Swedroe: Measure Value In Different Ways

Isolating the value factor through different metrics can be a diversifier.

A couple weeks ago, I took a look at some of the recent research on the value premium in large-cap stocks. A commenter on my post raised some issues that I thought were worth addressing separately and in-depth.

The first issue relates to a question posed by the commenter. He asks: “Do we have robust evidence that different (or more) fundamental sorts increase exposure to value?” He also provides the following answer: “The best way to ‘check’ this is to start with an out-of-sample measurement period (1940-1962, per James Davis) and sort large cap stocks (‘fundamental indexes’) on the following variables with HmL loads:

Book Value = 0.25

Cash Flow = 0.21

Dividends = 0.11

Sales = 0.15

Low P/B = 0.43

So over this period, P/B had the greatest exposure to value and the most important determinate of value load wasn’t a particular valuation sort at all, but instead using a ‘low price’ (see Low P/B) sort as opposed to just a (any) fundamental weighting scheme.”

The numbers presented did indeed show that P/B ratios had the highest loading on value (or HmL: high minus low). However, that’s exactly what we should expect since the various values presented are specifically HmL loadings. In other words, the figures are just saying that P/B is most like P/B, which is a tautology.

The statement that "P/B had the greatest exposure to value" is thus not quite correct. It has the greatest exposure to value as measured by book value, but so what? It’s just another tautology. Again, the loading on value will depend on how you measure value.

Let’s see what professors Eugene Fama and Ken French and Dimensional Fund Advisors (DFA) have to say on this issue:

• In their 2006 paper, “The Value Premium and the CAPM,” Fama and French write: “The evidence for a weak value premium in the largest size quintile depends on using B/M [book-to-market ratios] to identify value and growth stocks ... when E/P [earnings-to-price] is the value-growth indicator, the value premiums for all size groups are more than two standard errors from zero.”
• From DFA’s Annual Report for the year ending Oct. 31, 2013: “Value characteristics are a function of stock price relative to one or more fundamental characteristics such as book value, earnings, or dividends.”

We see that Fama and French, as well DFA as a firm, agree that book-to-market is just one measure of value, not the only one. (Full disclosure: My firm, Buckingham, recommends DFA funds in constructing client portfolios.)

When evidence was presented that B/M hadn’t worked for the largest stocks from 1963 through 2004, Fama and French agreed. But they also argued, as seen in the quote above, that E/P continued to work quite well, and therefore value in general still worked. While Fama and French do note that B/M doesn't always work, they’re more than happy to use alternate value measures.

As to whether there has been much of a large-value premium since 1990—which was the question addressed in the 2014 study, “Is There a Value Premium Among Large Stocks?”—anyone can check this for themselves using data available from Ken French’s own website.

Following are the average monthly returns in percentages (with the t-stat in parentheses) for the period in question, July 1990 to June 2013:

• HmL for small-cap: 0.55 (2.48)
• HmL for large-cap: 0.05 (0.27)
• HmL factor (average): 0.30 (1.57)

We see that Ken French’s own data confirms what Sandro Andrade and Vidhi Chhaochharia found—a very small and statistically insignificant value premium in large-cap stocks for B/M since 1990.

We can also see how other value measures fared.

Ken French’s data set provides portfolio returns sorted by B/M, E/P and CF/P (cash flow-to-price) for the full universe, without breaking out into small and large. Since these are value weighted, the large-cap stocks will have a much greater effect on the whole portfolio. For the same period, July 1990 to June 2013, we have the monthly returns (with the t-stat in parentheses) for the top 30 percent minus the returns for the bottom 30 percent of stocks ranked by each value metric:

• B/M: 0.16 percent (0.84)
• E/P: 0.33 percent (1.96)
• CF/P: 0.21 percent (1.23)

While B/M looks more attractive than the large-cap HmL (because it has some small-cap influence) it’s also statistically insignificant; is half the size of E/P; and is still less than CF/P. These results are in keeping with what the authors of the study found.

The data set for E/P and CF/P go back all the way back to July 1951. The following are the monthly premiums for the period July 1951 to June 2013:

• B/M: 0.31 percent (2.97)
• E/P: 0.47 percent (4.49)
• CF/P: 0.38 percent (3.56)

Over this full period, B/M looks better, but still not quite as good as the others. But the main point I want to stress isn’t that B/M is less useful; instead, it’s that B/M works well in some periods and situations, and poorly in others.

When it’s not working well, other value measures have still been quite effective. And that’s why many highly regarded people in the field of finance believe in diversification of value measures.

There’s one additional point worth covering.

As was noted, it’s true that using multiple value metrics will likely result in somewhat higher turnover. However, the increase in turnover isn’t dramatic—perhaps on the order of an incremental 10 percent.

For funds that can patiently trade—because they aren’t pure index funds, whose sole goal typically is to track the benchmark index—an increase in turnover of about 10 percent (particularly in large-cap stocks) wouldn’t have any material impact on the fund’s returns. This may hold especially true in this new era, where bid/ask spreads have shrunk dramatically in recent years.

The benefits of diversification are well known. And they apply not only to asset classes and factors (or sources) of returns, but to multiple measures versus a single measure of value.

The historical evidence suggests that using a multiple-metric approach in constructing value portfolios can add value. The reason is that, while all value metrics have high positive correlations, the correlations aren’t perfect. Thus, the use of multiple value metrics provides a diversification benefit.

Larry Swedroe is the director of research for the BAM Alliance, a community of more than 140 independent registered investment advisors throughout the country.