“Rule No. 1: Never lose money. Rule No. 2: Never forget rule No. 1.” – the gospel according to Warren Buffett. Although this sounds quite simple, there are two schools of thought on how to go about accomplishing it.
Firstly, the market timers believe the answer is to “buy low and sell high”. However, a second group (including many prominent academics) maintain this is easier said than done, with very few investors actually getting it right with any degree of consistency. They expound that the recipe for creating wealth is simply to follow the patient approach, saying that “it’s time in the market, not timing the market” that counts.
This gives rise to the all-important question: does one’s entry level into the market, i.e. the valuation of the market at the time of investing, make a significant difference to subsequent investment returns?
In an attempt to cast light on this issue, an interesting multi-year comparison of the price-earnings (PE) ratios of the S&P 500 Index (as a measure of stock valuations) and the forward real returns was done by my colleagues at Plexus Asset Management. The study covered the period from 1871 to 2006 and used the S&P 500 Composite Index (and its predecessors). In essence, a total real return index and coinciding ten-year forward real returns were calculated, and used together with PEs based on rolling ten-year earnings.
In the first analysis the PEs and the ten-year forward real returns were grouped in five quintiles (i.e. 20% intervals) (Diagram A.1).
The cheapest quintile had an average PE of 8,5 with an average ten-year forward real return of 11,0% p.a., whereas the most expensive quintile had an average PE of 21,6 with an average ten-year forward real return of only 3,2% p.a.
This analysis clearly shows the strong long-term relationship between real returns and the level of valuation at which the investment was made.
The study was then repeated with the PEs divided in smaller groups, i.e. deciles or 10% intervals (see Diagrams A.2 and A.3).
This analysis strongly confirms the downward trend of the average ten-year forward real returns from the cheapest grouping (PEs of less than six) to the most expensive grouping (PEs of more than 21). The second study also shows that any investment at PEs of less than 12 always had positive ten-year real returns, while investments at PE ratios of 12 and higher experienced negative real returns at some stage.
A third observation from this analysis is, interestingly, that the ten-year forward real returns of investments made at PEs between 12 and 17 had the biggest spread between minimum and maximum returns and were therefore more volatile and less predictable.
As a further refinement, holding periods of one, three, five and 20 years were also analysed. The research results (not reported in this article) for the one-year period showed a poor relationship with expected returns, but the findings for all the other periods were consistent with the findings for the ten-year periods.
Although the above analysis represents an update to and extension of an earlier study by Jeremy Grantham’s GMO, it was also considered appropriate to replicate the study using dividend yields rather than PEs as valuation yardstick. The results are reported in Diagrams B.1, B.2 and B.3 and, as can be expected, are very similar to those based on PEs.
Based on the above research findings, with the S&P 500 Index’s current PE of 18,4 and dividend yield of 1,8%, investors should be aware of the fact that the market is by historical standards not in cheap territory, arguing for luke-warm returns. Although the research results offer no guidance as to when and at what level the current bull market will run out of steam, they do indicate that it would be irrational to bank on above-average returns from these valuation levels. As a matter of fact, investors should expect higher volatility and even the possibility of some negative returns.
It is easy to understand why Grantham came to the conclusion that “the best case for caution and bearishness is value, which is a weak predictor of one-year returns, but a dynamic predictor of longer-term returns”.
June 5, 2007 at 5:13 pm
Its hard to argue against buying good companies cheaply will not produce admirable returns.
And while the market is clearly not terribly expensive here, neither is it terribly cheap — just as profits are decellerating.
That’s the 2nd biggest risk to the P/E argument — the biggest being the 10 year bond yield rising towards 5.25%!
June 10, 2007 at 8:43 pm
[...] investing for the long-term, the general stock market isn’t going to be the place to be. Prieur du Plessis has done some excellent research on P/E ratios and stock market returns going [...]
June 10, 2007 at 10:42 pm
Historical data can be of great value in attempting to project the future, but history doesn’t ALWAYS repeat itself.
June 11, 2007 at 4:04 pm
It is not a question of whether history repeats itself, it is more an issue of investors continuing to behave in the same manner business cycle after business cycle. From over confidence to overly pessimistic.
I would suggest given the above piece the headlines in the press at times when PEs are the most attractive are negatively correlated to investor’s sentiment and outlook and vice versa during times of high PEs and over-optimism as is the case today.
As is the case with every cycle before, when the last bear has capitulated, who is left to raise the tide of valuation?
This time is different, that’s what makes it the same. C.
June 11, 2007 at 9:35 pm
Interesting post, very reminiscent of some of the arguments made in the past by Hussman and David Dremen, but updated for our present U.S. post-bubble environment.
Thanks, Barry, for linking to it and thanks to Prieur and colleagues for writing and compiling it!
June 12, 2007 at 1:59 am
[...] US equity returns: what to expect “Rule No. 1: Never lose money. Rule No. 2: Never forget rule No. 1.” – the gospel according to Warren Buffett. […] [...]
June 12, 2007 at 7:27 pm
Bob Arnott has done a bunch of these analyses, many of which have been published in CFA Institute’s FINANCIAL ANALYSTS JOURNAL.
I love reading this kind of stuff, and it pays to re-read it to confirm or deny your market suspicions, but as a practical investment tool, I’m not sure it’s that relevant EXCEPT at major market turning points (i.e., 2000, 1994, 1982).
Some people say that today’s low dividend yield needs to adjust for buybacks which were practically nonexistent prior to 1982 (when Exxon I believe got one of the earliest OK’s to do buybacks from the IRS & SEC). But you’d also have to keep in mind buybacks are promises that can be easily broken (unlike dividends) and buybacks should be netted against equity dilution from options, M&A, restricted stock, etc.
Finally, I agree with Barry that you don’t want to ignore the 10-year moving to 5.25%. However, past valuations from decades ago — let alone 50-75 years ago — had risk factors that today are properly discounted as nil (i.e., Depression, stock market wipeouts from wars, etc). Myself, I think valuations from 1973 onward — when we went to floating exchange rates — should be segmented from earlier eras like Bretton Woods and World Wars I & II.
JMHO.
June 19, 2007 at 2:34 pm
I have done work on P/E’s since 1960, and was one of the first researchers to show the superiority of low P/E investing. However, there is one aspect of P/E’s that is confounding: how to normalize earnings. The market P/E should logically be low at earnings peaks and high at earnings troughs; thus, when we measure returns from a base of either high or low P/E’s, there is at least some element earnings cyclicality at work. This could mean that we are actually underestmating the beginning P/E effect on future returns. In today’s market, there is some evidence that profit margins and earnings are well above “normal”. If that is so, and future returns are low amid an earnings decline, we could get a double whammy of both lower earnings and lower valuation. Conversely, we could have earnings decline and a valuation increase as the market became pleased by, say, the small extent of the earnings decline. Robert Schiller has tried to accommodate this normalization problem by using a 10 year average of real earnings. That worked in 1998-2000 to show the irrational exuberance he wrote about. But if there have been some recent more permanent changes in real earnings, as opposed to only cyclical forces, the 10 year avg. won’t work. For example, outsourcing of manufacturing costs may have improved margins on a permanent basis from what they would be otherwise. Or, the fact interest costs have declined steadily as aproportion of total costs since the early 1980’s.
Sorry to have gone on so long, but I’m sure you have thought of all this. By the way, Steve Leuthold, of leauthold & Co., has done some interesting normalization work. He uses a 20quarter average, 16 quarters back and 4 quarters forward. That way he gets a weighting for recency.
June 19, 2007 at 6:35 pm
While the current P/E and dividend yield would indicate mediocre returns over the next ten years on “long” investments, they also present a highly attractive environment in which to “short”. If longs are in a Bear Market, then “Shorts” must be in a Bull Market.
From our perspective, a steep market decline is virtually assured, and a Market Crash is highly probable. Moreover, this study is an excellent, albeit rare endorsement for Bear Market Timing, coming from one of the pillars of the “value, buy & hold” camp.
“The best case for caution and bearishness is value, which is a weak predictor of one year returns, but a dynamic predictor of longer term returns” – Jeremy Grantham
It is easy to understand why Jeremy Grantham came to his conclusion, and we to ours.
June 20, 2007 at 3:24 pm
You say in this paper:
“Based on the above research findings, with the S&P 500 Index’s current PE of 18,4 and dividend yield of 1,8%, investors should be aware of the fact that the market is by historical standards not in cheap territory, arguing for luke-warm returns.”
If you are to be consistent with your paper, you should be worrying, not about the current PE, but the about the “Schiller PE” you used in the paper, that is, the real S&P price divided by the real ten year average earnings. That PE is much higher than 18.4, and while I don’t have the actual data at hand, it must be well in excess of 25. That’s a number that I think was exceeded only in the last period of irrational exuberance.
I note from Schiller’s data that the S&P 500 high in real terms was 1631 in August of 2000,compared with1 530 today so the real return (price only) since then is still negative by a fair margin.
Can your staff give us the updated Schiller data? His website has updated only through last August.
June 26, 2007 at 6:03 pm
[...] reading at Investment Postcards from Cape Town digg_url=’http://www.sgfunds.com/journal/?p=24′; digg_skin = ‘button’; digg_bgcolor = ‘#FFFFFF’; [...]
July 19, 2007 at 7:09 am
While the current P/E and dividend yield would indicate mediocre returns over the next ten years on “long” investments, they also present a highly attractive environment in which to “short”. If longs are in a Bear Market, then “Shorts” must be in a Bull Market.
From our perspective, a steep market decline is virtually assured, and a Market Crash is highly probable. Moreover, this study is an excellent, albeit rare endorsement for Bear Market Timing, coming from one of the pillars of the “value, buy & hold” camp.
“The best case for caution and bearishness is value, which is a weak predictor of one year returns, but a dynamic predictor of longer term returns” – Jeremy Grantham
It is easy to understand why Jeremy Grantham came to his conclusion, and we to ours.
http://www.exceptional-bear-market-letter.com/