Sunday, March 13, 2016

Are dividend stocks better?

There are continuous debates for and against dividend stocks and dividend growth stocks. I hope this article would settle the debates. If you’re making money with any strategy recently, stick with it. From this test, I conclude that dividend stocks and dividend growth stocks are worse than non-dividend stocks. Read it with an open mind. I accept confirmations and challenges but not “You’re wrong without explanation”.

Do not be biased and data fit to back up your conclusion. Ensure the test can be reproduced with identical results, so there is no cherry picking and no bias. Ensure the number of stocks and the number of tests are large enough so the results are statistically acceptable.

Here is my test procedure. It would be a sample test procedure for other strategies.

A test consists of selecting a number of stocks according to a specific criterion such as the 30 stocks giving top dividends. The performance of the test is defined as the average return of the specific number of stocks (30 in my test) after a period of time (a year in my test).

·         I have four tests for each period of a year: Dividend Stocks, Dividend Growth Stocks, Non-Dividend Stocks and All Stocks. Select the top 30 stocks for each test.  
·         There are 10 tests and the results are averaged. The first test starts in the beginning of a year and end at the end of the year.

·         The last ten years resembles the current market better than older dates. Hence, I start the test on 1-1-2005 and end on 1-1-2015.

·         The start date is Jan. 2 as Jan. 1 is a holiday. In some tests, it is Jan. 3 or Jan. 4 due to weekends. It is the same for end dates. The results are annualized (= Return * 365 /No. of days tested).

·         The data base is S&P 500. Typically they are the stocks of largest companies. It is the All Stocks.

·         Using educated estimates, I add 2% dividend yield to the performance of the S&P500 index, 5% to dividend stocks and 4% to dividend growth stocks. Testing other strategies, dividends may not be as important as these tests.

    Alternatively, I could use ^SP500TR from Yahoo!Finance. I calculated and tested estimates. It would be very time consuming and impossible not using estimates as the dividend yield changes every trade session.

·         Dividend growth stocks have the top dividend yields and dividend growth rate equal to or greater than 10%.

·         Non-dividend stocks are stocks without dividends. Just select 30 of them randomly to be consistent.

·         You can find performance reports on dividend ETFs or funds specialized in dividend stocks, dividend growth stocks or a combination. Compare the results with SPY. Use them to confirm or challenge my test results.

My test has a new set of 30 stocks every year, so an exceedingly good or bad year only affects one test, not all ten tests with the exception of some stocks moving up or down for many years.

I call it window of testing as opposed to what most funds advertise by setting $10,000 or so and let it rises and falls for a long period (say 10 years).

·         Be careful on tests using small stocks that tend to bankrupt more often. The chance of survivor bias would give them better results than the actual results.


The above tests can be reproduced from a historical database if it handles survivor bias the same as mine. It is no cherry picking and I have no bias towards any of the test strategies. The result is for educational purpose only. I am not responsible for any error.

Avg. One-Year Return
Beat All by
Dividend Growth
No Dividend
All stocks

·         From the above table, both dividend stocks and dividend growth stocks do not beat All stocks in this database.

·         Non-dividend stocks beat All stocks in this database by a sizable margin. They represent the companies plowing back their profits to development/research and/or buyback instead of giving dividends. I was surprised by the huge return.

·         You should change your tests according to what you normally do to reflect new trading. For example, you should select 3 stocks only instead of 30 and/or delete foreign countries. However, it would be cherry picking.

·         If your dividend strategy has better return than SPY, do not change your strategy. My tests here are simple without many other filters. Most likely you can improve your returns with better ROE, low Price /Cash Flow, low Debt/Equity…

·         An article (3/1/16) from MarketWatch indicated a different finding than mine; I checked out that DVY (a dividend ETF) did not perform that well. 

Improve the test if more time is available

·         Use 12 months instead of one month for each test. Hence you should have 120 tests less 11 tests due to not enough data for the last year (as of 2/15/2016).

·         Take out the top performer and the bottom performer in the 30 selected stocks.

·         More weight on the last five years than the previous five years. Alternatively, I have one test for each year in the first five years and two tests for each year in the last five years.

Survivor bias

My historical database does not handle the delisted stocks. When I see there are less than 500 stocks in the database for S&P 500, I know they just deleted the stocks taken out from S&P 500 that year. However, later on it includes new stocks added to the S&P 500 index. The adverse impact of bankrupt stocks is far higher than the acquired / merged stocks. For example, the return of the test not including Lehman Brothers makes it look far better. All the tests here look better than they actually are.

The bias can be reduced or even eliminated by:
·         Larger companies as in this test.

·     I use “All” stocks, which consists of the S&P 500 stocks without those that have been delisted.

·      The dividend stocks should have less survivor bias than non-dividend stocks. I did not compensate this in my test results.

·         Actually resolve the bias by including the delisted stocks if your database does not handle it.

From the following table, the impact of 21% (0% means no bias) difference (both including the estimated dividends) is not small. However, all the three strategies have the same impact so it should not be a show stopper.

Avg. Ann. Return
From All
S&P 500


For more of my reasoning, check out the book described next. It has 800 pages (6*9) for $9.99. It could be the best $10 you ever spend.

The above is an abstract from my book "Complete the Art of Investing" which is available from Amazon.

I challenged to have the best-performed article in Seeking Alpha history, an investing site, for recommending 5 or more stocks in one year after the publish date. The concepts for that article are discussed in this book.

No comments:

Post a Comment