Wednesday, August 20, 2014

The ABC’s and XYZ’s of Dividend Investing

Let’s face it dividends are no longer a novelty.  They were very much so when we began writing this blog in December of 2004.  Thus, over the last ten years, a time when stocks have experienced both ecstasy and agony, many investors have come to understand the value of dividends as central to their investing strategies.  We have written a number of articles on the ABC’s of dividend investing, so some of this blog will be repetitive, but hang in there, when we get to the XYZ’s you’ll hear some new thoughts and tactics.
ABC’s

The two data points we discovered in the late 1980’s that really got our attention and set us on the road to becoming pure dividend investors were the following:    
  1. Since 1960, dividends have produced nearly 40% of the total return of the Dow Jones Industrial Average (DJIA)
  2. DJIA dividend growth during that time has averaged about 5.7%, and price growth has averaged near 5.9%.
When we first saw these two data points, it created a kind of eureka experience for us.  The fact that dividends represented such a high percentage of DJIA total return said that if we paid attention to dividends they gave us a 40% head start on achieving an attractive long-term return.  In addition, the remarkable similarity between the long-term average DJIA price and dividend growth had all kinds of implications.  When we first saw it, we realized in an instant how important dividend growth was to total return. 

Yet, with the numbers being so similar, there was a humongous thought that seemed almost too good to be true:  if the average annual rates of growth of DJIA dividends and prices were so similar, were they also highly correlated.  If they were, that would help to explain the other 60% of the DJIA’s total return because if the statistical significance of the correlation between dividend growth and price growth were high enough it would give us a tool to predict price growth.  To be more specific, if what is known as the coefficient of determination (R^2) were high enough, then so goes the dividend, so goes the stock.

Minutes after this thought went careening though our minds the answer was appearing on our Excel spreadsheet.  The coefficient of determination, or R^2 was nearly .90 (one is a perfect correlation). Over the last 54 years, dividend growth had been able to predict nearly 90 percent of the annual price growth of the DJIA. This is a fact that few people who consider themselves dividend investors are aware of.

The chart for this statistical study of dividend growth and price growth is shown below.
Dow Jones Industrial Average Price vs. Dividend
Since 1960
But, just as fast as this eureka occurred, another finding diminished our grand hopes.  Even though the cumulative prices of the 30 stocks in the DJIA were highly correlated with their cumulative dividends paid over the years, there were few of the individual members of the Index that had as high a correlation as did the DJIA itself.  In fact, only about eight of the companies had statistically significant correlations between their prices and their dividends.  

With such a short list, we turned to the S&P 500.  The S&P 500 prices and dividends were not nearly as highly correlated as was the case with the DJIA.  Nevertheless, when we ran correlation analyses of the 500 stocks in the Index, we found nearly 125 that had very solid correlations between prices and dividends, and many of these stocks were giving us buy signals on the basis of this simple correlation analysis.

For many years, we puttered around with our price-dividend models with some success.  Then came the mid-1990s and all the correlations fell apart.  Stocks price were growing at a much faster rates than were dividends.  Modern Portfolio Theory was in full swing by then and the wizards of academia were saying that share-buybacks were a more efficient way to reward shareholders than cash dividends because the former would be taxed at the then lower capital gains rate, while the latter would be taxed as ordinary income.

For almost the next six years, our correlation models gathered dust.  They were saying that stocks were significantly over valued as early as 1997, well before the Tech bubble was in full view.  During this time, we relied on our dividend discount models to ascertain buy and sell signals and used predominatly stocks with above average dividend yields and solid dividend growth in our portfolios.   

This first part of the story is what we call the ABC’s of dividend investment.  It identificed the importance of dividends in total return and lead us to stocks with higher than average dividend yields and solid dividend growth. We used these tools successfully for many years; then things changed, and we went looking for another valuation model to pinpoint over and undervalued companies. Between 2000 and the end of 2002, we learned a very hard lesson: estimates of future growth aren't guaranteed, indeed, sometimes projected dividend growth become actual dividend cuts.  
XYZ's
In early 2002, with stocks down from their highs of 2000 nearly 40%, almost by accident, we stumbled again onto our old correlation models.  As we ran correlation analyses of the S&P 500, we found company after company was significantly undervalued, many by as much as 25%.  The big sell off had pushed stocks too low relative to their dividend growth during this time.  As the year wore on, our confidence that our old correlation model was telling us the truth about the market steadily increased.  The main reason was that companies in a wide range of industries, from banking to capital goods to energy, were hiking dividends at an impressive pace.  The market was ignoring the hikes, but these companies were signaling with their dividends that their business was good and improving. In addition, there was a lot of talk about President Bush’s tax break for dividends.

Since we had not used the correlation model in several years, we called in consultants and statistical experts to look at our models and tell us how much we ought to trust the data we were seeing.  All the consultants were quick to warn us that correlation does not equal causation.  In essence, something other than dividend growth might be the real driver of the very high correlations that we were seeing. Interest rates had been falling, so we added interest rates to the formula.  In fact, we added just about everything we could think of to the formulas to ascertain what else might be driving the high correlations which revealed that stocks were undervalued.
In the end, only dividend growth and interest rates remained in the formulas for most stocks, and the story they told was clear: stocks were cheap.  As a witness to how convincing the data were, both consultants became clients and remain so to this day.
The real XYZ story, however, did not become apparent until late 2003.  Gradually our weekly correlation run, which took four and half hours to complete, showed that the companies with the highest correlations between their dividend growth and price growth were not the high-yielding stocks that we had traditionally used, but the companies with lower yields and faster dividend growth.
That trend has continued through the present.  In most cases, even during the 2008-2009 banking crisis, many of these faster growing companies had continued to hike their dividends at double digit rates and their prices have responded in kind. 

The following charts show the price-dividend correlations for four stocks: Praxair, Union Pacific, ONEOK, and Nike. All of these companies with the exception of ONEOK are XYZ stocks. They have yields as low as 1.2% for Nike but have generated double digit dividend growth over the last ten years.  

The first stock is Praxair (PX). PX is an industrial gas company with a remarkable 20 year record of rising dividends and prices. The chart is a scatter plot which shows dividends along the horizontal axis and stock price on the vertical axis. Importantly, look at the mathematical formula at the top of the chart. On the second line is the R^2 of the fit between PX's dividend growth and its price growth. That right! PX's dividend growth has been able to predict 98% of its price growth over the last twenty years. At present, the graph is showing that PX is at about fair value. However, if we use estimated dividend growth for 2015, the stock is a bargain.


Praxair (PX)
      
The next stock is Union Pacific (UNP). UNP is one of the largest railroads in the U.S. It has about a 2% current dividend yield. It has not had quite the twenty-year performance as has PX, but again it has a very high R^2 at .94 and the stock is about fairly valued. As with PX, using next year's numbers makes it a bargain.

Union Pacific (UNP)
Next, for those of you who insist on higher yielding stocks, ONEOK (OKE) may be your choice. It has a 3.4% current dividend yield and has increased its dividend over the last five years at a high double digit rate. OKE is one of the largest pipeline companies in the U.S. Their management team has taken bold steps to solidify their position as an important player in the oil and gas fields. Their R^2 is .98. OKE appears to be about fairly valued, as well.

ONEOK (OKE)
The last stock is Nike (NKE). You may wonder how in the world NKE can be counted at a dividend stocks. Well it may not be an old fashion ABC dividend stock, but it definitely is an XYZ dividend stock. Its current yield is only 1.2% but its dividend has grown at near 14% per year over the last fiver years. Nike has a .95 R^2, but with the price nearly $10 above fair value is not a bargain. Neither does it become a bargain when we add next year's projected dividend hike. However, it then becomes about fairly valued and thus, we continue to hold the stock.

Nike (NKE)

So there you have it. The ABC's and the XYZ's as we describe them in our dividend world. The ABC's might be considered the more traditional higher than average dividend yielders with long histories of hiking their dividends at modest rates. The XYZ stocks are those with average or below average dividend yield but with much higher than average dividend growth. In addition, these stocks are highly correlated to their dividend growth, usually over the long term.


In general, we believe most traditional ABC type stocks are fairly valued, or even a bit overvalued. On the other hand, we are find many bargains in the XYZ stocks. We'll talk more about them in coming blogs. Happy hunting.


Clients and employees of DCM own all of the above listed stocks. Please do not use this information for stock selection purposes. Please consult your financial professional.