High Dividend ETFs: Who Needs 'Em? (ETFs: DVY, PEY; CEF: FVD)

Josh Peters at Morningstar recently wrote a good article on the appeal of investing for dividends. Peters noted the fact that many companies are paying out a smaller fraction of their accumulated earnings in the form of dividends than has historically been the case.

Peters is suggesting that dividends will likely go up in the future because of the income demands by Boomers and because many firms and the U.S. economy as a whole is more knowledge driven and less capital intensive than it used to be -- not to mention that corporate profits are very high. These are all good points. Further, after a number of years of high earnings, many firms are sitting on considerable amounts of cash that will either be used to fund share buybacks or dividends. These funds could also be deployed, however, to fund internal growth or acquisitions. Further, the incentive structure for executives are often tied to share price performance and paying dividends will reduce the potential for price appreciation. The value of employee stock options is inversely related to dividends.

Purchasing stocks with high dividends relative to stock price has historically been a winning strategy. This is another reason why many investors look for high dividend stocks. It remains to be seen whether high dividend ratios continue to be a good predictor future higher-than-normal returns, but the underlying logic is compelling.

A Set of High Dividend Stocks

One of the key issues in dividend investing is to consider risk and return together. In an earlier paper, we demonstrated that high dividend stocks might also be accompanied by high volatility. Sometimes, those higher dividends are achieved at the cost of a high-risk portfolio. Part of the high volatility in that earlier selection of high dividend stocks, a list put together by Ben Stein and Phil DeMuth, was undoubtedly due to their leveraged dividend strategy—such a strategy is often not sustainable. Phil Demuth and Ben Stein, in response to the criticism regarding leveraged payouts, came up with another set of stocks in which all members had high dividend ratios but with dividend payouts less than trailing earnings:

Stein and DeMuth 7% Solution Part II

These firms have average trailing twelve-month dividend yield of 7%. We decided to run these firms through the Quantext Portfolio Planner (QPP) to see how a portfolio of these firms would look. QPP allows the user to specify the period of history to use in an analysis and we started with using the trailing five years through January 2006. What we are looking for is a consistent dividend history. Two factors immediately appeared when we performed this analysis. TUES came up with an annual dividend yield of 0% in QPP, while Yahoo! Finance lists the yield at 6.2%. A closer inspection revealed that this discrepancy was due to the fact that TUES has only paid dividends twice—they have a very short history of paying dividends. This is simply too short a history to give me a lot of confidence, so I dropped them off my list. Looking at the rest of the firms:

5-Year Dividend Yields vs. TTM

The TTM (Trailing Twelve Month) Yield was provided to me by Phil DeMuth. QPP calculates historical annual dividend yield and the minimum rolling 12-month yield (above). Even this simple analysis is quite interesting in terms of providing some insight into yields. Obviously a low dividend yield can be due to a large increase in price as opposed to any change in dividends, and vice versa. When the TTM yield is much higher than the longer term statistics, it is natural to look to see whether the increased yield is due to a major decrease in price—as is the case with DLX.

The first thing that QPP showed me about these stocks, aside from the high dividend yield, is the high volatility associated with these stocks. NL, for example, has exhibited a standard deviation in annual return of more than 40% over the last five years, about 2.5 times the total volatility of the S&P500. Every one of these stocks has been more volatile than the S&P500, although their low Betas will help in dealing with the volatility. If you want to build a portfolio out of these stocks, you need to be prepared to deal with the volatility and manage it correctly.

Model High Yield Portfolio

In order to construct a portfolio out of these stocks that is attractive, in my opinion, you may as well reduce the redundancy in this set—too many individual REITS. Also, I decided to drop MMP because I want only components for my portfolio with a ten-year track record. Also, because these are high-volatility stocks, I want to throw some bonds into the mix. The bonds will also help to prop up the dividend yields. For the bond funds, I have chosen VBIIX and VBLTX although I would be just as happy with a couple of bond ETF’s. I chose VBIIX and VBLTX because they have almost a ten-year track record. The portfolio that I chose, without over-tuning it, is shown below, along with the historical results and the Monte Carlo simulation output. This portfolio has an average annual dividend yield of better than 7% over the past five years, but the total volatility over the past five years has been moderate. Further, this total portfolio has a Beta of only 33% which means that this portfolio is not strongly coupled to the overall market. While the S&P500 has exhibited an annual standard deviation in return of 14.38% over the past five years, this portfolio has had annual standard deviation of 10.97%.

Figure 1: Model High Yield Portfolio 2001-2006

Further, while this portfolio has obviously out-performed over the past five years, at a level that is not sustainable, QPP projects that this portfolio will generate an annual average return of 11.56% per year—far above the projected average annual return of 8.3% for the S&P500.

It is fair to ask whether this performance over the past five years is an anomaly. To address this issue, we also ran this simulation against data from July 1996 through January 2001. We started in July 1996 because this was the first full month of listing for the two Vanguard bond funds. This period gives us almost five years of market history prior to the most recent fiver year data shown above.

Figure 2: Model High Yield Portfolio 1996-2001

For the period from 7/1/1996 through 1/31/2001, this portfolio had an impressive average annual yield of 8.38%. While higher than the most recent five years, this is actually remarkably consistent with the more recent results. Several other points bear mentioning. First, given only data available up to 1/31/2001, the Monte Carlo model projects a future average return for this portfolio of 11.22% per year, with a standard deviation of 10.9% (see Portfolio Stats, above). Note that this projection is substantially below the actual results for this period of history in which this portfolio returned an average 18.89% per year. What is really notable here is that the actual average annual return on this portfolio over the next five years was 13.7%. The projected volatility (annual standard deviation) matches well between the projected value (Portfolio Stats in Figure 2) and the realized value (Historical Data in Figure 1). The consistency of the performance of this portfolio is particularly impressive given that in the earlier five year period, the S&P500 returned an average of 18.35% per year and in the second five year period the S&P500 returned an average of 1.67% per year.

For purposes of comparison, the model portfolio shown above has a trailing twelve month yield of 5.8%. The projected long-term average return of 11.5% per year is also well above the projected total return for the S&P500 of 8.3% per year.

Comparison to Dividend Focused ETFs/CEF

There has been a lot of recent attention focused on high dividend ETF’s and high dividend mutual funds. While most of these have very short histories, there are two ETFs and one CEF for which we have one year or more of data. Using the past twelve months (through February, 2006), we can compile statistics for these funds:

iShares Dow Jones Select Dividend Index (DVY)
First Trust Value Line Dividend Fund (CEF: FVD)
PowerShares HighYield Dividend Achievers (PEY)

These three funds have dividend yields of around 3% over the past twelve months—far less than the dividend yield in our model portfolio. Further, the projected standard deviation in annual return for these funds is very similar to the projected annual standard deviation in return for our model portfolio (which is 11.5%), but these funds have Betas of between 70 and 83%, while the historical and projected Betas for our model portfolio is in the markedly lower—more like 40% over the past ten years. That said, our model portfolio has exhibited a Beta of 74% using just the most recent twelve months. Short-term Betas like this are a poor estimate, but we really have no other simple choice for purposes of comparison with the funds because the funds have such a short history. Our model portfolio has exhibited considerably higher total volatility than these three funds over the past twelve months but our longer-term projections suggest this portfolio will exhibit standard deviations in return on the 11-12% level. This level of volatility might be considered in light of the fact that the long-term standard deviation for the S&P500 is around 15%.

Summary of the Model Portfolio

Of the five high-yield stocks in our model portfolio, two have seen large price declines in the past twelve months (DLX, NL) which help to account for their very high current dividend yields. That said, the overall portfolio shows a consistently high dividend yield in the two five year periods that we have examined—one in which the market as whole was going strong and one with anemic returns for the market as a whole.

The model portfolio has generated high dividend yields for an extended period of time—during both bull and bear market conditions. Let’s use the most conservative number that is close to the most recent twelve months--about 6% annual yield. This yield is far higher than the dividend focused ETFs/CEF.

Given that you can achieve this type of performance for a small group of stocks with a long and consistent dividend history, why would you bother with the dividend-focused funds? Further, because of the major declines in NL and DLX, these two stocks look quite attractively priced right now and these price declines have helped to drive up the dividend yield for current buyers. It is important to note, of course, that most investors will want more individual stocks or funds in a portfolio because investors in this model portfolio will be exposed to a fairly high degree of company-specific risk. Portfolio theory does not compensate for this effect. The ETFs/CEF have less company-specific risk because they invest in a larger set of firms, but they also charge their management fees and have a considerably shorter history to examine. It would be quite simple to expand this portfolio to include more individual stocks—potentially drawn from the same original population of firms suggested by Stein and DeMuth. From a portfolio standpoint, a sector concentration with multiple stocks requires some care because there will be correlation between portfolio components that exceeds the correlation implied by Beta. This can be handled quite easily in QPP and the forthcoming release adds a new tool to allow compensation for these effects in an easy way.

For the more conservative investors, we find that if you had 60% of the portfolio in the two bond funds and 8% each in the five stocks, you would still have an average annual yield over the past five years of more than 6% per year and a projected total annual return of 8.6% per year—better than the projected total return of 8.3% per year for the S&P500.

.......
View a .pdf version of this article with higher resolution images.

More information about quantitative tools for portfolio management can be found at the Quantext site.