Using Forward Earnings
Regular readers of "A Dash" know that I like to present sports analogies to help investors escape their biases about the market. Since we are at the start of a new football season, let us engage in some football forecasting.
Suppose that you have an opportunity to invest in a season-long football pool. You can take one of two approaches:
Plan A. Look at the exact records and statistics for each of the teams over the past ten years. You have perfect data for this period, so you cannot be criticized for your analysis. Your method of prediction is to choose the teams in each division that did the best over the last decade.
In short, you are using past performance over a set time frame to predict the future. You pick a time period long enough to escape short-term fluctuations and find long term team strengths.
Plan B. Look at how each team did last year. Look at scouting reports for each team for the upcoming season. Take the predictions from a group of scouts and use their average forecast as the starting point for your analysis. You are allowed to tweak it a bit, since some of the scouts may be biased locals.
In short, you rely mainly on interpretation of current information — coaching changes, injuries, retired players, new acquisitions, strength of schedule — to make a forecast for each team's record.
The Contrast
In one case you are using known data — rock-solid information. The problem is that you must then make guesses about how these trends may have changed. In the other case you use micro-level data which includes estimates and speculation. You are relying upon the expertise of the sources.
Please note that you cannot avoid making an estimate as part of your forecast. It is only the nature of the estimate that has changed.
The Market Application
There is a constant drumbeat of criticism about market valuation using forward earnings. The most common criticism, that estimates are too optimistic, is open to challenge. If the estimates are too high, why is the beat rate consistently in the 65% range?
The key choice is the same as our football example. The fans of the Shiller 10-year past earnings method take pride in having solid data. Then they make a wild guess about whether the trend will continue. Those praising this method point to a few notable successes, mostly times when P/E ratios were very low since interest rates were very high.
Those interested in forward earnings are taking the aggregate work of dozens of specialists. If you think they are a little high, you can feel free to add an error range. If you do so, you should look at past data — especially that of recent years.
Time Frame for Testing
Pundits choose time frames to suit their purpose.
Many critics of forward earnings use data from before 1980, mostly because that suits their purpose. Since we did not have forward earnings data before then, all of those sources are misleading at best and deceptive at worst.
There is also a logical question about comparisons from ancient times. Here is a surprise. The same critics who now complain about the effects of high frequency trading think it is completely appropriate to cite data from the 1930's or the 1970's.
Markets have changed dramatically as information became more widely available to all and the bid-ask spreads narrowed. In the 1970's, the average investor had no idea what the sell-side earnings estimates were and had to pay a huge commission to act even if he did. It is different now.
Conclusion
These points are blindingly obvious, yet widely ignored.
Here is an offer for anyone who thinks that using ten years of past data is the best method: Send me an email and I'll show you how to enter a nice football pool. The smart money will welcome you and Dr. Shiller.
Jeff,
We’ve discussed/debated this topic before, but I wanted to make a couple quick points, and I’ll follow up later with a more detailed comment.
“The key choice is the same as our football example. The fans of the Shiller 10-year past earnings method take pride in having solid data. Then they make a wild guess about whether the trend will continue.”
This isn’t accurate. Here is the thing. There is no need to make a “wild guess”. We have 100+ years of data that shows regardless of the environment trend rate EPS growth rate in the overall market is 6-7% (although there were a few decades of earnings stagnation). This really is indisputable.
Now if you have a really good argument/thesis for why the aggregate corporate EPS growth rate is going to be structurally higher for the next 5-20 years, then I AM ALL EARS because that would suggest the market is extremely undervalued here.
What I hoped you would address in a post on forward earnings, and did NOT, and what I wrestle with is getting the turning points right absent a crystal ball.
Look, we KNOW forward earnings estimates are going to be wrong at major inflection points because analysts generally just extrapolate the existing trend. Almost absurdly, P/Es based on forward estimates said stocks were “cheap” at the October 2007 peak and “expensive” at the March 2009 low when obviously the exact opposite is true.
So to me, the real challenge is realizing and applying that the market will trade off forward earnings estimates and the revisions 80% of the time, and you want to shift gears before the inflection point when forward estimates are about to start either getting whacked big-time like 2007-2008 or raised dramatically like 2009. The market is going to start moving either up or down long before the analysts start revising their estimates. What was the forward estimate in August 2008? What was the forward estimate in June 2009?
I hope you don’t take offense, but this note was mostly argument by analogy, and there are some holes in the analogy you use. The note is pretty light on any sort of data or statistical analysis. What might be interesting/useful/persuasive would be something like the forward P/E compared to bond yields and the subsequent 1-year, 3-year, and 5-year market performance and looking at it over both the 1980-2000 AND 2000-2010 time periods.
To be clear, this is all applicable to the overall market and individual stocks could differ from the overall market/economic environment.
Just curious, what do you think SPX earnings will be for 2011? Will the actual number be MUCH HIGHER or MUCH LOWER then the current forward estimate?
Mike C — The topic of overall market valuation is too complex for a single article. While I have written frequently on this subject, there are always new readers and people who have forgotten.
I have in mind a few new thoughts that will correct some of the misinformation currently circulating.
Meanwhile, I wanted to start off with a concept piece, encouraging people (including you) do step away from long-held conclusions (biases?) and take a fresh look.
The challenge –especially for you — is to explain why data from 100 years ago has any relevance for next year’s market. Also why it is better to look backward rather than forward. Also how you take the Shiller ratio and use it effectively for next year.
To summarize, a good comment engages on the terms set by the author. I choose what I think is most helpful at a given time — broad interest, current relevance, and overall helpfulness.
There is only so much I can do in each article, and they are not all about data. Some of them are about thinking. I was actually inspired by seeing Dr. Shiller on Kudlow last night. He does not use his data in the same way it usually cited.
Jeff
The key fact is that over timeframes above a year, both performance and prices tend to mean revert. This shifts the balance in favor of 10y-trailing and against earnings estimates. While we know that 10y-trailing are not always accurate predictors of future performance, on average they work pretty well because they take advantage of mean reversion over long timeframes. Meanwhile, we also know that estimates are, quite often, just plain wrong, as they chronically overestimate and chronically assume that the current trend will continue. In other words, they are premised (implicitly, although no “analyst” would admit it) on the assumption that the trend over the past few years will continue into the next few years. In other words, analysts are doing nothing more than projecting the relatively recent past into the future, but they’re assuming trend continuation while using timeframes at which mean reversion becomes applicable. We know empirically that the odds are against this.
Bottom-line: If you are a long-term investor, you are far better off using 10y trailing earnings (which do a superior job of predicting returns over the next 5-10y) then you are using earnings estimates. If you are a shorter-term investor speculator, then 10y trailing are inappropriate, but if you rely on forward estimates then you are going to constantly be behind the curve, running with the herd, which isn’t a good place to be either. Estimates are mainly useful as a guage of what the market is currently pricing in, not to help predict what will actually happen.
“Those interested in forward earnings are taking the aggregate work of dozens of specialists.”
There are some good analysts out there, but most are a joke. Career risk predominates, and to mitigate career risk you stick with consensus and simply project trend continuation. It’s fine to be wrong if everybody else is wrong in the same way, but suicide to go out on a limb alone (b/c even if you’re usually right, you’ll be caught wrong sometimes). That eliminates the potential value of estimates based on current information.
The smart-money buyside shops (e.g. hedgies) will tell you that they completely ignore price targets and estimates (except to gauge what market is pricing). Only the dumb-money buyside shops (e.g. pensions, most mutual funds) actually care about estimates.
“The most common criticism, that estimates are too optimistic, is open to challenge. If the estimates are too high, why is the beat rate consistently in the 65% range?”
Because the estimates are reduced until the they’re beatable:
http://www.ritholtz.com/blog/wp-content/uploads/2010/08/Analysts-bad-timing.png
Eric– The McKinsey data that every cites shows that estimates are somewhat too high for when forecast 2-3 years ahead. They are also too low at the time of the report.
At some point in time (one year ahead?) the estimates must be pretty accurate. That is why I am comfortable with 12 month forward earnings, but I don’t pay much attention to longer forecasts.
Thanks for the link. I am familiar with that report and have inquired with McKinsey about the data. I’ll write something more even if I do not get a satisfactory response from them.
Jeff
scrilla — One of my programs involves long-term investments, but I review them constantly and use active management. Even if I have a holding period of 18 months or so, I am interested in what will happen next year. I am not always right in my picks, of course, but I am certainly not behind the curve.
As to the herd — it is obvious that they are buying bonds instead of stocks with a forward P/E of 11 or so. I think of this gap as a measure of the current negative sentiment.
Thanks for your comment.
Jeff
Sorry to add this in a second comment, but it’s not surprising that the forecasts evolve towards approach the final EPS figures.
This isn’t a game on the part of the analysts to produce earnings “beats”, it’s just that they start out too optimistic about earnings growth.
“At some point in time (one year ahead?) the estimates must be pretty accurate.”
No, not true. Since 1975 they are consistently far too high on a one year forward basis. Look at the data which tracks estimates for ~2.5 years for each stock: two years leading up to the earns reports, then a bit longer until the final SEC GAAP-based filings are made.
Greg, RIA
Greg — Are you drawing your conclusions from looking at the squiggles on the chart, or do you have actual data? Note that 2009 is missing on the chart. The big discrepancies occur during recessions.
BTW, do you know what the vertical axis means. What is this EPS number? That is not how most people describe S&P 500 earnings.
Thanks for your comment, especially if you have a pointer to real data.
Jeff
Studies have shown PE ratios are mean reverting. Studies have shown PE ratios are close but not quite mean reverting.
Studies have shown PE ratios averaged over long terms have some use in predicting the market. Studies have shown long term PE ratios offer no predictive ability.
Earnings estimates always err on the high side. Earnings estimates merely track actual earnings lagged by a couple of quarters.
Gee, if we can’t even agree on simple concepts, where does that leave us?
Valuing the S&P 500 using forward operating earnings
“Now, to the issue of P/E ratios based on forward operating earnings. As noted above, it’s clear that forward operating earnings are generally much higher than the record level for trailing net earnings to-date, and of course, record earnings are always equal to or higher than raw trailing earnings.”
“Investors are used to the idea that “normal” P/E ratios are typically in the range of 14 to 16. But as Cliff Asness of AQR has repeatedly stressed, those norms are based on raw trailing earnings. If you calculate P/E ratios based on earnings figures that are higher, you clearly obtain lower P/E ratios.”
“As it happens, the long-term historical norm for the P/E ratio based on forward operating earnings would be about 12. Of course, that average of 12 includes the heights of the late 1990’s bubble. The historical average was just 10.6 prior to that point.”
Source: http://www.hussmanfunds.com/wmc/wmc070820.htm
————
“The two main failures of standard FOE analysis are that 1) analysts assume a long-term norm for the P/E ratio that properly applies to trailing net, not forward operating earnings, and; 2) analysts fail to model the variation in prospective earnings growth induced by changes in the level of profit margins, and therefore wildly over- or underestimate long-term cash flows that are relevant to proper valuation. By dealing directly with those two issues, we can obtain useful implications about market valuation.”
Source: http://www.hussmanfunds.com/wmc/wmc100802.htm
John Hussman then goes on to compare popular valuation models to actual S&P returns, and shows and explains why using forward operating earnings estimates with normal P/E ratios is demonstrably incorrect.
Cui bono. Who benefits from the creation of forward operating earnings as a measure of value?
Answer the above and you will understand why your argument is naive at best.
Really shocked to read this on your blog Jeff.
Paul — I’m shocked that you are shocked! Think about an individual stock. Take a stock like CAT, which I own. Do you look only at the past ten years of earnings, or do you try to look ahead. I read analyst reports, all of which have an earnings forecast. I make a decision about my own earnings forecast — but I am looking ahead.
That is what every investment manager does.
Somehow when it comes to market valuation people like you decide that everyone estimating earnings is lying. That has not been true for many years — at least in SOX era.
So I urge you not to be shocked. Instead you should put away your obvious existing biases as I challenged you to do. Explain why the earnings beat rate is so high.
The football pool is open for you!
Jeff
Steve – I read all of the Hussman articles. I strongly disagree with the conclusions from the selected segments you have quoted here.
At some point in my series I will take up some of these specific points, but it is difficult to deal with the meandering Hussman prose and ever-shifting theories. (What happened to peak earnings?) I’m not going to try to do it comprehensively in the comments.
Your final statement about comparing forward earnings to actual returns with popular valuation models is an accurate summary of Hussman, but he is wrong. He uses data from a time when there were no forward earnings forecasts. A truth squad is needed.
Hussman continues to look backwards and to ignore interest rates. I think he is wrong on both counts. I expect to continue to outperform his fund.
At some point in my series on valuation, I’ll attempt to summarize his wide-ranging arguments.
Meanwhile, when do you think the trailing earnings will signal a buy? Where do you suppose the market will be at that time?
Those are important questions to answer.
Jeff
The main thing for me is that the PE10 averages across winners and losers. One company may get a great idea, another may do something stupid. Of course. We know that. But PE10, and only similar long-term cross-company averages, looks for the theme within the noise.
(I mean, a “picker” would say “I don’t need those averages, I can spot winners.” To which I would say “tear it up.”)
I was actually inspired by seeing Dr. Shiller on Kudlow last night. He does not use his data in the same way it usually cited.
Jeff,
Can you clarify this point. It was on my to do list to watch this since you had mentioned it, and I finally had a chance:
http://www.cnbc.com/id/15840232/?video=1579464042&play=1
It was a pretty brief discussion about earnings and valuation towards the end with most of the discussion centered on housing. Shiller basically said he thought you should average earnings over a long period of time, but that one should own SOME stocks nonetheless even though he mentioned stocks were not “cheap” according to his metric.