I am not a big fan of dividend-oriented investing for most investors. However, income starved investors that are feeling the pinch from today's zero interest rate environment and are seeking alternative sources of income, and many have turned to dividend funds.
With this backdrop, I was disappointed to read a recent edition of Barron's, which featured an article that ranked top dividend funds ("Top Dividend Funds" appearing in the November 5, 2012 edition of Barron's). What was puzzling to me, and prompted me to write this article was Barron's decision to rank the dividend funds based upon their 3 year total return, including both dividends and appreciation in share price. While this measure is appropriate if your investment goal is long term growth, it is not appropriate if your objective is to provide a steady paycheck. And, if it is growth that you are most concerned about, why invest in dividend-rich funds?
I did some research on the 53 funds included in Barron's study and used an alternative measure to rank the performance of these funds that would be more appropriate for an income-oriented investor. The purpose of this paper is to present an alternative way of measuring the dividend performance of any portfolio. I am sure there are many others, but thought this would be worth sharing with the SA community.
Here are my rankings, alongside Barron's:
How I Ranked The Dividend Funds
In developing an alternative unit of measuring fund performance, I assumed that income-oriented investors would be most concerned about the following:
- The current level of dividend income
- The predictability, or volatility, of dividend income
- The rate of growth of dividend income
- Fund expenses (that creates a drag on fund performance)
While there is no single, perfect way to rank funds across all of four of these criteria, here is how I went about it:
- I performed a linear regression on each fund's most recent five years of historical "net dividends" (which I define as dividends less fund expenses), calculating the net dividends that would have been received each year for a given investment at the beginning of the five year period.
- Using this regression, I then estimated the expected net dividend in year 6 (essentially, extrapolating the regression line out one year).
- I calculated the standard error of the regression (i.e. how well or poorly the linear regression fit the data)
- Finally, I calculated the ratio of the estimated net dividend from step 2 above to the standard error calculated in step 3 above.
OK, now I'll explain this in plain English. If we plotted each year's dividend of a given fund on a graph, where the X axis was time and the Y axis was the dividend amount, the regression would be the straight line that would be drawn that "best fits" the plotted dividends over the five years. If the net dividends had an upward trend (i.e. it was growing), this would result in an upward, "positive" sloping line. In addition, those funds that produced higher average net dividends would have a line that was located higher above the x axis (or a higher "intercept" in regression speak).
Now, to explain the standard error component. This is a measure of how well the regression explains the observed data; said slightly differently, the more the plotted data varies from the straight line formed by the regression, the more "random" or volatile the net dividends.
Finally, taking the ratio of the estimated dividend in year six will consider (1) the impact of observed historical dividend growth (through the slope of the linear regression) (2) the impact of this historical dividend level (through the "intercept" of the linear regression, or how high above the x axis the line is located) (3) the impact of fund expenses (by reducing dividend for these expense when calculating a net dividend) and (4) the volatility of the dividends.
The ratio of the regressed dividend to the standard error can be thought of as being akin to a "Sharpe Ratio", but applied to the net dividend received instead of share price.
The rankings presented earlier are based upon this calculated ratio of estimated net dividend in year 6 (per the regression) to standard error; the higher the ratio, the higher the ranking.
Here are the results of the analysis (and, as usual, please remember that past results do not necessarily indicate future performance):
Historical performance of any investment is not necessarily indicative of future performance, so any analysis done on historical performance must be done keeping this basic fact in mind.
First, it is probably worthwhile to take a look at the #1 ranked fund on our list, the Principal Equity Income Fund (PQIAX), which has a significantly larger score than the other funds, just to illustrate what we were trying to accomplish in our ranking methodology. This fund was able to deliver consistent dividend increases over the five years of observed history and had the lowest standard error; in other words, it increases in just about a perfectly linear manner. This can be seen by inspecting the plot of the historical net dividends and the associated regression line:
So, as you can see Principal's fund was quite successful over this period of time in delivering consistent dividend increases (albeit, not at a stunning level of returns).
Looking at performance through this lens leads to some dramatically different conclusions regarding relative fund performance. Many funds that were ranked low by Barron's look much better for income investors. Federated's Equity Income Fund (LEIFX), Franklin's Equity Income Fund (FISEX), and American Century's Equity Income Fund (TWEIX) were ranked 35th, 41st, and 43rd, respectively, by Barron's, but all appear in our top six ranked funds. Conversely, some funds that were ranked high by Barron's did not fare as well under this alternative measure; Touchstone's Premium Yield Equity Fund (TPYAX), Rochdale's Dividend & Income Portfolio (RIMHX), TCW's Dividend Focused Fund (TGIGX), and Royce's Dividend Value Fund (RDVIX) were ranked 4th,8th , 6th, and 11th by Barron's but only a achieved a ranking of 38th , 41st , 46th, and 47th on our list.
Some other observations drawn from this analysis:
1. Ho-Hum Net Dividend Yields
After applying the regression, expected net dividends generated by the funds for year six are at most 2.7%, and the average for the top 15 ranked funds is only 1.38%, not much better than what could be realized through today's fixed income investments. Why is this? For starters, the average expense ratio for the top 15 funds is 0.91%, a significant drag on net dividend yield (reducing it by 40%). Second, the regression cuts through the "noise" from year to year dividend volatility to show its longer term trend; while some funds have attractive net dividends in some years, this analysis suggests that funds struggle to sustain this favorable performance over the long term.
2. Majority of Funds have Declining Net Dividends Trend
Of the 53 funds, 28 (or 53%) produced a negative slope from the regression analysis, meaning that they have net dividends that have a general downward trend over the observed five year period. Further, only 2 funds out of the 53 were able to produce increasing dividends every year over the observed five year period. This suggests that the probability of a net dividend increase/decrease from one year to the next is virtually identical to that of a coin toss, not statistically dissimilar from a purely random outcome.
3. High Dividend Yields Can Be Deceiving
Many of the funds with the highest average net dividends were not at the top of our list because of their high volatility. Federated's Stategic Value Dividend Fund (SVAAX) and Rochdale's Dividend and Income Portfolio (RIMHX), for example, were among the highest net dividend performing funds on the list, but only came in at 35th and 40th on our list, respectively, because of their inconsistent performance; in fact, both exhibited a decreasing dividend trend over the observed five year period and both were among the funds with the highest standard error.
4. Many Inefficient Funds
There are many funds on this list that generate lower expected net dividends and more volatility than their competitors - not a good combination! To better illustrate this, the graph below plots each fund against the standard error of the regressed net dividend (X axis) and the expected net dividend in year six per the regression (Y axis). Ideally, we want to have a fund with low standard error and high expected net dividend, placing them in the upper left part of the chart:
Note that this chart does not include 10 "outlier" funds that are even further away from the "efficient frontier" region (highlighted in yellow).
I have highlighted the top fifteen funds from our rankings, and as you can see they are all in the efficient frontier region, making them the most preferable of the funds in the rankings. Many of the remaining funds are far removed from this efficient frontier, suggesting that they are not delivering attractive net dividends for the risk that the investor is bearing.
This chart also illustrates another important point - that dividend fund performance cannot be succinctly summarized in a single measure. In particular, the fifteen top dividend funds each deliver different combinations of net dividend income and volatility; those more sensitive to volatility may prefer the Principal Equity Income Fund (PQIAX) (ranked 1st on our list) while those willing to bear more volatility in the hope of receiving higher levels of net dividend income may prefer the Columbia Dividend Opportunity Fund (INUTX) (ranked 7th on our list). One is not necessarily categorically better than the other; one offers the benefit of lower risk while the other offers the opportunity for more income.
Drilling down another level into the data, here is some additional information on various metrics derived in the analysis that may be of interest:
A Mixed Story on Expenses
The correlation of fees to our rankings is a bit mixed. While you do see some of the lower expense funds appearing in the upper part of our rankings (see the Vanguard Dividend Growth Fund (VDIGX) and Vanguard Equity Income Fund (VEIPX)), funds with modestly higher expense levels are also highly ranked, which may suggest that some active management may create a net economic benefit on net dividend performance.
What About ETFs?
For those of you interested in ETFs, I applied this analysis to a listing of the largest Dividend ETFs appearing in an article in Barron's on July 14th, 2012 ("Finding the Right Dividend ETF for You"). Here is a summary of the results:
In general, dividend ETFs had higher ratios of predicted net dividends to standard error than their mutual fund counterparts - seven out of the nine ETFs had ratios higher than the 16th ranked dividend mutual fund. This is partially due to lower operating expenses, but not entirely. To test this, I replaced the expense ratios of each ETF with the average expense ratio of the top 15 dividend mutual funds (which was 0.90%) and recalculated the ratios. The Vanguard Dividend Appreciation ETF ratio decreased considerably (from 19.23 to 9.77) but changed very little, suggesting that the difference in performance had more to do with dividend performance.
A deeper analysis of the root cause differences in performance between dividend ETFs and dividend mutual funds is beyond the scope of this article, but a cursory review of the fund strategies indicate that ETFs are much more inclined to use passive investment strategies, more oriented around dividend performance, and place less weight on capital appreciation.
- If you are considering investing in a dividend investment portfolio, it is important to measure performance in a way that is best aligned with your objectives as an investor. For the conservative, income-oriented investor, this means moving away from a single measure exclusively focused on total return (as Barron's did), and instead focussing on the level of dividend income, its volatility, growth, and related fund expenses.
- The conclusions you will reach on dividend fund performance changes dramatically when you shift from a total return to a net income view.