During the last six years, stocks have been, shall we say, unpredictable: surging bull market in 2007, bear market retreat in 2008-2009, then a powerful bull market over the last three years. As we celebrate the surprisingly good performance of stocks in 2012, we are now being buffeted by the annual year-ahead prognostications of the financial media and Wall Street strategists. Will they be any better at predicting 2013 than they were at predicting 2012? I don't think so.
Indeed, who correctly foresaw the wild ride we have been on over the last six years, ten years, twenty years etc.? Nobody I know. That is the reason I became a dividend investor approximately twenty years ago. It was then that I first learned that dividends had produced nearly 50% of the total return for stocks since the end of World War II. I also discovered that dividend growth was only about one-third as volatile as earnings or prices, and thus were far more predictable than either.
In summary, dividends offer a cash return, generate half of the total return of stocks, and are among the most predictable financial data for many companies because dividends are set by the board of directors.
All of us at Donaldson Capital Management are pleased that so many investors have discovered dividend investing. We think most of them will stick to it, and they will be rewarded for it over the years. I am troubled by one trend I see, however. Too many investors appear to be focusing on the current dividend yield alone. Our research has convinced us that a combination of dividend yield and dividend growth is the best all-season investment strategy.
Recently, I was speaking with my son, Justin, about assisting me with some analysis of the volatile stock markets we have been navigating in recent years. DCM has lots of data analysis equipment, such as Bloomberg and Value-Line, but we do not have what is called a data mining resource. Justin is a co-founder of a machine learning company, Big ML.com, in Corvallis, Oregon. He is a part of group of six tech scientists who are experts in the field of big data, or, as I still like to say, artificial intelligence. Justin always fusses at me when I use that term. He says he works in the field of machine learning.
BigML.com has a very accessible and easy to navigate website that can crunch huge quantities of data and generate a "bottom line," so to speak, on about any data set you can throw at it. A very interesting data mining study they have made is on who survived the Titanic disaster by sex, age, location on the ship. Some of the findings are intuitive. One group who disproportionately survived will surprise you.
Justin took me through a few examples of how to assemble the data I wanted to study and how to build a model at BigML.com. I have been practicing for months and I think I understand what is happening in the data mining process. Let me say I have heard many lectures from Justin about algorithms. They are the mathematical T's in the road that the computer "learns" to navigate by looking at all the data from many different angles all at once. There, let's see how much trouble I get into with him with that explanation.
This week, I did some modeling to determine what fundamental factors have been the best predictors of success for the 500 stocks in the Standard & Poor's Index. The fundamental data I used for each of the 500 companies was the annual growth rates of sales, earnings, and dividends over the last five years. I also included the average annual dividend yield, average annual price to earnings ratio, return on invested capital, and bond ratings. Below is a picture of the model I built using Justin's website. I am not going to try to explain everything that you will see. I will write additional blogs in the future that will help to further explain what is going on in the model.
The model (below), at first glance, looks like an upside down Christmas tree. Indeed, it is called a tree, a decision tree. The bulbs on the tree are called "nodes" and represent a group of companies in the S&P 500 with similar return and fundamental data characteristics.
Let's get to the good stuff. That is the heavy black line connecting the first four balls or bulbs from the blue ball at the top of the model. You might think of this black wavy line connecting the balls or nodes as the road to success. The model is not just picking the stocks that have performed the best over the last 5 years. It is identifying the best performing stocks that can be explained in association with their fundamental data.
On the right side of the model, is a color coded description of the nodes that the black wavy line is passing through, and the numerical levels that the model has identified as a dividing line between the better performing stocks and the less successful stocks.
Please move below the chart for a simple description of the model's determination of what kinds of stocks have been winners and why.
The model starts with 500 companies. The average annual return over the last five years of this group on an unweighted basis is +3.6%. The first dividing line, or node, between better performing stocks and the lesser performing stocks is market capitalization, or how big the company is. The model shows on the right of the chart that the dividing line is roughly at $4.3 billion in market cap. In general, stocks larger than $4.3 billion in market cap performed better than smaller stocks.
The next node is somewhat illogical that it shows up so early. The model is saying that stocks with dividend growth greater than a minus 25.65% did better than stocks with dividend growth higher than that level. That would seem to be a duh, but it is essentially kicking out most of the bank stocks, which were forced to cut their dividends dramatically in 2009 and fell sharply in price. The next node we cross through is 5-year average P/E, and the critical level is 12.80x. That is a slightly lower 5-year average P/E for our winning stocks than that of the S&P 500 as a whole, but not significantly.
At this point, we have eliminated about half of the original 500 stocks, but we don't know much yet about the nature of the winning stocks. All we really know is that big stocks did better than little stocks. The difference in P/E is not material, and the dividend growth node is not helping us at all.
That is about to change. The black, windy road to success just crossed into a node that reveals that stocks with average annual dividend growth of greater than 10.75% were big winners. This dividing line also cuts the number of remaining stocks to only 98.
In one of the most rambunctious times in the memory of most of us, big stocks that increased their dividends at least 10.75% per annum produced an average annual return of 9.83%, nearly 3 times the rate of return for the average stock in the S&P 500.
But wait, there is more. "We'll double your order if you order now." Forgive me. I have watched too much Holiday television. There is more and the defining characteristic of the next node, as shown on the right, is the dividend yield. I did not show the value of the node because it is at first a bit of a surprise. Stocks that have made it through this node have a dividend yield of LESS than 2%. Stocks that successfully passed through all the nodes thus far produced an average annual rate of return of 13.4%. But wait, that's not more, that's less. Yes it is, but that is one of the things we have been saying for the last decade: While dividend yield is important, it is not the main driver of success. Dividend growth is a better predictor of winning stocks if the growth is consistent and persistent. The problem with low yield, high dividend growth stocks is that high earnings growth is a must to continue the high dividend growth. Thus these stocks are much more volatile than higher yielding stocks with lower dividend growth.
There you have it. Over the last 5-6 years, a time as chaotic as any we have seen since the 1930s, the winning recipe in the stock market has been dividend-related. We are not surprised by this. This is the same trend we have observed over the last 20 years. It is also what the data reveals for the 40 previous years.
Importantly, as we said earlier, we discovered many of the dividend principles we still follow today in the early 1990s, a time when dividend investing was out of favor. We went looking for ways to deal with volatility after the crash of 1987. It was not until the early 1990s that we began to understand the importance of dividends. The following paragraph was contained in our quarterly letter mailed to our clients on January 15, 1993. One might say it sounds like we wrote it last week. If the truth be known, we probably wrote it or at least shared it with a client or a prospect today.
We believe that companies with growing dividends will ultimately draw the attention of Wall Street and institutional investors, as well. This is because a rising stream of income in a slow growth, low interest rate environment will, undoubtedly, become more and more valuable, causing prices to rise on such companies. Thus, in our judgment, an investment strategy aimed at increasing income is likely to produce capital growth, as well.