In a three-part blog series that got hundreds of comments, I attempted to articulate the arguments for why dividends are a valid basis upon which to select stocks. In the comments section of each article, Larry Swedroe weighed in on why he disagreed. Larry and I also exchanged a series of emails, and he was kind enough to send me the articles that he felt best made his case, as well as suggesting ways that I could test the meaningfulness of dividends in other ways. The discussion led to my writing an article in Advisor Perspectives titled Understanding the Controversy Over Dividend-Based Investing.
The Advisor Perspectives piece focused primarily on two pieces of evidence.
First, there is a research study titled Global Dividend-Paying Stocks: A Recent History, published as a white paper by DFA. Larry discussed this study in his column, as well as in some of the comments to my blog pieces. In a nutshell, the study looks at global equity portfolios selected from 23 developed economy markets on the basis of dividends, and compares these portfolios to the market portfolio of all stocks and to portfolios of non-dividend stocks over the period from 1991-2012. The portfolios were formed once each year and held for twelve months. The portfolios of dividend stocks had almost identical annual return as the market portfolio, but substantially lower risk. The market portfolio had a compounded annual return of 7.4% per year and the dividend portfolio had a compounded annual return of 7.6% per year. The market portfolio had annualized volatility of 18.4% per year vs. 16.8% for the dividend portfolio. In other words, the dividend portfolio provided the same annual return, with less risk than the market portfolio. The portfolio formed on the basis of non-dividend stocks also had a compound annualized return of 7.6%, but this was accompanied by annualized volatility of 27.2%. Larry expressed to me that he believes that the higher Sharpe ratio (the ratio of return to risk) for the dividend portfolios is simply due to the value tilt of the dividend portfolio. DFA did not perform a factor analysis in their study, nor did they form portfolios on the basis of value measures, such as low P/E or P/B, however, so we cannot really know if this is the case.
The second primary piece of research that I discuss in my recent article is a factor analysis of portfolios of U.S. stocks. These results, put together by Morningstar's Alex Bryan, are generated using a four-factor regression model that maps the returns of a portfolio to four distinct terms:
- Beta / market risk
- Size (small vs. large companies)
- Value (low P/B vs. high P/B)
The historical data for U.S. portfolios formed on the basis of high dividends and the historical values of these factors are all available from Ken French's awesome online data library. For those who are not aware of the history, Ken French and Eugene Fama wrote the papers that laid the intellectual foundation for DFA, and both remain closely affiliated with the firm. Performing a regression of a portfolio's returns on these factors over time allows us to see how much of the return is due to exposure to market risk (Beta), how much is due to a small-cap tilt or large-cap tilt, and how much is due to value (buying stocks that are trading at low price-to-book). The fourth factor measures the degree to which the performance of a portfolio is attributable to momentum effects. Two portfolios with the same factor weight on one of these factors can be said to have the same exposure to the corresponding term.
Using data from the early 1950s until the present, there is data for portfolios formed once per year from stocks in the highest 30% of dividend payers and held for twelve months. There is matching data for each of the four factors above. The regression calculation estimates the coefficient on each of these four terms. Bryan's results intrigued me so much that I ran the numbers myself and we compared notes. Our results matched exactly, as one would expect. I am also grateful to Mr. Bryan for discussing the results with me.
The first finding from the factor analysis of the high-dividend stock portfolios is that Alpha is zero. Alpha is average return that cannot be explained on the basis of the four factors. If Alpha had been positive, the portfolios of dividend stocks could have been said to outperform. If Alpha was negative, this would suggest that the dividend portfolios were somehow sub-optimal. Zero alpha simply means that the dividend portfolio has no consistent additional returns that cannot be explained by the four factors. This is not the end of the story, however.
Bryan also ran the regression on portfolios formed from stocks in the highest 30% of earnings-to-price (the reciprocal of P/E). High E/P portfolios have historically provided higher returns than high dividend portfolios, and this has been used to argue that dividends are a "weak form" of value investing. When the four-factor regression is run, however, the results tell a different story. The high E/P portfolio has a statistically indistinguishable loading on the value term as the high dividend portfolio, but the high E/P portfolio has markedly higher Beta. In other words, the high dividend portfolio has the same exposure to the value factor as the high E/P portfolio, but lower Beta. The dividend portfolio has lower returns than that high E/P portfolio only because it has lower Beta (less market risk).
The way that I look at Bryan's factor analysis is simply that portfolios of dividend-paying stocks are a low-Beta value strategy. That sounds good to me, given the substantial body of research that suggests that low-Beta stocks outperform (see here and here, for example).
Larry's response to these results, if I properly understand them, is that the factor analysis shows that dividends are nothing special because their contributions to return can be explained by the other factors. This is not, however, my main concern - nor is it the concern of other dividend investors. The point of these results is that portfolios formed on the basis of dividend outperform the broader market on a risk-adjusted basis. The factor analysis shows that these portfolios provide all of the exposure to the value factor as high E/P portfolios, with lower Beta. These portfolios are sacrificing nothing, except exposure to certain industries and sectors that pay no dividends. There is no evidence that these portfolios are sacrificing any return (or the Alpha would have been negative in the factor analysis). Dividend portfolios also tend to hold low-Beta stocks, and I find the research on the attractiveness of low Beta compelling. Even if you believe that dividend investors have an irrational preference for a "bird in the hand," as opposed to other value strategies, the factor results suggest that this comes at zero cost.
My take-away from reviewing these new studies on dividend-focused stock selection is that dividend portfolios are low-Beta value portfolios that also provide income. I, and many other investors, favor dividends for their consistency and predictability. From a total return standpoint, the consistency and predictability manifests in the form of lower total volatility and lower Beta. This does not mean that all investors should be income investors. In addition, as Larry has pointed out, there are times when dividend strategies get too trendy and overbought. On the other hand, even if dividends are simply a convenient marker for low-Beta value stocks, that's just fine because there is no penalty (e.g., there was zero Alpha) for using this marker to select stocks.
Disclosure: I have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it. I have no business relationship with any company whose stock is mentioned in this article.