Dogs of the Dow (DoD) Theory postulates that one can buy the highest dividend yielding stocks in the Dow Jones Industrial Average (NYSEARCA:DIA) and outperform the Dow over the following year. The theory is based on the notion that the highest yielding dividend stocks will be bid up as investors seek the higher dividends until the yields drop to a more reasonable range. Numerous authors have shown that this doesn’t always work. I reviewed a couple years on it in this article.
However, simply looking at the performance of the DoD stocks from different years is not really that compelling. DoD most likely does not work because there are many other factors at play that the drive the performance of dividend stocks. However, is there any merit to the notion of purchasing stocks when the dividend yields are high and perhaps selling stocks when the dividend yield is low? This is the question that this article will look at for a small selection of stocks and the SPDR S&P 500 Trust ETF (NYSEARCA:SPY):
|Ticker||Name||Price (June 15)||Current Yield|
|SPY||SPDR S&P 500 Trust ETF||127.02||1.73%|
|XOM||Exxon Mobil Corporation||78.66||2.44%|
|GE||General Electric Company||18.39||3.30%|
|KO||Coca-Cola Company (The)||64.97||2.90%|
|F||Ford Motor Company||13.15||NA|
I’ll review historical data to see if there is a correlation between dividend yield and the stock return over the next year. Returns will reflect the benefit of dividends. By fixing the stock and looking at returns over time, this analysis eliminates some of the variability challenges faced by DoD. Furthermore, by looking at an ETF that tracks an index it eliminates all the stock specific issues.
This analysis was completed using historical data from Yahoo Finance for both historical prices and dividends.
For each dividend payment, I looked at the associated yield at the ex-dividend date and then compared to return obtained over the following year (mathematically, I looked at the price 365, 366, or 367 days later). The yield was simply calculated by taking 4x that dividend and dividing by the closing stock price on the ex-dividend date. I used 4x instead of the trailing twelve months of dividends since Yahoo Finance dividends are actually adjusted for the stock splits. The 4x method works in this case since all the securities pay regular quarterly dividends with the slight exception of F which has not paid dividends in a while; however, its historical data is still valid.
By using dividend and split adjusted prices from Yahoo Finance, the simple return calculation would account for the intermediate dividends as well. The following chart shows the historical dividend yield for SPY:
Source: Derived from Yahoo Finance data
The first observation is that the chart is somewhat bumpy. The dividend yields for SPY are typically higher in December when the dividend payment is higher, possibly since some companies only pay a year-end dividend. This probably throws the overall analysis off a little.
The following chart shows the historical yield for XOM which shows a smoother yield as XOM generally pays consistently increasing dividends. This eliminates the higher year end quarterly dividend seen in SPY.
Source: Derived from Yahoo Finance data
The XOM historical dividend yield graph is interesting in that it shows that dividend yields mirror interest rates to some extent. In the early 80s when interest rates were extremely high, dividend yields were also high. This was also true in GE, F and KO. The next question is how do returns compare. The following graph shows the 1 year returns for SPY plotted against the initial dividend yield.
Source: Derived from data from Yahoo Finance.
The red line is the least squares linear regression line. The black oval circles some data points related to returns and yields just prior to the Great Recession. One can argue whether or not these data points or that entire period should be included in this type of analysis. It appears to be some sort of relationship for SPY yields and returns. The graph is sloped in the correct direction.
It turns out the correlation between yields and returns is about 34% which is an R-squared of about 12%. This is a positive correlation, but is an extremely weak one. To put it in comparison, people think of emerging markets as providing good diversification (i.e., not correlated) for SPY with correlation coefficients ranging from 50% to 85%. So a 34% correlation, while it is positive, is nothing to be excited about. If I exclude the period impacted by the Great Recession, the correlation coefficient climbs to 46% which is better, but again not really that great.
The following table shows the tickers and the correlations over the available data periods between dividend yield and the subsequent 1-year returns. SPY showed the strongest correlations.
Correlation Between Dividend Yield and 1 Year Return
|Ticker||Correlation (all history)||Correlation (up to YE 2006)|
Unfortunately, buying and selling stocks based on their dividend yields is probably not a good strategy. This also adds to body of evidence that Dogs of the Dow Theory is not a very good investing strategy. While the concept is not entirely without merit since the correlations are positive, it is simply not strong enough in my view to be followed. Also probably, as one would expect, the concept seems a little better when applied to ETFs representing more than one company.
So what should an investor do with this information? In my view, while it is not a good mechanistic strategy, a dividend yield that is substantially below a historical average for a security would be a warning sign while a dividend yield above its historical average might be a favorable sign. However, in either case additional research would be required to explain why the yield is deviating from normal.
Disclosure: I am long SPY.
Disclaimer: This article is for informational and educational purposes only and shall not be construed to constitute investment advice. Nothing contained herein shall constitute a solicitation, recommendation or endorsement to buy or sell any security.