December 12, 2006 Letters to the Editor

Updated: December 24, 2006.

Asset Allocation

Themeos asks an important question:

In calculating my bond/equity split, I throw the bond portion of my Balanced and Growth & Income funds into my bond allocation column. I'm wondering if this is an acceptable practice.

Your web site is one of my best finds this year. Many thanks.

HERE IS MY RESPONSE

First, thank you for your kind words. I appreciate them greatly.

Not only is your approach acceptable, it is EXACTLY right.
All of the studies that I know about use pure stock components or slices. For example, they use the S&P500 as a whole or a Large Capitalization Value slice or a combination of slices, where each slice is pure.

I am not aware of any study that includes a balanced holding as a single component. A balanced holding shows up as two or more slices.

Efficient Market Hypothesis

I received this letter from ElLobo on the M* forums.

I just wanted to tell you that I appreciate the work you do, and your posting style. I used to post quite a bit on the M* forums, but backed way down once I retired. I've since picked it up a bit, since I am recovering from two surgeries and the construction of a house!

Anyhow, thanks for pointing me to your site here. I'll need to work my way through it. I see you have several references to Pete Ponzo (Gummy). He and I go way back.

I had two questions concerning efficient markets that I thought I might post to you off the forum.

First, have you ever seen an equally weighted index, or have you ever calculated one? Similar to the market cap weighted indexes. I fully realize that an equally weighted index fund would probably be nigh near impossible to manage (rebalancing issues and all that). But I am interested in how such an index would have compared to a market cap index.

Secondly, have you ever seen F/F, CAPM, efficient market type work where only the capital change component of TR [total return, with dividends reinvested and including price changes] was considered for the return, rather than TR itself? That is, where dividend yields were ignored? Something sticks in my mind that perhaps all stock prices follow efficient market considerations, while dividends paid are reflected in the elusive value premium that may, or may not, exist?

Finally, where do you hail from? I'm out here on the left coast, in SW Washington State..

HERE IS MY RESPONSE

I post mainly at the Morningstar Income & Dividend Investing board. I post at other Morningstar discussion boards as well, including the Vanguard Diehards board.

Thank you for your kind words. I appreciate them greatly. I know that people recover from surgery. But I don’t know that they ever recover from building a house. [Humor, but it has an element of truth.]

I refer to Gummy’s (retired Professor Peter Ponzo) website often because his work is top notch. I wish you well.

The Efficient Market Hypothesis predicts that Capitalization Weighting of the entire market is always best.

There is a wealth relevant data in James P. O’Shaughnessy’s What Works on Wall Street. I have the Revised Edition, 1998. It is excellent for those who are comfortable with numbers. It includes processed raw data (detailed, easy to understand, partially reduced data) in a consistent format as well as conclusions.

James O’Shaughnessy compared two groups of 50 equally weighted stocks with each other and with the S&P500, which is capitalization weighted. The Large Stocks group had capitalizations greater than the database average, usually the top 16%. The All Stocks group had capitalizations greater than $150 million.

From page 6, Why Indexing Works:

“Indexing to the S&P500 works because it sidesteps flawed decision making and automates the simple strategy of buying the big stocks that make up the S&P500. The mighty S&P500 consistently beats 80 percent of traditionally managed funds by doing nothing more than making a disciplined bet on large capitalization stocks. Figure 1-3 compares the S&P500 with those for our Large Stocks..bought in equal dollar amounts. The returns are virtually identical..And it’s not just the absolute returns that are similar—the risk, as measured by the standard deviation of return, is also virtually identical for the two strategies..”

From chapter 4, pages 35, 37, 42 through 56, Ranking Stocks by Market Capitalization: Size Matters:

“As mentioned in Chapter 1, there is virtually no difference between stocks with market capitalizations above the Compustat mean (Large Stocks) and the S&P500..An investment in the All Stocks group did considerably better..”

BUT

“Most academic studies of market capitalization..are nearly unanimous in their findings that small stocks do significantly better than large ones..”

“The glaring problem with this method is that it’s virtually impossible to buy the stocks that account for the performance advantage of small capitalization strategies..”

“The only way to achieve these stellar returns [microcaps] is to invest a few million dollars in over 2000 stocks..”

“Stocks with market capitalizations between $25 million and $100 million, as well as those with market capitalizations between $100 million and $250 million, do outperform large stocks on an absolute basis but fail when risk [Sharpe ratio] is taken into account..”

“The big surprise is the performance of Market Leaders..,” which do well.

I have not seen source material related to the F/F (Fama/French), CAPM (Capital Asset Pricing Model) and the Efficient Market Hypothesis with the exception of a single briefing by Dr. Eugene Fama at the Morningstar Vanguard Diehard’s board.

My research recently drew my attention to the fallacy behind Capitalization Weighting. It fails to include dividends.

Capitalization Weighting of What?

You are correct in thinking about price-only behavior. Dr. Sharpe’s theorem applies to price-only returns. But the price-only return of the overall market must be inferior to the return that includes dividends.

I do not consider the terms “value premium” and “equity premium” to be useful concepts, per se. They cloud the issue. Efficient Market Hypothesis rhetoric makes matters even worse, assigning the word “risk” to higher returns regardless of cause.

[A good example of this is Rob Arnott’s “fundamental indexing.” The starting point for understanding the theory is to work with pairs of identical companies, whose stocks differ only because of randomness. Upon careful reflection, the advantage of fundamental indexing is readily apparent. It is not a matter of risk. It is not a matter of a value premium.]

Recently, I investigated the effect of predicting equity premiums as opposed to stock returns. I concluded:

“It is best to estimate real, annualized, total returns as a function of valuations (100E10/P). Estimating nominal returns does not do as well. Estimating equity premiums does poorly.”

Edited: Returns, Not Equity Premiums
Returns, Not Equity Premiums

I think that Dr. Eugene Fama’s briefing charts at the Morningstar Vanguard Diehard’s board were clear and straightforward. He uses methods similar to mine (regression equations extracted from historical data). From what I can see, his 95% “explanation” of the stock market is an R-squared value. If so, the award winning “explanation” that the Fama/French three factor model doesn’t do what it is supposed to do is in error.

That doesn’t mean that the briefing is correct in all regards. It doesn’t meet my New Standards for Financial Reporting. It contains no confidence limits. It excludes critical information related to timeframes. It glosses over a wealth of information related to variances and standard deviations. For example, it glosses over the fact that dividends reduce downside volatility.

New Standards for Financial Reporting

As far as I can tell, Dr. Fama’s briefing determined single year expected returns. My research shows that this removes almost all of the effect of valuations. In essence, his procedures obscure a wealth of opportunities. It is geared toward not finding opportunities. By not showing confidence limits at reasonable intervals, he makes predictions meaningless. [Yet, this is consistent with today’s Financial Reporting, which I consider atrocious.]

Finally, I am in Crestview, Florida, in the panhandle, near interstate 10. [The state does not always include the panhandle in its maps.] This is far enough inland to avoid the worst from hurricanes. Air conditioning makes Florida bearable in the summer.

Research Project

ElLobo of the M* forums followed up with this letter.

In response to my words: "You are correct in thinking about price-only behavior. Dr. Sharpe's theorem applies to price-only returns. But the price-only return of the overall market must be inferior to the return that includes dividends."

ElLobo wrote: Absolutely.

In response to my words: "I do not consider the terms "value premium" and "equity premium" to be useful concepts, per se."

ElLobo extended the discussion further:

I believe researchers have been mystified over the reasons for both a small cap and value 'premium' that exists in past return data. As I understand it, researchers have found that returns for small cap stocks have been a bit higher than those for large cap stocks, and returns for value stocks have been higher than those for growth stocks.

My point with my original question was this. Assume the price-only based returns (capital change component of TR [Total Return]) were fairly constant among asset classes. Then the full Monty TR [Total Return] would simply depend on the average dividend yield of each class, which is added to the price-only return. The data would then suggest that value companies tend to pay higher percentage dividends than growth companies (distribute earnings to investors, rather than retain earnings to fuel growth). Likewise, smaller companies tend to pay higher dividends (small companies are newer, hence higher dividends imply higher risks) than larger, more established companies.

So, a value 'premium' might be a 2% dividend yield, a small cap 'premium' might be a 2% yield. If the base price-only return was 7%, then the style box TR [Total Return] might be 9%, 8%,7% (LV,LB,LG), 10%, 9%, 8% (MV,MB,MG), and 11%,10%,9% (SV,SB,SG).

So my next question is, "Have you, or anyone else, looked at the average dividend yield of small cap versus large cap, and value stocks versus growth stocks?

The only use for discovering this (an explanation, really) is a possible Noble Prize in economics for you! 8-)

NOTE of Explanation: Morningstar “style boxes” are a 3x3 matrix that present information (such as allocation percentages and percentage returns) according to investment “styles.” The individual components are: Large Capitalization Value LV, Large Capitalization Blend LB, Large Capitalization Growth LG, Medium Capitalization Value MV, Medium Capitalization Blend MB, Medium Capitalization Growth MG, Small Capitalization Value SV, Small Capitalization Blend SB, Small Capitalization Growth SG.

ElLobo continued:

On another subject (rebalancing). I remember that any rebalancing bonus depends upon the correlation coefficient between the two assets being rebalanced. If perfectly correlated, minimum bonus. If perfectly inversely correlated, maximum bonus. My question is this: The cap [capitalization] change component of TR [Total Return] is the part that has a correlation coefficient associated with it. A correlation coefficient for the yield components of two assets should be zero, since yields are typically constant. Would this explain your statements that dividend portfolio shouldn't be rebalanced?

HERE IS MY RESPONSE

I think that you are on to something worth looking into. It would make a great research project.

I do not have the tools, nor do I have the data, to conduct such an investigation.

I view such a project as open ended. I see strong parallels with this web site. My research was begun in order to answer Rob Bennett’s question: “What are price adjusted Safe Withdrawal Rates?” On a minimalist basis, this might have been a handful of articles. Instead, my site now has 391 web pages.

Here are links that show how this research has developed:

Our Roots
May 2005 Highlights
May 2006 Highlights

We need to be careful about style boxes, definitions and possible overlap. Different people use different definitions.

I believe that James P. O’Shaughnessy’s What Works on Wall Street has helpful information. I believe that David Dreman’s Contrarian Investment Strategies: The Next Generation contains critically important information is Chapter 15: “Small Stocks, Nasdaq, and Other Market Pitfalls.”

Remember my quote from James P. O’Shaughnessy: “The glaring problem with this method is that it’s virtually impossible to buy the stocks that account for the performance advantage of small capitalization strategies..”

A significant fraction of the high returns associated with smaller stocks are artifacts related to liquidity. Some important studies simply split the difference between bid and asked prices, but the differences were huge. Some important studies assumed purchases that exceeded the available float (number of shares available for purchase) by a huge margin, yet they made no adjustment to prices.

A decent researcher with an active imagination could carry your idea into a host of areas fundamental to understanding stocks and other investments. Your project, stated in more general terms, is to incorporate dividends explicitly when it comes to any equity premium.

For example, a researcher might investigate your assertion about capitalization and dividend yields. Owners of a small capitalization company might favor higher dividend yields for the reasons stated. But large capitalization companies, especially mature large capitalization companies, are the ones who are able to pay higher yields.

A minimalist researcher might be content to report these differences in statistical terms. A more creative researcher will see lots of questions that need answers. Investors should demand higher dividend yields from smaller companies because their dividends are less secure. But investors may associate higher dividend growth (or better growth prospects) with smaller companies and pay for the opportunity. Or, public perceptions of the “well known, well established, proved” higher returns of Small Capitalization Value stocks may distort prices enough to lower dividend yields.

Regarding rebalancing: No. That is not the mechanism.

You are correct about probability distributions. Dividends affect the probability distributions. The main effect shows up as a shift in the mean, not directly related to the randomness term (covariance). Quantitative calculations of the rebalancing bonus usually assume a lognormal distribution, which is known to be wrong, but is often adequate. The presence of dividends introduces additional, subtle distortions to this lognormal approximation. As a practical matter, dividends introduce just one of many sources of error.

For those interested in seeing the mathematical equations and approximations, visit this link to Gummy’s (Professor Peter Ponzo) Web Site (Gummy Stuff).

Gummy's: The Rebalancing Bonus

Dividends shift the average return of an investment. For the rebalancing bonus to offer an advantage, various investments need to have similar (don’t be too precise) total returns. This is with dividends reinvested. If you have a reliable way to identify a better investment, you are better off increasing its allocation (subject to your tolerance to volatility).

I have always found it necessary to analyze rebalancing comparisons very carefully. Almost always, there is a lot of randomness. I have to compare what happens on the upside and what happens on the downside. My experience so far has been that rebalancing removes too much of the upside potential while offering very little downside protection. This is not true during times of high valuations such as today. However, it is better to reduce your stock allocation directly at such times (taking your lumps early, if necessary) as opposed to rebalancing (and having to take bigger lumps later).

For a great discussion on rebalancing, visit the Dividend & Income Investing board at Morningstar. Read conversation #1084 about Orders of Magnitude. Pay special attention to Russ’s (Sirschnitz) summary in post 8. [The whole conversation is worth reading.]

Morningstar

For a very quick overview, read Orders of Magnitude.

Orders of Magnitude
Edited: Orders of Magnitude

Letters to the Editor in 2006

Letters to the Editor in 2005