IS THE SHILLER PE OUT OF DATE?
Merrill Lynch seems to think so. In a recent note they listed the 5 reasons why the Shiller PE is unreliable. I think they build a pretty strong case (via Hedge Analyst):
Shiller’s PE misrepresents S&P 500 valuation
The current Shiller PE of 24x is 50% above its 1900-2010 average of 16x. While we would normally take concern with a valuation measure 50% above its long-term average, we encourage investors to ignore this one. We believe Shiller’s $55 inflation-adjusted 10yr trailing avg. EPS is not a fair representation of normalized EPS for 2010 and prefer our $90 Equity Time Value Adjusted (ETVA) 10 yr. avg. EPS. The current ETVA PE of 14.7x is almost a multiple point lower than the 50yr average of 15.6x and is well supportive of our 1400 year-end target.
Shiller’s PE understates normalized EPS
Shiller’s PE cannot be fairly compared across time because it neglects substantial shifts in dividend payout ratios over the last 110 years. Anytime the dividend payout ratio is not 100% EPS should rise with inflation plus the return on reinvested earnings. This is called an Equity Time Value Adjustment (ETVA). Shiller’s EPS does not fairly represent normal EPS because it assumes EPS only grows by inflation, which given a decade of high EPS retention, is flawed.
ETVA 10yr avg. EPS is a better proxy of normalized EPS
We believe our $90 ETVA 10yr avg. EPS estimate is a better representation of normalized EPS for 2010 and supports our 2011 normalized EPS estimate of $95. We advocate a 10yr PE based on 10yr avg. ETVA EPS because the 33% average S&P 500 payout ratio over the past 10 years is significantly lower than its
1900-2010 average of about 60%. While we expect the dividend payout ratio to rise to 38% through 2012 and prefer dividend growth stocks, the payout ratio will still stay well-below average.Compare today’s S&P 500 to 1960 onwards, not 1900
We prefer to compare today’s PE to averages since 1960. We do not advocate including the 10 years after 1914 in the long-term average 10yr PE as US companies benefitted from supplying Europe in WWI and experienced an exceptional swing in profits not likely to occur again. EPS tripled from 1914 to 1916. By 1921, in the deep post-war recession, profits fell to less than a fifth of the 1916 peak causing the 1921 PE to be 25.2x, but only 5.6x on average 10yr inflation adjusted EPS. Furthermore, the S&P 500 has only existed since 1957.
Pro forma overstates true EPS, but GAAP understates
While Shiller’s PE is based on GAAP EPS, we prefer to use pro forma EPS when making comparisons to EPS reported many decades ago. Since 2000, goodwill and asset write-downs increased owing to the elimination of pooling accounting for mergers and impairment tests for acquired goodwill. GAAP understates true EPS due to the downward bias of these balance sheet adjustments. We apply an accounting quality adjustment to our normalized EPS before using it for valuation to account for pro forma overstating EPS.
Source: Merrill Lynch/Bank of America
Sounds like it was in the late 90′… PE of 40x for the Nasdaq: everything ok, it’s new normal…
But i admit that comparing 1900 figures and 2010 figures is a little bit misleading.
The other readings such as Q ratio and GNP:Total market is still extended and is consistent with the Shiller data. Not sure how to rectify that, but I certainly don’t believe in a new normal in valuations….Maybe we have someone smarter than myself with thoughts on this…..
I agree TPC.
1) the others “PE-like” indicators say the same story (dshort, Hausman, …).
2) on the other side, technology evolution since 1900, and optimisation of all enterprise process can eventual lead to a (slightly) increase of the PE.
Not that I am an expert in this, but the ML arguments do not seem to be all that convincing after all, but sound more like justification attempts.
Dear Cullen, I read your blog everyday. I think your balanced representation of different views is very valueable, even though it sometimes leads to painful insights. Like this one, which seems to confirm your skepticism about using valuation metrics in investing. I couldn t find the link on Hedge Analyst, btw.
I ve been somewhat underexposed since last summer to stocks, partly because of the warnings from the Shiller/Hussman/Grantham trio, who as far as I understand all agreed that prospective 10 year returns weren’t particularly attractive anymore.
I am very curious to what extent this critique from ML also applies to the methodologies of Hussman and Grantham. If it does, and if their claims about overextended and soon to mean revert profit-margins are also flawed, I can finally turn into the raging bull that appartly almost everybody else is. Or?
Well, we still have a balance sheet recession of course.
Keep up the good work Cullen
“Q” has none of the flaws of PE10 and it is in full agreement.
Ignore Wall Street shills and prosper.
I think the key justification for them and fatal flaw in their logic is that a lower payout ratio HAS to imply that normalized EPS are higher than that implied by the Shiller method.
In effect, they are arguing that lower payouts = higher EPS retention = larger cash balances = more capital to deploy into CapEx/positive NPV projects = higher long term, terminal growth rates = higher normalized EPS = market cheaper than it appears.
If this were valid logic we should be able to look at companies within the market that have pursued similar capital structure policies and check share performance to see if it comfirms MLs view. While just one example, MSFT comes to mind – massively building cash balances over time have implied a LACK of future growth opportunities, not an abundance as MLs logic would suggest. MSFT, like the market in general over the past decade, has been rewarded with poor share price performance as such.
I would also counter ML by suggesting that even if I accept that normalized EPS have to be higher as payout ratios drop, building cash balances are indicative of declining ROE over time and that on the margins, each new project has higher risk/lower reward qualities than the last and thus, the market deserves a BELOW average multiple on those higher EPS.
And, to the extent Shiller has suggested 0% returns for the past decade and the market has actually delivered as much, with the metric also being in agreement with all the other valuation tools as others point out, it seems to have some credibility.
As always the sell side saying “This time is different”
Like an addiction, this propped up market is slowly drawing the masses into it, and risk is fading. The next plateau is high valuations feeling normal, and justified. Over time, and with new participants entering the market, we just continue to cycle from different this time to not different, and mean reversion. As humans, we’re doomed on this one.
On the other hand, JPY looks to topping this time. All in all, a falling JPY will fuel the carry-trade again. Take a look at the potential breakout on JPY/USD. This time is not different…
The biggest flaw to me is that it uses earnings from 10 years ago. Since the market is a leading indicator who cares about 10 years ago to say if its currently over/undervalued. I’d focus on forward earnings going back to 1945 – Post WWII.
The trouble with only looking at post 1945 figures are that they show a period in which the US was ascendent.
I believe that part of the reason that American market participants tend to be more bullish is that they believe that equities also outperform other assets classes over the long term. The problem with this is that they only look at the 20th and early 21st century when the US was, by far, the world’s economic out-performer.
With the US likely to be a laggard for the next decade or more, and with the US government and the Fed inflating away the US’s economic leadership, the outlook for US equities in real terms is likely to be bleak.
When the dollar stops being the world’s reserve currency, and it is a matter of when not if, the US will begin to look a lot like post-WW1 Britain.
LOL! What a joke. The payout ratio has declined from 60% to 30% in part because the earnings have declined from REAL to FICTION. Reported “earnings” have become increasingly meaningless as companies game the accounts. You can’t pay out fictional earnings which necessarily means the payout ratio falls. And if companies really are retaining more of their “real” earnings to “drive growth”, then whay has not the secular growth in corporate earnings increased? Give me a break. Merrill’s case for higher P/Es is as airtight as the hindenburg. Brought to you by the company that blew itself up on toxic MBS
I have not been able to find the Merrill piece. So, possibly I am missing something but…..
I am afraid this is not a compelling case. There are some points to be made, which is why combining various valuation models is important, but in the big scheme of things it leads to arguments about whether the market is 35% vs. 40% overvalued. Interesting, and over long enough time frames can make a big difference, but does not make a big enough difference for we with time frames of 10 to 20 years.
1. Higher retained earnings does not imply higher EPS growth rates. As Arnott, Asness, Steve Galbraith and others have demonstrated, quite the opposite. Lower payout ratios are correlated with lower EPS growth rates. They make a testable empirical observation and just hand wave past any evidence to back it up.
2. ” Anytime the dividend payout ratio is not 100% EPS should rise with inflation plus the return on reinvested earnings.” Nonsense on a stick. Okay, so please explain why earnings have not risen more than 2% above inflation over time? In fact, the trend growth is about 1% and has generally only been higher when starting from depressed margins (and lower when starting from elevated margins.) The payout ratio has not made a difference.
3. “We believe Shiller’s $55 inflation-adjusted 10yr trailing avg. EPS is not a fair representation of normalized EPS for 2010 and prefer our $90 Equity Time Value Adjusted (ETVA) 10 yr. avg. EPS.” Whether this bit of subterfuge is intentional or a product of ignorance is debateable, but the two numbers are not comparable in any way. Shiller’s number is not intended as a prediction of earnings in 2010. It is merely the number that when used is internally consistent with past numbers to allow you to compare levels of valuation used in the past. It is a 10 year average that of course will be significantly lower than the terminal year or terminal year plus 1. Their number is too high (at least adjusted for the cycle) in my opinion, but it is attempting to make an EPS prediction for 2010 and 2011. Shiller is not.
4. To wrap it up, if Merrill’s reasoning is correct sveral observations have to be true. a.) We have been in a 10-20 year period when earnings should have grown remarkably faster than in the past. They have not. b.) Earning have to be growing on a sustainable basis above both real and nominal GDP. I would argue that is impossible, and S&P 500 earnings growth has generally trailed Real GDP by about 60%! c.) methods that do not use ML’s assumptions that do attempt to project earnings in the future (including other methods by Shiller, which this method does not) should have performed poorly in making such predictions over the last ten years. They have not, in either predicting approximate real EPS or stock returns. ML’s method? I have no idea.
Wow! Ain’t the “new economy” great!
…will they never learn???
I agree with much of what Lance says, the big things for me are:
- I’d be fascinated to see more about their ETVA 10-year average, what it’s looked like historically. It’s not clear whether it’s a genuine cycle number or a point estimate that’s just been estimated/massaged — and I’d be curious as to when their ETVa has found the market expensive/cheap.
- I’m also skeptical of anything that chooses to use pro-forma EPS…it’s a little bit too far down the drinking the Kool-Aid for me. I mean, did we really learn nothing over the last decade!
- They only want to look at the period since 1960. This is pretty much a period that encapsulates the idea of the “great moderation” (even if it wasn’t fully implemented till the 1980′s) and (till the last couple of years) continuous credit expansion. I don’t think there is a more benign period in economic history to look at than 1957-2007.
i also would like to see someone post something more about etva. it looks like to me like etva’s relationship to schiller pe is comparable to ema is to sma.
if this is true, then of course you are going to get different values. you use the value (or blend) that you think is most relevant.
what am i missing…
What’s important is not so much the actual value of the Shiller PE, but rather the waveform it allows you to identify over long periods of time. If you stare at it long enough you will see half-periods of roughly 17 years up and 17 years down. Unless “things are different this time” the downward trend will re-exert itself at least one more time before things bottom around 2016.
As reported eps of 87 for 2011. Steepest yield curve . If u are short cover.if u are long lever up . Its 1999.
I largely agree with Lance’s comments above but another thing not mentioned are buybacks. Dividend payout rates have declined but buybacks have increased. To some affect companies are simply shifting dividend dollars to use for buybacks and not necessarily plowing the earnings back into the business. You really would need to look at dividends + buyback if you want to make the Retained earnings = growth argument they are making.
The data doesn’t support Merrill’s claim that lower dividend rate = higher EPS growth. The 3yr and 10 yr EPS growth rates have a positive correlation to dividend payouts rates in both the post-1900 and post-1960 time frames. The correlations are fairly weak but this means the higher the dividend rate, EPS growth tended to be higher. Exactly the opposite of what they are suggesting.
I have many problems with their analysis; typical sell-side, Wall Street search for justification IMO.
That is an obvious point I missed because the argument they used was so misleading I was pretty PO’d. Chad makes another obvious point as well, if their arguments were correct, they are really incorporated into Shiller anyway in terms of comparing to the past. Since Shiller isn’t making a projection in his P/E methdology, it is only relevant if you assume the next ten years will be dramatically different (in a positive sense for earnings) than the last ten years.
Which brings up another point. The commments about the past are irrelevant in making assumptions about future returns. Since the ratio is based on the last ten years, unless the next ten years is dramatically better the likely return ranges one would expect based on Shiller should hold as well as usual. The comments about pre 1960 become a red herring when you think about it. The numbers from 1910 have exactly no bearing on how this years P/E 10 is calculated. The only question is whether any particular 10 year period is likely to be dramatically different from the preceding 10 year period in terms of earnings. History shows (and they do not even attempt to make the argument otherwise, they just shift the terms they are discussing around as if they are) that there is little evidence over ten year periods the difference in earnings is ever major, and the few times when it is it mean reverts in subsequent years (for example post 1930 depression.) Even then, because markets tend to overreact, the shiller P/E works really well in identifying those “this time is different” turning points by predicting amazing or truly awful future returns due to investor overreaction.
The more I think about it, the worse their argument becomes, with a lot of it arguing points that are not germane to the question they are asking. I am actually disappointed I allowed myself to be run down the wrong hallways so easily.
Mr. Paddock’s analysis is spot on.
ML offers several faulty assumptions and subjective statements masquerading as observable data.
1. “Anytime the dividend payout ratio is not 100% EPS should rise with inflation plus the return on reinvested earnings.” OK, that is embedded already in the Shiller method, which simply averages the actual growth of real earnings. Those earnings reflect the historical return on reinvested capital not paid in the form of dividends to shareholders. That comes to about 1.8% by my math, dependent on the time series one uses.
2. “We prefer to compare today’s PE to averages since 1960.” Well, that is a neat trick of data mining. What is so special about 1960? How about post-Nixon gold standard removal. Why not 1960-1990 to exclude the tech/real estate bubble.
3. “Pro forma overstates true EPS, but GAAP understates.” This is patently and observably false. If one cared to look at the changes in book value over time, they would see that reported GAAP EPS overstates earnings, while operating EPS is a work of fiction altogether. Retained earnings that accrue to book value plus dividends paid are the only means of growing shareholder wealth.
I just took a really simple approach using Z Scores.
Whether you use from 1961 or 1871 to present, we are basically within .25 standard deviations of the mean.
The market doesn’t look outside of its historical range of valuation.
Disagree.
Using post 1871 data:
-Average CAPE = 16.4
-St Dev = 6.6
-Current CAPE = 24.2
-z-score = 1.19. Assuming normal dist, this puts the current CAPE in the upper 12% of historical valuations ( i.e., higher than 88% of all observations).
Using post 1860 data:
-Average CAPE = 19.4
-St Dev = 8.1
-Current CAPE = 24.2
-z-score = 0.60. Assuming normal dist, this puts the current CAPE in the upper 27% of historical valuations ( i.e., higher than 73% of all observations).
The second set of numbers should be “post-1960″ not 1860. My apologies.
I specifically mentioned I was keeping it simple.
No CAPE.
Just used the earnings for 2010 in the PE Ratio.
The best way to ball park aggregate market valuations is using a low-variability fundamental.
One way is to simply draw a trend line across a peak earnings trend growing at about 6%/yr, similar but not identical to what John Hussman uses. Then you can relate the market relative to the long term trend which mitigates the effects of a) the business cycle and b) profit margins. Unless there is some new math where companies can grow earnings faster than the economy, this peak trend line will “contain” earngings even at its highest cyclical peaks. It makes no judgment about macro issues.
Based on this method:
Peak Trend PE: 14.4x
Median: 10.5x
Expected 7-yr return 3.4%
(terminal reversion to median, annualized rate)
Interestingly as I am delviong deeper into their thesis, Merrill is explicitly claiming that S&P500 earnings will grow faster than the economy, though the time frame is unspecified. Moreover, the method uses mid cycle earnings, which in their view is a point where trend growth of 6% earnings can be counted on indefinitely. Of course, that trend only holds at peaks, and includes an inflation rate above 4% historically. The actaul regressed trend is substantially lower. So, the claim that using a 6% trend from mid cycle earnings (which means mid of the earnings growth cycle, downturns are not included in the cycle for this purpose) is representative is a pretty darned heroic assumption without some significant inflation and elevated margins indefinitely.
Whatever they are doing, I would like what they consider to be over valued. Should the S&P rise to the old highs of around 1,550 — a possible 16% move from here — if that isn’t overvalued I dont know what is.
But Wall Street never has a condemning word for the market, regardless of price or fundamentals.
Overthinking is as dangerous in trading as not doing homework.. 2009 and 2010 was the graveyard of some of the best traders and many one time hits of 2008 that litters the history of wall street…Roubini bot a Manhattan apartment…so if u don’t have a house buy one now .. if u have a house .. buy another
Forget earnings – they are creative accounting. Take figures you can’t fudge : $SPX and $UST. The ratio between them is currently 11. At the bottom of a bear market it will dip to circa 6 (5.5 in 2009). At the top of a bull market the ratio will reach at least 15. So if the parameters hold there is a 30% upside in the ratio. If stocks and yields share the upside equally a 15% gain on $SPX would equate to 1540 and a 15% gain on 10 year yields would equate to 4.11%.
Lance and Chad S. are right on in their comments. I wrote a more thorough critique of the Merrill piece which can be found here:
http://www.valuerestorationproject.com/2011/02/is-shiller-pe-outdated-more-thoughtful.html
Very interesting discussion. I was looking up discussion of Schiller 10 based on column this week’s Buttonwood column in The Economist (which mentions it as often)
-earnings growing faster than the economy. This is certainly possible in general, say in small European stock indices dominated by companies that do most of their business outside the home country. More S&P earnings come from outside the US economy than in 1960 or (if S&P had existed) 1901 or 1871. Which economy?
-Schiller 10 is not saying the stock market or economy function now exactly as in 1901 or 1966 (previous peaks mentioned in Economist column) but simply that (somehow normalized) earnings in the coming 10 years are not that likely to greatly exceed those in the preceding 10 years in general. I get that part. But there are also secular, or long cycle at least, rises and falls in PE, and it’s very unstatisfying to me to just say ‘they are 17 years up and 17 years down’ without explanation, an explanation that *would* depend more heavily on the character of economy and market being basically unchanged over long periods. ‘I could pick up the calf a few weeks ago, could do it today, so I’ll always be able to pick it up’. Sometimes things do become different.
-ML makes a good point about the period from 1914 to post WWI recession, extraordinary reasons for extraordinarily wide swings in profit which must have made Schiller 10 a less compelling measure during that period depending which big swings were included in the 10 year window. Schiller 10 now bascially says the current market PE has to be only 10 or ‘overvalued’, statistically because of the big crater in earnings of the recent crash. There are always people who’d only like the market at PE=10, I’m a bit underinvested myself and would prefer to be able to buy more at 10! But I don’t see a compelling reason the S&P *must* trade at (current PE) 10 because there were huge losses racked up in an almost unprecedented episode recently, when OTOH we have a quite historically reasonable ‘spot’ PE and we’re not at a (n obvious) cyclical peak in real business for companies; and 10yr treas rate 3.6% (again with a POV that says things *do* become somewhat different over time, sometimes, low i rates and low PE’s pairings of the distant past are not a wholly convincing answer, IMO).
Schiller 10 can’t be more than a very general guideline, seems to me. Wall Street sell side wants to sell, sure. Academics aren’t all necessarily bears, but they are over fond IMO of long term historical comparisons and overuse of the mocking ‘this time it’s different’ to hold off any criticism that they are comparing things that aren’t comparable because things really have changed, which sometimes they do. Moreover there are exceptions to the application of any rule drawn from longterm history. But do I personally expect high S&P returns this coming 10 yrs? no.