While reading a post at The Confused Capitalist discussing Ken Fisher's book, The Only Three Questions That Count I started thinking about how best to discuss the risks present in high normalized P/E years.
The following quote from The Confused Capitalist lead me to write this post:
Mr. Fisher's big opening statement in his book challenges a long-held market axiom that high PE's denote reduced returns for some period of time into the future. He states that the year immediately following a high PE year has virtually no statistical inverse correlation. While this might be true, it ignores the fact that there is heightened risk of a reduced return into the relatively near term future. Whether that lower return is realized in the immediate year following, Mr. Fisher's data suggests, no.
However, an investor (rather than a "speculator" or "trader") would be foolish to ignore the reduced odds of outperformance that this period provides.
After reading this, I decided I had to write a post directly discussing the risk present in extraordinarily high normalized P/E years – because such years are riskier than most years.
I don't mean to say that an intelligent investor (or more likely trader) can never have a good reason for buying during an obviously expensive year. I do, however, mean to say that anyone who blindly assumes the risks present in an expensive year are comparable to the risks that were present during a typical year in the 20th century is operating under a dangerous delusion.
Furthermore, while I prefer to focus on the long-term, I can not allow others to unthinkingly entertain the pleasing notion that the ill effects of high normalized P/E years are only felt in the long-run. The evidence directly contradicts this particular delusion. A one-year bet on the Dow during a high normalized P/E year is a risky bet quite unlike a one-year bet on the Dow in any other year.
For those who haven't read my series on normalized P/E ratios, let me explain how I worked with the data. I measured compound annual point growth in the Dow based on yearly averages for that index. For the sake of simplicity, dividends were ignored entirely – obviously, this omission benefits high normalized P/E years and serves to downplay the normalized P/E effect, because low normalized P/E years tend to have high dividend yields while high normalized P/E years tend to have low dividend yields. For a description of how I calculated normalized P/E ratios read On Calculating Normalized P/E Ratios.
With that explanation out of the way I can turn to the issue raised by The Confused Capitalist. Remember, I'm using 15-year normalized P/E ratios which differ somewhat from the P/E ratios you read about on a daily basis. Occasionally, the difference is quite large.
The poster child for such differences between P/E ratios and normalized P/E ratios is 1982, a year with a fairly ordinary looking P/E ratio of 14.26 but an absurdly low normalized P/E ratio of 6.88 – which just happens to be the lowest normalized P/E ratio on record. With the benefit of hindsight, we now know 1982 was a good year for long-term investors to buy into the Dow – so score one for normalized P/E ratios.
Of course, there's nothing magical about the 15-year normalized P/E ratio. It's just an indicator of cheapness. In 1982, the Dow's price-to-book ratio was also screaming "buy." When something's really cheap it tends to look cheap from a lot of different angles.
The angle I chose for this discussion is the 15-year normalized P/E ratio. I began by selecting the year with the highest normalized P/E ratio from each decade for which I have a full ten years to work with. That limited me to just six years – one from each decade from 1940-1999.
For the purposes of comparison, I also selected the year with the lowest normalized P/E ratio from each decade. This left me with two groups of six years each, a low normalized P/E group and a high normalized P/E group.
The low normalized P/E group had an average 15-year normalized P/E of 10.26 vs. 18.31 for the high normalized P/E group. The difference in average actual P/E ratios was somewhat smaller, 11.80 vs. 16.59. The difference in average price-to-book ratios was also quite large, 1.28 vs. 2.58. However, the difference in median price-to-book ratios was a bit smaller, 1.06 vs. 1.94.
Was there a noticeable difference in one-year performance between the two groups?
Yes. The six low normalized P/E years posted an average point gain of 17.76% while the six high normalized P/E years posted an average point decline of 3.81%. All six low normalized P/E years experienced point growth while five of the six high normalized P/E years experienced a point decline.
The very best performance among the six high normalized P/E years was a gain of 2.38%, while the very worst performance among the six low normalized P/E years was a gain of 5.57%. In other words, none of the six high normalized P/E years managed to outperform any of the low normalized P/E years.
All of this happened despite the fact that the low normalized P/E years averaged an earnings per share decline of 1.49% while the high normalized P/E years averaged EPS growth of 18.99%.
Five years out, their positions were reversed as the compound annual EPS growth of the two groups over a five-year period favored the low normalized P/E years, 9.77% vs. 5.18%.
These last two facts help explain the low normalized P/E effect. It's not just about cheapness – it's about mean reversion and illogical expectations running headlong into the brick wall of reality. As a result, high normalized P/E ratios are a sign of serious risk and strong earnings growth (even 19% a year) is no guarantee of safety. In fact, in the extreme years we looked at in this post, earnings declines were three times more likely among the group that posted great one-year returns.
For those curious about interest rates – yes, the low normalized P/E years did have a higher average interest rate than the high normalized P/E years. The two groups had AAA bond yields of 7.07% and 5.83% respectively.
As you've probably come to expect by now, the low normalized P/E years also saw stronger point growth over longer time periods than the high normalized P/E years. It was 17.76% vs. (3.81%) over one year, 10.63% vs. (0.99%) over three years, 11.85% vs. 2.38% over five years, 10.87% vs. 4.79% over ten years, and 9.67% vs. 5.16% over fifteen years.
So, at least according to this small group of twelve extreme years taken from six different decades, especially high normalized P/E years are especially risky and long-term investors aren't compensated for those risks in later years. In fact, both in the short-term and the long-term, buying the Dow in expensive years tends to lead to more risk and less reward than buying the Dow in cheap years.
Not exactly a shocking conclusion.
A Look At 15-Yr. Normalized Dow P/E Ratios