Geoff Gannon December 29, 2006

In Defense of Extraordinary Claims

About two weeks ago in a post entitled “We Have Some Bearish Bloggers Out There“, Bill Rempel wrote, “Personally, I’m in the ‘extraordinary claims require extraordinary proof’ camp.” I’d like to think I am too, because Bill is right – extraordinary claims do require extraordinary proof.

So, before making any extraordinary claims about future long-term market returns (i.e., predicting future returns that differ substantially from historical returns), I’d like to spend this post laying out the case for why current circumstances are extraordinary. After all, extraordinary times call for extraordinary claims.

Essentially, this is a post about why the present is unlike the past and what that means for the future.

In a previous post, I wrote:

Stocks are not inherently attractive; they have often been attractive, because they have often been cheap.

Unless they internalize this fact, investors risk assuming that historical returns that existed under special circumstances can continue to serve as a useful frame of reference, even when these special circumstances no longer exist.

Later in this post, I will discuss the possibility of a “paradigm shift” (i.e., a change in basic assumptions within the theory of investment) that began in 1995. The only other period in the 20th century which saw similar upheaval in investment thinking was the 1920s.

Common Stocks as Long Term Investments

That theoretical crisis (and the higher valuations that followed it) has often been partly attributed to a thin volume published in 1924 by Edgar Lawrence Smith. The book was called “Common Stocks as Long Term Investments” and it was based on a study of 56 years of market data (1866 – 1922).

Smith found that stocks had consistently outperformed bonds over the long run. Neither the data in support of this conclusion nor the logical explanation for this outperformance (public companies retain earnings and these retained earnings lead to compound growth) was wrong.

However, a few years after Smith’s book was published, the special circumstances of the past disappeared as stocks (which had historically had higher yields than bonds) saw their prices surge and their yields plunge. Soon, stocks had lower yields than bonds – part of the reason for their past outperformance (the initial yield advantage) was gone and the margin of safety which a diversified group of stocks had offered over bonds narrowed considerably.

Simply put, circumstances changed. John Maynard Keynes saw this possibility when he reviewed Smith’s book in 1925:

“It is dangerous…to apply to the future inductive arguments based on past experience, unless one can distinguish the broad reasons why past experience was what it was.”

That has been the objective of this little study from the outset. In this post, I will focus on how the circumstances of the present differ from the circumstances of the past.

I will also endeavor to demonstrate that historical returns were the result of special circumstances, which (logically) need not apply now or in the future. The historical data suggests these circumstances may yet return – and for the sake of net buyers of stocks, I hope the data is right and one day (soon) historical returns can once again serve as a useful frame of reference for the future.

Today, however, historical returns have about as much utility to the investor as the success rate of a procedure performed exclusively on 25 year-old men has for the surgeon who is preparing to operate on a 92 year-old woman.

There is nothing wrong with the data itself. But, there is something wrong with the assumption that data collected from one special case has predictive power when applied to another special case.

Cheap Stocks and Great Returns

Historical returns in equities have been great. However, it’s worth noting that throughout the period we’re referring to, stocks have often been cheap. How cheap?

Once again, here’s a graph of the Dow’s 15-year normalized P/E ratio for each year from 1935-2006:

PE ratio.jpg

From 1935-2006, the Dow’s normalized P/E ratio ranged from 6.88 – 30.84. The Dow’s average (mean) normalized P/E ratio for these years was 14.18. The median was 13.91.

Those figures include the 1995-2006 period, which I will discuss in greater detail later. For now, let’s start by taking a look at the period from 1935-1994.

Until 1995, the Dow’s normalized P/E ratio had ranged from 6.88 – 17.40. The average (mean) normalized P/E ratio from 1935-1994 was 12.31. The median normalized P/E ratio was 12.41. In other words, the Dow’s average 15-year normalized earnings yield was just over 8%.

I would estimate that in a little under 45% of all years, the Dow was priced such that long-term investors were effectively paying little or nothing for future earnings growth. Most market authorities would disagree with me on this point, because they would require an equity-risk premium.

Equity-Risk Premium – An Aside

This isn’t the place to have a long argument about the concept of an equity-risk premium. For now, I will simply say that you can not arrive at the conclusion that there is an equity-risk premium via deductive (a priori) reasoning. If you were locked in a room alone, you would never come up with the idea of an equity-risk premium. It is only in seeing the effect that you would seek out a cause.

You can only come to the conclusion that an equity-risk premium should exist by first knowing that it has existed. You have to work backwards from the effect to the cause. That’s troubling, because history consists of a series of special circumstances. It is non-repeatable.

So, the existence of a measurable aversion to stocks over some historical period does not necessarily lead to the conclusion that such an aversion is the result of a general principle (i.e., an inherent equity-risk premium). In fact, such a conclusion could merely be a contrived attempt to explain away an observable effect that has existed under certain circumstances – but needn’t always exist.

The equity-risk premium isn’t a general theory. It’s really little more than the acknowledgement that during the historical period being studied, market participants made choices that reflect an aversion to stocks compared to the choices an optimal return seeking automaton would have made.

It’s an interesting observation – but, it’s not a theory.

How Common Are Cheap Markets?

Returning to the question of how often the stock market has been cheap, I would estimate that during the period from 1935-2006, the Dow was priced to offer double-digit returns somewhere between 75% and 85% of the time.

Here, I don’t mean that the Dow did provide double-digit returns 75% to 85% of the time; nor, do I mean that past performance suggested it should provide such returns. Rather, I simply mean that valuing the Dow as an asset to be held until Judgment Day, would lead a clear-headed observer to conclude that double-digit returns were likely in about 75% to 85% of the years being considered.

I know this 75% to 85% number is a bit hard to swallow. So, if you don’t believe me, consider what Warren Buffett wrote on the same topic in his 2002 annual letter to shareholders:

“Despite three years of falling prices, which have significantly improved the attractiveness of common stocks, we still find very few that even mildly interest us. That dismal fact is testimony to the insanity of valuations reached during The Great Bubble. Unfortunately, the hangover may prove to be proportional to the binge.”

“The aversion to equities that Charlie and I exhibit today is far from congenital. We love owning common stocks – if they can be purchased at attractive prices. In my 61 years of investing, 50 or so years have offered that kind of opportunity. There will be years like that again. Unless, however, we see a very high probability of at least 10% pre-tax returns…we will sit on the sidelines.”

Buffett’s “50 or so years” of his 61 would translate into just under 82% of the time. He wrote that letter in early 2003. The four years since haven’t offered the kind of opportunity he looks for, while the seven years included in the study from before Buffett started investing did offer that kind of opportunity.

So, according to my math, that would work out to be a roughly 80% estimate from Buffett over the full 1935-2006 period. That estimate falls within the 75% – 85% range I cited based on the data.

I think this 75%-85% range is the best estimate you’ll find for how often the market has been so cheap as to offer double-digit returns when valued as an asset with a holding period of forever.

Unfortunately, I’m afraid a lot of investors don’t realize (or haven’t internalized) just how often the stock market has been really cheap. During the 1935-2006 period, stocks were priced as clear bargains in about 8 out of every 10 years. Buffett supports this conclusion with his assertion that stocks could be “purchased at attractive prices” about 80% of the time (50 out of 61 years).

If investors don’t start with an understanding of the fact that stocks have been so cheap so often, they won’t be able to put the historical data in its proper context. If you have a population that consists of 80% x and 20% y, is it reasonable to assume that data based on the entire population is a good reference point for your subject, if you know your subject is a y rather than an x?

In terms of valuation, 2006 (and thus 2007) is undoubtedly a minority year. Unfortunately, data based on a full population sometimes has little or no relevance when applied to a member of a minority group.

For instance, Turkey’s population is 80% Turkish and 20% Kurdish. My guess is that data based solely on the full population of the country (which would consist of 80% ethnic Turks) would tell you very little about any particular Kurd. Now, if you broke the data you had collected down into a Turkish group and a Kurdish group and used the Kurdish group to predict something about an individual Kurd – then, you might be on to something.

A More Detailed Look

From 1935-2006, the Dow’s normalized P/E ratio ranged from 6.88 to 30.84. The Dow’s average (mean) normalized P/E ratio for these years was 14.18. The median was 13.91. In half of all years, the Dow’s normalized P/E ratio fell between 10.53 and 16.43.

Here’s a breakdown of how common various normalized P/E ratios were from 1935-2006.

Normalized P/E of 5-10: 18 of 72 years or 25.00% of the time

Normalized P/E of 10-15: 28 of 72 years or 38.89% of the time

Normalized P/E of 15-20: 17 of 72 years or 23.61% of the time

Normalized P/E of 20-25: 5 of 72 years or 6.94% of the time

Normalized P/E of 25-30: 3 of 72 years or 4.17% of the time

Normalized P/E of 30-35: 1 of 72 years or 1.39% of the time

Fifteen Years Later…

For the years with a normalized P/E ratio between 5 and 10, compound point growth in the Dow over the subsequent fifteen years ranged from 4.01% to 15.69%. The average (mean) growth rate was 10.17%. The median growth rate was 10.03%.

For the years with a normalized P/E ratio between 10 and 15, compound point growth in the Dow over the subsequent fifteen years ranged from 0.92% to 12.28%. The average (mean) growth rate was 7.01%. The median growth rate was 8.17%.

For the years with a normalized P/E ratio between 15 and 20, compound point growth in the Dow over the subsequent fifteen years ranged from (0.14%) to 8.93%. The average (mean) growth rate was 2.19%. The median growth rate was 1.76%.

I’d love to show you the same data for the three highest normalized P/E groups. But, I can’t.

There is no fifteen year point growth data for years with a normalized P/E over 20, because the Dow didn’t record a year with a normalized P/E ratio above 20 until 1996. In fact, until 1995, the highest normalized P/E ratio on record was 17.40 – that high-water mark was reached in 1965. With the benefit of hindsight we now know 1965 was not an ideal year to buy stocks for the long-run.

Rising Multiples?

Today’s normalized P/E ratio is extremely high. So what? Hasn’t the normalized P/E ratio been rising over time, as investors have come to realize a diversified group of stocks held for the long-run is actually a low-risk, high-reward bet?

I’ll let you judge for yourself. I won’t even connect the dots for fear of biasing you.

Here’s a chart showing the Dow’s 15-year normalized P/E ratio for each year from 1935-1994:

dot 1994.jpg

Do you see a trend towards higher normalized P/E ratios over time?

I cut the graph off at 1995 for a reason. That’s the year everything changed. You’ll remember I said the Dow’s highest normalized P/E ratio had been 17.40 reached in 1965.

Although I didn’t include the data necessary to compute 15-year normalized P/E ratios for years before 1935, I do have enough data to know that the three “peak” normalized P/E ratio years during the 20th century were 1929, 1965, and 1999.

By “peak” years, I simply mean the three highest years that aren’t part of a chain of continuously higher normalized P/E years – unless they’re the highest year in that chain. Without this qualifier, the highest normalized P/E list would be monopolized by the years from 1995 – 2006. Each year in that group had a higher normalized P/E ratio than every year prior to 1995.

In other words, since 1995, the Dow’s normalized P/E ratio hasn’t just been above the mean, it’s been above the entire normalized P/E ratio range from 1935-1994. You can see that clearly in this graph, which shows the Dow’s normalized P/E ratio for each year from 1935 – 2006:

dot 2006.jpg

This graph is essentially just a continuation of the earlier graph. In fact, if you cover the points from 1995 – 2006, you can see the familiar outline of that graph with its long undulations and its frothy crest at 17.40. That bound was reached in 1965. In 1995, the Dow broke out of this upper bound and hasn’t returned since.

Terra Incognita

In this graph, it certainly does look like there’s a trend toward higher normalized P/E ratios. However, that trend only emerged over the last decade – not the last century.

In other words, the Dow’s normalized P/E ratio hasn’t been rising over time. It simply surged in the 1990s. That surge may be justified. However, it’s certainly a departure from the historical data. As a result, there’s no reason to believe historical returns from 1935-1994 have any utility whatsoever in predicting market returns in the new era that has emerged since 1995.

All the historical return data from before 1995 was based on lower normalized P/E ratios. Once again, I don’t mean the pre-1995 period had lower average normalized P/E ratios – I mean that no year from before 1995 had a normalized P/E ratio equal to or greater than any year from 1995 through today. Simply put, since 1995, market valuations have been in completely uncharted territory.

The only years with normalized valuations comparable to today’s occurred during the 1995-2006 period. So, referring to historical return data requires a choice between using data from recent years or using data from dissimilar years.

Paradigm Shift?

Is it possible that the surge in normalized P/E ratios beginning in 1995 was simply the culmination of a crisis within the investment discipline? Maybe normalized P/E ratios have reached “a permanently high plateau” now that a new paradigm has taken hold.

I won’t dismiss this argument entirely. There is some logic to it. After all, stocks have been an unbelievable bargain for most of the 20th century. Why should that continue to be the case? Eventually, won’t enough investors wise up to this fact and cause the so-called “equity-risk premium” to disappear.

If the normalized P/E ratio remains extremely high, there will be no need for stock prices to fall. Of course, these higher valuations must necessarily cause future returns to fall short of historical returns. But, there’s no logical reason why normalized P/E ratios must revert to the mean – future returns can be adjusted down, allowing current prices to remain high.

That’s true. In fact, the Dow could theoretically trade around a normalized P/E ratio as high as 40-50 without making stocks so unattractive as to completely eliminate them as a possible long-term investment (all of this assumes the equity-risk premium can disappear).

At around 50 times normalized earnings, the math gets terribly unforgiving. As a result, it’s hard to imagine any likely circumstances under which a market trading at close to 50 times normalized earnings could be a viable investment option – though it’s theoretically possible if long-term interest rates are very, very close to zero.

But, at lower normalized P/E ratios, such as 30 (and certainly 20) stocks could still compete with other investment opportunities. Stocks might lose most (or all) of their edge over other asset classes; but, stock prices wouldn’t necessarily have to fall – they could simply offer much lower returns than they had in the past. This could continue indefinitely – in theory.

I say “in theory”, because that seems a rather unlikely scenario. There is absolutely no evidence for it in the data.

Before 1995, the Dow’s normalized P/E ratio had ranged from 6.88 – 17.40. The average (mean) normalized P/E ratio from 1935-1994 was 12.31. The median normalized P/E ratio was 12.41.

So, a permanent jump to normalized P/E ratios above 20 would be quite a departure from the past. Could the leap be permanent? Could these new, higher normalized P/E ratios become the new norm?

Maybe. If we really are in a new era, the old historical return data isn’t relevant – it applies only to an era of low normalized P/E ratios. New, higher valuations must necessarily lead to new, lower returns. On the other hand, if we aren’t in a new era, the old historical return data is relevant – and normalized P/E ratios must fall.

Adjusting to the Norm

Normalized P/E ratios can fall in several ways. However, there are only two ways that seem reasonable given current conditions. Stock prices can either fall over the short-term or they can grow slowly (at less than 5-6% a year) over the long-term.

The data from 1935-2006 doesn’t provide much support to one route over the other. In the past, extraordinarily high normalized P/E ratios have been brought down to more normal levels through crashes and through stagnant markets.

The market can reach a more “normal” normalized P/E ratio by going down fast or going sideways for a very long time. During the 20th century, we saw normalized P/E ratios fall both ways.

To return to the 1935-1994 normalized P/E range, the Dow would need to trade around 10,135. That would simply bring it down to a valuation comparable to 1965.

To return to the average normalized P/E ratio for 1935-2006, the Dow would need to trade around 8,260. If the Dow were to trade at the average normalized P/E ratio for the 1935-1994 period, it would need to trade around 7,230.

Are any of these numbers likely destinations? The truth is stocks have probably been too cheap in the past and they’re probably too expensive today. Regardless, the Dow has been above 1965’s old normalized P/E high since 1995. So, for a little over a decade now, the market has been in uncharted territory. A normalized P/E ratio of 20-25 (today’s is about 21.50) is quite compatible with decent long-term returns for stocks relative to other asset classes.

However, such high normalized P/E ratios are not compatible with the kind of long-term returns seen during much of the 20th century.


Stocks are not inherently attractive; they have often been attractive, because they have often been cheap. The great returns of the 20th century occurred under special circumstances – namely, low normalized P/E ratios. Today’s normalized P/E ratios are much, much higher. In other words, the special circumstances that allowed for great returns in equities during the 20th century no longer exist.

So, don’t use historical returns as a frame of reference when thinking about future returns – and do lower your expectations!

Related Reading

On 15-Year Normalized P/E Ratios for the Dow

On Normalized P/E Ratios and the Election Cycle

On Normalized P/E Ratios and the Election Cycle (Again)

On Normalized P/E Effects Over Time

On Calculating Normalized P/E Ratios

On the Difference Between Actual Earnings and Normalized Earnings

On the Dow’s Normalized Earnings Yields for 1935-2006

I’ll have many more posts on this project in the days ahead. If you have any questions (or suggestions) about this project, please feel free to comment to this post – or, simply send me an email.