Every reliable historical indicator that I have been studying for the last eighteen months suggests that this market (SPY,DIA,QQQ) continues to be biased to the upside. In this article, I am going to argue that the history of one indicator, earnings growth, tells us that the market has much higher to go yet, roughly another 10% per annum until the decade is out (putting the S&P 500 near 4000 by 2021), but also that that growth may primarily come within the next three years.
In order to understand why that might be, we have to look at what everybody else has ignored (as far as I can tell) and dig into the history of earnings volatility.
When earnings explode from low levels in short, powerful bursts, as they did in 2009-2010 (about 800%), this tends to have an exceptional medium-term (between five and ten years) positive impact on stock prices, whereas steadier accretions in earnings tend to have an impact that is both weaker and shorter-lived (under five years).
The next chart indicates the connection.
(Source: All data in this article are courtesy of Robert Shiller)
First, notice the strongest seven-year growth rates (for convenience's sake, we can use the 100% region as a reference point). High points are in the early 1900s, 1917, late 1920s, early 1940s, early 1950s, mid- to late 1960s, late 1970s, l990, the late 1990s, 2006, and 2010. You might notice that there is an increased volatility, with peaks every few years now. We can expect another peak in the seven-year growth rate in 2016, unless anything comparable to the last crisis happens by then.
But, that is not the point I want to make quite yet. If you compare those earnings peaks with the performance of the S&P 500, you will notice that they tend to coincide both with the tops of the market and the bottoms. World War I, the late 1970s, and the 2000s all saw peaks in earnings growth and annualized returns were barely positive. The other peaks, in contrast, occurred when stock performance was in the double digits. I noted this before, but I was not completely sure as to why that was. I think I am much clearer on what accounts for that peculiar interaction now, and the hint is largely in the relationship between the seven-year growth rate and the one-year growth rate.
So, look again at all those instances in which the seven-year growth rate is peaking in conjunction with returns, and then look at the one-year rate. In nearly every bullish instance, the annual growth rate tends to flag as the seven-year rate rises. In fact, you could almost say that the seven-year and one-year growth rates over a given seven-year span are inversely correlated in those bullish instances. Look at the boom of the late 1990s, for example. Annual earnings do pick up at the very end of the bull market, but from the mid-1990s, there is a clear divergence in the seven- and one-year growth rates.
If you look at the one-year rates and compare them to stock returns, you may notice that about six years before the seven-year rate peaks, annual rates are jumping up while the stock market is bottoming out: 1895-6, 1922, 1935, 1947, 1995, and presumably, 2010.
What does all of that add up to?
In those seven-year bullish bursts in earnings, the cyclical growth was "front-loaded." Earnings exploded upward in the first year or two of that seven-year span, and then the earnings growth rate (that is, the more standard twelve-month change) both declined and stabilized. And, in each of those instances in which earnings exploded upwards, it was after a collapse in earnings.
In those seven-year bearish bursts in earnings, earnings growth was "back-loaded" in the cycle and was "spontaneous" rather than coming in response to a negative hit. In 1917, the peak in the seven-year rate was primarily due to the growth of 1916 and 1917, and did not come off of especially low levels. In the 1970s, earnings growth was relatively constant and did not come off of low levels. The 2000s are slightly more complicated, because volatility has created some overlap. The peak in 2006 was primarily due to the growth of 2004, and came off the high levels of earnings in 1999 (even though 2001-2002 experienced a negative shock). The peak in 2010 was due to the growth of that year, but also to the growth of 2004, although that was wiped out (and then some) in 2008-2009.
We can represent this relationship in a few ways.
In the chart above, the comparison was between the seven-year change in earnings and the twelve-month change. In other words, between the seven-year change and the change of the last annual segment of that seven-year span.
If we shift the annual changes over to the right six years, however, we can compare the seven-year growth rate with the twelve-month change of the first year of that seven-year period.
You will notice that in each of the bullish instances, the seven-year earnings growth rate (in this chart, I have annualized it to accentuate the relationship) peaks with the first year of that period. You may also notice that as the seven-year rates are rising in those instances, the annual rate from six years before drops, indicative of the tendency for the earnings performance in these bullish scenarios to kick off as a recovery from a negative earnings shock. Earnings collapse, rapidly recover, and then the growth rate stabilizes; from beginning to end this is roughly ten years.
You may also notice that we seem to be on the verge of a similar such episode. June 2007 was the previous cyclical peak in earnings, so it is virtually inevitable that the seven-year rate will rise in the latter half of this year--six years after earnings began to collapse--and barring a flock of black swans, will peak at a record level in 2016.
There is an even more interesting way to look at this, however, and that is the comparison between the seven-year growth rate in earnings and the P/E ratio. If we had not gone through the trouble of looking at earnings behavior over the last 140+ years, we would have had no reason to suspect that changes in earnings over any given duration would have any meaningful relationship with the P/E ratio at any given moment. But, as it turns out, P/E tends to expand during those periods of low and stable earnings, because stocks respond positively to those conditions. P/E falls when earnings rise off of already fairly high levels, because stocks respond feebly to that sort of earnings growth.
This dimension of the relationship between earnings growth and stock market performance we have described before, as in the following chart.
So, on average, the medium-term growth rate in earnings tends to be negatively correlated with the P/E ratio. Yet, look at some of the instances in which that relationship most obviously breaks down: 1901, 1928, 1997-1999. Also, 1914, 1924, 1945, and others. So what? In the first list (1901, etc), where the seven-year earnings growth rate rose with P/E, the following seven years (give or take) were marked by an increasingly persistent decline in returns. In the latter grouping (1914, etc), where the seven-year earnings growth rate fell with P/E, those instances were hit by sudden stock market crashes about seven years later (1921, 1929-1932, 1953).
If you take a seven-year rolling correlation between the seven-year growth rate and P/E and compare it to subsequent seven-year returns, you come up with some interesting results.
As you can see, the more negative the correlation, the higher the subsequent returns.
In other words, we can substitute P/E for the one-year earnings growth rate and stock performance to simplify the relationship between the medium-term growth rate and the market. But, even though we can substitute P/E for the way stocks react to annual changes in earnings, it is critical to remember that this is just a representation of a deeper, apparently fundamental relationship between stocks and earnings, although it does not exhaust that relationship. The significance is that studying variations in earnings growth gives us information that variations on the valuations metric--whether it be CAPE, the q-ratio, capitalization/GDP, price/sales, or even profit margins--do not provide.
I am sure readers will have noticed that the current correlation, although somewhat off of the record negative values of 2009-2011 remains at historically extreme levels, suggesting that stocks will continue to see exceptional appreciation over the following seven years, although it does not indicate how those returns will be distributed over that period.
The angle we started out with in the article, comparing medium-term growth rates with short-term growth rates, suggests that these types of bull markets do not tend to last longer than ten years and often achieve high returns through exceptional short-term performance. In early 1924, for example, the annualized return over the subsequent seven years was over 10%. So, even well after the crash had taken a major bite out of the capital gains of the late 1920s, that was a purchase that paid off. The same could be said for buying into the market in 1995; despite the abysmal performance of the stock market for the two years following the 2000 peak, there was an average return of about 10% over the 1995-2002 period.
It increasingly appears, from the metrics used in this article (and the previous one which looked at the yield curve), like the path to those types of gains are over a near-term (again, between now and 2017) parabola, although this also suggests that even if the parabola does not manifest itself, the medium-term returns will average out at virtually the same rate. To put it another way, using this type of analysis, the real question is not whether returns will be robust for the seven-year period from 2014-2021, but how those robust returns will be distributed over that period.
In upcoming articles, I hope to push the envelope somewhat further by exploring how we can use these relationships to time the market instead of using them to estimate medium-term returns.
In conclusion, whether we choose to compare short-term fluctuations in earnings with medium-term fluctuations or to compare those medium-term fluctuations with the P/E ratio, both of those comparisons suggest that this market still has considerable amounts of upside potential. The point of interest for me, however, is the fundamental foundation of these relationships, namely that the market is propelled upwards by short-term spikes in earnings in the immediate aftermath of crisis.
Disclosure: I have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it. I have no business relationship with any company whose stock is mentioned in this article.