The end of CAPE Fear? What happens to the Shiller CAPE ratio when we roll out the weak 2008/09 earnings?

In last week’s post on dynamic withdrawal rates, one of the withdrawal rules we actually liked quite a bit was based on the Shiller CAPE ratio. One disadvantage of any such rule: The CAPE is at a high level by historical standards, 29.30 to be precise as of this morning (March 22, 2017). Today’s CAPE-based withdrawal rates will be very stingy, only around 3% per annum.

So, what to do about our CAPE Fear? One reader recently made an interesting observation: The CAPE uses ten-year rolling S&P500 earnings. So, once we roll out the low earnings from the Global Financial Crisis (GFC) in 2008/9, average earnings should move up again and the CAPE should come down. But by how much? Probably not below 20. Still, how much of a decline in the CAPE can we realistically expect: 10%? 20%? We have to start a new Excel Spreadsheet for that. Let’s get cranking!

CAPE basics

The CAPE stands for Cyclically-Adjusted Price-Earnings (Ratio). Instead of calculating the stock index level over just one single year of earnings (= traditional trailing P/E Ratio), we use a ten-year average rolling window of real earnings. Inverting the number and calculating the CAEY (Cyclically Adjusted Earnings Yield) = 1/CAPE we have a measure that correlates pretty well with the real equity return going forward.

The CAPE definitely has a following in the personal finance and early retirement blogging crowd:

  • In our Safe Withdrawal Rate Series Part 3, we showed that the 4% rule works best when the CAPE was below 20 (see chart here). Failure probabilities were elevated when the CAPE is between 20-30 and unacceptably high for CAPE>30.
  • MadFIentist has a tool that calculates a safe withdrawal rate for an 80% Stock/20% Bond portfolio. It’s only 3.6%, based on a slightly outdated 2/1/2017 CAPE figure of 28.3, obviously lower now. I personally think that 3.6% is still quite aggressive. In Part 11 of the SWR study, I calculated the CAPE-based withdrawal rates with a multiplier b=0.50 and intercept a=1%, 1.5%, 2.0% and they would imply SWRs of around 2.68%, 3.18% and 3.68% for the three respective values of a. But the 3.68% SWR would be very aggressive. The two other parameter values for a had much better looking past performance.
  • Jim Wang at WalletHacks recently had a good post about the CAPE.
  • … and many more mentions in the FIRE community. Please let us know if we left out anything obvious.

One small complaint about the CAPE

Professor Shiller won the Nobel Prize. I’m just a guy with a blog, so who am I to complain about the little pet peeve I have with his approach? But for what it’s worth, here’s one complaint and a proposal for a quick and easy fix. Shiller uses earnings figures with a lag of up to 6-months. In his most recent spreadsheet (accessed 3/22/2017), the last quarterly earnings number is for September 2016 (Q3 2016 S&P500 EPS).

Shiller Table Screenshot 2017 03 22
Screenshot From Prof. Shiller on March 22, 2017: Earnings numbers are 6 months outdated!

What about the subsequent numbers? Well, fourth quarter numbers have not been finalized. We are already pretty late in the reporting season and almost all of the firms have reported their Q4 numbers. Let’s go to the index provider: S&P Dow Jones Indices, click on the item “Additional Info -> Index Earnings” and download a spreadsheet with not just past earnings but also earnings estimates for the current and future quarters. The “as reported” earnings for Q4 are $24.15 for that quarter, or $94.54 for 4 quarters rolling. There is no reason not to figure in that number already. Three of the four quarters are already 100% known and the final quarter is already 98.1% in. I also throw in the Q1 number of $99.69, trusting the analysts’ forecasts of another improvement in earnings. Also, notice that the Q3 figure of $89.09 exactly coincides with the final earnings number from Prof. Shiller. Q2 number from S&P ($86.92) is the same as the June number from Shiller, etc., so this is the exact EPS series that Prof. Shiller uses.

SP indices Table Screenshot 2017 03 22
Screenshot from S&P Dow Jones: Q4 earnings are already 97.8% in the bag. Why wait for confirmation of the final number? (accessed 3/22/017)

What happens in Shiller’s spreadsheet, when we add the 2016Q4 numbers as well as the 2017Q1 estimate? I added the Q4 number in December, the Q1 number in March and interpolated linearly in between (same method Prof. Shiller uses). See table below. The CAPE will be lower: 29.08 instead of 29.30. Not much but for those who are scared of the CAPE crossing the 30-line we’ve just gotten some extra breathing room! Thank your crazy Uncle ERN for that! For a change, this old curmudgeon is actually a little bit more optimistic than the consensus!

CAPE fear table 01
Adding 6 more months of earnings data: Impact on the Shiller CAPE.

So, equipped with that additional data, let’s plot the earnings vs. index since 2000, see the chart below. Because the index and earnings are on such different scales, let’s plot them in one chart, but with the index level on the left axis and earnings on the right axis. Also notice that the scales are such that a 2500 index level lines up with a $100 EPS level, so when the red and blue line coincide it would mean a 1-year-trailing P/E ratio of 25.

CAPE fear chart 01
S&P index vs. Annual Index Earnings. Adjusted to 3/2017 dollars.

The first thing I notice is that the 10-year rolling earnings number has stagnated for the last few years. And that’s even with the boost we get from adding the six additional months of strong earnings numbers that Shiller ignored. Unfortunately, this “rolling-out-effect” works both ways! Specifically, rolling out the 2007 earnings peak right before the GFC means that the 10-year rolling earnings number has been moving sideways for a while now.

What happens when the 10-year window no longer covers the 2008/9 crisis

Back to the original question. How far can we reduce the CAPE once the GFC is history? Just upfront: There is no single correct answer. Rolling out the old data from 2008 and 2009 means that we have to make assumptions about the new data coming in during the remainder of 2017 and then 2018 and 2019. How do we accomplish this? Simple: We consider a range of different assumptions and see what’s the range of possible CAPE paths going forward. If you don’t like our assumptions, let us know and we can run the whole simulation with your preferred assumptions.

Our baseline scenario:

  • From the S&P500 index level of 2,344 (=March 21 closing as used by Shiller) the index price level grows by 5% annually (p=5%). So the nominal return assumption for the S&P would be 7% if we add a 2% dividend yield.
  • Earnings will grow by 5% annually (e=5%).
  • CPI inflation will run at 2% annually.

Side note: This is Murphy’s Law of blogging: You write a blog post over the weekend and the market takes a dive on Tuesday and Shiller updates his spreasheet with the March 21 S&P500 closing. After I created all my charts and tables! I recomputed the numbers and created new figures this morning, so this is as fresh off the press as it gets!!!

In the chart below, the green dashed line will need a while to catch up. Even when we roll out the GFC, the green line will stay significantly below the red line (by definition because the green is a 10-year rolling average), but even below the red line, so we already know the CAPE will not slip below 25. That’s disappointing!

CAPE fear chart 02
Projecting forward EPS and index level at 5% growth rate annually: It will take a while for the 10-year rolling average EPS number to roll out the GFC!

Next, we plot the historical CAPE until March 2017 and then the projected path for different assumptions about index returns (p) and earnings growth (e), going all the way to 12/31/2027, see below:

CAPE fear chart 03
Projecting the path of the Shiller CAPE for different assumptions of index and earnings growth: Expect a further increase over the short-term, moderate decline by 2020 or 2021. Absent any major decline in the S&P index, the high CAPE ratios are here to stay!
  • Under the baseline (p=e=5%) we drop to about 26.5 in the year 2020 and settle around 27 later that decade. Still pretty high by historical standards.
  • When using a higher growth rate (p=e=6%) the CAPE will drop to slightly above 27 and then settle at above 28. Almost as high as today! There is no real relief from rolling out the GFC!
  • When both variables grow at a lower rate of p=e=4% we up just under 26.
  • If earnings grow 1 percentage point faster than the index (p=5%, e=6%) we get to slightly below 26 in the CAPE by December 2027.
  • If earnings grow 1 percentage point slower than the index (p=6%,e=5%) we will see only a temporary drop to below 28 but then a steady increase to well over 30.

Rolling out the Global Financial Crisis will not reduce the CAPE Ratio by much. A drop in the index level could do that, though that would be very, very unappealing for us in the FIRE crowd. A massive surge in profit growth would be the more pleasant way out of the high CAPE ratio. But hat are the chances for that? Maybe with a corporate tax reform, but don’t hold your breath for that.

A caveat. A big caveat!!!

All of the calculations here assume that there’s no recession and market correction before the year 2027. How likely is that? Well, the longest economic expansion in U.S. history was 1991-2001, according to the NBER. And those were ten years without an economic recession but with plenty of equity market volatility (Tequila Crisis, Asian Crisis, LTCM, Russian Default, etc.). I’m not saying that it’s impossible to have an 18-year expansion from 2009 to 2027, but it would be unprecedented.

Summary

For the last year or so, rolling out past earnings was actually bad for the CAPE. We replaced the peak earnings around 2006/7 with slightly depressed earnings from 2016 (weak energy sector profits). The stagnant 10-year rolling earnings meant that the Trump equity rally pretty much went one-for-one into a higher CAPE.

Rolling out the weak earnings numbers during the GFC will take until 2020 or even 2021. That may lower the Shiller CAPE somewhat, but don’t get your hopes up too high. Absent any big decline in the S&P500, high CAPE ratios are here to stay, somewhere in the range of 26-30. The only way the CAPE can decline again meaningfully without a nasty equity market crash is for earnings to grow faster than the price level. But even that will be a slow and agonizing process.

Are you concerned about today’s high CAPE? Please share your thoughts below!

 

Advertisements

44 thoughts on “The end of CAPE Fear? What happens to the Shiller CAPE ratio when we roll out the weak 2008/09 earnings?

  1. Great analysis as always ERN. This is why I did a simple hack to address the thorny figures in the CAPE carnival. The SWR you get from the failure edge case is closer to reality and is a simple guide for those nervous about high equity markets and historically low bond yields. Details here: http://tenfactorialrocks.com/hacking-the-retirement-calculators/

    I agree with you though removal of 2008-09 figures only modestly helps in CAPE ratio but it does take the edge off a bit so we are a little less irrational than a year before!!

    Liked by 1 person

  2. No sense of worrying about something you can’t control! What’s strange to me is that the economy as whole is doing pretty well despite all those bozo’s in Washington pointing fingers and name calling. Maybe we don’t need a Federal Government after all lol J/P.

    Honestly the only thing that really concerns me is the Student Debt bubble because that’s out of control. I’m not sure exactly how it could pull down the economy though. Hopefully financial companies don’t start selling CDS on student debt anytime soon lol.

    Liked by 1 person

    • Ha, that’s a nonchalant way of dealing with risks. I would say I fear mostly the things I can’t control, haha.
      But agree on the student loan issue. That will land on the balance sheet of the government (if we’re lucky) and I hope by then I’m in a lower tax bracket. 🙂

      Like

  3. Not terribly concerned about today’s high CAPE – as you note, there are two ways out of it and either one has its opportunities (a market drop seems like it could be a good thing long-term, but if someone’s just retired then I guess they would be peeved).

    Words of wisdom and humor re: Murphy’s Law of Blogging. I feel your pain and thanks for redoing the work!

    Liked by 1 person

  4. Yes, I’m very concerned. Many market indicators show a lack of value in the market. There aren’t many deeply discounted stocks, the market P/E is high, stocks are exceeding 200 day moving averages, etc. I’m still invested in the market but have an equity allocation on the lower end. People are bullish on stocks, but that makes me only more nervous. It reminds me a bit of 2000.

    Liked by 1 person

  5. Always a delight when you pop into my mailbox! Keep ’em coming!!!

    Big ERN, I need your professional help to keep my sanity and shed more light onto the mystery of CAPE!!!

    You see, I have an (arguably) unhealthy obsession with CAPE. On one hand it seems to have oracular powers to predict the future. On the other hand, the r^2 is around 1/3, leaving lots of wiggle room for other “factors in the zoo” to come to play. On one hand, it seems to defy EMH. On the other hand, others argue it does not. There’s a complete lack of solid theoretical underpinnings as to why 10 years hence is correct, but 1 year hence is not. Madness, thy name is CAPE!!!

    Getting more specificity, one issue that does not seem to surface much in the discussion is: CAPE is an accounting measure. Earnings are not free cash flow. Ah, the gospel of discounted cash flow for stock valuation; properly done (at least to get a good grade in business school) means you have to back out the non-cash transactions on earnings to arrive at free cash flow before you estimate the value of an asset. What’s this got to do with the price of tea in China? Well, it turns out that accountants just have not really every agreed on what exactly should be treated a non-cash transaction and what should not when it comes to earnings. Inventory accounting policies (LIFO, FIFO, average cost) impact the value of earnings. The use of “kitchen sinking” affects earnings. The so-called GAAP standards (Generally Accepted Accounting Principles) slowly change over time. This means that as you look back in time, there are discontinuities in the way accounting earnings are calculated. The FASB (Financial Accounting Standards Board) slowly, but inexorably releases SFAS (Statement of Financial Accounting Standards) rules that create changes (discontinuities) in how earnings are calculated. Apparently, just to show they can change whatever they want, SFAS rules are no longer to be called SFAS, but rather ACUs (Accounting Standard Update). SFAS 115, 142, and 144 issued in 1993, 2001 and 2001, respectively, have a measurable impact on accounting earnings (gee, that is within the time period we are trying to understand CAPE). Accounting earnings also include effects of servicing debt. If the debt levels of companies shift over time, earnings are affected (big hint: they have changed). Hmmm….the E in CAPE sure has a lot interesting things going on that make it seem a bit adrift of the time periods are weaving our CAPE.

    Now, how about we just ignore these pesky details (or pearls of wisdom) that were pounded into my head by some of the world’s brightest accounting and finance professors? Let’s just do a gut check by looking at the actual CAPE data as a good econometrician would. Start with a scatterplot. Does the CAPE’s forecasting ability remain “stable” over time. Hmmm…maybe decade to decade? Nope. http://jarrodwilcox.com/wp-content/uploads/2015/04/color.gif . http://jarrodwilcox.com/column/making-better-use-of-shillers-cape/ has additional commentary.

    But, wait a moment, we’re fitting a SINGLE line through all the data in the graph mentioned in the above paragraph and that is the CAPE??? Hmmm…it sure doesn’t look like a line fits the data very well.

    Other interesting views:
    Larry Swedroe argues many of these points: https://www.advisorperspectives.com/articles/2015/06/16/are-grantham-and-hussman-correct-about-valuations . Talk about floccinaucinihilipilification! In the end, he concludes the “new average” for the CAPE is 24.7.

    Jeremy Siegel has other interesting info: http://www.cfapubs.org/doi/pdf/10.2469/faj.v72.n3.1

    Liked by 1 person

    • Wow, thanks for that thoughtful comment! That’s essentially a guest post right there. Let me know if you ever want to write one!!!

      About the CAPE: For full disclosure, I personally and professionally prefer a forward-looking measure (earnings estimates, put into a DDM or REG model). I find it a bit fishy to look at 10-year rolling data. But I keep a strict firewall between what I do professionally (with all the data subscriptions) and what I do for the blog. For blogging, obviously, Shiller’s data is all I can use, especially when going that far back.
      I agree that earnings are not the perfect measure. Especially in the cross-sectional asset selection, other accounting measures would be much more appropriate. Too many companies are messing around with their accounting. For looking at the index level in the time series, though, I’m not too troubled. But I can also see that changes in accounting standards make the CAPE less comparable across time. But they won’t explain a CAPE of 29 today vs. a 16 average before.
      Thanks for the links to the Wilcox and Swedroe pieces! For me, the main reason for different intercepts is the scatter plot: different interest rate environments. 1960s, 2000s, 2010s had low interest rates that supported higher CAPEs. 1970s and 1980s required lower CAPE (=higher CAEY) because interest rates were higher and there was more inflation uncertainty. 1990s were the transition period.

      Also: whatever the way out of today’s CAPE of 29 may be, the implications for withdrawal rates are the same: withdraw less.
      option 1: If there is a big crash that brings the CAPE back to 16 (45% drop in the S&P) and from then on we will go back to the business as usual (6.5% real equity return average) you wouldn’t do so badly with a 3% initial SWR and a 4.5% SWR after the drop. A bit of a drop in the standard of living, but you’ll grow out of it.
      option 2: If we entered a low-growth, low risk-premium, low-interest new-normal then we still withdraw only 3%. in line with lower growth expectations.

      Anyway, thanks for this great comment and the links!!!
      Cheers,
      ERN

      Like

  6. As always great stuff. As someone else said my biggest fear with cape is really that I don’t trust earnings numbers over time. Also like you I have a hard time thinking you can solve the issues with looking at something predictively by p/e by essentially enacting a rolling ten year period, It just doesn’t really pass the smell test imho as the first thing you learn about the market is past performance does not guarentee future performance. Imho like operation planning you need to look at historic data combine with the inputs of how things will change in the future. In the operations world it’s the impact of marketing campaigns and competitor moves. In economics it’s technological changes, societal norms, and government movements.

    Liked by 1 person

    • Thanks, FTF! Well said.
      Like you, I believe that past performance (returns) is meaningless. But following past trends in earnings and earnings growth contains some information. So, the CAPE is still something I like to follow. Even though I’d prefer earnings forecasts, not a “backcast” 🙂

      Like

  7. Hello, good work on your research !

    A difficulty lies in that CAPE and/or other fundamental general market valuation measures, may not be suitable tools in providing definitive signaling and precision towards withdrawal rate adjustment / management of sequence of returns and asset balance drawdown risks.

    In the 21st Century, through improved computing power and the internet, Quantitative Investment Portfolio Science and quantitative finance research has evolved markedly. Within QIPS, the acquisition of predictive variables ( market invariants as M. McClung cites them ) that have shown robust statistically significant and non trivial outcomes over long test samples and the discarding of those that have not, are necessary in the design of objective, systematic investment strategies ( NDR and Leuthold Group to name two, have long been pioneers in this “field” ). The application of these predictive variable based strategies can be important in the tactical instruction towards the allocation into various classes of assets during diverse periods of low and high equity market risk. The use of academically proven risk premium factors via the use of low expense exchange traded fund products that represent these asset classes combined with tactical management, may offer an alternative investment process for investors.

    As conventional investment literature has repeatedly suggested the use of the SP500 index and domestic long duration investments ( and recently “total market’ indexes, International equities and bonds, REITS ) within retirement plan portfolios, the evidence suggests that if the design and management of an investment portfolio can be reconceptualized, then higher, risk adjusted alpha may be achieved. For example, the size and value factors ( small cap value ) have shown highest historical alpha premia produced and, extending down the alpha premia dimension, with the inclusion of emerging small cap, midcap growth, and large cap value universes, one can attain a more diversified portfolio consisting of thousands of companies vs. 500 in the SP500.

    An important aspect of attempting to use fundamental general market valuation measures ( such as CAPE ) is that deriving “absolute” levels of valuation from which to build a reasonably precise asset allocation / withdrawal rate adjustment decision heuristic has yet to be determined. Valuation levels and correlations to forward market returns can be “ambiguous” for long stretches of time. Thus a distinction must be made as to which variables or measures will provide confidence in this regard.

    The strategy presented in the link at bottom * illustrates the implementation of “non fundamental” tactical variables combined with risk premium factor based asset and “defensive” asset class selection. Decision heuristic using predefined transaction dates has eliminated the guesswork of the “when” to invest and a static, “defined” withdrawal rate helps shape a stable framework for income withdrawal expectations.
    This new paradigm approach of investing / income harvesting accompanied by asset balance growth may be of value to investors in both the accumulation and spending phases of the investment lifecycle.

    The “CAPE periods” spreadsheet tab ** displays 10 year returns produced by the strategy during high CAPE threshold periods. It appears that periods of high inflation and sideways, cyclical market movement have had the greatest impact on asset growth and sustainable income withdrawals.

    As per variable signaling, 2017 has been indicated as a Favorable risk profile year

    * http://tinyurl.com/znnqxdw
    ** http://tinyurl.com/gm2y5av

    Liked by 1 person

    • Thanks for sharing. I am not sure though why the CAPE10 is 31.325. I think this tool has a bug because they appear to have the same earnings data and index level as Shiller, but they get a wildly different CAPE10 figure. I suspect they messed up the calculation that CPI-adjusts the earnings.
      So: Caution about using tools on the web. 🙂

      Like

      • Got it, tools are sometimes worth what you pay. Still I’d imagine the point holds, comparing CAPE7 vs CAPE8 may give useful hints about the impact of removing the financial crisis without requiring future projections. It is not clear that 10 years is particularly magical for CAPE…

        Liked by 1 person

        • Agree! We can already see there the CAPE10 will be moving later (absent of a another recession) by using the CAPE7 or 8 today.
          The average length of the business cycle is 8 years, so why use CAPE10? My suspicion is that they wanted to capture the 1991 recession when they published their book in 2000, hence the 10-year rolling window.
          Economics/Finance/Statistics and sausage-making have one thing in common: Some of things going in there are not pretty to look at. 🙂

          Like

  8. I’m terrified of where the market is right now. I think too many people are sticking their head in the sand or too worried about politics to care about the market. I think they’re going to be in for a rude awakening. The market does not go up forever. I wouldn’t be surprised if Obamacare doesn’t get altered by the Republicans, that the market turns sour as they don’t think Trump will accomplish much of anything.

    Liked by 1 person

  9. Another great post ERN.
    This is on track to become the most comprehensive review of withdrawal strategies anywhere.
    Have you considered Michael McClung’s ‘Prime Harvesting’ strategy (‘Living Off Your Money’ by Michael McClung) for review?

    Liked by 1 person

    • I looked at the McClung methodology. At least what I could gather from google. My first impression is that I actually like it:
      1: Kitces and Pfau propose a rising equity glidepath in retirement. I agree with that concept and McClung achieves, at least on average.
      2: I like that you sell equities when they are likely expensive (after a long rally) and hold on to equities when they are cheap. You build in a bet on mean-reversion.

      I will put this on my To-Do list. Definitely something to investigate further. 🙂

      Like

    • Thanks for your thoughts. I haven’t checked Tobin Q for a while. But it looks like it’s overvalued just like the CAPE:
      http://mla.homeunix.com/q-ratio/
      And this goes to 2015, it’s probably more overvalued now.

      Also: It takes a model to beat a model. So, what’s your model that indicates the bull market has more space to go. For full disclosure: The high CAPE doesn’t predict a crash. The most likely scenario is merely a continuation of the bull market with a lower expected return.

      Like

      • The Q ratio is more stable than CAPE. The way “earnings” is calculated has slowly changed in the last 50 years (changing accounting rules), which is why the Q ratio is slightly more range-bound than CAPE.
        Our model incorporates an economic indicator part (economy+valuation).

        Liked by 1 person

  10. Hi ERN,

    Thanks for the informative post. A coworker pointed me to your Ultimate Guide to Safe Withdrawal Rates series and it has been really thought-provoking. For the first couple years of saving for early retirement, I focused on the saving and investing aspect rather than the exact target number, but now that those are dialed in and I am getting closer to the goal, I need to figure out exactly what the goal should be.

    In particular, I have found the discussion of future returns based on CAPE and SWR based on CAPE to be really interesting. However, one quibble that I have with your numbers is that they are S&P 500 specific. With the S&P 500 representing only 46.6% of global market cap (as of 3/31/17) this is limiting. Personally, I hold a portfolio that is split US/international based on market cap so I want to make sure that I account for this when looking at equity valuations and what is a SWR under present conditions

    Looking at Vanguard’s data for VTSAX and VTIAX, I see that VTIAX has a P/E ratio of 20.1 compared to VTSAX’s 24.9 as of 2/28/17. However, I haven’t been able to find any data on the CAPE for international equities. Have you found any such information? I noticed that in the comments of https://earlyretirementnow.com/2016/12/21/the-ultimate-guide-to-safe-withdrawal-rates-part-3-equity-valuation/, you answered a related question, but the link you gave to FinanciaLibre is a dead link.

    Liked by 1 person

  11. I am not on a path to early retirement, but expected US stocks returns do seem to be low and do concern me (I am UW US stocks for this reason). At the start of 2017 (So 2017-2026), I have them pegged at 3.2% CAGR. With bond rates where they are, I think the following decade may be frustrating for US investors. We shall see.

    Thanks for the great posts!

    Liked by 1 person

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s