In this article we will explore the ability of stock market valuations to predict both eventual investment returns and the risk of subsequent losses. This is a topic about which we've written before: bearishly prior to the 2008 crash, bullishly in late 2008 after the bulk of the crash had taken place, and then bearishly again -- prematurely, it would seem -- in late 2009. (These past articles are included for historical context; all necessary background will be provided below).
After discussing the relationship between valuations, potential returns, and risk, we'll apply this analysis to current stock market conditions. Finally, we'll end with a couple of "epilogues" regarding our methodology and addressing some criticisms of our chosen approach to determining market valuation.
Stock Valuations and Prospective Returns
The chart below is created using 50 years of monthly data for the S&P500 stock index. Two numbers are calculated for each month:
- The valuation of the S&P500 during that month.
- The total inflation-adjusted return that the market provided in the 7 years following that month.
(Two quick notes on our approach: first, we calculate market-wide valuation by comparing the price of the S&P500 to the ten-year average real earnings of the S&P500 companies. This tells us whether the S&P500 is -- based on the historical relationship between price and cyclically-adjusted earnings -- overpriced, underpriced, or somewhere in between. Second, we have adjusted inflation upward by .45% per year starting in 1997 in an attempt to compensate for the CPI methodology changes that took place around that time. See Epilogue 1: Methodology for many more details.)
Since we're trying to determine typical outcomes, we've grouped all the months into 5 categories of valuation, from cheapest to most expensive. Those are shown along the bottom of the chart. The blue bars indicate the median 7-year inflation-adjusted return experienced by investors who bought in each of the valuation categories:
The relationship could not be more clear. The lower the valuation at the time of purchase, the higher the typical subsequent return. And the difference was huge. Seven years after buying, the typical investor in the lowest valuation category had earned over 91% after inflation, while the typical highest-valuation category investor had earned just 6% -- a fifteen-fold difference!
Valuations and Risk
Market valuations also help to predict the likelihood of taking a long-term loss on an investment.
The chart below uses the same five categories of valuation over the past 50 years, from cheapest to most expensive. This time, the vertical bars measure the frequency of purchasing power loss: what percent of the time investors in each valuation category experienced a negative 7-year real total return (or, in plain English, how often the investment failed to keep up with inflation).
Here, the relationship is less predictive in the three middle-of-the-road valuation categories, all of which have a 7-year real loss frequency clustering around the 20%-25% range. But the difference is quite striking at the valuation extremes. Lowest-valuation category investors experienced negative real returns less than 1% of the time, while highest-valuation category investors endured negative 7-year real returns 40% of the time.
The conventional approach to risk is to define it purely as a function of volatility (the size of the changes in an asset's price over a certain period of time). A more volatile investment is considered to be more risky, and vice-versa. While many economists love this definition because it allows them to precisely define a so-called "risk" measurement to plug into their formulas, it is a completely inadequate definition for real-world investors who are more concerned with long-term results. The real "risk" that long-term investors should be worried about is not that of volatility, but of losing money (or, factoring inflation into it, purchasing power) on an enduring basis. The graph above shows that valuations are a critical element in determining the risk of enduring losses. This should be no big surprise -- it makes perfect intuitive sense that there would be more risk of eventual loss in buying an overpriced asset than an underpriced one.
Market Currently Priced for Poor Returns and High Risk
When we started writing this article, the price-to-10-year-earnings valuation of the S&P500 was over 22.8, putting the market at the low end of the most expensive category on the right-most side of the charts above. By the time we finished (we don't write very quickly), a correction had taken the ratio down to 21.6, sliding the market down into very top of the second-most-expensive category.
Week-to-week fluctuations aside, the S&P500 finds itself near the border between the most expensive and second-most expensive valuation categories, and thus comparatively overpriced. The historical record suggests that the US stock market is likely to deliver poor returns and an elevated risk of inflation-adjusted loss over the next 7 years.
Below is a look at our measure of market valuation starting during the 50-year window we analyzed above. Clearly, current valuations are comfortably on the high side compared to most of this historical timeline. The exception would be the decade or so of remarkably high values that began in the late 1990s, but the credit-fuelled conditions that allowed this outlier period to take place are unlikely to be repeated (and are, in fact, likely to go into reverse).
Just within the past three decades, the S&P500 has seen price-to-10-year real earnings extremes of less than 7 and more than 43. Investors at the year 2000 peak were willing to pay over 6 times more -- 6 times! -- for the same dollar of cyclically-adjusted, inflation-adjusted earnings as they were at the low in 1982. Far from being "efficient," as we discuss further in our conclusion, the stock market is prone to frequent and wild mispricings.
Just for fun, and in order to highlight the volatile but highly mean-reverting nature of market valuations over the long term, here's a look going all the way back to 1900:
Conventional wisdom recommends that investors should simply buy and hold stock and bond index funds. Investors can shift from stocks to bonds as their investment timelines shorten, but that re-allocation is based purely on the relative volatility of the asset classes involved. (A typical implementation of this approach can be found in the "target date" funds that now litter the 401k landscape). Valuations are not even considered in the mainstream buy-and-hold approach. Instead, investors are instructed to put their life savings into investments without making any attempt whatsoever to determine whether those investments are actually a good value based on the economic fundamentals.
This approach makes very little sense to us. Why would anyone buy an investment without first trying to understand whether its price is supported by the fundamentals? The conventional response is that it's impossible to determine such a thing. Markets are "efficient," we are told -- they reflect the sum of the knowledge of all market participants, and thus any market price must be inherently correct. This response fails to consider the fact that market participants are humans, many with investment-related career considerations and all with human emotions, and that they can be -- and very often are -- quite wrong in the aggregate.
The prior decade's two absolutely massive and easy-to-spot bubbles, first in technology stocks and then in housing, invalidate the idea that markets are efficient. So do the latter two charts in this article, which show that there have been extremely dramatic swings -- much larger than would ever take place in a truly efficient and rational market -- in what investors have been willing to pay for a dollar's worth of US company earnings. Notably, the indicator we've used to determine valuation can be calculated in real-time, meaning that investors can be aware of whether the market is currently overpriced or underpriced at any given moment.
For their part, the first two graphs in the article show that these highly inefficient swings in valuation are not only measurable, but can help investors to predict potential return and to ascertain how much risk is being taken to achieve that return.
We are not suggesting that anyone invest solely on the basis of numbers like these. There are always important macroeconomic forces to consider; notable examples (discussed further in Epilogue 2) include the potential effects of US over-indebtedness and the possibility that there has been a long-term shift in market valuations. And looking at the valuation of the entire S&P500 is a fairly blunt approach -- even in a market that is overvalued as a whole, for instance, there could well exist pockets of great value.
Nonetheless, taking the market's valuation into account is a tremendously important component of a prudent investing strategy. Contrary to the mainstream belief that one must take more risk to get a higher return, the charts above show that investors who heed the message of valuations are likely to achieve substantially higher returns while taking less risk. And the message being sent by valuations right now is clear: if history is any guide, investors in the broad U.S. stock market are taking a fairly high risk in order to achieve what will in all likelihood be sub-par long-term returns.
The valuation metric that we use for this analysis is the ratio of price to average 10-year inflation-adjusted earnings, also known as the Cyclically-Adjusted Price-to-Earnings ratio or "CAPE" for short. The standard price-to-earnings ratio is somewhat useful because it measures the price of the stock market against the earnings generated by the companies that comprise the market, thus determining how much investors are willing to pay for a given unit of earnings. The problem is that earnings can be very volatile due to the ongoing cycle of economic booms and busts. The CAPE addresses this problem by comparing the market's price to the average earnings (adjusted for inflation) over the past 10 years. This "cyclically adjusted" ratio is much more predictive of future returns than the standard price-to-earnings ratio.
We did make one adjustment to the CAPE calculations to account for changes to the consumer price index (CPI) formula beginning in the late 1990s. In 1996, the Boskin Commission issued the opinion that the CPI was overstating inflation, and the Bureau of Labor Statistics made some methodological changes as a result. The most thorough analysis we could find estimated that the changes to the CPI methodology in response to the Boskin report caused the CPI to calculate inflation at between .4% and .5% lower than it had previously. Since it's important to use a consistent definition of inflation over time, we have tried to compensate for this change by adding .45% (the middle of the estimated range) to annual inflation growth starting in 1997. We could just as easily have subtracted .45% from the annual inflation growth prior to 1997 with the same results, but adjusting the more recent figures was easier spreadsheet-wise.
This CPI modification impacts the post-1997 period in two ways. First, the higher inflation figure lowers inflation-adjusted returns during that time. Second, higher past inflation causes the CAPE valuations to be slightly lower than they otherwise would have been. This latter effect is counterintuitive because we are used to thinking in terms of inflation reducing something's value. But for our purposes, we aren't interested in what inflation has done to earnings since the time they were earned. We simply want to know what earnings were in each of the past ten years -- but in order to compare apples to apples, we have to restate those past earnings in today's dollars. A higher rate of past inflation will mean that today's dollars are worth less, which in turn means that past earnings will be higher when they are restated in current dollars. (Here's another way to think of this: imagine if you'd bought a house 20 years ago for $50,000 dollars, and that the house is worth $150,000 today. If you are trying to determine real price appreciation, you'd have to restate your original purchase in today's dollars. If inflation had totaled 50% over the 20 year period, the original price in today's dollars would be $75,000. But if inflation had totaled 100%, the original price restated in today's dollars would be $100,000. Higher inflation results in past home prices -- or past earnings-- being higher when restated in todays nominal, less-valuable dollars).
In our past work on this subject, we used 100 years of data in our calculations. This time around, we decided to use just 50 years to see if the many unique elements of that initial 50-year period -- the Depression, World War II, and a gold standard being the big ones -- skewed the results. As it turns out, the results were very similar, although the median valuation is higher during the more recent 50 years. We used the months from June 1954 through May 2004, the most recent 50-year period in which are able to estimate 7-year forward market returns
We made another change to our approach, as well. Knowing from our earlier research that valuations were only predictive of returns over longer timeframes, we'd previously calculated 10-year subsequent returns in order to determine outcomes for investors. We decided this time to change that to 7 years to see if a somewhat shorter time period would be enough time to let valuations do their magic. It was -- the numbers were of course slightly different, but not nearly enough to change our conclusions in any way. And why wait 10 years when you only have to wait 7?
Some pundits have questioned the value of the CAPE for various reasons. Some of the issues they raise are potentially good ones; some result from a complete misunderstanding of the purpose of the CAPE; and some are in between. We will go through all of these in turn. (Most of the anti-CAPE arguments below are paraphrased from a CNN/Money article written earlier this year).
Criticisms that completely miss the point
As context, it's important to understand the purpose of using 10 years of trailing earnings in the denominator. The premise is that on a market-wide basis (which is really a proxy for an economy-wide basis), earnings are mean-reverting. Booms and busts render earnings quite volatile in the short term, but over time they revert to the mean -- they tend to migrate back toward the long-term trend of sustainable growth. (We'll leave it at that for the purposes of this article; some justification of this idea and a couple charts can be found in another recent article of ours). The use of 10 year's worth of earnings allows us to smooth out these business cycles and to determine a normalized, sustainable level of earnings. This is crucial, because it allows us to compare the price of the market not to its current level of earnings, which may be distorted by an unsustainable boom or a temporary recession, but to its realistic long-term earning potential.
Most of the invalid criticisms of the CAPE stem from an apparent ignorance of the facts outlined in the prior paragraph. Some examples follow:
"You wouldn't buy an individual company based on its earnings from 10 years ago; therefore you shouldn't do so for the entire market."
This argument ignores the fact that while an individual company's earnings potential can change enormously in either direction over the long term, the same is not true for the economy as a whole. Economy-wide earnings are mean-reverting; individual company earnings are not. So arguing that a CAPE-like metric for an individual company wouldn't make sense (a premise we agree with) does not invalidate the use of the CAPE for the market as a whole.
"US companies earn a lot of profits from foreign sources now, so profits should grow faster."
This argument also ignores the idea that in a capitalist economy, an advantage that is available to most companies will be used by all of them in competition with one another. To the extent that more earnings become "available," as the argument implies has happened, more companies will step in to take some of those earnings for themselves, and this competition will drive down total earnings. This is how a free-market economy works, and it is why earnings are so ruthlessly mean-reverting. (We are reminded here of a quote from the venerable Jeremy Grantham: "Profit margins are probably the most mean-reverting series in finance, and if profit margins do not mean-revert, then something has gone badly wrong with capitalism.")
"Using 10 years of trailing earnings is backward-looking."
Looking at past data is not being "backward-looking," in the pejorative sense implied in this argument, if that past data helps us to predict the future. And in the case of earnings, as Mr. Grantham points out, it does. We are looking at past data to get a read on a level of earnings and earnings growth that can be sustained on an economy-wide basis. This is far more predictive than looking at current earnings, which can be buffeted by all kinds of temporary forces, or worse yet by looking at analysts' earnings forecasts, which are notoriously incorrect at turning points.
"Earnings from farther back in time should be weighted lower than more recent earnings."
This again misses the point that we are trying to determine a long-term sustainable earnings trend. Overweighting more recent earnings would throw us off if, as is often the case, recent earnings happened to be unsustainably high or low compared to the trend.
Criticisms that don't necessarily miss the point of the CAPE, but are just wrong
"The CAPE doesn't account for interest rates; today's low rates justify a higher CAPE valuation."
This is just a restatement of what is popularly known as the "Fed Model," which states that low rates justify higher stock valuations and vice-versa. But the premise of the Fed Model is provably incorrect. While investors have tended to tolerate higher valuations when rates are low, those low rates have done nothing to change the fact that buying at high valuations has tended to lead to poor returns. Valuations predict returns; interest rates offer no such predictive capability. Or, as Cliff Asness summed it up the excellent scholarly paper in which he thoroughly debunked the Fed Model: "Long-term expected real stock returns are low when starting P/Es are high and vice versa, regardless of starting nominal interest rates."
Criticisms that are potentially good, but turn out not to be an issue right now
"Our particular 10-year trailing period includes two severe recessions and earnings declines, thus understating trend earnings."
This one seems like it could be a valid point -- we did after all experience two brutal recessions in the past decade. It's possible that trailing 10-year earnings could be unusually low as result, in which case they would not actually be giving us a good read on future earnings potential.
It turns out, though, that this is not the case. In fact, the most recent 10-year trailing real earnings figure is actually slightly above its 50-year trend. While there were two big periods of earnings writedowns, these periods were sandwiched by bubble periods in which earnings were unsustainably high compared to the long-term trend. Of particular note was the housing bubble era from 2004-2006, when businesses were benefitting from a frenzy of debt-financed consumer spending and financial companies were booking spectacular, if illusory, profits. So we did have some big earnings declines, but these just offset the periods of unsustainably high earnings, with the result that 10-year earnings are now close to (but still slightly above) their long-term trend. This issue is certainly something to watch for in the future, but it's not causing any distortions to the CAPE right now.
Criticisms that are valid, but that we adjusted for
"Due to CPI changes in the mid-1990s, recent inflation is understated (or prior inflation is overstated), which would impact CAPE valuations and past real returns."
This is a great point. However, it's not an issue with the data presented in this article, because we have adjusted for changes in the CPI calculation as discussed in the Methodology section above.
Criticisms that may have merit, but whose validity will only be known in hindsight
"What if there has been a permanent shift upward in valuations, such that it's inappropriate to compare current valuations with those of decades past?"
Interestingly, for all the complaints about the CAPE, we haven't read this argument elsewhere. But when we discuss amongst ourselves the CAPE's ability to continue to predict returns, this is the one that troubles us the most. The phrase "it's different this time" is very deservedly the bane of value investors everywhere. And yet, once in a great while, there are very long-term shifts in relationships between prices and some of the fundamentals that underpin them. (The seemingly permanent shift downward in the US stock market's dividend yield starting in the late-1950s would qualify as one such example, in our opinion).
Might this be one of those times? Have investors come to accept a permanently higher valuation for stocks? One can't rule it out, but in our opinion, this isn't likely. With the benefit of hindsight, we can identify some factors that very likely contributed to the market's unusually heightened CAPE valuation over the past two decades. These include the following:
- The 1978 birth of the 401K, which likely helped to direct a lot more savings into the stock market.
- A secular decline in interest rates (remember, investors tend to accept higher stock valuations when rates are lower, even if it's a bad idea to do so).
- An enormous increase in borrowing, which financed more economic activity than there would have been in the absence of such borrowing.
- Retiring boomers will soon start draining their 401ks.
- We believe that interest rates are likely to rise substantially in the years ahead.
- We also believe that the US will soon enough hit a point at which it is forced to stop adding to its debt and start reducing it.
The last issue is particularly troubling in terms of future valuations. If values were higher than usual due to all our borrowing, then they might become lower than usual when we stop increasing our net borrowing and start to pay all that debt down. And a bond market crisis, should it come to that, would certainly not be good for stocks. If anything, the risks posed by our fragile debt situation would imply a lower valuation than usual, rather than the unusually high valuation currently in place.
Bonus non-CAPE-specific criticism
This final objection is not aimed at the CAPE, but at the general concept of modifying your investment allocation based on prevailing market conditions (one of which, of course, is market valuation):
"This is market timing; market timing doesn't work."
We've got two quick responses to this one.
First: what we describe here is not "market timing." In our view, market timing entails changing your investment strategy based on a specific prediction that the market will behave a certain way within a given timeframe. That's not what we're suggesting here at all. We are arguing that investors who pay attention to valuations have a higher likelihood of achieving higher long-term returns with lower risk of enduring loss. This approach is not dependent on correctly timing a particular market outcome. It's simply based on the reality that an underpriced investment is likely to eventually outperform an overpriced one.
Second: the charts in this article (along with our own investment track record) show that value-based investing -- whatever name you want to give it -- does work!