Earlier this week I posted about the current controversy around the cyclically-adjusted earnings in the Shiller PE, most notably the contention that the real earnings used in the Shiller PE are lower than they would otherwise be because of two serious earnings recessions.
Another question about the Shiller PE is how accurate it has been historically as a forecasting tool. Asness has backtested the performance of the market from various Shiller PE starting points from 1926 to 2012, finding as follows:
Asness observes:
Ten-year forward average returns fall nearly monotonically as starting Shiller P/E’s increase. Also, as starting Shiller P/E’s go up, worst cases get worse and best cases get weaker (best cases remain OK from any decile, so there is generally hope even if it should not triumph over experience!).
The Shiller PE at the time of Asness’s article was 22.2, and the current Shiller PE is 23.4. Both are squarely in the middle of the highlighted row:
If today’s Shiller P/E is 22.2, and your long-term plan calls for a 10% nominal (or with today’s inflation about 7-8% real) return on the stock market, you are basically rooting for the absolute best case in history to play out again, and rooting for something drastically above the average case from these valuations. This could happen. For instance, it could happen if total real earnings growth surprises to the upside by a lot for a very long time. But unless you are comfortable with forecasting that, or some other giant positive surprise, we believe one should give credence to the lower forecasted average returns from history. While market timing might not be the answer, changing your plans — assuming a lower expected market return, perhaps saving more or spending less, or making changes in your portfolio structure — are all worth serious consideration. I think the Shiller P/E is quite meaningful for planning.
Asness examines several other interesting market-level valuation metrics, finding that they tend to support the implications of the currently elevated Shiller PE, noting:
Some outright hucksters still use the trick of comparing current P/E’s based on “forecast” “operating” earnings with historical average P/E’s based on total trailing earnings. In addition, some critics say you can’t compare today to the past because accounting standards have changed, and the long-term past contains things like World Wars and Depressions. While I don’t buy it, this argument applies equally to the one-year P/E which many are still somehow willing to use. Also it’s ironic that the chief argument of the critics, their big gun that I address exhaustively above [from the earlier post], is that the last 10 years are just too disastrous to be meaningful (recall they are actually mildly above average).
He concludes:
While it’s indeed important to remember that no valuation measure is near perfect (I stress that in my initial table), I do believe that the Shiller P/E is a reasonable method, an unbiased method (it’s been 15+ years since it was created so nobody cherry picked it to fit the current period), and a method that is decidedly not “broken” based on today’s inputs. It has very limited use for market timing (certainly on its own) and there is still great variability around its predictions over even decades. But, if you don’t lower your expectations when Shiller P/E’s are high without a good reason — and in my view the critics have not provided a good reason this time around — I think you are making a mistake.
The current Shiller PE of 23.4 implies a real return of less than 0.9 percent per year for the next decade, with a best-case scenario less than 8.3 percent annually, and a worst-case scenario of less than -4.4 percent annually.
Read An Old Friend: The Stock Market’s Shiller P/E (.pdf).
Buy my book The Acquirer’s Multiple: How the Billionaire Contrarians of Deep Value Beat the Market from on Kindle, paperback, and Audible.
Here’s your book for the fall if you’re on global Wall Street. Tobias Carlisle has hit a home run deep over left field. It’s an incredibly smart, dense, 213 pages on how to not lose money in the market. It’s your Autumn smart read. –Tom Keene, Bloomberg’s Editor-At-Large, Bloomberg Surveillance, September 9, 2014.
Click here if you’d like to read more on The Acquirer’s Multiple, or connect with me on Twitter, LinkedIn or Facebook. Check out the best deep value stocks in the largest 1000 names for free on the deep value stock screener at The Acquirer’s Multiple®.
The reason this article makes it seems like this system doesn’t work is because right now according to the Shiller PE system we shouldn’t invest in the stock market and instead hold our earnings in fixed return bonds (or similar fixed income investments).
LikeLike
[…] Not surprisingly, when valuations have been at current levels or higher, future returns on the portfolio have been low or negative. When it comes to your investment portfolio, and most importantly your time horizon, a mistake can be extremely costly. As I quoted in “Shiller’s CAPE – Is It Really B.S.?:” […]
LikeLike
[…] Cliff Asness discussed this issue in particular stating: […]
LikeLike
[…] Cliff Asness discussed this issue in particular stating: […]
LikeLike
Shiller pe is real return, not nominal. Check the math on his website, irrational exuberance.
LikeLike
[…] Great Shiller PE Controversy: Are Cyclically-Adjusted Earnings Below The Long-Term Trend? and How accurate is the Shiller PE as a forecasting tool? What backtested returns does the current PE f…). He discusses it in some detail in this interview with Consuelo Mack on […]
LikeLike
Why on earth would you use nominal PE ratios instead of REAL Earnings Yields?
Do you think a PE ratio of 15 with 9% inflation is the same as a PE ratio of 15 with 1% inflation? CAPErs can normalize earnings over 10 years but not inflation?
I can’t believe anyone would be espousing that $1 of earnings at 1% inflation is worth the same as $1 at 12% inflation. I know Cliff doesn’t think so.
[I’m not going into the mean-reverting questions, either, although CAPE does NOT make the same assumptions with the operating/reported EPS ratio, which to be intellectually consistent you absolutely have to. But only if you’re concerned about this ‘magic formula’ being rigorous.]
CAPE would have kept you out of the US stock market for almost the entire decade of the 90s. That’s a mega-perverse outcome!
At least if one is going to use some formula to ‘value’ the markets we should examine the logical underpinnings and faulty assumptions implicit in every model.
It would have gotten you in the market in Oct 2008 and back out in Sept 2009 with zero gains. {I know it’s not a market timing model – but then trading into and then back out of the market in less than a year should not happen!}
The current valuation of CAPE is 1100 on the SPX [a 10% hike from 10 months ago, FTR]. That’s a 30% below today’s value for ‘fair value…’ I guess you need a drop to ~900 to be ‘cheap?’ That looks like bizarroWorld, as virtually all investors would think a 400-pt drop tomorrow would make the market ‘cheap.’
Here’s what the CAPE real earnings yield looks like using the long-bond [which is the best, unbiased estimate of US inflation expectations] over the past 120 years, as of last year:
The S+P looks to be around the 3rd cheapest since the late 1950s, after 1974 and 2008. Notice that the chart ‘works’ well – the market is expensive going into the Crash of ’29, and super-cheap at the bottom in 1933. It gets very cheap again in 1950, gets expensive in the late 60s, very cheap in the mid 1970s, and is most expensive in 2000. Not bad, huh?
Historically, when the cyclically adjusted earnings yield is 2-3 percentage points above the 10-year treasury yields, the S&P 500 index returns 58% in real terms during the following 10-year period. This could be higher or lower in the future, of course, but if we’re looking at historical time series….
LikeLike
CAPE is cyclically adjusted, and, hence, uses real earnings. This looks a lot like the Fed Model. Is that correct? Are you proposing we use the Fed Model? May god have mercy on your soul if you are.
LikeLike
CAPE does not use real earnings, it uses the nominal PE ratio. If you chart Real earnings yields, you get a completely different answer.
You are woefully mistaken if you think ‘cyclically-adjusted = ‘real.’ I can’t understand how you would make such a clear, vital error in understanding the difference unless you are being purposefully dense. Take the S+P earnings yield per annum and subtract the 10-yr UST.
This gives you the real yield, which is completely different from CAPEing it.
Simply take out the nominal PE ratios he uses and replace them with real yields. Also normalize for the Operating/reported EPS ratio ‘mean reversion’ if you believe in mean reversion.
No need to put words in my mouth, if I meant to say ‘Fed Model’ I would have.
Do you really not understand the difference between the real earnings of the Index, and CAPE? Wow, just wow.
LikeLike
Step 2 would then to be to graph the Real Earnings Yield of the S+P against itself over history, similar to CAPE. I did this in the link in the initial post above, showing it graphed over the decades. If you take 0.2 seconds to look at it, you will be comforted that it looks absolutely nothing like the Fed Model. JPMIM has discussed this at least once in the past several months.
Still waiting for the explanation of how a model that doesn’t distinguish btw a 19 PE with 1.5% inflation and a 19 PE with 12% inflation is as useful as one that does — obviously one stream of earnings is valued MUCH more highly.
After all, perhaps the biggest question in investing today is will US inflation stay at <= 2% and be quiescent, or will it morph into hyperinflation as Kyle Bass, et al, believe.
LikeLike
[…] https://greenbackd.com/2013/04/03/how-accurate-is-the-shiller-pe-as-a-forecasting-tool/ […]
LikeLike
Tobias, how are you thinking about interest rates in relation to the market and the implied multiples? Buffett discusses the impact of interest rates on equity valuations here:
http://money.cnn.com/magazines/fortune/fortune_archive/2001/12/10/314691/
“That gets to the first of the economic variables that affected stock prices in the two periods–interest rates. In economics, interest rates act as gravity behaves in the physical world. At all times, in all markets, in all parts of the world, the tiniest change in rates changes the value of every financial asset. You see that clearly with the fluctuating prices of bonds. But the rule applies as well to farmland, oil reserves, stocks, and every other financial asset. And the effects can be huge on values. If interest rates are, say, 13%, the present value of a dollar that you’re going to receive in the future from an investment is not nearly as high as the present value of a dollar if rates are 4%.”
“And then the reversal of those factors created a period during which much lower GNP gains were accompanied by a bonanza for the market. First, you got a major increase in the rate of profitability. Second, you got an enormous drop in interest rates, which made a dollar of future profit that much more valuable. Both phenomena were real and powerful fuels for a major bull market.”
Graham also realized that interest rates effected valuations, which is why his Central Value calculation brought them into play. Yes, he used 10 yr avg earnings, but he divided them by a number that was correlated with interest rates. The result is a multiple; a multiple shakes out if you will. But the multiple is higher when interest rates are lower, and lower when interest rates are higher. And for his calculation, which worked before WWII, he started adding the 50% arbitrary adjustment starting in the early ’60s as I have mentioned before.
Right now we have both of the factors Buffett mentioned going on. We have historically low interest rates and record profitability, both of which lead to higher multiples. I believe the increased after-tax profitability relative to GDP is driven by 4 things: 1.) foreign earnings which are additive and growing (it turns out that GDP does pick up foreign earnings), 2.) wages which have been declining since the peak of the tech bubble, 3.) effective corporate tax rates which have been declining since WWII, and 4.) interest expense which has been declining along with interest rates for a long time. The question is, will this all mean revert? There is no reason foreign earnings should. They are additive and growing. There is no reason that wages should. Although for them to rise, it would take a healthier economy which is bullish for equities. There is no reason taxes should at all. It is doubtful our politicians are going to be raising taxes on companies anytime soon. In fact, there seems to be a greater chance they will cut rates a bit to make us more competitive. For interest rates, yes they will mean revert to 5%-6% on the 10 yr, but that could take 6-10 yrs before we see that. And like wages, it would take a healthier economy for them to rise which is bullish for equities.
Any thoughts on this? And for what it is worth, if 2008 earnings are excluded from the Shiller PE, the multiple goes to about 20x. And if operating earnings are used instead of as reported, it drops further still. And if the growth of the American economy is taken into consideration, as opposed to viewing it as mature which makes a Shiller PE (10 yr avg earnings) analysis more useful, that changes things a bit too.
LikeLike
Buffett provides those examples to demonstrate that it is not the change in GNP that has the most impact on returns, but the starting ratio of total market capitalization to GNP. He sets out the interest rates and profit margins to provide some color and explain the elevated/depressed level of the market. Interest rates and profitability do not factor into his model, which has two inputs: GNP and total market capitalization. I do not dispute that interest rates are low, or that profit margins are at record highs, and forward estimates are higher still. In isolation, those data points would imply a market that is at the upper end of its range, and if we look at the TMC/GNP model proposed by Buffett, I don’t see how we can come to any other conclusion. The fact that the Shiller PE and Tobin’s q ratio roughly agree with the conclusion available from Buffett’s measure suggests to me that markets are expensive, and returns will be depressed.
You need to address Asness’s work in this post on the earnings in the Shiller PE that suggest real earnings are in fact above trend, even after including the disastrous 2008 earnings. It does not make sense to do as you have done and eliminate the worst year, then use operating earnings, and then compare it to the historical, unadjusted, “as reported” ratio. There are other ways to deal with outliers. For example, you could Winsorize the data, or you could use the median. You also need to be careful to compare apples to applies i.e. operating earnings to operating earnings, or “as reported” earnings to “as reported” earnings. Otherwise it’s not statistically valid.
LikeLike
At first I was thinking that this CAPE is pretty cool stuff – and I still do, but I’m also wondering, “what does the average retail investor do with this?”
I realise that going to cash when CAPE is above ‘x’ is perhaps not the right play.
I suppose you could take the approach of comparing the average to what else is available elsewhere, but even when you add (or subtract) just 1 standard deviation from the average, I’m left not sure that this is a good idea either.
Obviously if you’ve hit a time in your life when you are ready to invest in the market and there is a low CAPE, you would be pleased. But is the solution to a higher CAPE to keep saving cash?
There’s something in it – I love the monotony of the stats!
A question that occurred to me is, “what would these stats look like for a value investor?” After all, real life application is what matters in the end.
So, I propose a study:
Do the exact same study as Asness above (same cutoff points etc to keep consistent).
But this time, ONLY look at the value decile (using Price to Book or Price to Earnings – something standard to the literature). In other words, instead of the SP500 we only look at the value 50 (of the SP500).
We already know the value deciles beat the market. It’d just be interesting to see whether CAPE’s tell us anything different. Perhaps they wouldn’t.
Thinking about it, my guess would be a similar result – perhaps flattened out a bit.
What do you think?
Anyway, I do think one clear application is for those who do not invest solely in domestic markets. Global investors can use things like CAPE and dividend yields to find good countries to invest in (from a value perspective).
LikeLike
If we assume that value investors look for low Shiller PEs not at the market level but locally (i.e. low Shiller PE industries or deciles or whatever) then that makes sense. We did an eight-year average PE test in the book, and it did seem that there was a small advantage over the single-year metric. I know that some investors use it to identify interesting markets, and then use other tools to identify interesting stocks in those markets.
LikeLike
Sad to say I can’t recall, but have you any inkling on the performance of value strategies in periods of high shiller PE vs low? meaning is there any potential spread between growth/value that might increase/decrease during periods of high or low shiller PEs?
LikeLike
Amit, the short answer is I don’t know. The longer answer is that value is highly correlated with the market, but my observation is that it tends to outperform most in a secular bear when Shiller PEs are compressing over long periods, for example, 1966 to 1982, and 2000 to present. In a secular bull (1982 to 2000), value’s advantage over the market is less pronounced.
LikeLike
perhaps the financial repression period we are in (thanks for that link Tobias) has set conditions for a best case scenario…but I wouldn’t be on it. And, even if that best case scenario does come to fore – there’s a bigger reckoning on the way. Kind of like trading on the greater fool theory.
LikeLike
Agreed. For the folks at home, this is the James Montier piece to which Jim refers: “The 13th Labour of Hercules: Capital Preservation in the Age of Financial Repression” on the GMO research page here.
LikeLike
How was the Schiller P/e as a forecasting tool from the peak of the market in 2007 and from the bottom of the Market in 2009? Thanks
LikeLike
The May to October 2007 peak Shiller PE was ~27.5, which Asness’s chart puts at an average of 0.5 percent per annum for the decade. We’re at 14.4 percent total return since, which equates to about 2.9 percent per annum, and so the actual return has thus far outpaced the average return.
The March 2009 bottom Shiller PE was ~13.3, which Asness’s chart puts at an average of 9.1 percent per annum for the decade. We’re at 138.6 percent total return since, which equates to 23.7 percent per annum, and so again the actual return has thus far well outpaced the average return.
Before concluding based on those two points that the Shiller PE chronically underestimates returns, the caveat I would add is that we are mid-way through a market cycle, neither of those Shliler PE periods are complete (the March data point is ~4 years and 1 month old), and the Shiller PE is currently very elevated.
LikeLike
The Schiller P/E first went above 20 in around 1992. If you sold then, you would have been out of the market for the last 20+ years, other than for a brief period in 2009. Meanwhile the market has quadrupled.
Not sure it is a particularly useful metric.
LikeLike
In December 1992 the Shiller PE was 20.45. The average return from 20.45 according to Asness’s chart was 3.9 percent per annum for the subsequent decade. The annual total return for the S& 500 TR was 8 percent per annum for the period to November 2001, which was about ten months from the eventual bottom of the dot com bust in September 2002. Including that final date reduced annual returns to 2.5 percent, which is slightly below the average return. It’s roughly correct over longer periods (10 years +).
3.9 percent is not a signal to sell out, it’s a signal to anticipate 3.9 percent from the market. It makes it possible to compare the returns available to other opportunities, which might be better. For example, the ten-year treasury in late 1992 was ~7 percent.
LikeLike