Monday, December 5, 2011

Analyzing Dividend Investing

A very popular method of investing is to build a portfolio of individual stocks with a solid history of dividend payments. Dividend investors tend to believe that this approach will beat index investing, and index investors believe that dividend investors have sub-optimal portfolios.

Although dividend investing tends to be quite passive in the sense that it involves infrequent trading, it is at least a little bit active in the sense that dividend investors choose individual stocks. Any time you choose an active investing strategy, it makes sense to know how much your strategy is likely to underperform the market averages in the event that your strategy is little better than random.

So, if we begin with the premise that index investors are right and that dividend investors are likely to underperform, how much lower are their returns expected to be? To answer this question, I dug up an old paper by Meir Statman: How Many Stocks Make a Diversified Portfolio? In it he reprints a table of the number of stocks in a portfolio and its expected standard deviation from a book by Elton and Gruber. Here are a few lines from this table:

1 stock: 49.236%
10 stocks: 23.932%
20 stocks: 21.677%
All stocks: 19.158%

The volatility drag on compound returns is half the square of the standard deviation. So, here are the volatility drag numbers:

1 stock: 12.12%
10 stocks: 2.86%
20 stocks: 2.35%
All stocks: 1.84%

Compared to index investing where portfolios contain all stocks, smaller portfolios have the following excess volatility penalties on compound returns:

1 stock: 10.16%
10 stocks: 1.02%
20 stocks: 0.51%
All stocks: 0%

So, if a dividend investor owns 10 stocks and they are essentially randomly-selected, the investor is giving up 1.02% per year in returns. At 20 stocks, the penalty is only 0.51%.

Of course, dividend investors don’t believe that their stock choices are random. But even if they are wrong, a portfolio of 20 dividend-paying stocks from a wide range of different industries is only giving up about 0.5% per year compared to index investing (not counting taxes). Giving away returns is never a good idea, but many active investing approaches have higher expected losses to the index than just 0.5%.


  1. Taxes is somewhat important for eligible dividend.

    Tax on dividend needs to be paid annually up front, while capital gain(assuming buy and hold) defers until selling (hopefully decades later). The government has given dividend income a nice tax discount, so that for most tax brackets, it's about comparable tax rate to capital gains tax.

    I'm beginning to think that having to pay the tax up annually is a significant drag.

  2. @P2Sam: At least the tax issue is moot for registered accounts. You're right that paying taxes each year is not good. But, if an index investor needs the money to live, then he would be paying capital gains taxes each year anyway due to having to sell some units. Those who have taxable accounts and don't need the money immediately can use indexing to get the benefit of greater tax deferral than dividend investors. Index investors still collect some dividends (that will be taxed each year) but not as much as a dividend investor. Overall, there is a modest tax edge for the index investor.

  3. Is it plausible that moderately high-yielding stocks have less volatile stock prices because of the attractiveness of a steady payout? I would think this to be true. It is unlikely the stock price of a 5% yielder with a safe payout ratio will crash because the income from the stock will attract investors. This may offset the volatility drag you mention.

    At any rate, I seem to have learned the same thing along the way, the performance of a portfolio of 10 or 20 stocks can be quite similar to a broad index. I bet an investment in Canada's largest 15 companies would emulate the TSX index quite well.

  4. @Gene: Volatility measures are different depending what time period is used. I can believe that day-to-day volatility of dividend stocks may be lower, but for year-to-year volatility, if there are dividend cuts (or a threat of cuts), they will be volatility. I think everything is driven by company profits.

    I think you're right that "an investment in Canada's largest 15 companies would emulate the TSX index quite well." That said, if there is a gap of 0.5% to 1%, I'd rather have that money in my pocket.

  5. I am semi-retired (57 years old, restructured out of a job) who may work again if something interesting turns up. My wife is 55. In our taxable account, are we better off holding dividend ETFs (XDV and/or CDZ) instead of a broader index like XIC or XIU, considering that we would be retiring in next 5-7 years.

  6. @Be'en: The answer to your question depends on your current income level and cash needs. Going on the assumption that the two investment approaches will produce the same total returns, you need to figure out which approach would produce lower taxes (dividend taxes vs. capital gains taxes from selling units). If the dividend approach is the winner, you then need to figure out if the advantage is enough to overcome the higher MERs of the dividend ETFs. In the end I doubt that the difference is large. Of course there may be other factors unique to your situation. Good luck.

  7. Michael,

    You said volatility might take away 0.5% - 1% of market return from a diversified portfolio of stocks. Add to that the 2.8% dividend yield (currently, on the entire TSX), and you can see why dividend investors prefer this strategy. I can accept "paying" 1% to "earn" 2.8%. As you said, I'd rather have the money in my pocket.

  8. @Robert: This estimate of a 0.5% to 1% drag takes into account dividends (for both the dividend stocks and the general index). So, even with your higher dividends you're left underperforming the index. To beat the index over the long term your stocks would have to outperform.

  9. Interesting post Michael.

    Given I'm close to owning 25 different stocks, not all in the same proportions mind you, I have a 'drag' on my portfolio albeit a low one because of this diversification.

    That said, hopefully this active part of my retirement portfolio will have more winners than losers so it should outperform the index and if not, be close, so the 'drag' is not pronounced.

    CDZ 5-yr. performance is about 5%, XIU 5-yr. performance is about 2%. If an investor (like me) holds 10+ stocks of CDZ, as a proxy, the 'drag' could be slim to none as compared to indexes.

    Could. :)

  10. @Mark: If your stock proportions are not to far off being proportional to market cap, then your drag is probably close to 0.5%. I've made the bet that the gap between XIU and CDZ won't persist for the long term. I guess we'll see over time what really happens.

  11. The 0.5% performance drag oddly is very similar (albeit slightly higher) than the average MER in many couch potato portfolios.

    The paper you reference also selects the stocks randomly for the volatility measurements, but part of the argument for dividend paying stocks is that they can be less volatile other components of the index, therefore the volatility penalty would presumably also be lower than the examples presented in the paper. the drag could be lower than the 0.5% and even closer inline to the average MER of an index portfolio.

    I know you don't touch on the actual selection, but this would likely have a much greater influence on the dividend portfolio vs. index portfolio debate.

  12. @Sampson: My own portfolio's average MER is below 0.25% and most dividend investors have fewer than 20 stocks. So, I'd say there is still a respectable gap here. I also have some doubts that dividend stocks would have much lower long-term volatility. The consistent dividend may dampen short-term volatility, but I'm not sure about the long term. Certainly dividend investors tend to pick stocks that have been stable in the past. Whether that stability extends into the future is the interesting question.

  13. Hrm, I may be missing something here, but don't most Index ETFs charge and MER (low though it may be) AND pay a quarterly distrubtion which is taxed in non-reg accounts (higher than dividends since it's from dividends and cap gains)??

    So... if you could emulate the index by picking 20 dividend stocks and the trade comissions are a low enough % of your portfolio, it seems like the way to go.

  14. @Anonymous: It sounds like the scenarios you are comparing are a dividend investor with 20 stocks in a taxable account who spends the dividends (after tax) vs. an index investor using a taxable account who spends the dividends plus some capital (to bring his spending to the same level as the dividend investor). Tax-wise, the index investor has an edge in the early years because the bulk of the sold units would be returned capital. Over the years the index investor would benefit from deferred taxation on the capital gains. A further advantage is that capital gains taxes are lower than dividend taxes for higher-income people. But even for middle income people, I'd give the tax edge to the index investor because of the tax deferral. Lower income people are unlikely to need a taxable account; they could just use a TFSA.

    Ignoring taxes for a moment, we then have to compare MERs to the 0.5% volatility drag on a portfolio of only 20 stocks. It is certainly possible to build a portfolio with an average MER below 0.5%. The only remaining arguments in favour of dividend investing that I can think of right now are that it is more fun and many people believe that dividend stocks will outperform the general market. However, I don't believe they will outperform over the long run.

  15. Michael James,

    Finally someone brought up taxes payable every year which of course is compounding!

    So money (outside an RRSP or TFSA) is lost to taxes payable, plus lost opportunity costs...the money that could have been made had one not paid taxes.

    The end game (at retirement) is how to spend one's money while paying less taxes and not worrying about rates of return.

    I put up a calculator which in a nutshell has Person A with more money but trying to live off the interest and keeping the principal intact. Person B has less money but spends more money and pays less taxes because more money he takes out is his own capital. Over time this money get spent to zero.

    He has life insurance as a back-up (if he dies too soon he enjoys his money, his partner get the insurance proceeds tax free...if he lives for a long time he can go into his cash value insurance or sell other assets like his home tax that point he is at least 85).

    This is sort of a twist of the insured annuity I wrote on Million Dollar Journey which a 65 year old male was getting a high rate of return guaranteed. The problem is of course he gave up control of his money as well.
    With Person A vs Person B, Person B has control of his money and pays less taxes and gets to spend more money in retirement without the worry of the market.



  16. @Brian: This is a little off the topic of this post, but your calculator does not account for the substantial premiums one would have to pay for a $2,000,000 cashable insurance policy. Until you account for all factors, it's difficult to make a meaningful comparison.

  17. Hi Michael James,

    You are right in my calculator you may notice person B has say $500,000 less money than person A.

    The software I use integrates the age and the amount. Another words there is many moving parts. The software on my website is a simple comparison assuming Person B was dumb enough to buy permanent life insurance and less money than person A.

    If you want, e-mail a suggested rate of return you would expect over say 20 -30 years the person age and also assuming the person wants/needs some type of life insurance over those years. I can e-mail you a detailed picture comparing the two with the goal of having the person with life insurance in a better postion to spend and enjoy his money, with less risk and better protection.

    For example a 35 year male buys a whole life insurance an pays $23,620 per year for twenty years or $472,400.

    The death benefit is is at age 60 $2,078,168 and the cash value is $946,362 at age 65 the death benefit is 2,505,797 and the cash value is $1,290,839 at age 85 the death benefit is $4,927,460 and the cash value is $3,785,184

    This continues to grow. Person A has more money, but his goal is to be concerned with rates of return and trying to preserve capital.

    Person B is to spend and enjoy his money over 20 years. He has the insurance to fall back on and can still use the cash value tax free if he is still around after age 85.

    If markets are poor sooner or later Person A has to pull out more money and has no back-up plan.

    The game plan is not to use the insurance as an investment but really it is asset insurance. This allows one to spend and enjoy other capital pay less taxes and not worry about rates of return and with less risk.

    The model most people use is accumulate lots of money (with is good) but then spend as little so it lasts a lifetime. The problem is rate of return are random and usually happen at the worst time in older people's lives.

    Hope this helps.

    Feel free to contact me if you have other questions.


    ps. the reason I brought this up is I have a client who works at BCE and his shares (almost $800,000) is making his taxes increase ever higher because they are used to buy yet more shares.
    This means he has two choices, pay for the taxes out of pocket or sell shares to pay taxes!

  18. Brian: $23,620 per year from age 35 to 55 does not add up to $472,400 at age 65. Based on the 5% real return default value in your calculator, this adds up to $1,272,000. So, it would seem that person B should be starting with only $728,000 to make the comparison fair.

  19. "The volatility drag on compound returns is half the square of the standard deviation"

    Michael, I've come back to it a few times over the past few days, and I just can't wrap my head around this point.

    In the paper you linked to, the lower-volatility index approach achieved higher returns by increasing the equity allocation so that the volatilities matched (using leverage, or in the real world, decreasing the bond allocation). But I'm not sure that's a choice a real-world investor would make. I.e., if they were deciding whether to chose 25 stocks at random or the whole index, I don't think they would choose a different equity allocation.

    So without that factor, I'm not seeing how the increased volatility of a subset of stocks (setting aside the question of whether an investor can/will choose a subset with lower volatility) leads to lower returns.

    My thinking: if you take the universe of 500 stocks, after whatever time period, say 30 years, you have some return. And if you're projecting that return, then you have error bars associated with that: let's say after 30 years I expect to make 500% +/- 100%. If I instead chose to invest in a subset of 25 stocks, assuming the same fees, then I'd expect to also have 500% as my mean return, but with wider error bars, say +/- 200%. I'm not seeing though how I can expect to have less return (0.5%/yr compounded or whatever) by choosing a subset of 25 stocks if the average of all such subsets gets me the market average.

    Am I missing something with rebalancing perhaps? Like that the universe of 500 stocks is not in fact the average of all possible 25-stock combinations?

  20. @Potato: What you're missing is the difference between expected returns and expected compound returns. This is best illustrated with an extreme example. Suppose each year your money triples or goes to zero with 50/50 odds. The expected return each year is +50%. Compounded for 10 years you expect to to your money increase by a factor of (1.5)^10 or about 57. In reality, there is a 1 in 1024 chance of getting 3^10 times your money and a 1023 in 1024 chance of ending up with nothing. If we look at expected compound return, the analysis is very different. Mathematically, you are averaging the logarithms, or equivalently, taking the geometric mean. This analysis says that the expected compound return is -100%. Looking at compound returns gives more likely outcomes and discounts wildly high outcomes.

  21. I'm going to have to go off and do some reading again, but in my ignorance I think I'll do a blog post with a picture or two to see if I can get my thought process across more clearly.

    Basically I'm coming at it from a sampling point of view: the mean of my samples should converge to the mean of my population. One way to reconcile that that popped into my head is that samples taken at the beginning of the period may not be representative of the population by the end of the investment period.

  22. @Potato: What you say about sample mean approaching the population mean works for an arithmetic mean, but not a geometric mean. And it is the geometric mean that gives a more reasonable picture of what is likely to happen to an investor. With the geometric mean, the mean and mode are the same (based on a normal distribution), but the arithmetic mean is above the mode.

  23. To help wrap my head around it and/or demonstrate what I mean, I’ll take a small-scale model that can be worked on with paper & pencil (or you know, Excel).

    Let’s say that I have a universe of 6 stocks, with tickers of A - F. Let’s say I start with $1 in each for my overall index, and then I go into Excel and give each a random return, like so:
    A 1.77
    B 1.28
    C 2.92
    D 0.84
    E 1.86
    F 0.11

    My “index” would have started at $6, and finished at $8.78, with a fair bit of volatility in returns between stocks, and we can also assume over time as that evolved, though that's abstracted away here (I chose a range of 0 - 3 with a flat distribution for the randomize function). Now lets say that I wanted to make an active portfolio with 3 stocks and put $2 in each. I could pick ABC and have ended up at $11.94, or DEF and ended with just $5.62 left. In all, there would be 20 possible 3-stock portfolios to make up from this universe of 6 stocks. They are:

    A B C 11.94
    A B D 7.78
    A B E 9.82
    A B F 6.32
    A C D 11.06
    A C E 13.1
    A C F 9.6
    A D E 8.94
    A D F 5.44
    A E F 7.48
    B C D 10.08
    B C E 12.12
    B C F 8.62
    B D E 7.96
    B D F 4.46
    B E F 6.5
    C D E 11.24
    C D F 7.74
    C E F 9.78
    D E F 5.62

    The [arithmetic] mean of these 20 potential portfolios is $8.78, exactly the same as the overall index. Which is what I was expecting.

    So you're saying that this is my mistake, that the expected return of picking one of those 20 portfolios at random is the geometric mean of the 20 portfolio values? This is the point where my brain explodes.

  24. @Potato: It's not that the expected return is the geometric mean. The idea is that the geometric mean gives a better view of the typical result. It's not easy to see what I'm talking about with just one year. Try running it for 5 or 10 years. You will get a few big returns and many smaller returns with the mode at the geometric mean rather than the arithmetic mean.

    Alternatively, go back to my triple-or-nothing example. The expected return is (1.5)^10, but the actual result is zero 1023 times out of 1024 and 59049 one time out of 1024. For planning purposes, zero is the better assumption.