Seeking Alpha

Low P/E ≠ Low Volatility (At Least, Not All The Time)

|
Includes: NFLX, NOBL, SPY, VZ
by: Investing Doc
Investing Doc
Long only, value, growth at reasonable price, long-term horizon
Summary

A fellow contributor opines that low P/E ratios are desirable attributes in stocks for retirees as these will be associated with less volatility. The capital asset pricing model suggests otherwise.

I performed a study of S&P 900 companies to examine P/E ratios and subsequent volatility rates to see if there was empiric evidence for either thesis.

The results were both common-sensical and surprising.

Earlier this month, a fellow contributor, George Schneider makes the argument that lower price-to-earnings ratios lead to lower volatility, and therefore ought to be more attractive to retirement investors (who ostensibly would like to have some steadiness in their portfolios in their twilight years). He cites as part of his evidence for this assertion the high volatility displayed by Netflix (NFLX) compared against the almost bond-like behavior of steady dividend grower Verizon (VZ). George ends his piece concluding that retired investors would do better to “leave the higher-P/E, higher-volatility stocks to the young folks” while they “relax, sit back and calmly collect our dividends.” Multiple commenters also chimed in, suggesting that higher P/E ratios lead to lower future expected returns.

I had a couple of immediate questions upon reading this. The first was a theoretical one: most equity valuation models that I’m familiar with associate higher volatility (high-beta) stocks with lower price-to-earnings ratios, precisely the opposite of what George argues. In the comments, 07Ghost suggests that selection bias might be at play in George’s thesis, by noting that there are any number of counter-examples of highly-valued, low-volatility stocks out there. In a reply, George notes that while exceptions exist, the association of lower P/E ratios and lower volatility ought to be “a rule… we can use [as] one sign post in our research.” But is this really true? Pricing models such as discounted cash flows that rely upon the capital asset pricing formula:

Cost of equity = Risk-free rate of return + Beta x (market rate of return – risk-free rate)

suggest an inverse relationship between valuation and volatility. Future dividends and cash flows are discounted using cost of equity as a divisor; as the reader is aware, this implies that expected returns of higher-volatility stocks will be discounted at a higher rate, and that these stocks will in turn receive lower valuations. If this theoretical basis holds, then how does George arrive at his conclusion? Is there empirical evidence to support his claim?

Another commenter points out that sector selection probably accounts for much of George’s reasoning, by noting that higher volatility issues with high valuations are often found in highly cyclical sectors, and that underlying volatility is associated with cyclical changes in earnings. Indeed, George himself seems to agree that conservative investors would do well to ignore highly cyclical sectors and focus instead the less-exciting, staid sectors to generate income.

But is this a case of trying to have it both ways? If lower valuations are indeed associated with lower volatility, then shouldn’t this also hold true across sectors? How about within sectors? And what would a sector-limited focus imply for actual performance? Basically, how would an approach focused primarily on avoiding unpredictability fare compared to a more diversified portfolio in terms of overall volatility, dividend income, and total return?

Question #1: What empirical evidence supports a correlation between volatility and P/E ratios?

Though I’m sure I could find references on the internet that answer this question for me, this is an investing website, not a surrogate for a search engine, and I wanted to explore this issue on my own. So I collected historical financials (including dividends) for the S&P 900 (SPY) from Morningstar, including current and prior Dividend Aristocrats (NOBL). Combined with these financial data as well as historical prices from Google Finance, I calculated yearly median P/E ratios as well as median trailing 3-year betas for each issue. I then calculated the average correlation between a stock’s P/E ratio and subsequent volatility over the next 2- and 5-year periods:

Overall correlation between trailing PE and subsequent volatility was a decidedly equivocal 0.03, arguing against a positive relationship between the two. Lower PE’s (5-50) indeed had a lower average subsequent beta (1.14) compared with higher PE’s (50-95, average beta 1.2). However, much of this appeared to be driven by lower betas in the 10-20 range. In fact, trailing PE’s in the 12-13 range were associated with the lowest subsequent betas of all, with subsequent beta increasing with the square of the distance to this value (overall correlation 0.88):

So—at least over the past 10 years—there does appear to be a relationship between volatility and price/earnings ratios, but the relationship isn’t exactly linear. Rather, stocks that fall into a narrow range (about 12-13) tend to have the lowest subsequent volatility, and stocks outside of this range—priced either for significant growth or decline in earnings-per-share—have progressively higher volatility.

This empirical finding is hard to square against both George’s claim and the capital asset pricing model, but fits intuitively. Stocks that are priced for significant declines in profitability might either confirm those fears or surprise to the upside; either way, prices are likely to shift significantly. Rather than attempting to suggest a simple relationship between P/E ratios and volatility, one might say instead that extremes in market sentiment are often followed by extremes in price movements.

Question #2: How do valuation ratios relate to volatility within sectors?

Of course, the reader will point out that this analysis was carried out across sectors, which carries its own problems. While this helps the analysis with generalizability, it introduces potentially confounding factors. Most obviously, different sectors have different capital expenditure demands that affect cash flows; as a result, one would reasonably expect the technology sector to have higher valuation ratios than, say, industrials. But different sectors often have different degrees of exposure to cyclical factors as well. Does the relationship mentioned above—that extremes in market sentiment are followed by extremes in price movements—obtain to individual companies within sectors? Or might sector characteristics be sufficient to explain any sort of relationship between valuation and volatility?

I ran the same analysis as before, but this time by individual sector. Here, the results were dramatically different. For example, here are trailing PE’s listed against their subsequent betas for the Industrials Sector:

And again, for Technology:

And again, for Basic Materials:

These distributions all look terribly different. The Basic Materials sector, for example, exhibits a nearly bimodal distribution, suggesting a high degree of cyclicality (with lower volatility seen with higher valuations, perhaps reflective of cyclical earnings troughs). Interestingly, the Industrials sector has a similar, if perhaps more muted, bi-modal distribution as well, while Technology exhibits a strongly positive trendline, with higher volatility being associated with higher valuations. The differing behavior suggests, sensibly, that different sectors behave in different ways, and while that isn’t exactly ground-breaking, it’s interesting to me to see this play out in the data.

What about the thesis from just before that progressive distance from an inflection point results in progressively higher volatility? Does that still apply within sectors as well? The answer, it seems, is a solid “yes, but it depends,” as certain sectors adhere to this finding more than others:

Industrials

Healthcare

Technology

Consumer Cyclical

Utilities

Financial Services

Basic Materials

Real Estate

Consumer Defensive

Communication Services

Predictability

94.23%

64.46%

48.67%

72.39%

80.04%

91.01%

14.91%

92.16%

86.95%

-77.43%

Median PE

19.50

21.05

23.75

18.85

17.20

15.90

19.90

15.11

20.00

22.50

The table above shows calculated “predictability,” that is to say, the correlation between subsequent volatility ratings and the square of the distance from a given price-to-earnings ratio to a sector median. The data suggest that the thesis from Question #1 is valid. Most sectors are relatively predictable, though sectors exposed to external factors (Basic Materials) or prone to disruption or rapid change (Healthcare, Technology) are—unsurprisingly—much less predictable.

But outside of these sectors, the thesis still seems to apply strongly, and even Healthcare and Technology showed moderate to strong predictability. (Communication Services, it must be noted, displayed a negative predictability rating, which is to say that it displayed a strong trend towards lower volatility with increasing distance from the median P/E—but it should be stressed that its n was a mere 35 companies.) The bottom line: For most sectors, it’s not looking for valuations on an absolute basis that is the most helpful in predicting volatility, but on a basis relative to a sector median.

Interim conclusion:

While it can’t be said that lower valuations necessarily equal lower volatility, it certainly appears that George was onto something: that investors interested in achieving some predictability in their portfolios might do well to consider certain sectors—presumably if their valuations are not extreme. The most predictable sectors appear to be Industrials, Financials, Real Estate, and Consumer Defensive; the least predictable appear to be Basic Materials, Technology, and Healthcare.

This shouldn’t be read as a recommendation to go out and buy these sectors currently, as their predictability also suggests that more extreme, expensive valuations will be more likely to result in significant downside when prices correct. Moreover, this analysis hasn’t quite answered the third question of whether or not focusing on more predictable sectors leads to outperformance, whether in total return or income. Is it worth pursuing high-predictability or is there a "predictability premium" that has to be paid? And would it be worth considering such a strategy now given current valuations? Those questions I’ll attempt to answer in a follow-up article.

Disclosure: I am/we are long SPY. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.