CAPEs together.  As you can see, the lines overlap perfectly.


The three different versions of the CAPE overlap because they are ultimately doing the same thing mathematically, though in different ways.  Given that they are identical to each other, I’m going to focus only on the Total Return EPS version from here forward.  I’m going to refer to the version that conducts buybacks at fair value prices as “Total Return EPS (Fair Value) CAPE”, and the version that conducts buybacks at market prices as “Total Return EPS (Market) CAPE.”  I’m going to refer to Shiller’s original CAPE simply as “Shiller CAPE.”

The following chart shows the Total Return EPS (Market) CAPE alongside the Shiller CAPE, with the values of the former normalized so that the two CAPEs have the same historical average (allowing for a direct comparison between the numbers).


(Note: in prior pieces, I had been comparing P/E ratios to their geometric means. This is suboptimal. The optimal mean for a P/E ratio time series is the harmonic mean, which is essentially what you get when you take an average of the earnings yields–the P/E ratios inverted–and then invert that average.  So, from here forward, in the context of P/E ratios, I will be using harmonic means only.) (h/t and #FF to @econompic, @naufalsanaullah, @GestaltU_BPG)

The current value of the Shiller CAPE is 27.5, which is 93% above its historical average (harmonic) of 14.2.  The current value of the Total Return EPS (Market) CAPE is 30.3, which is 71% above its historical average (harmonic) of 17.8.  Normalized to matching historical averages, the current value of the Total Return EPS (Market) CAPE comes out to 24.2.

At current S&P 500 levels, then, we end up with 27.5 for the Shiller CAPE, and 24.2 for the Total Return EPS (Market) CAPE, each relative to a historical average of 14.19. Evidently, the difference between the two types of CAPEs is significant, worth 12%, or 250 current S&P points.

But there’s a mistake in this construction.  To find it, let’s take a closer look at the chart:


From the early 1990s onward, the Total Return EPS (Market) CAPE (the red line) is significantly below the Shiller CAPE (the blue line), suggesting that the Shiller CAPE is overstating the market’s expensiveness, and that the Total Return EPS (Market) CAPE is correcting the overstatement by pulling the metric back down.

What is driving the Shiller CAPE’s apparent overstatement of the market’s expensiveness? The obvious answer would seem to be the historically low dividend payout ratio in place from the early 1990s onward.  All else equal, low dividend payout ratios push the Shiller CAPE up, via the increased growth effect described earlier.

But look closely.  Whenever the market is expensive for an extended period of time, the subsequent Total Return EPS (Market) CAPE (the red line) ends up lower than the Shiller CAPE (the blue line), by an amount seemingly proportionate to the degree and duration of the expensiveness.  Note that this is true even in periods when the dividend payout ratio was high, e.g, the periods circled in black: the early 1900s, the late 1920s, and the late 1960s.  If the dividend payout ratio were the true explanation for the deviations between the Shiller CAPE and the Total Return EPS (Market) CAPE, then we would not get that result.  We would get the opposite result: the high dividend payout ratio seen during the periods would depress the the Shiller CAPE relative to the more accurate total measures; it would not push the Shiller CAPE up, as seems to be happening.

The converse is also true.  Whenever the market is cheap for an extended period of time, the subsequent Total Return EPS (Market) CAPE (the red line) ends up higher than the subsequent Shiller CAPE (the blue line), by an amount seemingly proportionate to the degree and duration of the cheapness.  We see this, for example, in the periods circled in green: the early 1920s and the early 1930s through the end of the 1940s.  The deviation between the two measures is spatially small in those periods, but that’s only because the numbers themselves are small–single digits.  On a percentage basis, the deviation is sizeable.

The following chart clarifies:


So what’s actually happening here?  Answer: valuationnot the dividend payout ratio–is driving the deviation.  In periods where the market was cheap in the 10 years preceding the calculation, the Total Return EPS (Market) CAPE comes out above the Shiller CAPE. In periods where the market was expensive in the 10 years preceding the calculation, the the Total Return EPS (Market) CAPE comes out below the Shiller CAPE.  The degree above or below ends up being a function of how cheap or expensive the market was, on average.

The following chart conclusively demonstrates this relationship:


The bright green line is the difference between the Total Return EPS (Market) CAPE and the Shiller CAPE as a percentage of the Shiller CAPE.  When the bright green line is positive, it means that the red line in the previous chart was higher than the blue line; when negative, vice-versa.  The pink line is a measure of how cheap or expensive the market was over the preceding 10 years, on average and relative to the historical average. When the pink line is positive, it means that the market was cheap; when negative, expensive.  The two lines track each other almost perfectly, indicating that the valuation in the preceding years–and not the payout ratio–is driving the deviation between the two measures.

What is causing this weird effect?  You already know.  The share buybacks associated with the Total Return EPS (Market) CAPE are being conducted at market prices, rather than at fair value prices.  The same is true for the dividend reinvestments associated with Shiller’s proposed Total Return CAPE and with the version I presented in August of last year; those reinvestments are being conducted at market prices.  That’s wrong.

When share buybacks (or dividend reinvestments) are conducted at market prices, then periods of prior expensiveness produce lower Total Return EPS growth (because the dividend money is invested at unattractive valuations that offer low implied returns).  And, mathematically, what does low growth do to a CAPE, all else equal?  Pull it down.  Past periods of market expensiveness therefore pull the Total Return EPS (Market) CAPE down below the Shiller CAPE, as observed.

Conversely, periods of prior cheapness produce higher Total Return EPS growth (because the dividend money is reinvested at attractive valuations that offer high implied returns).  And what does high growth do to a CAPE, all else equal?  Push it up.  Past periods of market cheapness therefore push the Total Return EPS (Market) CAPE up above the Shiller CAPE, as observed.

Looking at the period from the early 1990s onward, we assumed that the problem was with the Shiller CAPE (the blue line), that the low dividend payout ratio during the period was pushing it up, causing it to overstate the market’s expensiveness.  But, in fact, the problem was with our Total Return EPS (Market) CAPE (the red line).  The very high valuation in the post-1990s period is depressing Total Return EPS (Market) growth (the expensiveness of the share buybacks and dividend reinvestments shrinks their contribution), pulling down on the Total Return EPS (Market) CAPE, and causing it to understate the market’s  expensiveness.

The elimination of this distortion is yet another reason why the buybacks and dividend reinvestments that form the Total Return EPS (or any Total Return Index used in valuation measurements) have to be conducted at fair value prices, rather than at market prices.  Conducting the buybacks and dividend reinvestments at fair value prices ensures that they provide the same accretion to the index across all periods of history, rather than highly variable accretion that inconsistently pushes up or down on the measure.

Now, a number of readers have written in expressing disagreement with this point.  To them, I would ask a simple question: does it matter to the current market’s valuation what the market’s valuation happened to be in the distant past?

Suppose, for example, that in 2009, investors had become absolutely paralyzed with fear, and had sold the market’s valuation down to a CAPE of 1–an S&P level of, say, 50. Suppose further that the earnings and the underlying fundamentals had remained unchanged, and that investors had exacted the pummeling for reasons that were entirely irrational. Suppose finally that investors kept the market at the depressed 1 CAPE for two years, and that they then regained their senses, pushing the market back up to where it is today, in a glorious rally.  In the presence of these hypothetical changes to the past, what would happen to the current value of a Total Return EPS CAPE that reinvests at market prices?  Answer: it would go up wildly, dramatically, enormously, because the intervening dividends that form the Total Return index would have been invested at obscenely low valuations during the period, producing radically outsized total return growth.  What does high growth due to a CAPE? Push it up, so the CAPE would rise–by a large amount.

Is that a desirable result?  Do we want a measure whose current assessment of valuation is inextricably entangled in the market’s prior historical valuations, such that the measure would judge the valuations of two markets with identical fundamentals and identical prices to be significantly different, simply because one of them happened to have traded more cheaply or expensively in the past?  Obviously not.  That’s why we have to conduct the buybacks and reinvestments that make up the Total Return EPS at fair value.

The general rule is as follows.  When we’re using a Total Return index to model actual investor performance–what an individual who invested in the market would have earned, in reality, with the dividend reinvestment option checked off–we need to conduct the hypothetical reinvestments that make up the Total Return index at market prices.  But when we’re using a Total Return index to measure valuation–how a market’s price compares with its fundamentals–then we need to conduct the hypothetical reinvestments at fair value prices.

The following chart shows the Total Return EPS CAPE properly constructed on the assumption that the buybacks and reinvestments occur at fair value prices:


As you can see, the deviation between the two measures comes out to be much smaller. Normalized to the same historical average, the current value of the Total Return EPS (Fair Value) CAPE ends up being 25.9, versus 27.5 for the original Shiller CAPE.  The difference between the total return and the original measures comes out at 5.7%, a little over 100 current S&P points (versus 12% and 250 points earlier).

Surprisingly, then, properly reinvesting the dividends at the same valuation across history more than cuts the deviation in half, to the point where it can almost be ignored.  As far as the CAPE is concerned, when it comes to the kinds of changes that have occurred in the dividend payout ratio over the last 144 years, there appears to be little effect on the accuracy of Shiller’s original version.  The entire exercise was therefore unnecessary. Admittedly, this was not the result that I was anticipating, and certainly not the result that I was hoping to see.  But it is what it is.

It turns out that Shiller was right to reject the dividend payout ratio argument in his famous 2011 debate with Siegel and Bianco:

“Mr. Shiller did his own calculation about the impact of declining dividends on earnings growth and concluded that it is marginal at best, not meriting any adjustment.” — “Is the Market Overvalued?”, Wall Street Journal, April 9th, 2011.

If the subsequent foray into Total Return space caused him to change that view, then he should change it back.  He was right to begin with.  His critics on that point, myself included, were the ones that were wrong.

Now, this is not to suggest that we shouldn’t prefer to use the Total Return version of the CAPE over Shiller’s original version.  We should always prefer to make our analyses as accurate as possible, and the Total Return version of the CAPE is unquestionably the more accurate version.  Moreover, even though the changes in the dividend payout ratio seen in the U.S. equity space over the last 144 years have not been large enough to significantly impact the accuracy of the original version of the CAPE, the differences between the payout ratios of different countries–India and Austria, to use an extreme example–might still be large enough to make a meaningful difference.  Since the Shiller CAPE is the preferred method for accurately comparing different countries on a valuation basis, it only makes sense to shift to the more accurate Total Return version.  Fortunately, that version is simple and intuitive to build using Total Return EPS.

Admittedly, there is some circularity here.  In building the Total Return EPS Index on the assumption of fair value buybacks, we used the Shiller CAPE as the basis for estimating fair value.  If the Shiller CAPE is inaccurate as a measure of fair value, then our Total Return EPS index will be inaccurate, and therefore our Total Return CAPE, which is built on that index, will be inaccurate.  Fortunately, in this case, there’s no problem (otherwise I wouldn’t have done it this way). When you run the numbers, you find that the choice of valuation measure makes little difference to the final product, as long as a roughly consistent measure is used.  You can build the Total Return EPS Index using whatever roughly consistent measure you want–the Total Return CAPE will not come back appreciably different from Shiller’s original. What drove the deviations in the earlier charts were not small differences in the valuations at which dividends were reinvested, but large differences–for example, the difference associated with reinvesting dividends at market prices from 1942 to 1952, and then from 1997 to 2007, at prices corresponding to three times the valuation.

Now, there are other ways of adjusting for the impact of changing dividend payout ratios. Bianco, for example, has a specific technique for modifying past EPS values. As he explains:

“The Bianco PE is based on equity time value adjusted (ETVA) EPS.  We raise past period EPS by a nominal cost of equity estimate less the dividend yield for that period.”

I cannot speak confidently to the accuracy of Bianco’s technique because I do not have access to its details.  But if the method produces a result substantially different from the Total Return EPS CAPE (which it appears to do), then I would think that it would have to be wrong.  When it comes to changing dividend payout ratios, the Total Return EPS CAPE is airtight.  It treats all periods of history absolutely equally in all conceivable respects, perfectly reducing them to a common basis of 0% (payout).  Because it reinvests the dividends at fair value (the historical average valuation), every reinvested dividend in every period accretes at roughly the same rate, which corresponds to the actual average rate at which the market has historically accreted gross of dividends (approximately 6% real).

If our new-and-improved version of the CAPE is appropriately correcting the dividend payout ratio distortions contained in the original version, then the deviation between our new-and-improved version and the original version should be a clean function of that ratio (rather than a function of other irrelevant factors, such as past valuation).  When the dividend payout ratio is low, our new-and-improved version should end up below the original version, given that the original version will have overstated the valuation.  When the dividend payout ratio is high, our new-and-improved version should end up above the original version, given that the original version will have understated the valuation.

Lo and behold, when we chart the deviation between the two versions of the CAPE alongside the dividend payout ratio, that is exactly what we see: a near-perfect correlation (91%), across the full 134 year historical period.


The blue line shows the difference between our new-and-improved version of the CAPE and the original version.  The red line shows the trailing Shiller dividend payout ratio, which is the 10 year average of real dividends per share (DPS) divided by the 10 year average of real EPS.  We use a Shillerized version of the dividend payout ratio to remove noise associated with recessions–especially the most recent one, where earnings temporarily plunged almost to zero, causing the payout ratio to temporarily spike to a value north of 300%.

The fact that the two lines overlap almost perfectly indicates that the deviation between our new-and-improved version and the original version is a function of the factor–the dividend payout ratio–that is causing the inaccuracy in the original version, rather than some other questionable factor.  That is exactly what we want to see.  It is proof positive that our new-and-improved version is correcting the distortion in question, and not introducing or exploiting other distortions (that, conveniently, would make the current market look cheaper).

Now, to be clear, the secular decline in the dividend payout ratio seen across the span of U.S. market history has not substantially affected the accuracy of the original Shiller CAPE.  However, it has substantially affected the trend growth rate of EPS.  So, though it may not be imperative that we use the Total Return version of the CAPE when measuring valuation, it is absolutely imperative that we use the Total Return version of EPS when analyzing earnings trends and projecting out future earnings growth.

We are left with the question: if the distortions associated with the dividend payout ratio are not significant, then why does the Shiller CAPE show the U.S. equity market to be so expensive relative to history?  We can point to three explanations.

  • First, on its face, the market just is historically expensive–even on a non-Shiller P/E measurement.  Using reported EPS, the simple trailing twelve month P/E ratio is roughly 20.5, which is 53% above its historical average (harmonic) of 13.4.  Using S&P corporation’s publication of operating EPS, the simple trailing twelve month P/E ratio is 18.8, which is 40% above that average.
  • Second, the accounting writedowns associated with the 2008-2009 recession are artificially weighing down the trailing average 10 year EPS number off of which the Shiller CAPE is calculated.  Prior to 2014, this effect was more significant than it is at present, given that the 2001-2003 recession also saw significant accounting writedowns.  The trailing 10 year average for the years up to 2014 therefore got hit with a double-whammy.  That’s why the the increase in the Shiller CAPE in recent years has not been as significant as the increase in market prices (since December 2012, the CAPE is up roughly 30%, but prices are up roughly 50%).  2014 saw the 2001-2003 recession fully drop out of the average, reducing the CAPE’s prior overstatement.
  • Third, as the chart below shows, real EPS growth over the last two decades–on both a regular and a Total Return basis–has been meaningfully above the respective historical averages, driven by substantial expansion in profit margins.  Recall that high growth produces a high CAPE, all else equal.


These last two factors–the effects of accounting writedowns and the effects of profit margin expansion–will gradually drop out of the Shiller CAPE (unless you expect another 2008-type recession with commensurate writedowns, or continued profit margin expansion, from these record levels).  As they drop out, the valuation signal coming from the Shiller CAPE will converge with the signal given by the simple ttm P/E ratio–a convergence that is already happening.

We conclude with the question that all of this exists to answer: Is the market expensive? Yes, and returns are likely to be below the historical average, pulled down by a number of different mechanisms.  Should the market be expensive?  “Should” is not an appropriate word to use in markets.  What matters is that there are secularsustainable forces behind the market’s expensiveness–to name a few: low real interest rates, a lack of alternative investment opportunities (TINA), aggressive policymaker support, and improved market efficiency yielding a reduced equity risk premium (difference between equity returns and fixed income returns).  Unlike in prior eras of history, the secret of “stocks for the long run” is now well known–thoroughly studied by academics all over the world, and seared into the brain of every investor that sets foot on Wall Street.  For this reason, absent extreme levels of cyclically-induced fear, investors simply aren’t going to foolishly sell equities at bargain prices when there’s nowhere else to go–as they did, for example, in the 1940s and 1950s, when they had limited history and limited studied knowledge on which to rely.

As for the future, the interest-rate-related forces that are pushing up on valuations will get pulled out from under the market if and when inflationary pressures tie the Fed’s hands–i.e., force the Fed to impose a higher real interest rate on the economy.  For all we know, that may never happen.  Similarly, on a cyclically-adjusted basis, the equity risk premium may never again return to what it was in prior periods, as secrets cannot be taken back.

Posted in Uncategorized | Leave a comment

Using Total Return EPS to Decompose Historical S&P 500 Performance: Charts from 1871 to 2015


In this piece, I’m going to do five things:

  • First, I’m going to clarify the purpose of Total Return EPS, what it’s trying to accomplish.  In a single sentence, the purpose of Total Return EPS is to convert dividends into EPS so that the fundamental sources of return can be added together into one single term whose past growth rate can be analyzed and whose likely future growth rate can be projected.
  • Second, I’m going to explain why the trend growth rate of Total Return EPS for the S&P 500 (~6%) is roughly equal to the historical average return on equity for the U.S. corporate sector (~6%).  The explanation will include a proposed theory for why return on equity generally reverts to the mean, and also for why it may not revert to its prior mean in the present environment.
  • Third, I’m going to address a question that a significant number of readers have asked: why does Total Return EPS assume that buybacks occur at fair value, rather than at market prices?
  • Fourth, I’m going to show how actual total return can be “decomposed”–i.e., separated out–into three contributing components: (1) Total Return EPS growth, which consists of regular EPS growth plus the return from reinvested dividends (or hypothetical share buybacks), (2) the return contribution from the change in valuation–in this case, the change that occurs in the ttm P/E ratio from price to sale, and (3) the return contribution from interim deviations from fair value–a neglected source of return that arises from the valuations at which dividends are reinvested (or at which shares are hypothetically repurchased), and therefore the rate at which they accrete.
  • Fifth, I’m going to present charts of these components for the S&P 500 from 1871 to 2015, on time horizons of 10, 20, 30, 40, 50, 60, and 70 years.

Three Options: Dividends, Expansion, Repurchases

When the corporate sector earns profit, it can do one of two things: distribute the profit to shareholders as dividends, or reinvest the profit.

  • When it distributes the profit to shareholders as dividends, the shareholders get a direct return–a direct deposit of money into their accounts.
  • When it reinvests the profit, the shareholders get an indirect return–”growth.” The profits earned in future periods, and the future dividends that can be paid from them, increase in size.  In an efficient market, this increase coincides with an increase in the market prices of shares, allowing shareholders to realize a return by selling.

Looking closer at the second option, the corporate sector can reinvest profit in one of two ways: by using it to fund business expansion, or by using it to repurchase equity (or debt). Both options produce growth in earnings per share (EPS).

  • When the corporate sector uses profit to fund business expansion, it adds new capital that it can use to produce and sell additional output to the economy, from which additional income can be earned.  It grows the EPS by growing the E.
  • When the corporate sector uses profit to repurchase equity–for example, by buying back shares on the open market and then cancelling them–it grows the EPS by shrinking the S.  (Note: The corporate sector can also use profit to repurchase or retire debt.  We can view this option as roughly equivalent to the repurchase of equity. Both options entail a reduction in the number of outstanding claims on the corporate sector, rather than an increase in the size of the corporate sector’s operations).

What we have, then, are three destinations for corporate profit: (1) the payment of dividends, (2) investment in business expansion, and (3) the repurchase of equity (or debt).  The first option entails a direct return, a direct deposit of money into shareholder pockets.  The second and third options entail an indirect return, achieved through growth in EPS.

EPS Growth: In Search of a Trend

What we want to know is the “trend” (or “normal”) rate of growth of EPS.  Knowing that trend rate would allow us to roughly estimate the likely future trajectory of EPS, given its position relative to trend.

To illustrate, suppose that the trend rate of EPS growth is 4% per year, but that EPS over the last several years hasn’t grown at all, or worse, has fallen substantially.  We would then expect future EPS growth to be higher than the trend rate, higher than 4%, as EPS catches up.  We would expect there to have been some kind of stunting process–say, a depression in profit margins–that explains the underperformance relative to trend, and that entails the potential for future outperformance, to be unleashed in an eventual mean-reversion.

The problem, of course, is that when we look at the historical data, we do not find a stable, reliable trend growth rate in EPS.  Instead, we find a trend growth rate that has increased substantially over time.  The following table shows averages of rolling 10 year annualized real EPS growth rates for the S&P 500 for the periods 1871 to 1930, 1930 to 1990, and 1990 to 2015, with each period beginning and ending in January:


As  you can see in the table, the average rolling growth rate seen from 1990 to 2015 is four times the rate seen 100 years before it.  And note that this rate is the growth rate of GAAP EPS.  It include the effects of the questionable accounting writedowns that took place in 2003 and 2009.  If we use a corrected version of EPS that excludes those writedowns, the rolling average growth rate for the period increases to 4.76%–more than six times the rate achieved 100 years before it.

The reason for the increase in the trend growth rate of EPS is no mystery.  EPS growth varies inversely with the share of profit that is paid out as dividends.  That share has fallen over time.  The following table shows average payout ratios for the periods in question:


Now, there’s a legitimate question to ask here.  How much EPS growth should a given reduction in the dividend payout ratio actually produce? Is the observed increase in the growth rate, from 0.72% to 3.16%, commensurate with the observed decrease in the average payout ratio, from 69% to 52%?  Sadly, there’s no way to know.  Because we have no way to know, we can’t be sure that the reduced payout ratio is the only factor, or even the main factor, driving the increased growth.  For all we know, there could be some other factor driving it, a factor that will substantially impact growth going forward, either positively or negatively.

Return on Equity and Fundamental Total Return

We can distinguish between two sources of shareholder total return.  The first source is “fundamental”, and arises from the payment of dividends and the growth of EPS (or some other relevant fundamental).  The second source is “nonfundamental”, and arises from changes that occur in the valuations of assets between the time of purchase and the time of sale.  If you buy an asset and it pays you a dividend, or its price goes up in response to growth in relevant fundamentals–sales, earnings, net asset values, and so on–that is a “fundamental” return.  If you buy an asset and its price goes up independently of any type of growth in fundamentals–somebody just offers to pay you a higher price for the asset, because they want it more than the next person–that is a “nonfundamental” return.

The item that follows a consistent trend over time is not EPS growth per se, but the fundamental total return that accrues to shareholders–the return that dividends and EPS growth combine to produce.  Let me now explain why that return follows a consistent trend.  Bear with me.

The fundamental total return that accrues to shareholders is a function of the return that corporations generate on their equity, on the amount of capital that was invested to form them.  That return, after all, has to go to someone; it goes to the shareholders, those that made the investment, that put the capital in.

Now, return on equity (ROE) is mean-reverting.  When ROE is high in a given sector or industry, new investment flocks in, seeking to capture the high return.  The new investment leads to excess capacity, increased competition, weakened pricing power, and a reduction in profit that pulls the ROE for the sector or industry back down. When ROE is low in a given sector or industry, new investment stops happening.  The reduction in investment leads to an eventual undercapacity, reduced competition, increased pricing power for the remaining firms, and an increase in profit that pushes the ROE for the sector or industry back up–assuming, of course, that the goods and services being produced are actually wanted by the economy.  If they are not wanted, then the ROE for the industry or sector will go to zero, which is where it belongs for those that make unwanted things.

We’re currently seeing a textbook case of this process play out in the energy sector.  The economy needs a certain amount of oil.  The prior market price of oil–$75+–reflected the marginal cost of producing that amount, plus the extra “oomph” that speculation probably added.  But then efficient new drilling techniques were developed.  The strong profits that these techniques could earn with oil at $75+ led to an investment boom.  The investment boom eventually created an overcapacity that has pushed the price of oil down and that has dramatically lowered the return.  New investment has therefore dried up–and will stay dried up–until an undercapacity develops that increases the price enough to make the return attractive again.

This process of mean-reversion functions at its fiercest in the energy sector, where the good being sold is a pure commodity, and where there are few barriers or “moats” to block out competition and new entry.  But it applies in a general sense to all sectors and industries, and to the aggregate corporate sector as well.

(Caveat: If an economy evolves in a way that entails an increase in the number of barriers and “moats” in place to block out competition and new entry–i.e., in a way that makes it harder for new capital to partake in the high returns that existing capital might be enjoying–then the “mean” that the return on equity reverts to might increase accordingly. It remains an open question as to whether the new technology economy, with its tendency to produce winner-take-all scenarios in which the first mover is forever protected from competition–think $MSFT, $FB, $GOOG, $AAPL, and so on–has provoked such an increase.  I suspect that there is at least some of that effect at play in the much-discussed increase in ROEs and profit margins that we’ve seen take place over the last 20 years.)

Now, because the fundamental total return to shareholders is a function of the ROE, and because the ROE is mean-reverting, the fundamental total return to shareholders–paid out to them in dividends and growth–is similarly mean-reverting. Its mean-reverting nature is the reason that it follows a reliable trend over time.

Some might find this point hard to grasp–it’s admittedly hard to explain. To get a better feel for it, just think about the fundamental return that accrues to energy sector shareholders–shareholders in companies like $XOM and $CVX.  Can you see how the process that produces mean-reversion in energy sector ROEs would also produce mean-reversion in the fundamental return that $XOM and $CVX shareholders receive over time?  The same operating environment that allowed those companies to generate outsized fundamental returns–outsized EPS growth and outsized dividends–when oil was $75+ is what pulled in all of the new investment that fueled the current overcapacity, the squeeze to find buyers of all of the output, that is now pushing those same fundamental returns back down, hedges notwithstanding.

If there had been no new oil to drill, then that would have been a very powerful “moat”, and the high returns that these companies enjoyed might have been sustainable.  But when there is a new discovery that opens up ample new supply with the promise of a high return to anyone with capital that wants to make an investment in it, the high returns–to the new entrants and the existing players–simply will not last.

Ideally, then, we would ditch the effort to find the trend growth rate in EPS, and would instead focus on finding the trend in the metric that actually follows a trend–fundamental total return.  The problem, of course, is that fundamental total return is tied up in two distinct types of terms: EPS growth and dividends.  To properly analyze the trend in that return over time, we need a way to convert the terms into the same type of term, so that they can be added together to produce a single term, a single index.

The optimal way to solve the problem is to convert the dividends into a type of additional EPS, and then add the additional EPS to the actual EPS.  Then, we will end up with one single term that grows over time at a consistent trend rate, whose position relative to trend we can examine and make informed future projections based on.

That is precisely what the technique in the prior piece tries to do.  It tries to convert dividends into additional EPS by hypothetically assuming that dividends are diverted into share buybacks.  It then adds the additional EPS from the hypothetical share buybacks to the EPS that actually occurred, so as to form the unified, all-in-one term being sought: Total Return EPS.

The Equivalence of Reinvested Dividends and Share Buybacks

Now, from a total return perspective, it doesn’t matter whether a corporation chooses to distribute its profit as dividends, or use its profit to buy back shares.

  • If it pays out dividends, the dividends will be reinvested (that’s at least the assumption that “total return” indices hypothetically make).  The reinvestments will cause the number of shares that each shareholder owns to grow.
  • If it buys back shares, its outstanding share count will shrink, and therefore its earnings per share (EPS) will grow.  Mathematically, the growth in its EPS will roughly equal the growth in the number of shares that the shareholder would have come to own via dividend reinvestment.  If the market is efficient–meaning that it properly prices value–then the shareholder will end up no better or worse off, at least on a pre-tax basis (after tax, of course, is a different story).

Another way to express the point: When a corporation buys back shares with money that would otherwise have gone to dividends, it is effectively doing the dividend reinvestments for the shareholders.  It is accumulating shares in their names, as opposed to paying money out to them for them to accumulate shares on their own, independently of the company.  In truth, the two are not perfectly equivalent–share buybacks are actually slightly more accretive than reinvested dividends, for mathematical reasons that are too tedious to try to explain.  But they are close enough.

For any equity market, then, we are free to interchange dividends and share buybacks at will.  We can rebuild total return indices on the assumption that all dividends are hypothetically replaced with share buybacks, or that all share buybacks are hypothetically replaced with dividends–the replacements, if properly constructed, will have no perceptible effect on the total return.

The technique used to build the Total Return EPS index exploits this convenient equivalence.  It assumes, hypothetically, that for all of history, all dividend money that actually got paid was not actually paid, and was instead used to buy back shares.

When we examine Total Return EPS over history, we find that it does follow a reliable trend, as theory would suggest.  Consider the following table, which shows average rolling 10 year EPS and Total Return EPS growth rates over the periods identified in the previous table.


With Total Return EPS, we see a far more consistent growth rate.  That growth rate is roughly on par with the corporate sector’s average historical return on equity, some number close to 6%, as theory would again suggest (FRED).


Buybacks: Why Fair Value Prices?

The hypothetical buybacks that are used to form the Total Return EPS are assumed to occur not at market prices, but at fair value prices–prices that correspond to an average valuation across history.  A number of readers have tweeted and e-mailed in, asking why we make this assumption.  Why not assume that the buybacks occur at market prices instead, and save the confusion?

The reason is simple.  We’re trying to build an index that captures the fundamental total return that corporations generate for their shareholders through the profits they earn, which they deliver to their shareholders in the form of dividends and EPS growth.  If we were to conduct the buybacks at market prices, then that return would fluctuate based on the market’s valuation–a nonfundamental factor that has nothing to do with those profits.

In conducting the hypothetical buybacks, we are effectively converting dividends into a type of EPS (that gets added to the regular EPS to form the Total Return EPS).  That is, instead of paying out the dividends, we are using them to shrink the S, which effectively adds more EPS (makes EPS bigger by reducing the denominator).

Now, the valuations at which the buybacks occur represents the effective rate of conversion between dividends and EPS.  A low buyback valuation will convert dividends into a large amount of EPS, as the dividends will buy back a large number of shares. Conversely, a high buyback valuation will convert dividends into a small amount of additional EPS, as the dividends will buy back only a small number of shares.

To capture the true fundamental total return, and not add nonfundamental noise associated with where market prices just so happen to be, we need to ensure that the same rate of conversion between dividends and EPS is applied in all periods.  We need to ensure that each dividend, adjusted for size, adds the same amount of relative EPS as every other, regardless of when it happens to be paid.  That’s why we assume that all buybacks occur at the same valuation: “fair value”, which in the previous piece, we defined in terms of the historical average of the Shiller CAPE).

The assumption that the buybacks that underlie Total Return EPS occur at fair value may seem trivial and unimportant, but it makes a meaningful difference.  This difference will become particularly significant when we try to use Total Return EPS to build a new-and-improved Shiller CAPE.  If we use market valuations for the buybacks in Total Return EPS, the new-and-improved Shiller CAPE will end up skewed, giving an unnecessarily inaccurate picture of the market’s true valuation.  I plan to discuss the point in more detail in a later piece on the Shiller CAPE.

An Important Clarification: The Buybacks are Hypothetical

Let me now make an important clarification, based on some of the questions I’ve received. Regular S&P 500 EPS has continually grown throughout history.  Its growth has been driven by both business expansion (real economic investment that adds capital and increases output–EPS growth driven by growth in the E) and the repurchase of equity (buybacks, acquisitions, mergers, and so on–EPS growth achieved by shrinkage of the S).

The Total Return EPS doesn’t modify any of that growth.  Rather, what Total Return EPS does is add the additional growth that would have been produced if the dividends that were paid out had instead been used, completely hypothetically, to buy back shares (or fund acquisitions, mergers, and so on).

The following schematic makes the point more clear:


Some readers have asked: in constructing Total Return EPS, why do you assume that the buybacks occur at fair value prices, when, in reality, corporations like Apple and IBM are buying back their shares at market prices?   This question misses the point.  When I talk about conducting buybacks at fair value prices, I’m not referring to those buybacks, the ones that actually happened in reality, or that are happening now.  Their effects have already shown up, or will show up, in regular EPS growth.  The buybacks that I’m referring to, the ones associated with the construction of Total Return EPS, are hypothetical buybacks–buybacks that didn’t actually happen, that aren’t happening, but that we assume happened or are happening, in lieu of dividends, so as to convert the dividend return into a type of EPS growth.

Interim Valuation: A Neglected Driver of Returns

Our assumption that the hypothetical buybacks occur at fair value highlights a crucial fact about returns that often gets missed.  Valuations matter to returns not only in relation to terminal prices–the price at which you buy and the price at which you sell–but also in relation to interim prices–the prices at which your dividends get reinvested (or, in this context, at which your CEO buys back shares in your name).  As we will later see, this effect is not small, not negligible, even though we might intuitively expect it to be.

In a future piece, I’m going to explore the impact further.  For a quick teaser, consider the following surprising result.  From 1871 to 2015, the actual annualized Total Return for the S&P 500–including the return from changes in valuation–was 6.89%. If, from 1871 to 2015, everything had been kept the same, except that interim prices had been permanently pushed up to a Shiller CAPE equal to the current value of 27.5, with the dividends reinvested at those high prices, rather than at the much cheaper prices that were actually realized, the total return would have been only 4.78%.  That’s more than 200 bps–almost a third of the historical total return–lost to this mechanism.

Remember this fact the next time you find yourself assuming that a policymaker-coddled market that always stays elevated, that never crashes or corrects, would somehow be a good thing for buy-and-hold investors.  It would not be.  The winners in such a market would actually be the impatient, weak-willed, market-timing-prone people who sell to buy-and-hold investors, when those investors go to reinvest their dividends (or when corporations go to buy back shares, which is all they seem to want to do these days).  Those people would never again have to sell at unfair prices, never again have to foot the bill for the bargains that buy-and-hold investors–the Warren Buffets of the world–have historically enjoyed.

In the next section, I’m going to present the theory that underlies the decompositions that will follow at the end of the piece, so that others can reproduce the results themselves.  If you’re not interested, feel free to fast forward to the end, where the charts are presented and discussed.  To briefly summarize, I’m going to arrive at the following two equations:

(6) Total Return EPS Growth = EPS Growth + Return Contribution from Dividends Reinvested at Fair Value

(7) Total Return = Total Return EPS Growth + Return Contribution from Change in P/E Ratio + Return Contribution from Interim Deviations from Fair Value

Along the way, I’m going to explain what each term means, and how each term is calculated from the data.

Decomposing Equity Total Returns: The Theory

In his 1981 magnum opus, Robert Shiller eloquently delineated the fundamental components of equity total return:

“Once we know the terminal price and intervening dividends, we have specified all that investors care about.” — Robert Shiller, “Do Stock Prices Move Too Much to be Justified by Subsequent Changes in Dividends?”, 1981 

We can translate this point loosely as follows:

(1) Total Return = Price Growth + Return Contribution from Reinvested Dividends

We can express price growth in terms of growth in EPS (a fundamental that gets decided by economic processes) and the return contribution from the change in the P/E ratio (a value that gets decided based on the brute forces of supply and demand mixed together in equilibrium with investor beliefs about what is a fair, appropriate, justified, responsible, sufficiently-rewarding price to pay).

(2) Price Growth = EPS Growth + Return Contribution from Change in P/E Ratio

Substituting (2) into (1) we get:

(3) Total Return = EPS Growth + Return Contribution from Change in P/E Ratio + Return Contribution from Reinvested Dividends

Now, let’s look at this last term, Return Contribution from Reinvested Dividends.  We can express this term as the combination of (a) the Return Contribution from Dividends Reinvested at Fair Value and (b) the Return Contribution from Interim Deviations from Fair Value.  The return contribution from reinvested dividends is the return that would have accrued if they had been reinvested at fair value, plus the “extra” return (positive or negative) that has arisen from the fact that, in reality, they were not actually reinvested at fair value, but were reinvested at higher or lower valuations, producing a lower or higher return.

We end up with:

(4) Return Contribution from Reinvested Dividends = Return Contribution from Dividends Reinvested at Fair Value + Return Contribution from Interim Deviations from Fair Value

Combining (3) and (4) we get a total return equation with four components:

(5) Total Return = EPS Growth + Return Contribution from Dividends Reinvested at Fair Value + Return Contribution from Change in P/E Ratio + Return Contribution from Interim Deviations from Fair Value

Now, to substitute in Total Return EPS, we recall that Share Buybacks and Reinvested Dividends are the same thing.  This means:

(6) Total Return EPS Growth = EPS Growth + Return Contribution from Dividends Reinvested at Fair Value

Inserting (6) into (5) we get:

(7) Total Return = Total Return EPS Growth + Return Contribution from Change in P/E Ratio + Return Contribution from Interim Deviations from Fair Value

These two equations, (6) and (7), are the equations that we are going to visually plot. Before we can do that, however, we need to find a way to quantify the terms in each equation.

We do that as follows:

    • (Regular) EPS Growth: Trivial.  We calculate the annualized % change between the starting and finishing values of (regular) EPS.
    • Total Return EPS Growth: Again, trivial.  We calculate the annualized % change between the starting and finishing values of Total Return EPS.  Directions for how to build the Total Return EPS index can be found in the previous piece.
    • Return Contribution from Dividends Reinvested at Fair Value: We take the difference between Total Return EPS Growth and Regular EPS Growth.  This difference equals the contribution from reinvested dividends (or, alternatively, the contribution from share buybacks–they are the same thing).
    • Return Contribution from Change in P/E Ratio: We take the difference between price growth and EPS Growth.  This difference just is the return contribution from the change in the P/E ratio.

Now, to get the final term, the Return Contribution from Interim Deviations from Fair Value, we need to build a new index.  Call that index the “Total Return EPS with Purchases at Market Prices” index.  This index is identical to the Total Return EPS index, except that the buybacks are conducted (or the dividends reinvested) at market prices rather than at fair value prices.

    • Return Contribution from Interim Deviations from Fair Value: Take the difference between the annualized growth of “Total Return EPS with Purchases at Market Prices” and the annualized growth of Total Return EPS.  This difference just is the added return that comes from buying back shares (or reinvesting dividends) at market valuations that do not always average to fair value.

Charting the Decomposition

In the following charts, I’m going to decompose–i.e., separate out–the historical S&P 500 total return into three contributing components: Total Return EPS Growth (purple), Return Contribution from Change in P/E Ratio (orange), and Return Contribution from Interim Deviations from Fair Value (blue).  I’m going to further decompose Total Return EPS Growth into two components: Return from Reinvested Dividends (identical to Return from Hypothetical Share Buybacks) (green) and Regular EPS Growth (yellow). Recessionary periods for the U.S. economy will be shaded in gray.

The decompositions will be conducted on the returns at time horizons of 10, 20, 30, 40, 50, 60, and 70 years, from 1871 to 2015.  For each time horizon, there will be two separate charts (miniatures shown below), with the first chart decomposing the total return, and the second chart decomposing the Total Return EPS.  As with all numbers in this piece, the growth rates and returns are real, properly adjusted for inflation.

10 Years


A brief discussion on how to read the chart.  The x-axis has two dates.  The upper is the starting date for a period, the lower is the ending date–in this case, 10 years later.

Consider the slice of the chart that begins with 1989 and ends with 1999.  I’ve boxed it in red below:


The purple, the Total Return EPS growth, was roughly on par with the historical average, around 6%. What this means is that from 1989 to 1999, the sum of the return from dividend reinvestments (or hypothetical share buybacks–same thing) at fair value and the return from regular EPS growth amounted to 6% per year.

The orange, the Contribution from the Change in P/E ratio, was enormously positive, adding more than 10% to the return.  Of course, that’s consistent with what we remember. In 1989, valuations were reasonable; in 1999, they were in a bubble.  The transition from normal valuations to bubble valuations produced phenomenal returns for shareholders. In hindsight, of course, nothing was actually “produced”–returns were simply pulled forward from the future, stolen from those that bought in at the end.

The blue, the Contribution from Interim Deviations from Fair Value, was actually negative, subtracting approximately 1% from the return.  This also checks with what we remember. From 1989 to 1999, valuations were substantially above average.  There were only a few very mild corrections that took place–certainly nothing resembling a crash.  For the most part, the market just went straight up.  The above average valuations depressed the return from reinvested dividends relative to the alternative of a market at fair value (which is what Total Return EPS is indexed to).

Notice that as we move to the right in the chart, towards starting dates in the early 1990s, the Contribution from Interim Deviations from Fair Value gets even more negative, approaching -2% per year.  To understand why, recall that the market in the late 1980s and early 1990s was actually valued fairly attractively.  When we move to the right, those years drop out, and get replaced by the acute phase of the tech bubble, when the market was radically expensive.

The thin black line is the actual total return, which almost exactly equals, within a few bps, the sum of the contributors, as it should.  Note that I’m calculating the actual total return not by summing the contributors, but by building an entirely separate total return index, using the normal methods for doing so.  The chart can therefore be taken as empirical proof that the decomposition is analytically correct–the numbers, calculated by separate methods, add up perfectly, as they should.

The chart of the decomposition of Total Return EPS, shown below, follows the same structure:


The chief thing to notice in the Total Return EPS chart is how the mix of the return has shifted from green (reinvested dividends, or alternatively, hypothetical share buybacks) to yellow (EPS growth).  This shift will become more clear and compelling as we move to longer time horizons, where the interfering cyclical noise will get smoothed out.

20 Years



30 Years



40 Years



50 Years



60 Years



70 Years




On longer time horizons, we see certain patterns crystallize.  The Total Return EPS, shown in purple, converges on a trend growth rate slightly below 6% annualized.  The green–the dividend (buyback) return–shrinks, while the yellow–the return from regular EPS growth–expands, keeping the sum of the two–Total Return EPS growth–on trend.

The shift from green to yellow is the shift in corporate preference visualized–away from dividends and towards growth.  When we try to conduct trend analysis on regular EPS–the yellow–we inevitably miss this shift, and therefore arrive at faulty conclusions.  What we need to analyze instead is the black line, the sum, the Total Return EPS, which has held to its trend comparatively well over the long-term.

In earlier periods of the charts, frequent market cheapness contributed meaningfully to the return.  But over time, as the market has become more efficient, less prone to violent downturns and crashes, that contribution has faded.  Notice that the blue–the contribution from interim deviations from fair value–is much thinner now than it used to be.  In charts of shorter time horizons (10 or 20 years, for example), it has even gone negative.  The shift to a negative contribution reflects the secular increase that has occurred in the market’s valuation, the valuation at which dividends are reinvested.  If, going forward, the market successfully holds steady at its currently elevated valuation, successfully avoiding the pull of downturns and crashes, then the blue will stay negative, and total returns will underperform the historical average accordingly–simply by that mechanism, never mind the others.

Investors need to understand that they can’t have it both ways: they will have to either accept historical levels of volatility, which will allow them to reinvest their dividends at cheap prices every so often (and allow their CEOs to buy back shares and acquire companies at those same prices), or they will have to accept lower than normal historical returns.  The growing corporate preference for buybacks (and acquisitions, and mergers) as a low-risk, tax-efficient alternative to risky capital expenditure will only exacerbate this impact.

At present, nearly 100% of current S&P 500 EPS is being used to fund dividends and buybacks–a trend that looks set to continue.  Going forward, interim valuations–which will influence the returns that those dividends and buybacks produce–are therefore likely to be even more impactful than they were in the past.  If valuations remain where they currently are–at levels that would qualify as historically expensive even on the uncertain assumption that profitability will remain at record highs–future returns are likely to suffer accordingly.

Posted in Uncategorized | Leave a comment

Introducing the Total Return EPS Index: A New Tool for Analyzing Fundamental Equity Market Trends

rshillerIn late December of 2010, with the S&P 500 pushing through the mid 1200s on the heels of QE2 exuberance, my favorite financial economist–the great Robert Shiller–made what will likely turn out to be very inaccurate prediction.  To be fair to Shiller, it wasn’t really a “prediction”, but more of an “estimate.” He estimated that the S&P 500 would trade at 1430 in the year 2020.

Wait, did he mean 2430?  No.  He meant 1430–the level the market was at in late 2012, just before it went on its epic run.  He estimated that the market would be there eight years later, in 2020.

His estimate could still end up being right.  But, at this point, it will take a lot of luck. The trough of a 30% market downturn–from current levels, not higher levels–will have to occur exactly in 2020.  Either that, or the market will have to suffer an even larger downturn, and then recover to the target by 2020.  Neither of these chance occurrences seems to have been intended as the basis for the original estimate.

Rather than chuckle at the wrongness of the estimate, we may want to consider its implications for the current market.  Here we have an excellent financial economist, one of the few to have accurately identified both the tech bubble and the housing bubble in real-time, estimating that in five years the market will be 30% below where it is today.

What is that telling us?  Maybe it’s telling us that nobody–not even the “experts”–knows what’s going to happen in the market, and that all of this “analysis” mumbo-jumbo is just a front that people offer up in defense of stances that they’re already entrenched in for other reasons (emotional, dispositional, moral, ego-related, career-related, business-strategy-related, because they’ve already gone on record, cemented an identity, staked a reputation, and need to be “right”, etc.)

But maybe it’s telling us something else.  Maybe it’s telling us that this market has been stretched very thin, well beyond the limits of defensible valuation, such that well-reasoned prior estimates for where prices would now be are missing the mark by miles. If true, that’s great news for those that were buying equities in 2010.  It’s not great news for those that are buying equities now.

Historical EPS: The Trend is Not Your Friend

In the CNBC interview, Shiller explained the basis for his estimate:

“The problem with the traditional price earnings ratio is that earnings are just too volatile from year to year…  We’re talking ten years out.  So I’m going to go back to 100 years.  The growth of real inflation directed earnings is surprisingly low.  From 1890 to 1990 it was only 1.5 percent a year…  I take earnings and I extrapolate them out at 1.5 percent from where they–S&P 500 earnings–are now and then I apply a price earnings ratio of 15, which is the historical average for 1890 to 1990.”  – Robert Shiller, CNBC, December 31st, 2010 

The best way to understand Shiller’s logic here would be to refer to a chart of S&P 500 earnings per share (EPS) relative to its historical “trend.” The CEO of Business Insider, Henry Blodget, recently highlighted a nice version of such a chart, put together by GMO’s James Montier.

montierdaChris Brightman of Research Affiliates presented a similar chart in a 2014 piece entitled “The Profits Bubble.”  I’ve written over the chart in maroon:


The logic works like this.  We assume that EPS oscillates around, and eventually reverts to, a long-term trend (the gold line in the first chart, the black line in the second).  To estimate EPS for a date in the future, we determine where the trend will be on that date (that is, we assume an eventual reversion to it).  We then apply a “normal” P/E multiple to the EPS estimate to get an estimated price.

As of 2Q 2010, reported S&P 500 EPS was $67.10, almost perfectly on the historical trend.  So Shiller extrapolated the trend growth rate–1.5% real, plus 2% inflation–out 10 years.  He then applied a 15 P/E ratio.  At 1.5% + 2% = 3.5% growth per year over 10 years, $67.10 becomes $95.  15 times $95 is roughly 1430–Shiller’s 2020 price target.

Before I share my chief concern with this logic, let me say that I agree with Shiller’s (and Blodget’s, and Montier’s, and Brightman’s) thesis that long-term returns from current prices will likely be disappointing, meaningfully less than the historical average of 6% real. I would much rather be with them in that debate than with those that are arriving at optimistic conclusions by naively extrapolating their own recent experiences, pretending that markets are always as kind and rewarding as they have been to U.S. investors over the last 6 years.  As a strategy, extrapolation can work, but it doesn’t generally work at the tail end of a tripling of prices that has left the market historically expensive on pretty much every measure available.

That said, Shiller’s trend-based argument is flawed.  The historical trend growth rate of real S&P 500 EPS that he uses–1.5%–does not apply to the current market.  The reason is straightforward, and centers on the impact that secular changes in the dividend payout ratio have had on growth.  EPS in the current market grows at a much faster pace than it did in the past, because a much larger share of current profit is devoted to growth-generating reinvestment, with a much smaller share devoted to the payment of dividends.

Now, here’s the problem.  We don’t have a clear, reliable, uncontroversial way to account for the impact that changes in the dividend payout ratio have had on earnings growth over time.  We are each then left to estimate the impact for ourselves.  Not surprisingly, those of us that want the market to go down estimate it to be a small impact, and therefore ignore the risk that it might undermine trend-based arguments.  Those of us that want the market to go up estimate it to be a large impact, and dismiss trend-based arguments altogether. Both approaches are wrong.

In this piece, I’m going to attempt to solve the problem.  I’m going to introduce a new type of EPS index, called the “Total Return EPS” index.  The Total Return EPS index represents what EPS would have been if corporations had never paid any dividends at all, but had instead used all of their dividend cashflows to buy back their own shares (or acquire or merge with other existing companies–there’s no difference in this context).  When regular EPS is converted to Total Return EPS, growth trends can be analyzed without concern for the distortive impact that changes in the dividend payout ratio have had over time, because the dividend payout ratio for all eras gets reduced to the same common denominator, 0%.

In previous pieces (on foreign share, financial share, technology share, sectoral balance of payments, and wealth redistribution), I’ve argued that corporate profits, though overextended, are not as overextended as they might initially appear to be.  A study of Total Return EPS confirms this argument. Regular S&P 500 EPS–shown in the Blodget/Montier/Brightman charts above–is a whopping 75% above its historical trend. But Total Return EPS is only 28% above its historical trend.

After introducing the Total Return EPS index, I’m going to teach the reader how to quickly and easily build it using Shiller’s publically-downloadable spreadsheet.  I’m then going to use it to redo his 2020 S&P 500 estimate, based on the data that was available in 2010.  In contrast to the bearish 1430 number that he arrived at, my estimate will come out to around 2150–still bearish, but not absurdly so, and significantly more likely to hit the mark, at least in hindsight.

It turns out that we can use Total Return EPS to do all sorts of interesting things: accurately predict future growth based on position relative to trend, decompose and visualize historical returns in terms of their contributing components, estimate profit margins during periods in the late 19th and early 20th century when the data necessary to calculate them was not available, construct new-and-improved Shiller CAPEs that allow for valid comparisons across history and across countries, and many more.  But those would be too much to discuss in one piece, so I’m going to save them for later.

Here, I’m simply going to introduce the concept, and use it to generate a new 2020 S&P 500 estimate.  Before I do that, I’m going to briefly outline the factors that have driven the secular decline in the dividend payout ratio.

Secular Decline in The Dividend Payout Ratio

An early dividend announcement from the Bank of North America.  Chartered in 1781 at the urging of Alexander Hamilton, the Bank of North America was the first Central Bank of the United States. It merged to form Wachovia Bank, and now exists as a part of the present day Wells Fargo ($WFC).

The market’s dividend payout ratio, which is the percentage of earnings that goes to dividends rather than to growth-generating reinvestment, has declined significantly over the last century.  In the chart below, I use the “Shillerized” version of the dividend payout ratio to illustrate the point.


Why has the dividend payout ratio fallen so much?  We can point to at least three interconnected reasons:

First, the investment world is not as endeared to the idea of receiving dividends as it used to be, and therefore corporations are not under as much pressure to pay them.

“Do you know the only thing that gives me pleasure?  It’s to see my dividends coming in!” — John D. Rockefeller, quoted by John Lewis in Cosmopolitan, 1908.

Shareholders are perfectly comfortable, and often prefer, to see their idle excess funds deployed into other accretive activities–business expansion, share buybacks, acquisitions, mergers, and so on.  The increased liquidity and efficiency that the modern market affords ensures that any wealth created in these activities will immediately accrue to the benefit of shareholders in the form of higher stock prices.  In earlier periods, that was not necessarily the case.

Second, SEC rule 10b-18, originally issued in 1982, gave corporations that engage in share repurchases “safe harbor” from claims of market manipulation, removing a key legal risk that had otherwise discouraged the practice.  With that risk removed, share buybacks became a virtual no-brainer, an efficient and perfectly legal way of distributing capital to shareholders without causing them to incur unwanted tax liabilities.  When share buybacks displace dividend income, the money that continuing shareholders would otherwise have had to pay in dividend taxes stays inside the asset, where it is able to compound over time.

But what about situations where the market is expensive? Shouldn’t they encourage a shift away from buybacks, and towards dividends?  No.  Given the way that most people invest–with the “dividend reinvestment” option checked off–any dividends that are paid in lieu of buybacks will lead to the same outcome.  They will be used to purchase shares at the same expensive market prices–it’s just that in the dividend case, the purchases will take place outside the company, rather than inside.  The only relevant difference will be in the tax, which will be paid now rather than later.

Third, increased reliance on stock options as a form of performance-based employee compensation has created obvious reasons for corporate managers to prefer internal reinvestment to dividends.  When corporate managers pay dividends, the wealth contained in those dividends leaves the company, and no associated EPS growth is generated from it. Their claim on the wealth, which they hold through their stock options, is therefore lost, and any compensation they would have received for the EPS growth that the wealth would have produced, they do not receive.  But when they reinvest the wealth internally, it stays inside the firm, preserving their stock-option-based claims on it, and creating a direct boost to EPS that brings them that much closer to their performance targets.

It should come as no surprise then, that buybacks, acquisitions, and mergers now dwarf dividends as a destination for excess funds.  The following chart, borrowed from Michael Mauboussin’s fantastic piece on capital allocation trends in the U.S. corporate sector, illustrates the phenomenon.  The amount of capital recycled into mergers, acquisitions, and share repurchases is now roughly six times as large as the amount of capital returned to shareholders in the form of dividends.  100 years ago, that amount would have been an imperceptible fraction of the dividend amount.


Constructing the Total Return EPS Index

According to S&P corporation, trailing twelve month (ttm) reported EPS for the S&P 500 is $102.77.  What would it be today, if, starting in 1871, all of the dividends that were paid out to shareholders had instead been used to repurchase shares (or acquire or merge with other companies)? That’s the question that the Total Return EPS index is trying to answer–not only for today’s date, but for all dates in market history.

To get an answer, we start with Shiller’s familiar spreadsheet, which contains average monthly prices and reported earnings for the S&P 500 and its pre-1957 ancestry, with data obtained from S&P corporation and the Cowles Commission.  We can use this data to create a log chart of EPS over time.  The EPS is shown below in red, with the exponential trendline shown in black and recessionary periods shaded in gray.  Recall that exponentials look linear on a log scale.


This chart is essentially the Blodget/Montier/Brightman chart shown earlier.  As we see in the chart, current EPS is way above trend–by almost 75%.  If, right now, it were to fall back to trend, it would have to fall all the way down to $60.  At current prices, the index would then be valued at roughly 35 times earnings.

Now, let me make two brief points about the chart:

First, the chart shows inflation-adjusted EPS.  Every earnings data point charted in this piece is inflation-adjusted to today’s dollars.  When studying equity markets over the long-term, we absolutely have to inflation-adjust.  To not inflation-adjust would be to retain a source of noise and volatility in the data that we cannot accurately model or predict. Any patterns that we subsequently identify would automatically be called into question as potential cases of coincidence-exploitation and data-mining.

Second, the massive drops seen in 2003 and 2009 were the result of the application of accounting standards that were not applied to prior eras and that do not reflect true earnings performance.  In a perfect world, we would correct the chart accordingly, replacing the drops with more accurate estimates of earnings for the affected periods.  In some contexts, we will make such a correction, substituting operating EPS for reported EPS.  But in other contexts, we won’t, because it’s not necessarily needed, and will create false grounds for bears to dismiss what we’re trying to show.

In terms of analyzing the current equity market, it’s not crucial that we correct for the drops, because we’re not looking at Shiller-type averages of past EPS.  We’re only looking at trailing twelve month (ttm) EPS.  From the perspective of the ttm period ending in February 2015, the accounting-driven implosions of 2003 and 2009 have long since dropped out of the metric.  That said, when you look at the chart and attempt to glean patterns and trends from it, remember that the large drops in those periods are visual distortions, and are not supposed to be there.

The following chart shows regular EPS with the 2003 and 2009 periods corrected.  A mix of Bloomberg’s and S&P’s operating EPS series is inserted in the place of reported EPS starting in January 1998.


If you look closely, you will notice that the EPS values in this chart and the uncorrected chart fail to adhere to a consistent long-term growth trend.  From the 1870s all the way through to the 1930s, there was essentially zero EPS growth.  From the 1930s through to the 1990s, there was more EPS growth, but still not all that much. Then, starting in the mid 1990s, EPS growth exploded, achieving in a 20 year period what had earlier taken 60 years to achieve.  How can we fit all of these periods onto the same trendline?  We can’t, which is why the chart is ugly and uninspiring–a straight line forced onto a highly malformed time series.

To turn the time series into something more visually compelling, we need to build an index that puts all periods of history on an equal footing with respect to the tradeoff between dividend income and earnings growth.  If we do that, everything will fall into place.

Enter the Total Return EPS Index.  The Total Return EPS Index models what EPS would have been if the dividend payout ratio had been 0% at all times in history, with all dividend cash flows instead used to repurchase shares (or acquire or merge with existing companies).  The Total Return EPS index reduces all periods to that common denominator, so that different periods can be accurately compared with each other, and a true underlying growth trend legitimately extracted.

Now, there’s one crucial, counter-intuitive caveat that we need to insert.  The index should not assume that shares are bought back at market prices.  For if it makes that assumption, then periods of history where stocks were cheap will show abnormally high growth, driven by the high rate of return that the share buybacks, conducted at cheap prices, will have produced.  Conversely, periods of history in which stocks were expensive will show abnormally low growth, driven by the low rates of return that the share buybacks, conducted at expensive prices, will have produced. We’re trying to determine the historical trend in EPS growth.  Valuation–the cheapness and expensiveness of stock prices in any given period–has nothing to do with that trend.  We therefore need to remove its effects from the index.

Instead of assuming, as most “total return” indices do, that shares are bought back at market prices, our new index will therefore assume that they are bought back at prices that reflect the same underlying value across history.  Such an assumption is needed to ensure that all periods of history are placed on an equal footing, to allow for accurate comparisons between them.

But how shall we measure value across history?  The best available tool–far from perfect, but adequate to the task–is the Shiller CAPE.  We’re going to assume, then, that shares are bought back at prices that correspond to a constant Shiller CAPE across history–in this case, a Shiller CAPE equal to the historical (geometric) average of 16.59, which we will arbitrarily refer to as “fair value.”

Unfortunately, for the 1871-1881 period, there aren’t 10 years worth of trailing data on which to calculate a Shiller CAPE.  So, for that period, we’re going to have to use the simple ttm P/E ratio.  We will assume that shares are bought back at a ttm P/E ratio equal to the historical (geometric) average of 15.54.

Now, to the spreadsheet itself.  Here’s the concept.  We start by creating a column that contains an arbitrary starting share count. We make the initial share count, in the first month of the series, equal to 1.0 (or whatever number you want).  We then shrink the share count each month by the amount of shares that the monthly dividend would have been able to purchase at fair value prices (again, “fair value” as measured by the Shiller CAPE for periods after January 1881, and the ttm P/E ratio for periods before January 1881.)

The images below show the technique in practice from 1871 to 1881.  In each month, we shrink the share count by the amount of shares that the dividend for that month would have been able to buy, assuming a fair value ttm P/E multiple of 15.54.


In January of 1881, the first month where a Shiller CAPE reading is available, we shift to that measure of fair value.  In each month, we shrink the share count by the amount of shares that the dividend for the month would have been able to buy, assuming a Shiller CAPE of 16.59.


What we end up with is a gradually contracting share count that tracks what would have happened to the actual share count if all dividend payouts had instead been used to repurchase shares.  Dividing the actual reported EPS by that share count, we get the Total Return EPS, which is what the EPS would have been under a 0% dividend payout ratio.


But if all dividends had instead been used to repurchase shares, then the market’s price on the subsequently higher earnings would have been higher. We therefore need to calculate a new price index, a price index that reflects what the price would have been, on the assumption that the market had valued the now-higher earnings in the same way, with the same P/E multiple.  To do this, we divide the actual price by the contracting share count, just as we divided the actual EPS by that share count.  We end up with the Total Return Price Index, which goes together with the Total Return EPS index (we will use this price index in a future piece to calculate a new-and-improved Shiller CAPE).


The following chart plots the uncorrected Total Return EPS on a log scale.  At present, it’s roughly 28% above its historical trend, versus 75% for the uncorrected regular EPS.


The following chart plots the corrected Total Return EPS on a log scale.  At present, it’s roughly 21% above its historical trend, versus 58% for the corrected regular EPS.


Notice, in particular, the dramatic improvement in the consistency of the trend.  The growth from the 1870s to the 1930s, for example, is on par with the growth from the 1930s to the 1990s, which is on par with the growth from the 1990s to now.  In the charts of regular EPS shown earlier, that was not the case.  The first period, when the dividend payout ratio was high, had virtually no growth; the second period, when the dividend payout ratio was lower, had moderate growth; the third period, when the dividend payout ratio was lowest of all, had substantial growth.

Annualized, the historical trend in Total Return EPS growth comes out to around 5.5% per year for the uncorrected case, and 5.7% per year for the corrected case–both close to the generic 6% number that represents the U.S. corporate sector’s historical return on equity.  This is not a coincidence.

A New 2020 Estimate

I’m now going to redo Shiller’s 2020 price estimate using the corrected version of Total Return EPS.  As of June 2010, nominal EPS was $69.43.  Total Return EPS was 7.5% below its historical trend.  Boosting nominal EPS by an amount sufficient to bring Total Return EPS back to trend, we get $75.  So, if EPS in 2010 had been $75, Total Return EPS would have been perfectly on its trend.

The Total Return EPS, as constructed, grows at an average rate of 5.7% per year. Inflation adds roughly 2%, and the dividend, using the yield at the time, subtracts roughly 2%.  So we get 5.7% + 2% – 2% = 5.7% as the expected nominal EPS growth rate. Applying that expected growth rate out over a 10 year period to the starting $75 number, we end up with $131.  That’s the EPS estimate that this trend-based method produces for the year 2020. Multiply by 15 and we get 1965, versus Shiller’s overly bearish 1430.

Now, the 15 number that Shiller used as the “average” ttm P/E ratio represents the average ttm P/E ratio from 1890 to 1990.  An average from a later, more relevant slice of history would probably be more appropriate.  The (geometric) average ttm P/E ratio from the S&P 500′s inaugural year of 1957 to present is 16.46.  Using that number, we arrive at a year 2020 S&P 500 price estimate of 2156 (which, interestingly, is roughly the same as the year 2020 estimate–2154–that I arrived at using a more direct method in a piece from 2013.)

These estimates assume a reversion of both earnings and P/E back to their properly-measured historical trends.  Such a reversion may not occur.  But if it were to occur, it would not be anything historically extreme–just a drag on EPS growth that brings final EPS to a value ~15% to ~20% below where it would be if it grew at the trend rate from today forward, coupled to a downshift in the P/E ratio from the current 18+ back to the average of around 16.5.

It may be hard to believe that such a reversion could ever take place in this teflon, never-lose, never-give-a-bear-a-damn-thing U.S. equity market, but go ahead and #timestamp that estimate, we’ll come back to it.  A number of impeding forces are likely to push back on U.S. equity performance between now and 2020, that have only recently started to rear their heads.

Posted in Uncategorized | Leave a comment

Janitor to Multimillionaire? Not In This Market

Last week, CNBC profiled the inspirational story of Ronald Read, a gas station attendant and janitor from Vermont who amassed an $8MM fortune simply by investing portions of his small salary into high-quality, dividend-paying U.S. equities.

Judging from the familiar names that Read is reported to have owned–AT&T, Bank of America, CVS, Deere, GM, GE, and so on–alpha generation doesn’t appear to have been the main driver of his success (though it was probably a contributor).  Over the last 40 years, those names have roughly tracked the large cap index, as you can see in the chart below.


The main driver of Read’s success appears to have been discipline: saving diligently and investing efficiently, and sticking with the process over a lifetime.  That’s good news, because it means that anyone can potentially do what Read did.  No special stockpicking talent is required.

Would it be possible for a present-day investor in Read’s employment circumstances to amass the fortune that Read amassed, if that investor were to apply Read’s discipline to a U.S. equity strategy?  The actual numbers were crunched, and the surprising conclusion was yes.


Per the analysis, if an investor wanted to amass an $8MM fortune, she would need to put around $300 per month into the market.  Compounded at an 8% rate over 65 years, the ensuing investment pile would be worth around $8MM today.  Now, $300 per month is a lot of money for someone working a minimum-wage job.  But it’s not a prohibitive amount. $10 per hour times 40 hours per week times 4 weeks per month is $1600.  On a pre-tax basis, $300 is roughly 20% of that amount.

But there’s a mistake in this analysis.  It makes no adjustments for inflation. Sixty-five years ago, $300 per month was a very large amount of money–150% of the monthly wage of the average non-supervisory worker, which was itself almost double the federally legislated minimum wage.  No janitor on earth would have been able to afford it.

Now, to be fair, the 8% return assumption used in the analysis might refer to a real, inflation-adjusted return, rather than a nominal return.  If so, then the analysis is based on a return assumption that will almost certainly prove to be wrong.   There have only been a few periods in U.S. history where a long-term buy-and-hold investor could have earned 8% after inflation.  Needless to say, those periods–marked by widespread skepticism and obliviousness towards equities, and therefore extremely cheap equity valuations–had very little in common with the current market environment.

The following chart shows the annualized real total return of the S&P 500 for 65 year periods beginning in 1871 (endpoint 1936) and culminating in 1950 (endpoint today). All dividends are reinvested at market prices.  The 1932 bear market low is circled in red for reference.


As you can see, the real total return only breached 8% on two occasions: first, during the troughs of the Great Depression (specifically, from April of 1932 into 1933, and then after the 20% correction that ended in September 1934, which took the index back to its January 1932 level), and second, during the selloff that occurred in the months after the Pearl Harbor attack (with a final bottom in April of 1942, which again took the index back to its January 1932 level).  The market’s valuations at the lows of those periods, as measured by the Shiller P/E ratio, were 5.5 and 8.5 respectively, versus 27.2 today. Notably, in the case of the Great Depression, the return over the subsequent 65 year period was artificially boosted by the valuation extreme seen at the endpoint–1934 plus 65 years conveniently equals 1999 (party like it’s).

To get a better answer, let’s redo the analysis.  How much would an investor have to put aside each month to have $8MM at the end of a long investment lifetime, one that begins today, say, at age 25, and that ends at Ronald Read’s final age of 92?  Rather than apply an arbitrarily chosen annualized real return assumption (and then argue over whether the assumption will prove to be accurate going forward), we can just run the simulation directly, using the numbers that the market actually produced during Ronald Read’s investment lifetime.

What we’ll need to do, then, is calculate the amount, in 2015 dollars, that an investor would have had to have put into the S&P 500, each month starting in 1948, to end up with $8MM today.  Comparing that amount to the current salaries of low-wage workers, we will be able to see whether or not the strategy is realistically affordable.

To have produced an investment pile worth $8MM today, on the assumption that all investments occur in a tax-sheltered account, with no transaction fees and no management fees, an investor would have had to have put aside $524 each month in 2015 dollars.  $524 is a little over 30% of the gross monthly minimum wage.  The investor would therefore have had to have found a way to live on a bit less than $1100 per month, before taxes.  Such a frugal feat would have been extremely difficult to accomplish; in the typical real-life situation, where there are actual mouths to feed, it would have been impossible.

And note that taxes are not negligible here.  Even if we assume a 0% income tax (netting out the income transfers that the worker would be eligible to receive), the combined 7.65% that the worker would have to pay in social security and medicare taxes (on the gross wage) would reduce the after-tax, after-investment monthly income to a number below $1000.

Of course, as we move up the income chain, the prospects get brighter.  The average hourly wage for non-supervisory workers in the U.S. is around $21, more than double the federally legislated minimum wage.  $21 per hour comes out to around $3400 per month. Taking out income and payroll taxes, and assuming no income transfers received, the after-tax income falls to around $2800.  It’s reasonable to think that a determined worker could manage to take $524 out of that amount each month.  And so even if the numbers don’t work out for janitors and gas attendants, they might still work out for average employees.  The redemptive, feel-good conclusion that motivated the profile is therefore preserved: that Wall Street can, in fact, make average people, who have no special stockpicking talents, into multimillionaires.

But there’s another problem.  In order for present investors to be able to earn the kinds of returns that would transform contributions of $524 a month into an $8MM sum over a 65 year period–the kinds of returns that the market actually saw from 1948 until today–the market will have to do that dirty thing that it did multiple times from 1948 until today, the thing that those of us who are invested in it pray for it not to do: CRASH.

It’s true that $524 per month invested in U.S. equities from 1948 to present would have produced an investment pile worth $8MM.  But those monthly contributions, and the quarterly dividends earned along the way, would have been invested and reinvested into the market at the ultra-cheap valuations that the many crashes, corrections and bear markets seen from 1948 onward produced.  Make no mistake: those valuation depressions were absolutely critical to the attractive returns that investors employing the strategy would have earned–and that Ronald Read did earn. Without them, the returns to the strategy would have been far less impressive.

Right now, we are not anywhere near the kinds of depressed valuations that helped to power the outsized returns of the post-war era, not even on measures that the most extreme of market optimists might otherwise embrace.  And so, unless the market sees meaningful downside (or long, frustrating water-treading) in the years ahead, downside (or water-treading) that makes stocks genuinely cheap again, investors employing Read’s discipline should not expect to achieve his lucrative outcome.

What if stocks from 1948 to 2015 had never crashed, never corrected, never entered ugly bear markets that depressed their valuations?  What if their valuations had instead always equaled today’s valuation, 27.2 on the Shiller P/E?  How much would an investor’s monthly contributions have had to have increased, to have produced the same $8MM in the final tally?  It turns out that we can give an exact answer to this question.

We know what the S&P 500′s earnings were at each point in time from 1948 to 2015.  We can therefore calculate what its prices would have been at all times during the period if its valuation had stayed constant at today’s level, a Shiller P/E equal to 27.2.  We can then calculate what size of contribution would have been required to produce the same $8MM investment pile, assuming that the contributions had been invested, and the dividends reinvested, at those hypothetical prices.

The following chart shows the actual S&P 500 price index alongside a hypothetical price index which postulates a Shiller P/E ratio that always equals the current value–27.2. Note that the y-axis is log scale, and all prices are nominal.


As stated earlier, to end up with $8MM by making regular investment contributions (and reinvesting dividends) into the blue line (the actual S&P 500 over the period), an investor would need to commit $524 (in 2015 dollars) per month.  To end up with the same fortune by making regular investment contributions (and reinvesting dividends) into the red line (the hypothetical S&P 500 that never crashes, whose valuation is always equal to today’s valuation, a Shiller P/E of 27.2) an investor would need to commit $1831 per month–almost four times the earlier amount!

The average non-supervisory worker, who makes $2800 per month after tax, cannot afford to plow $1831–65% of her net income–into the stock market.  And so unless stock valuations are going to retreat markedly from current levels (or spend very long periods of time going nowhere), investors shouldn’t expect to be hearing about Ronald Read stories in the next go-round.

Now, the Shiller P/E might not be the best way to measure valuations in the current market.  But even on valuation measures that are more generous, the simulation still points to a substantial increase in the required contribution.  Per S&P Capital IQ, the current trailing twelve month P/E ratio on operating earnings is around 18.5. To have generated $8MM from a market that always held at that valuation, an investor that started in 1948 would have had to have contributed $1005 per month–roughly double the earlier amount.  And note that in using a one year earnings measure to gauge valuation, we are implicitly assuming that profit margins over the next 65 years will average out to their current record-high levels–possible, but a very risky assumption to make.

In all likelihood, present and future valuations will prove to be more important to returns over the next 65 years than they were to returns over the last 65 years.  The reason why is that the growth of the population is set to slow, with the average age set to increase significantly.  Slowing population growth and a significant upward shift in the average age, towards the elderly, implies reduced aggregate demand growth, and therefore a reduced need for expansive corporate investment.  Corporations, if they want to do well for their shareholders, will therefore have to shift their capital allocation strategies away from traditional capital expenditures towards what might be perjoratively described as “capital recycling”–the payment of dividends (which get reinvested into the market) and the conduct of share buybacks (which are essentially identical to reinvested dividends in terms of their effect on total return).  But the rate of return that dividends (reinvested at market prices) and share buybacks (repurchased at market prices) produce is strongly dependent on the valuations at which the reinvestments and repurchases occur.  It follows that as the corporate sector shifts towards capital recycling as an allocation strategy (a shift that is already well underway), valuations–not only at the moment of purchase, but also during the entirety of the holding period–will become increasingly important to the market’s return prospects.

To illustrate the point, imagine a demographically and technologically stagnant future world where dividends are punitively taxed, and where capital expenditures, in excess of depreciation, are neither needed nor profitable–a world where such expenditures do nothing but fuel competition, deflation, and profit margin shrinkage–put simply, a world that is Japan.  If corporate managers in such a world are good stewards of capital, they will deploy 100% of their earnings into share repurchases. EPS growth will then be entirely determined by how much the repurchases contract the S, the share count.  But the amount by which a given repurchase event contracts the S, the share count, is determined by the repurchase price–and therefore, the valuation.  And so, in such a world, expensive valuations will depress EPS growth, and by extension, investor returns.  The simulation above bears this out, with reinvested dividends as the proxy for share buybacks.

Crucially, a large portion of the EPS growth that was realized over the last 65 years was the result of expansive capital formation, the net building of new things that produced new profits, and an EPS that grew by the E.  In a futuristic Japanese world where that driver of EPS growth is removed, and where all EPS growth results from repurchases, a shrinking S, the returns to an always-expensive scenario will be that much weaker.

If you’re an investor with a short time horizon, you should want valuations to stay high, or even better, go higher, into a bubble, so that you can get the most out of your holdings when you cash them out.  But if you’re a disciplined investor that is in this for the long term, particularly a 20-something, 30-something, or even early 40-something, with a lot of income yet to be earned, you should not want valuations to stay where they are.  You definitely should not want them to go higher, into a bubble.  Instead, you should want the opposite of a bubble, a period of depressed valuations–the lower the better.

Granted, a rapid downward move in the markets, towards valuations that are genuinely cheap, would entail the pain and regret of mark-to-market losses on present holdings.  But that pain and regret will only be short-term.  In 20 or 30 or 50 or 65 years, the paper losses, by then evaporated, will have been long since forgotten, having proven themselves to have been nothing more than opportunities to compound wealth–monthly contributions, reinvested dividends, and share buybacks–at high rates of return.  As Ronald Read’s example shows, the compounding adds up over time, allowing disciplined investors–even those of modest means–to build surprising fortunes.

Posted in Uncategorized | Leave a comment

Technology and Finance: Drivers of a Profit Margin Explosion

In this piece, I’m going to show that the profit margin expansion seen in the U.S. corporate sector over the last two decades has been driven largely by gains in the financial and technology sectors.  I’m then going to examine arguments for and against the sustainability of this shift.

Profit Margin Contributions By Sector

The following chart shows the aggregate net profit margin of publically-traded U.S. equities with market capitalizations greater than $200MM from January 1964 to October 2014.  Recessionary periods are shaded in gray.


The next chart separates out the contribution to the aggregate profit margin by sector. For each sector, the colored area represents the individual earnings of the sector divided by the total revenues (sales) of all sectors.  Note that the sum, the black line, is just the aggregate profit margin shown in the previous chart.


Notice the rising contribution from the financial sector (light purple) and the technology sector (olive green), and the falling contribution from the other sectors in aggregate:



In January of 1964, financial and technology sector earnings contributed 0.49% to the aggregate profit margin, which was 6.60% at the time.  Today, they contribute almost seven times that amount, 3.42%, to an aggregate profit margin of 8.09%.

Changes in  Sectoral Revenue Contributions

A better way to think about what has happened here is to think in terms of sectoral revenue contributions: revenues of individual sector as a percentage of the total revenue of all sectors.  The following chart shows the evolving revenue contributions of each sector, from January of 1964 to October of 2014.


As the chart illustrates, the revenue contribution of the combined financial and technology sectors–the amount of total revenues that are revenues from those sectors–has increased substantially over time.  In January of 1964, the revenue contribution was 5.41%.  Today, it is 23.83%–almost a quarter of the total.



This change in revenue contribution matters because technology and financial sector revenues tend to be earned at higher profit margins than the revenues of other sectors: historically, 7.74% for technology and finance, versus 5.19% for the rest.  And so if the revenue contribution from the financial and technology sectors has increased, then we should expect the “normal” profit margin of the aggregate corporate sector–if there is such a thing–to have increased as well.  The targeted “mean” in a “mean reversion” will have shifted upward, rendering the ensuing picture less bearish.

(Note: the revenue contribution of healthcare, shown in brown, has also increased substantially over time.  But, to the surprise of many, current healthcare profit margins, at 6.2%, are below their own historical average, and only slightly above the historical average of the aggregate corporate sector.  They are not appreciable contributors to current profit margin elevation).

It turns out that we can correct for this shift, creating an “adjusted” profit margin that accounts for the effects of changing sectoral revenue contributions.  What we need to do is take the average historical revenue contribution of each sector, and compute what the aggregate profit margin would have been, at each point in time, if each contribution had been equal to its individual historical average.

The following table shows the historical average revenue contributions of each of the 10 GICS sectors:


So, historically, energy revenues have represented 12.78% of total revenues, materials revenues have averaged 7.80%, health care revenues have averaged 5.46%, and so on. What we want to know is, what would the profit margin of the aggregate corporate sector have been at each point in history if the revenue contribution of each individual sector had been equal to its average?  This “adjusted” profit margin will filter out changes that have been driven solely by shifts in sector size and contribution, and will thus provide a more accurate picture of the aggregate profit margin to use when making historical comparisons.

At this point, Bulls are probably hoping that I pull out a chart showing that when profit margins are properly adjusted in this way, that they end up not being historically elevated. Sorry, not quite.  As the chart below shows, the adjustment doesn’t make much of a difference.


The unadjusted profit margin (blue) is 45% above its historical average, versus 35% for the adjusted profit margin (green).  Relative to the respective averages, the adjusted profit margin is only about 10% “less elevated” than the unadjusted profit margin.  This difference is worth something, no question–but it’s not enough to eliminate profit margin concerns outright.

The reason that the adjustment doesn’t make the kind of difference that we might otherwise expect is that the profit margins of the financial and technology sectors have themselves expanded dramatically in recent years.  The truth is that profit margin increases within the financial and technology sectors, rather than increases in their contribution to total revenue, have been the primary drivers of the aggregate profit margin increase.

The following chart shows the profit margins of the combined finance and technology sectors (red) alongside the profit margins of the combined other sectors (blue):


As you can see, profit margins in finance and technology have exploded.  Combined, they are running at almost twice their historical averages–86% above, to be precise.  The profit margins of the combined other sectors are hardly elevated at all–only around 18% above their historical averages.

The following charts show the profit margins of the finance and technology sectors individually:



Notice that a large chunk of the move is recent–a phenomenon unique to this specific cycle–especially in the technology sector.

Explaining the Rise

It goes without saying that finance and technology, which together represent over 42% of current U.S. corporate earnings, are two sectors that we should keep a close eye on going forward.  Changes within them have driven the profit margin expansion of the last several years, which itself has driven the bull market, having made possible a “goldilocks” scenario in which earnings have been able to grow robustly despite slow top-line growth and almost non-existent inflation.  The slow top-line growth and almost non-existent inflation has pushed the Fed into an aggressively easy monetary stance that has served as fuel for persistent P/E multiple expansion, with more and more investors ditching the misery of zero-yield cash and bonds to join the market advance.

The finding that the profit margin expansion has been driven largely by changes inside the finance and technology sectors sheds doubt on other stories that have been offered as explanations.  Weaker labor unions, increased access to cheap foreign workers, a rise in earnings taken in from abroad, lower corporate taxes, more effective corporate tax avoidance schemes, and so on–these explanations fail to make sense of the fact that profit margins haven’t increased nearly as much in sectors outside of finance and technology. Whatever the correct explanation for the current state of profit margins ends up being, it needs to be one that applies with some preference to finance and technology, which is where the most dramatic shift has taken place.

What, then, is the explanation for the rise?  Why have profit margins in finance and technology increased so dramatically over the last several years?  Will the increase hold up?

A Bullish Angle

On the finance front, bulls can make a compelling argument that the financial sector’s contribution to the profit margin increase is likely to be sustained.  The increase in the financial sector share of total revenues has been driven by higher debt levels across the economy–that change will almost certainly prove to be secular.  At the same time, the increase in profit margins within the financial sector has arguably been driven by the drop in short-term interest rates (funding costs for financial institutions), which is a change that is also likely to be secular.  Note that the last time that financial profit margins were at their current levels was in the early 1960s, when short-term interest rates were low.  The Fed tightening cycle that lasted from the late 1960s through the 1980s seems to have been what pulled them down, as they fell much more precipitously during that period than the profit margins of any other sector.  They only began to regain their prior levels in the mid-to-late 1990s, as the Fed shifted to an easier monetary stance.

On the technology front, bulls can make a similarly compelling argument that the revolutionary technology of the information economy, which has only been fully fleshed out in the last decade, has been the game-changer, having created an increasingly “winner takes all” environment in which it has become more and more difficult for potential new entrants to credibly compete with the first-mover.  If they are right, then profit margin mean reversion–in the technology sector and in any other sectors that manage to piggy back on the dynamic–would seem to be less likely.

A Bearish Angle

Of course, bears can offer compelling counter-arguments to this optimism.  On the finance front, they can point to the fate of the yield curve–which, assuming the Fed follows through on its normalization plans, does not look good.

When the Fed cuts rates for the first time, the long end of the curve usually stays put. People continue to expect an eventual return to normalcy, and price the long end accordingly.  The result ends up being a steep curve that boosts financial sector profit margins.  But when the Fed cuts rates and keeps them cut, for a period that seems to drag on forever, because the economy never seems to get hooked into the kind of genuine inflationary expansion that would justify a tightening cycle, the market eventually figures things out. Investors realize that long-term rates need to be lower, and pulls the long-end down accordingly, at the expense of financial sector profitability.

Eventually, the Fed will raise the short-end–if not simply out of a desire to restore some normalcy to monetary policy.  When that time comes, the long end will again be slow to respond–this time slow in the opposite direction, slow to rise, given the anchoring and inertia of market participants who, by then, will have grown accustomed to the idea of secularly low interest rates.  The result will be a yield curve that gets flatter and flatter with each hike, and a financial sector whose profit margins get squeezed.  That seems to be exactly where we are currently headed, and it is not bullish.

In the most recent earnings data from banks, we’ve seen a consistently weak trend: flat YOY revenue growth and negative YOY EPS growth, brought on by increased competition, particularly among smaller banks, increased regulatory expenses, and reduced profitability due to a flattening yield curve.  Loan growth, which would otherwise represent the bright spot, is not making up for the reduced profitability.

On the technology front, bears can make a similarly compelling argument.  “Tech” is the most cutthroat and competitive of all sectors.  Historically, it has produced subpar returns for investors (ranked number 6 out of 10 sectors), likely due to the way in which disruption and competition have worked to break down dominant positions within it. When we look at the seemingly impenetrable empires of the $AAPLs and $GOOGs and $MSFTs and $FBs of the world, it can be tempting to think that the truly massive levels of profitability they currently enjoy will be forever secure–but this kind of thinking is not supported by history.

There’s a particularly interesting and relevant analogy that bears can raise in this context, one that involves a different sector: commodities.  The historical evidence on the real return potential of spot commodities is overwhelming: there is no real return potential, spot commodities do not offer real returns.  For proof, consider the following 130 year chart from Dylan Grice.  Notice the black line languishing stupidly at the bottom:


But then again, over the last decade, we saw a massive boom in commodity prices around the world.  As always happens, compelling stories emerged to explain why the boom had occurred and why it would almost certainly hold up–insatiable demand growth from China, India, and other emerging markets, an increasingly constrained supply that fails to grow, even in response to large price increases, and so on.  If you had told people in 2007 or 2008, or in 2010 or 2011, that these were just stories, and that there would eventually be a painful reversion to the mean just a few years later, very few people would have taken you seriously.  Nobody in the commodity complex at the time was seriously entertaining the possibility.

But now here we sit, in a healthy economic expansion, with oil trading below $45 (!!), the same inflation-adjusted price that it traded at 30 years ago, near the lows of the last oil downturn.  A 30 year period of zero real returns for this and other spot commodities has once again vindicated the apparent lesson of history: that spot commodities do not produce real returns.  Now, to be clear, I don’t expect profit margin bearishness to receive the same degree of vindication–but some caution and humility are certainly in order, given the possibility.

Posted in Uncategorized | Leave a comment

Intrinsic Value: Interest Rates, Inflation, and the Forgotten Concept of the Time Value of Money

In the previous piece, I offered a definition of the investment concept of “intrinsic value.” Intrinsic value is the value that the owner of a security realizes from holding the security, rather than selling it.

To determine the intrinsic value of a given security, we can apply a simple test.  We posit that the security that cannot ever be sold, but must be held until maturity.  We then ask ourselves: what is the maximum price that we would be willing to pay, or alternatively, the maximum amount of cash that we would be willing to exchange, to own the security?  That amount of cash must equal the intrinsic value of the security, the value that accrues to us simply from owning it, otherwise the exchange would not make rational sense.

In this piece, I’m going to explore the set of fundamental considerations that would impact a rational agent’s assessment of the intrinsic value of different types of securities.  The analysis will seek to clarify “the way things ought to be” in financial markets–the way they would be if everyone invested rationally, based solely on the intrinsic value contained in the investment opportunities presented.

To be clear, “the way things ought to be” in financial markets is not the way things actually are, particularly with respect to long-dated assets, assets whose maturities are too far out in the future to “wait for.”  Market participants that trade and invest in long-dated assets do so based not on estimations of “intrinsic value”, but rather on estimations of how the prices of those assets will evolve over the short-term to medium-term, a few months to several years, which is the limit of human look-forward capacity, and the time horizon on which investor performance is measured.  Investors are not able, personally or professionally, to seriously consider longer time horizons, on the order of decades or even centuries, even though that is often how long it takes for the “intrinsic value” of long-dated assets to play out.

Investors worry about the “fundamentals” of long-dated assets not for their own sake, but because the fundamentals influence the prices, through non-fundamental perceptual and behavioral channels.  The fundamentals serve as subjective inputs into the minds of investors, factoring into the rule-based calculations that drive actions and outcomes in the market: “X is happeningit probably means Y.  From a portfolio standpoint, the right move is probably for us to do Z.”

Cash, Bonds, Stocks, Other

Investors are confronted with a range of different types of assets in which to hold their wealth.  We can simplify this range into four categories: Cash, Bonds, Stocks, and Other. To determine the “intrinsic value” of assets in each category, we need to express them in terms of cash, which is the basis for measurement.

Cash is just cash, money, whatever must be accepted by law to repay debts, public and private.  The prices of all assets are expressed in terms of it, therefore the intrinsic value of one unit of it is one.  Bonds are a finite collection of more-or-less guaranteed cash payments, usually consisting of small cash payments for a time (coupons), followed by a large cash payment at the end (return of principal).  Equities are an infinite collection of non-guaranteed cash payments (dividends, or rental payments on the use of capital, land, housing, and so on).  The “other” category consists of unproductive assets, assets that do not generate a cash flow–think, gold bars.  These assets have very little intrinsic value, and are almost always purchased with the intent of eventually offloading the investment onto someone else.

The dividends that shares of equity pay out to their owners tend to grow at a rate that exceeds the rate of inflation.  The reason is twofold:

  • First, the dividends are backed by corporate earnings, and are paid out as a percentage thereof.  On a unit basis, corporate earnings equal price minus cost.  Inflation–a change in the price index–acts to increase both of these entries equally, therefore it acts to increase their differenceearningsequally as well.
  • Second, not all of the earnings are paid out as dividends.  Some of the earnings are used up in the purchase of growth.  The amount of growth purchased adds additional earnings, and therefore additional dividends, to the numbers of the future.

Now, to be clear, funding costs–for example, dilution–can cause per-share earnings to not keep up with inflation, particularly when the corporate sector is inefficient in its use of the proceeds.  In the present context, we will assume that corporations fund their growth internally, without increasing share count (an assumption that has proven valid in recent experience), rendering the issue of dilution moot.

Leaving the “other” category aside, we are left with two types of assets whose intrinsic value we want to measure: bonds and shares of equity.  So we return to the critical question: how much of each type of asset would we be willing to exchange for cash, if we could not ever go back on the exchange?

A bond is just a future stream of small cash payments (coupons), followed by a final payment (repayment of principal).  How much cash, held in hand right now, would we be willing to trade for that future stream?  The answer, for us, is the intrinsic value of the bond.

A share of equity is a future stream of small, growing, inflation-linked cash payments, without a maturity date. How much cash, held in hand right now, would we be willing to trade for that future stream?  The answer, for us, is the intrinsic value of the share of equity.

The Time Value of Money

Money now obviously is not the same as money later.  Money now is better, for a myriad of reasons, not the least of which is that it can be used now, at the option of its owner. Money later, in contrast, cannot be used until it is received.

The difference between money now and money later is the “time value of money.”  Looking at the current state of bond prices in the developed world, we might think that money has no time value to current investors.  After all, long-term bonds in the developed world trade at prices with implied yields approaching zero.  An investor who lends his money to a government in the developed world for five, ten, twenty, even thirty years, gets essentially nothing in return–nothing except the original money that was lent out, for a net nothing.

But to conclude that money has no time value to current investors would be a huge mistake.  The reason that current investors are willing to lend their money to governments at zero rates of interest for prolonged periods of time is that they know they can easily get out of the loans by selling the associated securities back into the market.  For all intents and purposes, in a liquid market, where investors are confident that they will be able to sell their investments at or near cost, the “time value of money” loses relevance.  The assets become the functional equivalents of “money now”, given that they can be converted into “money now” at the push of a button.  In purchasing the asset, the investors don’t have to “part” with their money, therefore they don’t have to put a price on the cost, to them, of “parting” with it.  If they did, the price dynamics observed in developed world bond markets would be very different from what they are.

Take any institutional fund that is currently eager to lend its money to the Swiss or Japanese governments for decades at near-zero interest rates, and tell that fund that it will have to hold the associated debt security until maturity–that it will not, under any circumstances whatsoever, be allowed to get the money back by selling the security to other investors (or engaging in any “tricks” that might simulate a sale, such as posting the security as collateral for a loan, or selling short a similar security).  You would quickly see the time value of money come back into play, in a dramatic way.  To be sure, it is not zero.  Not even close.

When word gets out that a financial institution is in trouble and is facing a liquidity crisis, its customers rush to redeem their money.  The main fear that drives their behavior isn’t the prospect that the money will be lost forever–the investors know they will almost surely get it back, after everything has been cleaned up, years later.  Their most pressing worry is actually the prospect that the money will get stuck inside of a black hole in the interim–a confused, entangled “what belongs where?” scenario, a court battle involving complicated and drawn-out litigation–and that the customers therefore won’t be able to access the money for months, years, maybe decades.  Ouch.  Again, we see the importance of the “time of value of money”–when it is actually at stake.  In a stable, liquid market with a confident bid, it is not at stake, and so it doesn’t factor in.  But things can easily and quickly happen to put it at stake, which is why long-term assets–assets with maturities on the orders of many years, decades or centuries, that cannot realistically be “waited out”–are prone to violent bouts of volatility, when confidence in the presence of future bids near the current price is lost.

The Exercise

Cash held in the banking system carries essentially all of the benefits of cash held in hand, with a number of additional perks and conveniences.  For this reason, individuals usually choose to hold their cash in banks, in the form of deposits.  The banks normally pay interest on the deposits, which they fund through the income they generate on their loans. Without getting into the details, central banks in modern financial systems have the ability to adjust the rate of interest that banks, in the presence of market forces, have to pay on their deposits.  Expectations with respect to the future path of this rate of interest have a substantial effect on the “intrinsic value” of all other assets, because all other assets must compete with it.

To illustrate, let’s do the exercise.  You have $100,000 in wealth, and three modalities in which to store it:

(1) Cash: You can hold the wealth as a cash deposit in an insured bank, and earn interest on it.  While in the bank, you will be able to spend it on consumption.  Depending on the nature of your deposit, you may have to wait a few weeks to spend it, maybe a month or two, but you can afford to wait that long.  To keep things fair, let’s suppose that if you choose this option, you can only spend the money on consumption–you cannot go and invest it in the other options later.  You will have to make the “investment” decision now, and you will have to stick with it for good, at least as far as that money, the $100,000, is concerned.

The interest rate that you will be paid will be determined by the central bank, chosen so as to bring the rate of inflation–and any other macroeconomic target that the central bank might have–onto target.  In periods where there are strong inflationary pressures, the rate will be increased, so as to incentivize you and others to hold your cash rather than spend or invest, and to disincentivize others from borrowing it to spend or invest.  The same is true in reverse: in periods where inflationary pressures are weak or negative, the rate will be reduced, so as to disincentivize you and others from holding your cash, get you to spend or invest it instead, and to incentivize others to borrow it to spend and invest.

(2) Bond: You can buy (miniature) 10 year treasury bonds.  Each bond pays guaranteed fixed interest payments of $60 per year, followed by a large principal repayment, $1,000, in 10 years. Importantly, you cannot sell one of these miniature bonds after you have bought it.  You must hold it to maturity.  The money that is figuratively “contained” inside it will be locked up, unable to be used by you in any way, until then.

(3) Equity: You can buy diversified shares of equity, say, the 500 companies of the S&P 500.  The share pays $40 per year in dividends, the dividends grow anywhere from 1% to 4% per year, plus inflation, with a sharp recessionary drop every several years that is eventually fully recovered.  Crucially, the shares have no maturity.  You will never get the original principal back–what you will get back is an ever-growing stream of dividends, which over the very long-haul, will exceed what you put in by many orders of magnitude.

We have not yet stated the per-share price of the bond and the equity.  The goal here is for you to seriously think about the options, as if they were presented to you right now, and identify the maximum price that you would be willing to pay for each share, the maximum amount of cash that you would be willing to permanently exchange for them–which, for you, is the “intrinsic value.”

As a rational agent, what do you need to know in order to determine the “intrinsic value” of each type of security?  For starters, you need to know, or estimate, the concrete specifics of the payout stream. How much is the bond going to pay?  How much is the cash going to pay?  How much is the equity going to pay?

For the bond, you already know the entirety of the future stream–$60 per year, plus $1,000 in ten years. The stream carries no uncertainty in its payouts.  But knowing that alone is not enough.  You also need to know the nominal interest rate that cash in the bank will pay you over the next ten years.  It will not make rational sense for you to pay a price for the bond that implies a return that is any lower than that, any lower than what you can get in cash, because cash also affords you the precious ability to have the money and use it, which the bond does not afford  you.  Therefore the bond needs to be priced to pay you more.

Now, we know that the central bank is going to set the cash interest rate so as to control inflation.  So the true variable that matters here is the future neutral nominal interest rate, the nominal rate of interest that the central bank will have to set going forward, given the structural dynamics of the economy, in order to keep inflation and any other target that it might have–employment, foreign exchange control, financial stability–on target.

In truth, of course, you’re going to demand even more than the expected neutral nominal rate, you’re going to demand a premium to compensate you for the time value of money, the cost of losing the ability to use your money.  How much you demand will be determined by the amount of value that money in hand has for you in comparison with money promised in the future.

How different, for you, is “money now” from “money later”?  The answer will obviously depend on the myriad of complex psychological and economic factors that define your unique personal situation. How much do you value the comfort and security of having access to your money, the ability to use it if you should want to use it, or need to use it?  How much more valuable is that kind of money to you, in comparison with money that will be locked away for a long period of time, inaccessible to you?  How many things are there in the economy for you to buy right now that might tangibly increase your happiness, or the happiness of those you care about?  How old are you, and to what extent is the money needed to fund your desired consumption expenses?  If the money is needed, will the coupon or dividend payments that will accrue on it if it is permanently locked away in a bond or an equity be large enough to fund those expenses?  If so, then you may be fine with seeing it locked away, given that you can get by on the infinite payouts that will accrue.  What are your expectations with respect to inflation?  Inflation eats away at the future purchasing power of money.  High inflation therefore widens the difference between “money now” and “money later”, given that it makes “money later” into “less money.”  All of these variables, and a number of others, will factor into your estimation of the “time value” that money has for you.

To summarize the bond case, then, we’ve identified two variables that matter to the intrinsic value of a long-term treasury bond: (1) the expected neutral nominal interest rate on cash over the life of the bond, which sets the minimum floor for what you can rationally accept from the bond, given that you have the alternative of holding cash, and (2) the time value of money, which you ultimately have to specify for yourself, given the unique psychological and financial details that characterize your individual situation.

For the equity case, the evaluation is more complex.  We need to estimate the future growth of the dividends, and by extension, the future growth of the earnings out of which they will be paid (and which will pay for their future growth).  In the scenario, we set a range of 1% to 4% after inflation, but that’s a huge range–any information that pushes the number in either direction is going to be very important.

We can separate the growth of dividends into two components: real per-share growth, and inflation.  The first component is determined primarily by the health and dynamism of the underlying economy, and by the efficiency and capital allocation skill of the aggregate corporate sector. The second component is driven by culture, demographics, supply constraints and policy.

The two factors that were relevant to the intrinsic value of the bond–the expected neutral nominal interest rate and the time value of money–are just as important to the intrinsic value of the equity.  As with the return produced by holding the bond, the return produced by holding the equity competes directly with the alternative of holding cash in the bank and collecting the future neutral nominal interest rate.  Similarly, holding the equity instead of the cash entails loss of a large amount of money that would otherwise be accessible.

What we end up with, then, are four variables that determine the “intrinsic value” of the equity: (1) the expected neutral nominal interest rate, (2) the time value of money, (3) the expected future rate of real per-share growth, and (4) the expected future rate of inflation.

Now, here comes a critical move.  We can combine (4) and (1) into a single variable, the expected neutral real interest rate.  Going forward, what real interest rate, after inflation, will the central bank have to set in order to maintain inflation, and any other targeted macroeconomic variable, on target?  That rate is critical, because it expresses the difference between (a) inflation, a crucial component of the nominal growth that the equity payouts will exhibit, and (b) the nominal interest rate that the cash holdings will earn.

The Fed Model

The Fed Model is a popular a method of measuring equity valuations.  The model assesses valuation by comparing the earnings yield on equities to the long-term government bond yield. When equity earnings yields are substantially higher than the long-term government bond yield, equities are said to be cheap.  When equity earnings yields are not appreciably higher than the long-term government bond yield, equities are said to be expensive, or at least neutrally priced.

In practice, the Fed Model has caused a number of analysts to push back on the growing consensus that the US stock market is expensive, while Emerging Market stock markets are cheap.  These analysts acknowledge that earnings yields in the US are lower than in the Emerging Markets (or alternatively, that P/E ratios in the US are higher than in the Emerging Markets), but they point out that we cannot talk about yields and P/E ratios in a vacuum.  We have to compare them to the available alternatives, the attractiveness of which are captured by prevailing interest rates.

But this way of thinking is partially wrong.  It ignores the fact that interest rates are typically set at low or high levels in response to low or high levels of another variable that matters greatly to equity returns–inflation.  Why has the US  central bank set the interest rate at a low level? Because the US does not have enough inflation. Why has the Brazilian central bank set the interest rate at a high level? Because Brazil has too much inflation. The low inflation in the US contributes to an environment of low nominal earnings and dividend growth, and therefore low nominal total returns, all else equal (and note that all else is not equal, in this case).  The high inflation in Brazil (or Argentina or Zimbabwe) contributes to an environment of high nominal earnings and dividend growth, and therefore high nominal returns, all else equal.  The Fed Model fails to capture and factor in the impact of this crucial difference.

If we’re going to connect P/E ratios to interest rates, as the Fed Model tries to do, the interest rates that we should use are real interest rates, interest rates that take out expected future inflation, which is a significant component of nominal equity returns. When we do this, we see that a number of emerging markets with high interest rates and high P/E ratios, such as India, deserve to have high P/E ratios, because their real interest rates are very low, if not outright negative (making cash and bonds that much less attractive in comparison with inflation-linked equities).  Similarly, a number of countries with low interest rates, such as Japan under pre-Abenomics deflation, deserve to have low P/E ratios, because their real interest rates are high (making cash and bonds that much more attractive relative to inflation-linked equities).

Foreign Equity Investing

This dynamic extends quite elegantly to the realm of foreign equity investing. To use the example of Brazil, Brazilian equities currently sell at very low P/E multiples–at last check, around 8-9 times, with correspondingly high dividend yields and substantial room for P/E multiple expansion over the long-term.  For this reason, many US investors, frustrated with the lack of attractive options at home, have explored the country as a potential investment opportunity.

Suppose that you are a US investor that wants to capture the return potential of the Brazilian equity market.  But you want to capture that return in dollar terms–the terms of your own currency.  If the Brazilian market goes up 200% over the next 10 years, you want the value of your Brazilian investment, in your own currency, the Dollar, to achieve that same return.  The only way that you can make this happen is by hedging the currency.  You would go simultaneously long the Brazilian stock market, and short the Brazilian currency, the Real.  Then, your return in dollars would exactly mimic the local currency return of the Brazilian stock market.

But there’s a problem. The cost of shorting the Brazilian currency is the Brazilian interest rate; you will have to pay that interest rate to whomever you borrow the currency from in your short.  Right now, the rate is quite high, north of 10%.  That 10% will represent a significant drag on your returns.  For this very reason, it’s impossible for you to create a dollar-denominated investment that will exactly track with the Brazilian stock market. The best you can hope to do is create an investment that tracks with the Brazilian stock market minus 10% per year.

Not all is lost, of course.  Your investment might still produce an attractive return, even in the presence of the high carry.  The interest rate in Brazil is high, 10%, but that’s because inflation in Brazil is very high–well north of 6%.  The 6% inflation is going to add to the nominal growth in Brazilian earnings and dividends.  When combined with the high dividend yield, and the significant multiple expansion that is likely to occur as sentiment improves, the return that the investment might be able to make up for the 10% carrying cost.

What we need to do in a valuation analysis is combine these two numbers–the inflation and the interest rate–since they offset each other in terms of their effects on the return. The inflation adds to the return, and the interest rate–which is the carrying cost–subtracts from it.  The combination of the two,  of course, just is the real interest rate, which, you will recall, is what we found to also be a critically important variable in the determination of the intrinsic value of domestic equities.

The real interest rate in Brazil is 10% minus 6% which equals 4%–on the high side globally. For this reason, Brazil probably should have a lower P/E multiple than the developed world, where zero or negative real interest rates have become the policy norm.

Posted in Uncategorized | Leave a comment

What Is Intrinsic Value, And Who Decides It?

James Osborne of Bason Asset Management recently published an excellent critique of the investment concept of “Intrinsic Value.”  I urge readers to take a minute and go check it out.  In this piece, I’m going to try to tackle a question that James poses.

That question: what is intrinsic value, and who decides it?

No Selling Allowed 

Here’s my answer.  The “intrinsic value” of a security is the maximum price that an investor would be willing to pay to own the security if she could not ever sell it.

Three points:

(1) All I am doing here is defining the term.  You can define the term in another way if you wish, but then you will be talking about something else.  When I use the term “Intrinsic Value”, I am talking about the maximum price that an investor would be willing to pay to own a security if she had to hold it until “maturity”, i.e., for the entirety of its natural life as a security, which, for an equity security, means forever.

(2) According to the definition, the “intrinsic value” of a security is different for different individuals.  That’s to be expected.  Value, like beauty, is a judgement made by the individual–it exists only in the mind of the individual, the eye of the beholder.

(3) The definition fits with the literal meaning of the word intrinsic–”inherent, innate, inborn, inside the thing itself.”  The value that an asset has, inside itself, cannot be a function of the quantity of other useful things that other people happen to be willing to exchange it for in a market.  Rather, the value must remain present even when no trading is allowed.

Intrinsic Value: A Thought Experiment

So let me now ask you a question.  Suppose that I have a security to sell you.  The security works as follows.  It pays out $10 in dividends per year.  The dividends grow at a real (inflation-adjusted) rate that ranges anywhere from 3% to 5% per year.  Of course, over the short-term, the dividends can grow at different rates–sometimes they can even grow at negative rates, i.e, fall.  But, crucially, over the very long-term, they always recover. They always regain their 3% to 5% long-term growth trendline.

Now, to avoid uncertainty, let’s assume that the aforementioned features of the security are guaranteed by the full faith and credit of the U.S. government.  So there is essentially zero risk that the security will not behave  in the way that I just described.  The question: what is the maximum price that you would be willing to pay, in cash, to own the security, if you could not ever sell it?  Alternatively, what, for you, is the “intrinsic value” of the security?

Take a moment and consider the question as if the proposition were really there for you to take. What is the maximum price that you would be willing to pay? I’m not going to mention any number as a starting offer, because I don’t want to influence your answer.

I’ve posed this question to a number of individuals, both inside and outside the financial industry.  Almost everyone answers with a price that is less than $250.  Note that at a price of $250, the security would offer a 4% yield, fully protected from inflation, with 3% to 5% real per annum growth added on top.  Not bad.  That valuation is seen as minimally necessary to compensate the investor for the cost of forever parting with his principal.

The S&P 500: A Growing Stream of Dividends

If you’re particularly clever, you’ve probably noticed that the security that I’ve described here is basically the S&P 500 stock index divided by four.  The S&P 500 presently pays an annual dividend of around $40 per year.  Not all of the earnings of the companies in the S&P 500 are paid out to shareholders as dividends–some are spent (read: “used up”) on capital expenditures and asset acquisitions.  That is precisely why the dividends are able to grow over time at a rate that exceeds the rate of inflation, i.e., the rate at which the prices of all things in the economy, including the prices of the goods and services that corporations sell, changes.

The reason that I divided the S&P 500′s dividend by four is to prevent the current price, 2070, from creating a false anchor in the mind that influences the “intrinsic value” intuitively ascribed to it.  So take your earlier price, the maximum price that you would pay for the $10 per year security, and multiply that price by four.  That’s your final price, the intrinsic value that you ascribe to the S&P 500.  I ascribe around 800.  You might ascribe 1000.  Or maybe 600.  Certainly not the present price of 2070–unless you’re crazy.

If you doubt the logic here, ask yourself: what is the S&P 500, intrinsically, apart from all of this baseball-card-trading that we engage in when we play in markets? That’s the question that you will have to confront if you decide to make a genuine, non-redeemable investment in the security, that is, buy the security without having the ability to sell it.  The answer: to you, it is just a growing stream of dividends, nothing more.

Now, how reliable is the assumption that the dividends will grow over the long-term at a rate that exceeds the rate of inflation?  Pretty reliable.  The historical reliability of the assumption is demonstrated in many centuries of actual data, not only in the US, but in other capitalist economies. That reliability is supported by the inherent diversification of the index–we’re talking about many different companies from many different industries, rather than a single company that might one day go bust.  But even if we view the dividend stream as not growing reliably over the long-term, or posit a larger uncertainty around the growth than the previous 3% to 5% range allowed, that will only pull the “intrinsic value” lower–and the lowness relative to the current market price is precisely what I’m trying to emphasize.


As you can see in the chart, the historical real rate of growth of per-share dividends for the S&P 500 has been significantly less than the stipulated 3% to 5%.  It’s actually been closer to 1.4%. But there’s an important factor at work.  Most of the S&P’s history was dominated by periods in which only a small portion of earnings were consumed on the purchase of future growth.  Most of the earnings were delivered directly to shareholders in the form of dividends.  Corporate managers have since evolved a preference for earnings reinvestment, and so the current dividend stream tends to grow faster, though it is smaller than it could be, or would have been in the past.

Now, to create a full analogy with the “intrinsic” (i.e., can’t-sell) S&P 500, let’s add a final caveat to the security.  You, as the owner, get to determine the payout ratio.  We can think of the payout ratio as a dial that you can adjust, any time, at will.  You can opt for less dividends now, and more dividend growth, or for less dividend growth, and more dividends now.  If you want, you can even choose to have all of the annual earnings that back the security–in the case of the S&P 500, around $110–paid out to you in dividends. The cost of doing this, of course, is that the earnings and dividends will stop growing.  They will grow at a real rate of 0%.

You might think that the S&P’s current price of 2070 is a reasonable price for a growthless security that pays out $110 a year.  But check that thought.  Suppose you had $100,000 in cash sitting around, earning nothing.  Suppose further that there is nothing else on earth that you can invest it in but the “intrinsic” (i.e., can’t-sell) S&P 500, priced at 2070.  Would you really be willing to part with all of that money, permanently, in exchange for a perpetual payment of only 5.3%, $5,300 per year?  Not very many people would be.  Some people wouldn’t even be willing to accept 10%$10,000 per  year, or even 20%, $20,000 per year.

Even if never used, the simple ability to get your money out of the security and back into your pocket is worth a ton.  That ability represents the difference between your being willing to pay 2070 for the security, on the trust and confidence that you will only be one mouse click–one sell order–away from getting your money back, and being willing to pay only 600, on the stipulation that you will be stuck holding the security until maturity, i.e., for the rest of its life, which means the rest of your life.

Now, to be clear, I’m not saying that it’s irrational for you to be willing to pay 2070 for the security.  I’m saying that, built into your willingness to own it at that price, is a (largely justified, at least for now) expectation that you will be able to sell it, on demand, to someone else at a price near that price–hopefully, at a price higher.  That expectation makes the security dramatically more attractive to own than it would be if it were just what it is intrinsically–a simple stream of growing dividends that had to be held as such.

The Dividend Discount Model

In 1956, Myron Gordon and Eli Shapiro developed the dividend discount model of equity valuation.  On this model, the intrinsic value of an equity security is the sum total of all of the security’s future payouts, from time now until time infinity, discounted back to the present.  Gordon and Shapiro showed mathematically that when discounted at a required rate of return r, an infinite stream of dividends with a starting annual dividend level of d and an annual growth rate is worth a present price p, roughly equal to,

(1) p = d / (r – g)

Rearranging to solve for r, we get,

(2) r = d / p + g

which neatly says that the rate of return that an equity security produces for its owner equals the dividend yield (d / p) plus the dividend growth (g).  Note that if we want, we can make the dividend yield equal the earnings (e) yield, but then g will go to zero, so we will end up with,

(3)  r = e / p

which simply states that the rate of return that an equity security produces for an investor equals the earnings yield, provided that all of the earnings are paid out in present dividends.

Now, the dividend discount model is just a way of formalizing the intrinsic valuation process. The ambiguity and subjectivity in that process remains–the model simply places the ambiguity and subjectivity inside the convenient term r, the discount rate, which is the rate of return that investors demand in order to exchange cash now for cash later.  You can make the intrinsic value of a security be anything you want, any number from 0 to infinity, depending on the discount rate that you choose to impose.  And why must you choose to impose any one discount rate over any other?  As an investor, it’s your call.

To illustrate the power of the discount rate, let’s do the actual exercise for the S&P 500. The index pays a dividend of $40 per year, and the dividend grows at a real rate of 3% per year.  Pick your discount rate, and make sure that it’s a real discount rate, adjusted for inflation).  If you pick a 7% discount rate, then the S&P 500 is intrinsically worth  $40 / (.07 – .03) = 1000.  If you pick a 5% discount rate, then the S&P is intrinsically worth $40 / (.05 – .03) = 2000.  The difference in rate is only 2%–but the difference in price is 100%.

Ask yourself: what makes a 7% discount rate any more “privileged” than a 5% discount rate?  To cover the “losses” associated with converting a present cash sum into an infinite stream of future cash payments, why must an investor demand 7% rather than %5?  Or 5% rather than 3%?  Note that all of these rates are above the loss-adjusted rates that can presently be earned in other asset classes.  Indeed, the risk-free dollar benchmark is priced (in dollars) to deliver an inflation-adjusted rate close to zero, even on very long time horizons.

Robert Shiller and Equity Volatility

In the 1981 masterpiece that earned him a Nobel Prize, Robert Shiller posed the question: “Do Stock Prices Move Too Much to be Justified by Subsequent Changes in Dividends?” He empirically demonstrated that the answer was yes, and concluded that markets cannot be described as “efficient.”  Notably, he used a definition of the term “efficient” that his efficient market hypothesis (EMH) opponents, chiefly Eugene Fama, would never accept. But that is a topic for a different piece.

Shiller blamed the large discrepancy between realized price fluctuations and realized dividend fluctuations on the irrational psychological and emotional forces that drive investor behaviors.  His underlying point–that behavior drive the market–is obviously true, but I think there is a more elegant, less demeaning way to frame it.

The intrinsic value of a security–the price that investors would be willing to pay to own it on the stipulation that they would have to hold it indefinitely–is dramatically different from the price that investors are willing to pay knowing that they can easily sell it to others (without incurring a large loss.)  That difference is the value of liquidity.

When liquidity is present, backed by trust and confidence in the stability of the market, investors tend to view their equity holdings as if those holdings were identical to cash in the bank.  They don’t price in the lost liquidity associated with making a genuine, non-redeemable investment in something, an irreversible conversion of present cash into long streams of small future cash payments.  Nor should they price in that lost liquidity–it’s not lost.

If lost liquidity were a reality that had to priced into stock market investments, equities would trade at valuations that are significantly lower than the valuations at which they currently trade. Instead of being willing to pay 2070, and probably higher in the coming months, for the $40 per year, 3% to 5% real growth security that the the S&P 500 represents, investors would demand a far more attractive price–probably a price below 1000.  For some investors, a price as  low as 200 might not even be enough.

Suppose you have $1,000,0000.  I ask you how much money you have, in your name. You will answer $1,000,000–whether that money is invested in an S&P 500 index fund, or whether it’s sitting in the bank.  If it’s in the index fund, it’s not money–but, for you, it might as well be, because you have access to a simple, on-demand means of exchanging it for money, a stable, reliable market in which you can exchange it.  Because of that access, you are able to derive all of the psychological and consumptive benefits of having the money, even though it is not actually in your possession.

But now let me put you in a situation where you might not be able to get that $1,000,000 out of the S&P 500 for a very long time, maybe for the rest of your life, because the markets are crashing and are going to stay crashed.  If I tell you “Don’t worry, the dividends that underlie the true value of your investment will be unaffected by the crash”, will that be much consolation to you?  Will it relieve you of the sense of loss?  Obviously not.

As an investor, you lose your money not when the prices of your investments fall, but when you exchange money for them in the first place.  When you initially buy in–that is when the money is no longer yours.  Crashes force you to view the condition of no longer having access to the money as a genuine loss, a genuine sacrifice, because it removes the people who are otherwise there to give the money back to you.

Now, to the main point of this piece.  Equity prices are volatile–much more volatile than the earnings-backed dividend payments that render them intrinsically valuable, i.e., valuable in themselves–because the trust and confidence that forms the basis for liquidity in the market can be fickle and unreliable.  The market can pull its liquidity in a heartbeat, and sometimes does pull it. When the liquidity is pulled–when investors conclude that they aren’t going to be able to sell at the prices they paid, at least not for a long time–their desire to be invested falls dramatically, as it should, given that they did not enter into their investments on the stipulation that they they would be permanently stuck inside them.

We can think of market prices as hovering between two poles: (1) the “intrinsic value” price, the maximum price that investors would be willing to pay to own a cash flow stream if there were no liquidity, no ability to sell the investment, ever, and (2) the “bubble” price, the maximum price that investors would be willing to pay own a cash flow stream, however paltry, if they were certain they would be able to get out of the investment without losing money, and hopefully be able to get out of it making money, realizing a return simply from the trading process.  These poles are separated by many thousands of points, many hundreds of percent.  The market swings between them based on factors that seem to have little to do with the long-term earnings and dividend prospects of companies because the factors influence the tender trust and confidence that investors have in the market’s future stability and trajectory, a trust and confidence that ultimately separates the two poles from each other.

Interestingly, on this way of thinking, the main reason why bond prices are less volatile than equity prices is not that the coupon payments of bonds are more reliable than the dividend payouts of well-diversified equity indices.  Rather, the reason is that bonds have a maturity date, a date where you can get your money out of the investment even if no one is willing to buy it from you.  That difference makes all the difference in the world.

If you buy a brand new 10 year treasury bond, and its price plunges in a panic, the worst that will happen to you is that you will be stuck holding the security for 10 years.  At 10 years, you will be made whole on your investment, regardless of what the market decides to do with the price.  Having to wait 10 years is certainly not as costly as having to wait forever, as one would have to do with an equity security that no one wants to buy.

Importantly, as the maturity date–the “finish line”–of a bond gets closer, it becomes easier to find others willing to risk a loss of liquidity in the security, given that that the “finish line” represents their “finish line” as well.  The security becomes easier to buy and just hold to maturity, which essentially ensures that there will be liquidity–a price reasonably to close to fair value–driven by confident investors that are willing to buy regardless of whether they think there will be yet others willing to buy from them.

For equities, however, there is no maturity, no “finish line.”  There is no amount of time that you can wait inside the investment in order to be guaranteed of being made whole on it.  Being made whole on it requires other people to want to buy it from you–without their interest, which depends crucially on their trust and confidence in the reciprocal process, their sense that others will be willing to buy from them at some point, you cannot be made whole on the investment, at least not in any finite amount of time.

Now, to be clear, if you hold an equity security for a long enough period of time, you can get your initial investment back in dividends.  But that’s not the issue.  You are not in the investment to get your initial money back, a return of your initial capital.  You are in the investment to get an appropriate return on your capital–you cannot get such a return in any finite amount of time without other people to sell to.

To use an example, if  you hold an equity security with a growing 2% yield, you will get your money back after 30 years or so.  But you will not get the return on your money that holding a security for 30 years demands.  In contrast, if you hold a 30 year government bond for the same amount of time, the entirety of its term, you will get such a return, which is why the two types of securities–bonds and equities–are fundamentally incomparable as instruments.

Valuation: Why It Matters

As investors in the real world, we do not invest in securities on the assumption that we are going to hold them forever, and therefore realize their “intrinsic value.”  Rather, we invest in them with the specific expectation of being able to sell them to other people at higher prices than we paid, thereby realizing a return.  We expect this return to be realized in a reasonable amount of time–months, years, maybe decades–certainly not any longer. A dividend stream can help pad our returns over those horizons, but the prices at which we sell the securities ultimately determine them.

We should worry about valuation, then, not because it determines the dividend return that we will receive on our investments, but rather because valuation is a factor that influences the perceived attractiveness of the security to other potential buyers of the security, those to whom we will sell, who we should view as our customers.  It is their perception of an equity security’s valuation–not our opinion of the reality–that will determine the price that they will be willing to pay for it.

That’s why it can be misguided for investors to focus on valuation metrics that no one uses. The value of an attractive valuation is that the valuation will be attractive to other potential buyers of the security–not that it is attractive to us, using our own pet methods.

Now, to be fair, let me add some nuance to the point.  Attractive valuations can either be obvious–readily seen by all–or they can be hidden.  When they are obvious to all, manifest in the classic “P/E” heuristic that investors use to quickly assess valuation, there will usually be some other factor–some set of fears–that is causing investors to not want to buy, despite the low P/E.  If we know that those fears are misguided and will eventually subside, then we can buy now, at the low P/E, and sell later, at what will by then be a more normal P/E.

Alternatively, the value may be hidden by the present P/E ratio.  It is then that unconventional metrics–metrics such as price-to-book, price-to-sales, enterprise value to EBITDA, Shiller CAPE, and so on–can be useful.  Such metrics can point to situations where the P/E is high, but high because of an abnormally low present E, rather than an abnormally high P.  Knowing that the E will eventually rise in a way that the market is not presently expecting, we can buy a cheap security that the market does not yet know is cheap, and then sell it when the value becomes evident to all, at which point the price will already have been pushed up.  Note that we can do the same in reverse, using unconventional metrics to stay away from expensive securities that appear cheap, appearing cheap because their Es have been artificially inflated by unsustainable trends–fads, bubbles, and so on.

Posted in Uncategorized | Leave a comment

Dilution, Index Evolution, and the Shiller CAPE: Anatomy of a Post-Crisis Value Trap

In the first century, the historian Plutarch introduced a famous philosophical paradox.  The paradox goes like this.  A ship–”The Ship of Theseus”–was returning home to Athens from Crete.  As it sailed, the wooden planks that made up its structure gradually decayed.  The sailors kept the ship afloat by replacing the decaying planks, one by one, using fresh wood that they were carrying onboard.  Eventually, the sailors replaced all of the wooden planks that made up the the ship’s original structure, so that the new form of the ship had no material in common with the old form.  The question followed: was the ship the same ship through the change? If so, what made it the same ship, rather than a new ship, a different ship?

“For they took away the old planks as they decayed, putting in new and stronger timber in their places.  The ship became a standing example among the philosophers of the logical question of things that grow: with one side holding that the ship remained the same, and the other contending that it was not the same.” – Plutarch, Theseus, 75 A.C.E.

Approximately 1500 years later, the philosopher Thomas Hobbes took the paradox further.  He asked us to imagine the following.  All of the old, decayed wood of the original Ship of Theseus is gathered up from scrap and used to build a new ship.  There are then two ships: one ship that is spatially continuous with the original Ship of Theseus, whose material has been fully changed out, piece by piece, and another ship made from the scrap material of the original Ship of Theseus.  Which of these ships is the true Ship of Theseus? 

The “Ship of Theseus” problem frequently arises in the world of music fandom.  Consider, for example, the 1970s soft rock group, the Little River Band, which produced famous hits such as “Reminiscing” and “Lonesome Loser“.  To this day, the Little River Band remains together.  But there are no current members of the band that were in the band when it was originally formed.  All of the founding members, those who sang the hits as we are used to hearing them, have been swapped out.  A “Ship of Theseus” question thus arises: is the band that currently goes on tour as “The Little River Band” the true Little River Band, or is it the equivalent of a cover band, singing the same songs, while only pretending to be the original? To add the Hobbesian twist, what if the original members of the Little River Band were to come together to form a new band, a cover band of the Little River Band.  Would this new cover band be the true Little River Band, since it contains the founding members?  Or would it be a mere replica, since it is not continuous with the original?

You’re probably asking yourself what relevance this paradox has to finance, or to anything. But now here’s a question for you. Suppose that we have an index of stocks that represents the equity market of a given country, an index that we use, without further questioning, to draw conclusions about important topics such as the country’s valuation and expected future performance. What would happen if, like planks on the Ship of Theseus, or members of the Little River Band, most or all of the individual companies in the index were to be removed, replaced with new companies?  Would the index remain the same index? Or would it become a different index?

The question of “sameness” and “difference” is inherently metaphysical, and therefore has no answer.  But there is a more practical question that we as investors have to be concerned with.  That question is this.  Given radical changes in the constituents of an index, is it appropriate to use the index’s historical metrics–its historical earnings, growth rates, valuations, profit margins, returns on equity, and so on–to draw conclusions about what the index’s future performance is likely to be?

Ireland: The Perfect International Value Play?

Looking out over the long-term, it’s going to be very difficult for US investors to receive the “normal” 10% nominal annual equity returns that they have received historically.  Literally everything will have to go right.  Profit margins and returns on equity will have to stay elevated, contrary to the tendency of mean-reversion.  Multiples will also have to stay elevated, which means that interest rates will have to stay low.  But low interest rates are a consequence of weak economic growth and weak inflation.  How are companies going to consistently produce strong earnings per share (EPS) growth–the kind that would be needed to underpin 10% total returns for shareholders over the long-term–in an environment of weak economic growth and weak inflation?

Up to now in the current recovery, and really over the last 10 years, profit margin expansion and share buybacks have been the primary drivers of EPS growth for U.S. equities.  They are the reasons that strong EPS growth has been possible amid the persistent softness in economic growth and inflation (softness that has depressed the corporate top-line, but that has also provoked zero interest rates and an elevated P/E multiple).  Can profit margin expansion and share buybacks continue to be robust drivers of EPS growth, indefinitely, even as shares become more and more expensive for corporations to buy back, and as the income imbalances between capital and labor, the rich and everyone else, get closer and closer to the limits of economic and societal stability?  There are good reasons to think not.

Because long-term equity returns in the U.S. are likely to be sub-par, many investors have turned to foreign equity markets for better opportunities.  Where is the value in the equity world right now?  According to the Shiller CAPE, a popular technique for measuring value across economic cycles, the value is in Europe, specifically, the distressed countries of the Eurozone.

In my view, out of all of the countries of the Eurozone, the most interesting from an investment perspective is Ireland.  As a country, it has all of the features needed for strong long-term equity performance, features that many of its cousins in the Eurozone lack: a productive, highly-skilled, flexible labor force, capital-friendly, pro-business government policies, and a young, growing population in a demographic sweet spot.  To complete the investment case, Irish stocks are apparently very cheap, with the Irish index sporting a Shiller CAPE under 10.

It would seem, then, that Ireland is set up to produce spectacular returns.  But there’s a problem.  If you look closely at the actual names that make up the Irish index, you will be hard-pressed to find significant value.  The following table shows the constituents of the ISEQ 20, Ireland’s benchmark, sorted by market cap weighting as of March 2014:


Most of these companies enjoy above-average valuations.  That’s to be expected, as the companies are high-quality.  Glanbia?  Kerry Group?  Smurfit Kappa? Aryzta?  These are growing, thriving businesses.  They deserve to be priced as such.

There are two potential cases of deep value in the index: The Bank of Ireland and the building material producer CRH.  But these companies together only make up 29% of the index capitalization.  The majority of the index–71%–is composed of companies that are not deep value.  How can an index represent deep value when 71% of its constituents are not deep value?  How can an index trade at a CAPE below 10 when 71% of its constituent companies sport CAPEs significantly higher than that number?  Where exactly is the low CAPEness coming from?

Enter the Ship of Theseus paradox.  It turns out that many of the companies presently in the ISEQ 20 are new entrants, having replaced the financial roadkill that died off in Ireland’s massive housing bubble and subsequent banking crisis–roadkill that includes Anglo-Irish Bank, Allied Irish Banks, and so on.  That roadkill is gone, forever, having either been nationalized or diluted into oblivion.  But, crucially, the earnings that it generated during the bubble, from 2004 to 2008, is still part of the ISEQ’s earnings per share during those periods, and is therefore getting credited in the CAPE calculation for the index.

The following table shows constituents of the ISEQ 20 as of March 2014 alongside the constituents as of January 2007.


As you can see, there has been substantial turnover in the index.  The current ISEQ 20 has one major commercial bank, The Bank of Ireland, which represents ~9% of the index’s total capitalization and 0% of the index’s current earnings.  The ISEQ 20 of 2007, however, had three major commercial banks, which together made up ~40% of the index’s total capitalization and an even greater share of its earnings at the time.

The Shiller CAPE is a tool for detecting hidden value.  During cyclical weakness, a company’s classic trailing-twelve month (ttm) P/E ratio will be abnormally elevated by temporarily depressed earnings, and will therefore give an inaccurate picture of the company’s future earnings potential.  To get around this problem, we use the Shiller CAPE, which compares the company’s price to the average earnings that the company generated over the previous 10 years.  The 10 year average of earnings gives a more complete picture of the earnings that the company can be expected to produce in the future, when conditions return to normal.

At the index level, the same logic applies.  We compare the price of the index with the average of the index’s earnings over the previous 10 years, to get an accurate picture of the earnings that the index can be expected to produce when conditions return to normal.  In this case, however, there’s a really big problem.  The index has undergone a radical makeover.  It’s not the same index anymore.

Conceptually, it doesn’t make sense to expect a normalization in Ireland’s economic condition to catapult the ISEQ’s earnings back to the levels seen from 2004 to 2008, when the ticking time bomb of a highly-leveraged banking system was the engine of profit growth. Of the three banks that made up the majority of the index’s earnings at the time, two are no longer in the index, and the other is an unrecognizable version of its former self, having undergone a massively dilutive recapitalization.

The Shiller CAPE: The Dilution Distortion

It turns out that there is an even more significant illusion being produced here.  To illustrate the illusion, I’m going to present calculations of the Shiller CAPEs of individual companies in the Irish index.  Note that the data, the majority of which is taken from GuruFocus, may contain minor errors, specifically related to the capitalization and share count of the companies, given how complex the changes have been since the crisis. Regardless, the calculations are adequate to illustrate the underlying process at play.

Consider the Bank of Ireland, whose CAPE calculation is shown in the table below:


Notice the column “Shares (MM).”  As you can see, there’s been a huge explosion in the shares outstanding of the Bank of Ireland, obviously related to the massively dilutive recapitalization that the bank was forced to undergo in conjunction with the financial crisis.

Let’s think about how this dilution might impact the CAPE.  In an extreme dilution, a company’s share price will fall by orders of magnitude–appropriately.  In the case above, the price fell from over 900 to 15.  In the CAPE calculation, the appropriately-collapsed price will be compared with the company’s past earnings per share, earnings that were earned when the share count was orders of magnitude smaller than it currently is.  The dilution-depressed current price per share will thus get measured against an artificially inflated past earnings per share, a number that in no way reflects the company’s future earnings potential. Users of the metric will therefore walk away with a completely false picture of the company’s valuation.

It turns out that there is an additional illusion associated with the dilution.  One would expect the crisis that caused the dilution to have produced a period of negative earnings that will get averaged into the Shiller CAPE, negating at least a portion of the artificial earnings excess of the boom.  To be sure, in the case of the Bank of Ireland, those negative earnings did come through.  Crucially, however, they were “registered” during the same reporting period as the dilution.  The losses were therefore diluted over an artificially large number of shares, producing a relatively small per share loss (relative to the large per share gains that were enjoyed during the boom).

To make the point more clear, let’s get specific.  To arrive at a per-share basis, the $1B to $2.5B that the Bank of Ireland earned each year in the pre-crisis period is being divided by the pre-dilution number of shares, approximately 24 MM.  But then, after the crisis, the subsequent $4B loss is being divided by the post-dilution share count, a number ranging from 100 MM to 750MM.  The result is what you see above.  The Bank of Ireland appears to have earned $50 to $100 dollars per share per year during the boom times, and to then have lost only around $20 per share during the entirety of the bust.  When you average these per-share numbers together to compute the average earnings, you get a deceptively high average, and therefore a deceptively low Shiller CAPE.

Now, to eliminate this distortion, what I’ve done in the table is calculate the CAPE on an absolute basis in addition to on a per-share basis.  Instead of comparing the price per share to the average real earnings per share over the last 10 years, the “Absolute CAPE” compares the current market capitalization to the average real net income over the past 10 years, with both numbers unadjusted for share count.  On this absolute basis, the CAPE for the Bank of Ireland rises from a ridiculously cheap 0.40 to a seemingly expensive 20.62.

The same distortion emerges to an even greater degree in the case of Allied Irish Banks, whose CAPE calculation is shown in the table below:


Uncorrected for the dilutive distortion, Allied Irish shows a CAPE of 1.99.  But the absolute CAPE is actually steeply negative, indicating that, on an absolute basis, the bank lost more money in the crisis than it earned during the boom.  Is that surprising? It shouldn’t be–over the long-term, bubble-bust finance is not a good business.  You eventually get completely wiped out.

There are similar distortions associated with the way in which troubled company’s tend to exit the index.  Troubled companies often get delisted and removed from an index before all of their losses have been taken, allowing the index to escape from the losses scot-free, even though the corresponding gains were registered without hindrance.  Worse yet, with some types of indices, even when the troubled company does remain in the index to register its losses, the publishers of the index don’t count the losses in the index earnings, because the losses represent one-time, non-recurring events.

I would have included a CAPE analysis of Anglo-Irish Bank, but they represent a prime example of an exit distortion, having been nationalized in 2009.  From 2004 to 2008, they earned a profit typical of the other banks, on the order of around $1B per year.  But then, in 2010, well after they had been nationalized and removed from the ISEQ, they took a cool $15B impairment loss, a loss that, if registered in the index, would have more than wiped out any profit that they contributed during the boom.  How convenient.  The ISEQ is able to count Anglo-Irish’s highly artificial profits earned during the boom, but then when the bust comes along, and it’s time for Anglo-Irish to drop its turd, the bank is already long gone from the index.  Its turd gets dropped in a black hole, leaving the ISEQ’s earnings unaffected.

Now, let’s look at the CAPEs of some of the larger non-financial companies in the ISEQ. First, CRH:


As the table shows, there’s clearly some value in CRH.   But it’s nothing to write home about. Notice that the company enjoyed the same dilutive effect that the banks enjoyed, albeit to a much smaller degree.  The absolute CAPE is 3 points higher than the per share CAPE, owing to the fact that the share count has increased by almost 50% over the period.

CRH is easily the cheapest non-financial stock in the Irish stock market, and yet it’s CAPE isn’t even below 12.  We should therefore be extremely suspicious when we see the Irish stock market as a whole register a CAPE below 10.  Trivially, a country cannot have a CAPE lower than the CAPE of its cheapest current constituent.  If it does, then something has necessarily gone wrong in the analysis.

Here is the CAPE for Ryanair, Ireland’s premier airline, which makes up about 9% of the ISEQ:


Ryanair sports a CAPE of 36.45–hardly a case of deep value.  Notice that its per share CAPE is actually higher than its absolute CAPE.  The reason is that it’s been shrinking its shares, rather than growing them.

Here is the CAPE for Kerry Group, a food producer in the ISEQ:


Again, a very high CAPE, on par with the CAPE levels that you might see in a growing U.S. company.  The per share CAPE is roughly the same as the absolute CAPE because there’s little change in the share count.

U.S. Banks: Similar CAPE distortions?  

The U.S. banking sector is often cited as the cheapest sector of the U.S. equity market.  It may be the cheapest sector–I’m not going to argue that point.  But the CAPE should not be what leads us to this conclusion.  The CAPE is not a conceptually valid way of measuring value in a post-crisis environment where share counts have appreciably changed.

The same distortion that we saw in the CAPEs of the Bank of Ireland and Allied Irish is present in the CAPEs of America’s junky financial analogues.  Consider, for example, the CAPE of Bank of America ($BAC), calculated below:


As you can see, $BAC suffered significant dilution in the crisis aftermath, simultaneous with its post-crisis writedowns, creating a distortion in its per share CAPE.  The per share CAPE comes in at 8.14, when the absolute CAPE is 19.29.

The following table shows the CAPE calculation for another junky financial, Citigroup ($C):


Again, the same distortion is present, to an even greater degree, given the greater dilution. The per share CAPE is 4.77, whereas the Absolute CAPE is 18.25.

Energy Companies: A More Benign Distortion

It turns out that the Shiller CAPE also creates distortions in the valuation of energy companies.  The reason is that energy companies generate earnings off of a depreciating asset base.  The appropriate way to value them is not to look at their past earnings, generated on assets that are now used up, but to conduct a discounted cash flow analysis of the future earnings that they will generate on their current asset base, as that base depletes away.

Consider, as an example, the case of Total, the integrated French oil company.  The following table shows Total’s Shiller CAPE:


As you can see, Total trades at a very attractive CAPE relative to the market.  It also trades at an attractive ttm P/E ratio–and always has.  The reason that it trades at an attractive CAPE and ttm P/E ratio is that its past earnings are not directly relevant to its current value.  What is relative to its current value is the ratio of its price to the discounted sum of its future earnings, earnings that will be generated as its finite oil reserves are drilled out of the ground and sold.  How plentiful are those reserves?  What is their quality?  How expensive will drilling them out of the ground be?  From a valuation perspective, these are the questions that matter.

The intrinsic value of an asset is the discounted sum of its future cash flows.  If you have a company with recurring cash flows generated off of a surviveable asset base, then it makes sense to use trailing metrics like the CAPE, the ttm P/E ratio, and the ttm dividend yield to approximate the value.  But if you have an energy company with an asset base that depletes every time product is pumped out and sold, an asset base that is difficult and costly to replace through new discovery, then these metrics will not provide an accurate picture of the value.

Discounted at 10%, the net present value of Total’s proven oil and gas reserves is $47B. The company trades at a market capitalization of $139B, with an enterprise value, including net debt, that is even higher.  On those numbers, Total is hardly a case of deep value. To the contrary, it appears to be overvalued–by at least 200%.  Before we jump to that conclusion, however, let’s consider a few points:

  • 10% may be too large of a discount rate to apply to the assets in the present interest rate environment.  Of course, lowering the discount rate won’t fully alleviate the apparent overvaluation.  Even at a 0% discount rate, the net present value of Total’s oil and gas assets is only $105B (and that’s before the recent oil price drop).
  • Total is an integrated company, and generates profit from refining and marketing in addition to production.  The profits associated with its refining and marketing arms have to be included in the valuation analysis, just as its net debt has to be included.
  • Proven reserves are often only a conservative estimate of the quantity of oil and gas that an energy company has access to and will be able to produce and sell over time. Given that the company trades at a 50% premium to its undiscounted proven oil and gas reserves, the market probably expects Total’s unproven reserves to be significant, possibly even larger and more valuable than its proven reserves.

The point, however, is that an analysis of the cash flow that will be generated out of Total’s current oil and gas assets, and not an analysis of the cash flow that it generated last decade, off of assets that have long since been converted into carbon dioxide, is what will determine the price that oil and gas investors will be willing to pay to own the company, and the price that they should be willing to pay.  That’s why Total trades at depressed P/E and CAPE multiples.  P/E and CAPE multiples simply are not relevant considerations in the oil and gas valuation process.

Reasons to Be Skeptical of European Value

If you examine the indices of the country’s in Europe that are allegedly offering investors deep value, you will notice that these indices are heavily allocated to financials and to energy as sectors.  In cases such as Ireland where the indices are not heavily allocated to the financial and energy sectors, there’s little deep value to be found.

The following table shows the CAPEs of important European countries, borrowed from Star Capital’s fantastic interactive website, alongside the country allocations to the financial and energy sectors in the respective MSCI indices.


As we see, the countries have low CAPEs, but they also have lopsided allocations to the financial and energy sectors.  In fact, there’s an apparent pattern: the higher the allocation to the financial and energy sectors–especially the financial sector–the lower the CAPE.

The heavy exposure to the financial sector substantially increases the risks of distortion, particularly given the credit bubble and subsequent crisis that Europe experienced. Greece, with a whopping 56% allocation to financials, and an unrecognizably low CAPE of 3.5, is particularly suspect in that respect. Where is its ultra-low CAPE coming from?  My guess: not from healthy business selling at attractive prices, but from crashed-out zombie banks that are distorting the index.

Of all of the regions in the world, Europe offers what is clearly the worst fundamental backdrop for investment.  The continent is overregulated, with inflexible labor laws and a generally business-unfriendly political climate, at least in certain countries.  The continent’s household, corporate, and financial sectors are heavily-indebted.  The population is in clear demographic decline.  The different countries that make up the continent have different cultural and competitive dynamics, yet are all trapped in a single currency union.  The exchange rates between the countries are therefore unable to naturally adjust so as to bring payment balances into line.

As if these structural headwinds weren’t enough, the monetary authority in Europe is a joke.  It has no ability to do anything to stimulate the European economy except talk.  The northern bloc won’t allow it to do anything more.  How long have we been attending to these meetings, listening to Mario Draghi tell us about the things that he might one day do?  At every meeting, the date of eventual action is pushed off to the next meeting, or beyond.  Nothing ever happens.

Markets love monetary policy, but in truth, monetary policy has little to offer in a situation like this, where households and corporations are deleveraging, and where the population and the workforce are shrinking.  In addition to supply-side labor reforms, what Europe needs is aggressive fiscal policy.  Fiscal policy has the ability to directly and reliably increase aggregate demand.  If aggregate demand is strong, real investment will start making economic sense (it doesn’t make economic sense right now).  Real investment will therefore increase, creating new sources of employment and income, fueling further increases in aggregate demand, incentivizing additional real investment, and so on, in a virtuous cycle.  For such a cycle to reliably take hold in a world that faces the kinds of headwinds that Europe faces, there needs to be an aggressive commitment on the part of policymakers to take whatever fiscal actions are necessary to keep aggregate demand strong–to intentionally and unapologetically run the economy hot, even if this means dropping freshly printed euros from a helicopter in the sky.

On that front, Europe could not possibly be worse off.  The weaker countries that need aggressive fiscal stimulus have no ability to borrow in their own currencies.  To conduct fiscal expansion, they have to get the permission of a separate country, Germany, a country with an obsessive fear of inflation and government debt, that does not have to share in any of their pains.

We can celebrate the fact that Mario Draghi said something, and that markets around the world rallied, but we should not let superficial price action blind us to the fact that the project of the Eurozone is an unsustainable mess.  The union is going to have to eventually dissolve, or at least undergo a substantial makeover.  Such a change is sure to bring turmoil to European financial markets, whether it comes next year, 5 years from now, or 20 years from now.  European investors deserve to be appropriately compensated for the risk.

Are they being appropriately compensated?  It’s not clear.  From a Shiller CAPE standpoint, it looks like they are being compensated, but that’s likely to be a result of the high financial and energy sector exposures that European indices contain.

Interestingly, U.S. investors can find the “deep value” that allegedly exists in Europe right here at home, in their own backyards.  All they have to do is go to the sectors that dominate European indices–financials and energy.  If they want a low CAPE, they can buy low-quality U.S. banks that were forced to recapitalize in the credit crisis–$BAC and $C, for example–or large cap integrated oil companies that trade on the productivity of their underlying oil and gas assets, rather than on P/E ratios–$CVX and $XOM, for example. These companies sport Shiller CAPEs that are just as low as the deep value companies of Europe.  There’s hardly a difference, for example, between the Shiller CAPE of a $BAC and that of a Banco Santander ($SAN), or the Shiller CAPE of a $CVX and that of an Eni Spa ($E).  The numbers are essentially the same.

Solutions to the Problem

As a metric, the Shiller CAPE is still useful.  It just needs to be employed with caution in countries that are coming out of large credit booms and busts, particularly those that have heavy exposure to financials, or that have had heavy exposure to financials in the past.  I’m therefore going to conclude the piece with some proposals for how investors might be able to avoid, or at least work around, the CAPE distortions that these countries give rise to.

One way would be to get under the hood of the indices themselves–examining how they’ve changed over time, how much dilution has taken place, what specific crisis-related losses have and haven’t been counted in the earnings numbers, and so on–adding whatever adjustments may needed to allow for an accurate valuation analysis to take place.  Unfortunately, this would be a difficult task.  The data is hard to find, and would take a very long time to piece together.

Another approach would be to use indices that intentionally exclude financials, and possibly energy companies as well. Unfortunately, none of the major index publishers produce ex-financial or ex-energy indices–for Europe or for any country.  Investors would have to build them directly, which would again be very complicated and time-consuming.

A more practical approach would be to evaluate the countries using ttm valuation measures, as a sort of “second check” on the Shiller CAPE metric.  The ttm P/E ratio is often criticized for only providing a picture of the last twelve months.  But that’s actually an advantage in this context, as it eliminates “The Ship of Theseus” problem. When you look at the ttm P/E ratio for an index, you can be sure that the “E” that you are looking at in the denominator is associated with the same companies as the “P” that you are looking at in the numerator. As we saw in the case of Ireland, you cannot always be sure of this fact when you use the Shiller CAPE on an index.

One good valuation metric to use, backed up by significant academic research, is the ttm enterprise value to ebitda (EV/EBITDA) ratio.  The advantage of ttm EV/EBITDA is that it includes net debt, which should be part of any valuation analysis, and also that it eliminates many of the non-recurring non-cash charges that tend to distort earnings, particularly around recessions.  The disadvantage, of course, is that it doesn’t count depreciation, and therefore it causes companies that have high depreciation costs, such as energy companies, to look artificially cheap.

If strictly non-cyclical measures are preferred, two additional ttm metrics that can be used to “second check” the Shiller CAPE are the ttm price to sales (P/S) ratio and the ttm price to book (P/B) ratio.  Like the P/E and EV/EBITDA ratios, these metrics only look at the prior year, and therefore avoid the “Ship of Theseus” problem.  At the same time, they solve the problem of cyclicality, given that sales and book values do not significantly fluctuate across the business cycle.

The problem with P/S and P/B ratios is that they tend to be different for countries that have different sectoral compositions.  Naturally, countries with higher allocations to high margin and high ROE sectors will tend to exhibit higher P/S and P/B ratios than those dominated by low margin and low ROE sectors.  We don’t necessarily want to penalize them for that in the analysis.  Additionally, for the P/B ratio, not all countries writedown their assets using the same standards.  European companies, for example, did not take the “goodwill” writedowns that U.S. companies took during the crisis.  For that reason, their P/B ratios tend to be lower, as explained in this analysis from KPMG.

A clean way around this problem would be to normalize the P/S and P/B ratios of different country indices to reflect the different sectoral compositions that those country indices exhibit and to reflect an application of the same writedown accounting standards.  Then, an apples-to-apples comparison between countries would become possible. Unfortunately, such a project would be too difficult and too time-consuming to put into motion.

When we check Ireland’s CAPE against its ttm P/S and P/B ratios, we quickly notice that our prior suspicions were correct: Ireland is not a case of deep value.  The country trades at a P/B ratio of 2.3 and a P/S ratio of 1.4, both of which register as expensive in comparison with the rest of the globe.  To be clear, Ireland may still be an attractive long-term investment opportunity–it probably is, given its many strengths–but the reason has nothing to do with its apparent status as deep value.

Fortunately, when we check the CAPE of the more-distressed PIIGS countries–Portugal, Italy, Greece, and Spain–against their respective P/S and P/B ratios, the countries continue to register as cheap.  It’s probably true, then, that the countries represent deep value–specifically, deep value concentrated in the financial sector, and to a lesser extent, the energy sector.  With respect to Greece, however, the P/B and P/S ratios, at 1.0 and 0.5 respectively, are not as cheap as would be expected given the 3.5 CAPE, which is almost half that of the closest competitor. Something is likely wrong with that number.

A final solution would be to not discriminate at all on the basis of country borders. If we’re looking for international value, let’s look for international value, in whatever country it happens to be located.  By looking strictly at individual companies, we can eliminate the need for indices altogether, bypassing the “Ship of Theseus” problems they create.

On that theme, there are a number of well-run international ETFs that take valuation factors with solid historical track records and apply them in foreign markets to locate attractive individual company opportunities.  Examples include (1) Cambria’s $FYLD, an international version of the successful $SYLD, which invests in companies that have a high shareholder yield, (2) Invesco’s $IPKW, an international version of the successful $PKW, which invests in companies that are buying back significant quantities of their own shares, and (3) Valueshares’ $IVAL, a not-yet-launched international version of the recently launched $QVAL, which invests in companies that exhibit attractive ttm EV/EBITDA ratios and that pass various quality screens.

Posted in Uncategorized | Leave a comment

Not Everyone Sucks at Investing

Judging from the financial headlines, we live in a world where everyone sucks at investing.

Hedge funds?  Consistent underperformers: this year, last year, the year before that, the year before that, the year before that.  Every year, it seems.  Just google “hedge funds” and “underperform”, to see the flurry of giddy articles that pop up.


Individual Investors?  Again, consistent underperformers.  They get excited at the tops, they panic at the bottoms, they do everything exactly backwards to the maximum extent possible.  The published numbers here are quite ugly: according to Dalbar’s 2013 QAIB publication, the average individual equity fund investor has earned a pathetic 3.69% annualized return over the last 30 years, versus the S&P’s 11.11% (note: the QAIB report may contain distortions).

How is such consistent underperformance possible?  The answer, we are told, is behavioral. Investors, of both the professional and the layman stripe, tend to herd.  They prefer to do what everyone else is doing.  And so they end up buying when assets are in high demand, at the worst possible prices, and selling when assets are out of favor, again at the worst possible prices.

There’s an obvious problem with this narrative, which you’ve probably already noticed. For every party in a trade, there is a counterparty–for every buyer, a seller, for every seller, a buyer.  There must, then, be an outperforming counterparty to the underperforming average investor, and the undeperforming average hedge fund, and the underperforming average day trader, and the undeperforming average endowment, and whoever else underperforms on average.  Someone had to be smartly selling to those groups in 2000 and 2007, for example, when they were frantically trying to get in, and smartly buying from them in 2003 and 2009, when they were desperately trying to get out.  Who–what group–is that someone?  And why doesn’t the financial media ever celebrate its achievements?

I’m glad to be able to tell you that I am a member of that group.  Over the last 15 years, I have compounded my own capital at a 35.9% annual rate, profiting handsomely from the ill-timed and ill-advised decisions of “average” individual investors, mutual fund managers, and hedge fund managers alike.  And don’t be fooled; it’s not just me.  A lot of us do quite well, thankfully.

Of course, everything I just said is a bald-faced lie.  So don’t worry.  I’m not better at life than you.  But how did it make you feel to read about someone else’s spectacular performance?  Probably not very good.  That’s why the media prefers the “everyone sucks” headline.  It makes for fun, satisfying, ego-pleasing reading.

The truth is this.  Investors in aggregate are the market.  Before frictions (fees, transaction costs, etc.), they cannot underperform.  Nor can they outperform.  For they would be underperforming and outperforming themselves, and that is obviously impossible.  Now, if we arbitrarily divide the market into different categories of participants–individual investors, hedge funds, pension funds, corporations, and so on–then it would be possible for some categories to consistently underperform others (note that this would create tension for the efficient market hypothesis–pure negative alpha is, in fact, a type of alpha). But, necessarily, the other categories would be outperforming.

What category of investor, then, is consistently outperforming the market, against the consistent underperformance of hedge funds, individual investors, and other losers?  You will be hard pressed to find an answer.  An obvious candidate would be the corporate sector, which has, in recent years, taken large amounts of equity out of the market through share buybacks and acquisitions, effectively forcing the rest of the market to be net sellers.  The problem with citing corporations as the clever counterparties, however, is that corporate managers exhibit the same herding tendencies as the rest of the market.  According to Z.1 data, they too prefer to buy high and sell low, having bought heavily around the 2000 and 2007 peaks, and having sold at the 2003 and 2009 troughs.


Part of the problem here is that we arbitrarily treat the S&P 500 as “the market”, the benchmark for evaluating performance.  But the S&P 500 is not a reasonable benchmark to use, since investors in aggregate do not allocate the entirety of their portfolios to U.S. equities.  Indeed, investors in aggregate cannot allocate the entirety of their portfolios to U.S. equities–if they tried, prices would go to infinity.  The strategy of devoting an entire portfolio to U.S. equities, which may look brilliant right now given the recent performance, would necessarily become a bad idea (if it isn’t already a bad idea).

The appropriate benchmark for performance evaluation is the global asset market, which includes all global assets: stocks from all countries, bonds from all countries, real estate in all countries, and, importantly, cash from all countries (commercial paper, government bills, bank deposits, and so on).  Over the long-term, some groups will surely outperform this market.  If the efficient market hypothesis is true, we should expect it to be those groups that choose to accept the most risk in the choice of what they own.  If the efficient market hypothesis is not true, then we should expect it to also include those groups that possess skill, that manage to own the right assets at the right prices at the right times.

Similarly, some groups will surely underperform the global asset market, because those groups choose to take on less risk than the global asset portfolio contains (making it possible for other groups to take on more risk), because those groups lack skill (making it possible for other groups to demonstrate skill), or because those groups stupidly accept unnecessary frictions–management fees uncompensated by skill, overtrading with high commission costs across large bid-ask spreads, and so on (making it possible for financial middlemen to earn a living).  But the point is, with performance properly measured, it’s not possible for everyone to consistently underperform.  Not everyone sucks at investing.

Posted in Uncategorized | Leave a comment

Valuation from All Angles: S&P 500, Russell 2000, and the 10 GICS Sectors

(Much thanks to the must follow @ElliotTurn for valuable help and feedback in the development of these charts and tables)

In this piece, I’m going to present a series of charts and tables that seek to efficiently convey the state of U.S. equity valuations from all available vantage points–that is, “from all angles.”  Note that a convenient slideshow aggregating the tables and charts together is presented at the bottom.

S&P 500:

The following “ttm” chart shows trailing-twelve month (ttm) values and ratios from 1996 to 2014 (click on the chart to enlarge):


(Legend: The squares show the following metrics (1 to 20, left to right, top to bottom): (1) real price returns and real total returns (with dividends reinvested at market prices), (2) trailing-twelve month (ttm) dividend yields, (3) ttm price to earnings (P/E) ratios and fwd P/E ratios based on analyst estimates, (4) ttm enterprise value to earnings before interest, taxes, depreciation and amortization (EV/EBITDA) ratios, (5) ttm price to ebitda (P/EBITDA) ratios, (6) real ttm sales and book value growth, (7) real ttm dividend growth, (8) real ttm EPS growth, (9) real ttm EBITDA growth, (10) annualized inflation rates for the prior 6 years and long-term government bond yields, (11) ttm price to sales (P/S) ratios, (12) ttm dividend margins (ttm dividends as a % of sales), (13) ttm EPS margins (ttm EPS as a % of sales), (14) ttm EBITDA margins (ttm EBITDA as a a % of sales), (15) interest, taxes, depreciation and amortization (ITDA) as a % of EBITDA (which gives a picture of how much the earnings are being reduced by those expenses at any given time–very important), (16) price to book (P/B) ratios, (17) ttm dividend payout ratios (ttm dividends divided by ttm EPS), (18) ttm EPS return on equity (ROE) (ttm EPS divided by book value), (19) ttm EBITDA ROE (ttm EBITDA divided by book value), (20) real net debt (debt minus cash and liquid assets, i.e., the difference between enterprise value and price).  The dotted black line in each chart shows the metric’s average for the period.)

The following “Shiller” chart shows different types of Shillerized valuations from 1996 to 2014:


(Legend: The squares show the following metrics (1 to 15, left to right, top to bottom): (1) shiller P/E ratio (real price divided by the of average real ttm EPS seen over the prior 6 years (10 leads to too much information loss), (2) price to peak earnings (P/PkEPS) ratio (real price divided by the highest ttm real EPS earnings reading seen over the prior 6 years), (3) Shiller EV/EBITDA (using 6 years), (4) enterprise value to peak EBITDA (EV/PkEBITDA) ratio (using 6 years), (5) Shiller price to EBITDA ratio (using 6 years), (6) real shiller EPS (average of real ttm EPS over the prior 6 years), real peak EPS (highest ttm EPS seen over the prior 6 years), (7) real Shiller EBITDA (average of real ttm EBITDA over the prior 6 years), (8) real peak EBITDA (highest ttm EBITDA seen over the prior 6 years), (9) real ttm Sales and real ttm Book value, (10) – (14) margins and ROEs for all Shiller and peak metrics (Shiller EPS / sales, Shiller EPS / book value, Peak EPS / sales, Peak EPS/ book value, Shiller EBITDA / sales, Shiller EBITDA / book value, peak EBITDA / sales, peak EBITDA / book value, (15) asset turnover, i.e., sales / book value.)

The following table presents data from the above charts in numeric form.


(Legend: The upper left quadrant shows valuation metrics as of the close on 11/13/14 and the average for the period (along with the delta between the present value and the average). The upper right quadrant decomposes the returns into dividends, growth in fundamentals, and changes in valuation for three different fundamental bases: price to sales, Shiller P/E, and Shiller P/EBITDA.  Note that the “true ROE” of the corporate sector is the return that it would produce in a given period if valuation were held constant during that period.  Thus the true ROE equals the dividend return plus the return due to growth in the given fundamental (which will necessarily equal the growth in the price if the valuation relative to that fundamental stays constant).  The lower left quadrant shows margins and ROE as of the close on 11/13/14 and relative to the average for the period.  The lower right quadrant shows valuation metrics relative to the long-term government bond yield.)

I will now present the same charts and tables for the the Russell 2000 and the 10 GICS sectors–Consumer Discretionary, Consumer Staples, Energy, Financials, Healthcare, Industrials, Materials, Technology, in that order–in slideshows.

Slideshow: TTM Charts

Here are all of the “ttm” charts (ttm valuation ratios, growth, margins, ROEs, inflation, government bond yields, etc.) in a slideshow (going from upper left to lowe right: SPX, R2K, Discretionary, Staples, Energy, Financials, Healthcare, Industrials, Materials, Technology).  Click on any image to start the slideshow there:

Slideshow: Shiller Charts

Here are all of the “Shiller” charts (Shillerized data) in a slideshow (going from upper left to lowe right: SPX, R2K, Discretionary, Staples, Energy, Financials, Healthcare, Industrials, Materials, Technology).  Click any image to start the slideshow there:

Slideshow: Tables

Here are all of the tables (going from upper left to lowe right: SPX, R2K, Discretionary, Staples, Energy, Financials, Healthcare, Industrials, Materials, Technology).  Click on any image to start the slideshow there:

In a subsequent piece, I will present the same charts and tables for 17 different countries.  I will also present tables that rank the sectors and countries by the different valuation and growth factors.

Posted in Uncategorized | Leave a comment