I think CAPE has a problem predicting returns (SPY,DIA,QQQ,CAPE) in periods that are marked by deflationary pressure and volatile earnings. Insofar as we are living in a period of deflationary worries and volatility, and despite the fact that CAPE's "prediction" of low returns over the next decade may pan out, it will probably be "wrong" about the next five years. The history of the relationship between earnings, stock performance, P/E ratios, and deflation hints that this bull market has yet to run out of steam. In this article, I will focus on the relationship between earnings and stock performance and then relate that to the episodes in which CAPE fails to predict returns.
Earnings and Stock Market Performance Since 1871
So, how do earnings relate to stock returns? It is not easy to summarize that history, but let's see if we can't reduce it at least to a few useful generalizations.
In the following chart, I compare annualized earnings growth rates over every annual interval from one to thirty years with concomitant stock returns, and I compare how well they correlated with one another both before and after World War II.
Click to enlarge(Source: All data in this article are from calculations based on data graciously provided by Robert Shiller).
Over the very long term, it would appear that earnings growth and stock returns are highly correlated. In other words, if you were able to predict earnings growth over the following thirty years, you could have profited from that insight. Except after World War II. After World War II, there is no obvious relationship between earnings growth and stock returns. This is one reason I see CAPE and the notion of "irrational exuberance" as problematic. Markets seem to have been much less irrational before the dollar became preeminent. But, this chart also would, at first blush, seem to call into question the relevancy of earnings forecasts, even if they are always correct.
On second thought, though, that long-term relationship between earnings and returns is awfully robust. And oddly, at the 28-, 29- and 30-year intervals, the correlation is even stronger over the very long term (1871-2014) than it was over the first half of the period (1871-1945). Although I am not a mathematician by any stretch of the imagination, I think I can guess why that is.
The following chart gives us a clue: earnings growth and stock return rates tend to be inversely correlated as long as earnings growth is reliably positive, but when earnings are prone to negative shocks, stocks and earnings are apt to be positively correlated with one another. Since World War II, earnings growth has remained relatively reliable (roughly 6% growth per year), but when the earnings crisis hit in 2009, all of a sudden the market was panicking like it was 1907.
The market, despite the transformation in relationships that is so obvious in the bar chart above, is still acutely vulnerable to earnings crises. That has not changed. The market is nuts, but it is not crazy.
How Do We Know When A Rise In Earnings Is Bad?
As is so often the case, that really raises as many questions as it answers, though. For example, how should we think about a boon in earnings? What if we somehow knew that earnings would grow 10% per annum on average over the next decade? Apparently, we should view this boon positively or negatively depending on whether "earnings growth is reliably positive" and whether "earnings are prone to negative shocks"!
That might seem like a blatant contradiction, but if you look at the ten-year growth rates in the previous chart again, you might notice that the subsequent, normalized ten-year earnings growth rate in 2002 was about 13%. That is, during the 2002-2012 period, earnings grew on average 13% a year, which was one of the best performances ever. And yet, right in the middle of that period (i.e., 2009), we experienced the single greatest earnings shock ever.
What I am trying to say is that magnificent earnings growth can also coincide with extreme volatility and be "prone to negative shocks." So, when we think about the conditions under which exceptional earnings growth may be good or bad, we might not have to look too far away for a way to resolve the contradiction. It could be the volatility of that growth.
To put it another way, 5% growth over a ten-year period would be more bullish than 10% growth, if the volatility in that growth rate were significantly higher in the latter scenario. This has happened before. From 1982-1992, earnings grew about 1% on average, but returns ran at a 14% clip. In 2002-2012, as we just noted, earnings grew 13% on average, but returns were 2% per annum. The difference was that, although in both instances, there were some years in which earnings declined, in the more recent episode, even though the decline was brief and did not impact the overall growth rate, it seemed to trigger this relationship that we had not seen since World War II.
A simpler way of putting this is that it was not sufficient to have accurately predicted the 2002-2012 growth rate in order to predict returns; one would have had to have accurately predicted the 1999-2009 growth rate, as well. If you had known in 1999 that the subsequent growth rate would be -15% on average (or made a similarly gruesome guess), and you were aware of the mercurial nature of the relationship between earnings and returns, you might have predicted that stocks were going to underperform in the ten years to 2009, and that the recovery in earnings after the crisis would be where most of the gains in the 2002-2012 period would have come from.
But, what does the future hold now? Earnings margins are back to record highs, but will they gently come down from those heights or plunge back into the abyss? Is there any way to predict these sorts of earnings shocks?
Can We Predict Earnings Crises?
That is actually the question I had set out to answer when I first started researching this particular article, and although I could not find any especially satisfying methods to do so, I had what I thought were some interesting leads. They have forced me to reconsider some of the relationships that we talked about in the previous article. Perhaps the most interesting hint was the high correlation between P/E and subsequent earnings growth.
Look at the following chart. It shows the correlation between P/E and subsequent earnings growth over every annual interval from one to thirty years. If you use the log of P/E, the impressiveness of the short-term relationships is diminished rather significantly, but when I looked beyond the correlation, it got me to thinking. It seems to me that P/E has managed to achieve that high degree of correlation partly through the volatility of earnings growth. I don't want to overdo it, but it seems to me that P/E tends to do a better job of predicting earnings growth when that ratio is very high. And, since earnings tend to be stable over the long run, the only way that a high P/E could manage to predict high earnings growth was through volatility. In other words, earnings would have to dip significantly first in order to achieve a high short-term gain. I also know that high P/E multiples tend to manifest themselves during periods of low inflation. Moreover, in an article earlier this year, I pointed out that CAPE's ability to predict returns seemed to really only hold when earnings were reliably positive, and I also noted that the "guarantee" of long-term positive earnings growth appeared to be linked to the positive inflation "guaranteed" by the Fed since the Depression.
So, it got me to thinking that deflation might be a critical element here. The big problem with drawing that relationship out, however, is that inflation has changed so much over the last 150 years. It is both higher and more stable than it was under the gold standard, because of the rise of the service economy. Many people appear to be under the misapprehension that the Fed has masterminded inflation stability, but that is not at all the case. Commodities (DJP,GSG,RJA) are just as volatile now as they were in the nineteenth century or when Adam Smith wrote about relative prices in the 1770s (and silver (SLV,GLD), actually, is much more so). The only thing that has changed is the composition of the consumer price index, which is now linked to fluctuations in labor costs rather than goods prices. If the Fed can really be credited with price stability, it is precisely to the degree to which the Fed can be credited with moving us from a manufacturing economy to a service economy.
The Role of Deflation
The point is, making a statistical comparison between post-Nixon Shock inflation and gold standard inflation is difficult to do. We could generally characterize the pre-Fed gold standard as being a period of "low inflation," even though there were instances in which inflation was in the double digits. In any case, keeping those difficulties in mind, here is a comparison of how well seven-year returns were predicted by CAPE alongside the rate of inflation. I think there is a connection there, but I am not sure I would have noticed it had I not already been thinking about P/E and inflation for a long time already.
Here is another view using a rolling ten-year correlation of CAPE and ten-year returns and five-year rolling correlations involving five-year returns.
In any event, even if readers should disagree with my notion that deflation has something to do with this, it will at least be useful to mark those instances in which CAPE was poorest at predicting returns: prior to the 1920s and then again in the late 1990s. Even if I am wrong about the deflation connection (actually, especially if I am wrong about the deflation connection), CAPE's uneven performance suggests we should be thinking about what makes the market tick.
In other words, insofar as all of these pre-1945 conditions (deflation, CAPE's unsteady performance, earnings crises, a positive correlation between earnings and stocks) have started popping up here and there over the last twenty years, we should perhaps be wondering what metrics are reliable when looking forward.
Let's look just once more at that comparison between earnings growth and returns. In the following chart, I have put the two series on different axes and flipped the returns to draw out the post-war inverse relationship. This relationship has hung on whether or not earnings were reliably positive. But, something else is going on. By the 1990s (keeping in mind that these charts are showing subsequent returns), that inverse correlation was already breaking down, way before the earnings shock of 2009. Earnings grew rather strongly in the 1990s, but stocks did not slow in any meaningful way. As we all know, as earnings grew, so did the P/E multiple in the 1990s in one final fit of "irrational exuberance."
In other words, the market was already returning to the pre-war relationship well before the earnings shock occurred. So, we can assume that the market was already aware of the earnings crash that was coming fifteen years later or that something else was going on that was altering those relationships. Of course, I think the link between this 1990s phenomenon and the pre-War situation was low inflation. The 1990s is when the fear of deflation became palpable, and I don't think this was coincidence. Perhaps we could think of the link as being more complex, perhaps low-yield and low-inflation, but the theme is broadly the same.
This is probably a good place to pause.
In this installment, we have primarily focused on coming up with a few general propositions about the relationship between earnings growth and stock performance. To recap, we can say that:
1. If there is one imperishable rule governing the relationship between earnings and stocks, it is that the stock market does not like earnings shocks. Not on any day of the week.
2. When earnings can be characterized as volatile and susceptible to earnings shocks, stocks are highly correlated with fluctuations in earnings.
3. Where earnings growth is relatively stable, stock performance tends to be inversely correlated with earnings.
4. There appears to be a link between the appearance of earnings volatility and the emergence of a low-yield, low-inflation environment.
5. CAPE's failures to predict returns also appear to occur during periods of increased earnings volatility (somewhat more precisely, an earnings volatility that is sufficient to induce a significant fall in earnings) and low inflation.
I imagine some readers will want to munch on some of these observations and conclusions before I push on to the next stage of the argument. For those who simply cannot bear the anticipation, I will leave you with the charts below.
In my previous article, I talked a lot about the relationship between earnings and returns as mediated through P/E, whereas in this article we spent a good deal of time comparing earnings and stock prices directly, and then tried to relate that to CAPE's failure to predict markets. In the next article, we are going to stitch this together somewhat by comparing CAPE's relationship with returns and P/E's relationship with earnings.
Earnings vs CAPE
If you look at the relationship between CAPE and subsequent five-year returns and then at the relationship between unadjusted P/E and concurrent seven-year earnings growth, you might notice a certain symmetry.
Where CAPE fails to predict five-year returns, the relationship between P/E and earnings growth breaks down. Our present discussion suggests that this is unlikely to be mere coincidence; moreover, for those who read my previous article, I strongly suspect that this is an exploitable relationship. And, based on the total analysis of the historical relationship between earnings and all the other factors we have talked about, I tend to think that the next five years will be unambiguously bullish "despite" the high and rising CAPE ratio.
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.