Dividend Investing

Strategy for Long-term Outperformance

First draft: April 22, 2012

Michael Clemens*
michael.clemens@yahoo.dk / mhc@bankinvest.dk

Abstract
One place of particular recent investor interest has been dividend investing. Dividend investing has

historically outperformed both the broader market and value investing, and at the same time shown lower

risk. Dividend investing ‘overlap’ with both value investing and low-volatility investing, but is an

independent investment style of its own. The specific reasons for the outperformance of dividend investing

are a reduction in the agency costs associated with high free cash flows and a probable systematic

mispricing (undervaluation) of high dividend paying stocks. Dividend investing sees fewer years with losses,

which according to prospect theory adds significant utility to investor experience. The outperformance of

high dividend yield stocks has been robust over the 1928-2011 timeframe. Since the underlying reasons for

this outperformance are more behavioral in nature than institutional, chances are that history will repeat

itself.

Key words: Dividend investing, agency cost of free cash flows, systematic mispricing of high dividend paying
stocks, value premium, beta puzzle, investor utility.

* Senior portfolio manager at BankInvest Asset Management A/S, Sundkrogsgade 7, DK-2100 Copenhagen, Denmark.

Electronic
Electroniccopy
copyavailable
availableat:
at:https://ssrn.com/abstract=2056317
http://ssrn.com/abstract=2056317
What is Dividend investing?

Dividend investing is generally seen as an investing strategy according to which investors buy stocks with

above average dividend yields. By implication therefore, dividend investing is a subset of the broader value

investing universe. Often an element of dividend persistence and sustainability are also included in

dividend portfolio selections. Exhibit 1 illustrates how investment managers following a dividend investing

strategy usually find their stocks among those value stocks with a perceived higher quality (in terms of

earnings and dividend sustainability). Hence, traditional value stocks based on reversion-to-the-mean

arguments on profit margins are typically not part of a dividend investing fund. Since dividend investing

also shows lower risk than ‘the market’, there is a certain overlap with ‘low-volatility investing’.

Exhibit 1: Dividend Investing in Morningstar Categories

High Low-Volatility Investing

Quality

Dividend
Investing

Traditional Value Investing Universe

Distressed
Value
Investing
Low
Quality
Deep Value Value Blend Growth Deep Growth

Morningstar Style Universe

Note: Illustrative only.

This paper sets out to explore if dividend investing is a sustainable investment strategy over the long term.

Electronic
Electroniccopy
copyavailable
availableat:
at:https://ssrn.com/abstract=2056317
http://ssrn.com/abstract=2056317
Prior studies on the performance of dividend yield strategies

The literature on the performance of dividend yield as an investment strategy may be divided into two

broad categories: the return prediction literature, which is more academic in nature and the more

practitioner oriented long-term return studies of various dividend strategies. Into the first category fall

studies such as Fama and French [1988], who found an increasing predictive power of dividend yield as the

forecast horizon extended from months to years. More recent work of Ang and Bekaert [2006] among

others concludes that dividend yield is not (anymore) a good predictor of subsequent return. One reason

may be the increased use of share buy-backs as a way to return cash to shareholders, such as proposed by

Robertson and Wright [2006], leaving the dividend yield to explain a lower proportion of total shareholder

cash returns. The second category of literature is more concerned with long-term returns of ‘simple’

investment strategies, such as ‘The Dogs of the Dow’ for example. ‘The Dogs of the Dow’ is a simple

investment strategy according to which, at the start of the year, investor buys the ten stocks in the Dow

Jones index with the highest dividend yield. The next year, the same procedure is repeated etc. Combining

the data from Hirschey *1999+ and O’Higgins *2000+, a time series spanning 38 years is created.

Exhibit 2: ‘Dogs of the Dow’ – US evidence 1961-1998

Ten Highest Dividend Dow Jones Industrial

Yield Stocks Average
Period 1961-1998 1961-1998
Average Annual Return, % p.a. 15.1% 12.5%
Geometic Return, % p.a. 14.0% 11.4%
Annual Volatility 15.6% 15.4%
Return / Risk 0.90 0.74
Beta 0.91 1.00
Jensen's Alpha 3.6% 0.0%
% Positive Years 87% 79%
% of 10 Year Periods With Outperformance 90% 10%
Minimum Annual Return -15.9% -22.3%
Largest Drawdown -15.9% -32.2%

Electronic copy available at: https://ssrn.com/abstract=2056317

Looking at raw or risk adjusted return, the ‘Dogs of the Dow’ dividend investment strategy has been a

success over the 1961-1998 time frame in the US. More interesting, however (given its lower level of

diversification) is that the 10 stock high dividend portfolio actually has seen overall better risk metrics than

the 30 stocks in the Dow Jones index.

Exhibit 3 show international evidence for ‘Dogs of the Dow’ strategy and the general picture is one of

outperformance both in terms of raw returns and in risk-adjusted terms.

Exhibit 3: ‘Dogs of the Dow’ – International experience

Market Period Index Raw Return Risk Adjusted Return Source
Australia 2000 - 2006 ASX 50 Higher Higher Alles and Sheng [2008]
Canada 1987-1997 TSE 35 Higher Higher Visscher and Filbeck [2003]
Canada 1987-1997 TSE 300 Higher Higher Visscher and Filbeck [2003]
China 1994-2009 A shares Higher Higher Wang et al [2011]
Finland 1988-2008 OMSH25 Higher Higher Rinne and Vähämaa [2001]
Germany 1961-1998 DAX Higher Higher Kotkamp and Otte [2001]
Germany 2001-2010 DAX Higher Higher Nilsson [2011]
India 1996-1999 BSE-30 Lower Lower Sahu [2001]
Latin America 1994-1999 - Higher Higher Da Silva [2001]
- Argentina 1994-1999 Local Higher Higher Da Silva [2001]
- Brazil 1994-1999 Local Lower Lower Da Silva [2001]
- Chile 1994-1999 Local Higher Higher Da Silva [2001]
- Columbia 1994-1999 Local Higher Higher Da Silva [2001]
- Mexico 1994-1999 Local Higher Higher Da Silva [2001]
- Peru 1994-1999 Local Higher Higher Da Silva [2001]
- Venezuela 1994-1999 Local Higher Higher Da Silva [2001]
Nordic Region 1991-2005 Nordic 75 Higher Higher Dahlstedt and Engellau [2006]
Poland 1997 - 2007 WIG Index Higher Higher Brzeszczynski and Gajdka [2008]
Sweden 1987-2009 OMXS30 Higher Higher Andersson et al [2010]
UK 1984-1994 FTSE 100 Lower Lower Filbeck and Visscher [1997]
UK 1994-2007 FTSE 100 Higher Higher Brzeszczynski et al [2008]
US 1973-1991 Dow Jones Higher Higher O'Higgins [1991]
US 1946-1995 Dow Jones Higher Higher McQueen et al [1997]
US 1964-1997 S&P 499 Higher Higher Domain et al [1998]
US 1964-1986 S&P 500 Higher Higher Domain et al [1998]
US 1987-1997 S&P 500 Lower Lower Domain et al [1998]
US 1961-1998 Dow Jones Higher Higher Hirschey [1999]
US 1973-1998 Dow Jones Higher Higher O'Higgins [2000]
Notes: List is not meant to be exhaustive.
Some surveys covers only a few years and caution should be taken when drawing conclusions on a limited sample size.
Returns are pre transactions costs and taxes.

Electronic copy available at: https://ssrn.com/abstract=2056317

A recurring explanation for the outperformance of the ‘Dogs of the Dow’ strategy is that the universe from

which stocks are selected is the blue chip universe, significantly reducing the risk of buying financially

distressed stocks. Hence, the strategy then becomes a dynamic asset allocation strategy based on

dividend/value investing principles, or as argued by Domain et al [1998] and Rinne and Vähämaa [2011],

effectively a De Bondt and Thaler *1985+ ‘winner-loser’ strategy with a one-year formation period. Not only

is the ‘Dogs of the Dow’ investable (blue chip universe), the strategy usually also has a relatively low

portfolio turnover (annual rebalancing) of between 2,96, 3.83 and 4.5 stocks per annum (McQueen et al

[1997], Alles and Sheng [2008] and Brzeszczynski et al [2008]). Some caveats before getting too excited.

Firstly, while the strategy is clearly investable, it may not be representative of the typical mandate of

professional asset managers given the relatively narrow universe of companies/industries in the DJIA.

Secondly, critical voices such as Hirschey [1999] claim that the apparent success of the strategy may be due

to data snooping and/or simple data errors in some of the surveys. While Hirschey [1999] do find a higher

gross return and a higher risk-adjusted gross return for the strategy, he nevertheless ends up calling it a

‘myth’, once transaction costs and taxes are taken into account. However, the fact that the strategy has

been replicated with success outside the DJIA/US, however, does add to the credibility of the strategy.

Why dividend investing outperforms

Dividend investing is a unique style investment strategy in its own right. Certain overlap with traditional

value investing and low-risk investing, however broadens the ‘sources’ of long-term outperformance.

The existence of the value premium

While known for many years, see for example Nicholson [1960] and Basu [1977], it was not until Fama and

French [1992] it became widespread accepted that ‘value’ outperforms ‘growth’ and therefore also ‘the

market’. Without going into detail on the vast literature on the origin of the value premium, the literature

boils down into three categories:

Electronic copy available at: https://ssrn.com/abstract=2056317

 Data snooping argument. This argument stem from the early responses to Fama and French [1992],

among which we find Black [1993], who argues that the value premium may be specific to their

particular data set and hence a result of data snooping or a spurious relationship. Subsequent

research which used other datasets from other markets and other periods, have proven this

argument wrong.

 Higher risk argument. This argument is attributable to Fama and French [1993] and claims that

value stocks are more risky than other stocks and that this particular risk element is not captured

by the traditional capital asset pricing model. Hence the development of the F&F 3-factor model.

Much research, however, has provided evidence against the higher risk argument. Fuller et al

[1993] found that the value portfolio in their data had a beta of 1.00, while the growth portfolio

had a beta of 1.08. Risager [2005], examining the Danish stock market did find higher volatility for

value stocks than for growth stocks, but this was entirely due to a higher upside volatility, while

downside volatility was lower for value than for growth. Andrikopoulos and Daynes [2004]

examining UK stocks in the 1988-2002 period conclude “…value is less risky than growth, using the

standard risk measurers beta and standard deviation”. Finally, Chan and Lakonishok [2004] show

that in the period 1979 to 2002, the volatility of the Russel 2000/3000 Growth indices were

markedly higher than the volatility of the Russel 2000/3000 Value indices, at the same time value

outperformed growth by 6% p.a. in the period.

 Behavioural biases argument. The main article in this line of argument is Lakonishok, Schleifer and

Vishny [1994] and later updated by Chan and Lakonishok [2004]. The argument generally goes that

investors naively extrapolate recent patterns in earnings and overlook reversion to the mean

tendencies (representativeness). This tendency will drive the price of growth stocks up way too

high and imply subsequent underperformance. Vice versa when a company disappoints, investors

will naively extrapolate the negative pattern trend in earnings and overlook the mean reversion

Electronic copy available at: https://ssrn.com/abstract=2056317

 tendency. This will drive the price of value stocks way too low and imply subsequent

outperformance. The LSV [1994] arguments can trace their roots back to the winner-looser effect

described in De Bondt and Thaler [1985]. Jeremy Grantham of GMO (Grantham [2012]) argue that

career risk in the investment management industry creates herding and momentum effects, which

drives the excess volatility in stock prices compared to underlying fundamentals (cf. also Shiller

[1981+). It is this excess volatility that creates the ‘value opportunities’ (however defined) in the

market.

While there is hardly disagreement that over longer periods, value has historically outperformed growth,

the reasons for this outperformance are less clear. Evidence for the ’higher risk equals higher return’

argument is at best mixed, which leaves the behavioral argument of naïve extrapolation / winner-looser

effect as the main argument.

The beta puzzle

The beta puzzle is the term used for the empirical observation that low-beta/low-volatility portfolios

outperform high-beta portfolios. The literature on the beta puzzle dates back to the late 1960’s and early

1970’s with contributions from Black et al [1972], among others. The early studies focused on whether or

not beta was able to predict differences in return such as predicted by the capital asset pricing model.

Here, the focus is more on the underlying reasons (Cause + Transmission Mechanism in Exhibit 4) for the

puzzle.

Exhibit 4: Beta puzzle: Causes, Transmission Mechanism and Effect

'Cause' 'Transmission Mechanism' 'Effect'
Leverage constrains, Behavioural Time-varying beta's Long-term outperformance of
biases, Limits to arbitrage low beta stocks

Observing how/when the outperformance of low beta stocks is formed, Sefton et al [2011] argue that low-

risk portfolios have time-varying betas. In up-markets (unstressed markets), betas are less dispersed and

Electronic copy available at: https://ssrn.com/abstract=2056317

low-beta portfolios does not underperform markedly. However, in down-markets (stressed markets), beta

dispersion becomes wider and low-beta portfolios outperforms markedly. Effectively this makes a low-beta

portfolio a style-timing strategy and explains the long-term outperformance. As mentioned, this is

probably not an underlying cause but rather what is called ‘Transmission Mechanism’ in Exhibit 4. Going

through the recent literature, three underlying ‘causes’ for the beta puzzle seems to emerge: Leverage

constrains, Behavioral biases and Limits to arbitrage.

 Leverage constrains. This ‘classical’ argument assumes that market participant face various

leverage constrains. For the long-only investors, high-beta stocks provide an alternative way (to

leverage) to add market exposure to the portfolio. This will result in the price of high beta stocks

being bid up above the equilibrium otherwise assumed in a non-constrained world. Hence, in a

capital asset pricing model (beta) perspective, the expected return of high-beta stocks becomes

lower. Cowan and Wilderman [2011] takes the leverage argument a bit further and argue that

buying high beta stocks is buying upside leverage but with downside protection (as compared to

simple leverage) – like buying the market and a call option. The call option usually comes at a cost,

which according to their arguments is ‘transferred’ to low-beta stocks which then ‘collect’ the

premium. Hence the long-term performance difference between high-beta and low-beta stocks.

 Behavioral biases. The behavioral biases come in different forms and shapes. Baker, Bradley and

Wurgler (2011) argue that the preference for lotto (which bid up prices for high risk stocks, leading

to subsequent underperformance), representativeness (using companies such as Microsoft and

Google as representatives for high tech stocks, ignoring the low base rate of success) and

overconfidence (overconfident investors are more likely to disagree; and assuming that high-beta

stocks involve larger uncertainty than low-beta stocks, overconfident investors typically make

larger errors with high-beta stocks than with low beta stocks), all contribute to the beta puzzle.

Grice (2012) provides an elegant argumentation for the preference for lotto bias based on insight

from cognitive psychology. Behavioral portfolio theory argues that private individuals construct

Electronic copy available at: https://ssrn.com/abstract=2056317

 their portfolios in two layers: Layer one is designed to avoid poverty and layer two is designed to

aim for richness. Assuming like Blitz and Vliet (2007) that individuals in layer one acts rational and

risk-averse (i.e. basically the market portfolio) and that in layer two, risk-seeking behavior is seen,

risky assets will be in excess demand. This is related to the preference for lotto argument but put

into a portfolio context.

 Limits to arbitrage. Blizt and Vliet (2007) and Baker, Bradley and Wurgler (2011) among others

argue that the strong focus on short-term performance, benchmarking and tracking error

measurement makes portfolio managers unwilling to exploit the undervaluation of low-beta stocks

in a ‘CAPM-dominated industry’.

Less agency waste and probably mispricing of high-dividend yield stocks

Besides the arguments attributed to why value stocks and low-risk stocks outperforms over the long-run,

(at least) two dividend-specific arguments explain the long-term outperformance of dividend stocks.

Less value destruction when paying dividends (compared to the alternatives)

Traditional financial theory assume that the market is efficient, that capital structure of the firm is

irrelevant and that the dividend pay-out decision does not matter for stock performance; the latter

argument put forward by Miller and Modigliani [1961]. However the assumptions behind these theories are

as unrealistic today as they have always been. Hence, the theory may in the worst case be misleading. One

of the main pillars of traditional finance theory is the assumption of efficient markets. Even if this was the

case, it is by no means certain the managers of the companies which stocks are listed on ‘the market’ acts

rational, i.e. tries to maximize long-term shareholder wealth. This is why the agency costs of free cash

flows, cf. Jensen [1986] helps explain the outperformance of dividend investing strategies over the longer

term.

Electronic copy available at: https://ssrn.com/abstract=2056317

The basic argument in Jensen [1986] is that the principal-agent conflict is most important in firms with high

free cash flows. Since dividend paying firms tend to be larger, more mature and more cash generative

firms, the agency cost of free cash flow theory is especially relevant for investors following a dividend

investing strategy.

To reduce the risk of managers ‘mismanagement’ of the free cash flow, Jensen argues that increased debt

may reduce this risk (banks will now supervise management and ensure a more rational use of cash flows).

Jensen also argues that debt may substitute for dividends in order to achieve this aim. However, Jensen

seems to ignore the situation where a company is already at its optimal capital structure, and if free cash

flows are still large at this point, debt is no longer a substitute for dividends. Committing managers to a

reasonable recurrent dividend pay-out is therefore the ‘next step’ in reducing the principal-agent conflict in

companies with high free cash flows. This is more in line with the arguments put forward by Easterbrook

[1984]. The final variable in reducing agency costs is (non-trivial) personal equity ownership by managers;

cf. Crutchley-Hansen [1989].

Exhibit 5: Principal-Agent Conflict on Free Cash Flow

Interest payments and debt repayments Banks

The ba nk wi l l moni tor ma na gement, from whi ch
s ha rehol ders wi l l a l s o benefi t

Firm Free Cash Flow Dividends Shareholders

Pa i d di vi dends a re ri s k free for s ha rehol ders

Retained earings / cash flow

Us es : M&A's , s ha re-buy ba cks , NPV<0 projects
Herei n l i es the ri s k to s ha rehol ders

At the optimal capital structure, (net) debt repayments are generally not advisable as this would increase

the overall cost of capital. Hence, in this situation, the use of free cash flows boils down to a choice

between M&A’s, share buy-backs or dividends (or some combination hereof).

10

Electronic copy available at: https://ssrn.com/abstract=2056317

The main problem with M&A’s is that M&A volume is usually highest at the peak of the cycle, where profits

are strong and credit easily available. However, this is usually also the time where stock prices are at their

highest and making an acquisition at the peak of the stock market plus paying a traditional M&A premium

is a recipe for financial disaster. Hence, it is no surprise that most of the literature on M&A’s suggest that

from a financial perspective, most M&A’s fail. Agency issues involve managers expanding firm size just for

the sake of size, as management compensation is often related to a measure of ‘size’.

Share buy-backs are typically driven by high profits and managers optimism about the near/medium-term

future. On an aggregated level, share buy-back volume tends to be highest when profits (and stock prices)

are high. When profits are low, manager optimism is typically also below average, and share buy-back

volumes are scaled back. However, this is probably also the time where stock prices are below their long-

term trend. Hence, in aggregate, the average share buy-back price is made at above average stock prices,

diluting long-term shareholder value. Agency issues may also play a role in companies’ willingness to buy

back stocks at apparently any share price: If managers in some way are rewarded on an EPS basis, meeting

EPS targets at the expense of value creation may become more important for the manager.

Unlike share buy-backs, dividends are more ‘sticky’. Hence, ordinary dividend pay-outs are often lower

when profits (and stock prices) are high. When profits are low(er), dividends show their stickiness as

managers often not adjust dividends to cyclical ups and downs but only to permanent changes in earnings

levels. Hence, dividends are relatively lower and re-investing opportunities back into stocks are also lower

when profits and stock prices are high. On the other hand, dividends are relatively higher and re-investing

opportunities into stocks are also higher when profits and stock prices are low. This ‘optimal timing’

relationship between stock prices and dividend re-investing may be one reason for the outperformance of

various dividend indices (which generally assume re-investing of dividends into the same stock).

11

Electronic copy available at: https://ssrn.com/abstract=2056317

Comparing M&A’s, share buy-backs and dividends payments as alternative use of free cash flows, it seems

that the ‘waste’ or value destruction of shareholder funds is less pronounced when paying dividends as

compared to other uses.

Systematic mispricing of high dividend paying stocks

Traditional valuation ratios make high dividend paying stocks look more expensive than no/low-dividend

paying stocks, although this is an ‘optical illusion’.

A simple example will illustrate this: Assume two identical firms: Firm A with a dividend pay-out of 100% of

profits and Firm B who accumulates cash on the balance sheet (or pays down debt). At T = 0, profits is 100

and P/E is 10.0x for both firms, so the share price of both firms is 1,000. Assuming no underlying growth,

after one year, profits in Firm A is still 100 and P/E is still 10.0x. However, Firm B has earned an after-tax

interest income on the retained earnings/cash flows. Assuming an after-tax interest income of 5.0%, the

net profit of Firm B in year 1 is 105 (100 + 5.0% x 100) and the P/E ratio falls to 9.5x. In year 2, the forward

P/E of firm A is still 10.0x, while the forward P/E of firm B is 9.1x, apparently 9% ‘cheaper’ than Firm A.

Exhibit 6: High Pay-out Firms Have Higher Forward Valuation Multiples

Firm A (100% pay-out) Firm B (0% pay-out)
T=0 T=1 T=2 T=0 T=1 T=2
Net profit 100.0 100.0 100.0 100.0 105.0 110.3
Underlying net profit growth, % 0.0% 0.0% 0.0% 0.0% 0.0% 0.0%
Pay-out ratio 100% 100% 100% 0% 0% 0%
Dividends 100.0 100.0 100.0 0.0 0.0 0.0
Retained earnings 0.0 0.0 0.0 100.0 105.0 110.3
After-tax interest rate 5.0% 5.0% 5.0% 5.0% 5.0% 5.0%
After-tax interest income on retained earnings 0.0 0.0 5.0 5.3
Price @ T = 0 1,000 1,000 1,000 1,000 1,000 1,000
P/E @ T = 0 10.0 10.0 10.0 10.0 9.5 9.1

Not only P/E multiples but also EV/EBITDA and similar multiples are impacted by differences in dividend

pay-out ratios. The general rule is that, the higher the dividend pay-out ratio, the higher the forward

12

Electronic copy available at: https://ssrn.com/abstract=2056317

valuation ratio and vice versa (depending on the interest rate and P/E ratio, share buy-backs may induce

even lower valuation rations than just cash accumulation/debt repayments).

This systematic pattern of high dividend-pay-out firms appearing ‘expensive’ may induce some portfolio

managers to underweight or even avoid high dividend pay-out firms in their portfolios, hence paving the

way for long-term outperformance of this type of stocks.

Empirical section

This section aims to illustrate that, over the long run, dividend investing has outperformed the market, and

at the same time shown lower risk.

The first data set is taken from the database provided and maintained by Professor Kenneth French of

Dartmouth College (henceforth F&F database). The database provides broad coverage of US listed stocks

and has been widely used by academics and no further description of the data is deemed necessary here.

The second data sets used are provided by Bloomberg and are widely used global indices for either ‘the

market’, ‘value’ or ‘dividend investing’.

F&F database

Of the many portfolios in the F&F database, the D/P (dividend / price) portfolios is of interest here. Focus is

on annual returns on the portfolios sorted by quintiles. Recall that the portfolios are rebased annually at

end-June according to D/P based on dividend per share from the previous year and prices at end-June.

Exhibit 7 shows that over the whole 1928-2011 period, the best performing quintiles are the high D/P

quintiles. The best performing quintile is the fourth quintile and not the fifth quintile, suggesting a non-

linear relationship between D/P and (subsequent) return. Note also that the ‘dividend effect’ is investable,

given that the difference in returns between the fourth quintile and the simple average return are highest

for the value weighted portfolios. This is not a trivial issue for practitioners.

13

Electronic copy available at: https://ssrn.com/abstract=2056317

Exhibit 7: Risk and Return Characteristics by D/P Quintile since 1927, based on F&F data set
Portfolios sorted on D / P No Quintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5 Simple Q4 - AVG
Annual data 1928 - 2011 Dividends Lowest D/P Mean D/P Highest D/P Average
Value Weighted Portfolios
Geometric Return, % p.a. 8.1% 8.6% 9.5% 9.4% 11.5% 10.7% 9.6% 1.8%
Annualized Volatility, % p.a. 34.0% 22.9% 19.5% 20.9% 21.6% 24.6% 23.9% -
Return / Risk 0.24 0.38 0.49 0.45 0.53 0.43 0.40 -
Equal Weighted Portfolios
Geometric Return, % p.a. 12.8% 11.7% 12.6% 12.8% 13.5% 12.8% 12.7% 0.9%
Annualized Volatility, % p.a. 46.0% 27.8% 24.4% 24.6% 26.4% 27.1% 26.0% -
Return / Risk 0.28 0.42 0.51 0.52 0.51 0.47 0.49 -
Note: The difference between the Value Weighted and the Equal Weighted returns may in part be due to the so-called ‘small-cap
effect’, which is also an element in the F&F 3-factor model.

Finding the best return for the fourth quintile is specific for the D/P data set. Analyzing other F&F data sets,

both the BE/ME and the E/P data sets have a positive relationship between the valuation variable and the

accumulated total return over the period where data is available. However, it is not unusual to find that the

highest dividend yield strategy does not yield the best total return. Patel et al [2012] analyzing total return

for dividend deciles within the S&P 500 finds that the best performing deciles over the 1980-2011 period

were the 8th and 9th deciles with the 10th decile (highest dividend yield) somewhat lower but still above the

market return. They argue that very high dividend yield is a sign of financial distress, which may explain that

the 10th decile underperforms the 9th and 8th decile.

A possible explanation for the non-linear relationship between D/P and subsequent return may stem from

the so-called dividend signaling hypothesis, according to which managers signal permanent changes in

earnings levels through (lagged) changes in dividends. The seminal work in this field is Linter (1956) but also

M&M [1961] acknowledged that in imperfect markets, share prices may react to changes in dividends. The

empirical work on dividend increases/decreases seem to be in agreement that share prices (excess return)

follow the same direction as the change in dividends, cf. Al-Malkawi et al [2010]. However, the reaction is

asymmetric with a larger negative share price reaction/drift when dividends are lowered than the similar

positive reaction when dividends are increased. Since the risk of dividend decreases is highest in the fifth

D/P quintile, it is possible that the reason the fifth D/P quintile underperforms the fourth D/P quintile is due

to a relatively larger number of dividend decreases in the fifth D/P quintile than in the fourth.

14

Electronic copy available at: https://ssrn.com/abstract=2056317

In terms of volatility risk, there seems to be a tendency for a ‘smile-effect’, i.e. highest risk in the lowest

(incl. no dividends) and highest D/P quintiles. This is not an unusual pattern. Basu [1977] found a similar

‘smile’-effect when investigating beta-risk for his P/E-based portfolios, with the lowest beta also found in

the fourth quintile (second lowest P/E). This may suggest that investors who want exposure to the value

effect (whether P/E or D/P driven), but who also do not want to take the additional risk associated with the

most extreme value stocks may benefit from focusing on those stocks with high, but not highest dividend

yield and/or those stocks with low but not lowest P/E.

In the case of the D/P portfolios investigated here, the risk-return trade off is actually best for the fourth

quintile. Other F&F portfolios sorted on BE/ME or E/P also suggest that the best trade-off between risk and

return is not in the fifth quintile but in the third (BE/ME portfolios) or fourth quintile (E/P portfolios). This

also suggests that investors may benefit from improved risk-return tradeoff by eliminating the most

extreme value stocks/portfolios from one’s investable universe.

To check the robustness of the results, the full 84-year period is divided into three sub-periods, see Exhibit

8. What is interesting is that the fourth D/P quintile is the best performing quintile in two of three sub-

periods for both the value and equal weighted portfolios. In the ‘investable’ value weighted data set, the

best quintile in the last sub-period is the fifth quintile with the fourth quintile as second best. Hence, the

main conclusions from the full-period data set are to a great extend also found in sub-sets of the data

suggesting a high robustness of the results.

Exhibit 8: Return by D/P Quintile. Three sub-periods since 1928, based on F&F data set
Portfolios sorted on D / P No Quintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5 Simple Q4 - AVG
Geometric Return, % p.a. Dividends Lowest D/P Mean D/P Highest D/P Average
Value Weighted Portfolios
1927 - 1955 7.9% 8.6% 8.2% 8.3% 10.7% 9.3% 8.8% 1.8%
1956 - 1983 8.6% 9.0% 9.2% 9.1% 11.9% 10.7% 9.8% 2.1%
1984 - 2011 7.7% 8.3% 11.1% 10.9% 11.9% 12.1% 10.3% 1.5%
Equal Weighted Portfolios
1927 - 1955 15.0% 10.4% 10.6% 9.9% 10.4% 11.2% 11.2% -0.8%
1956 - 1983 14.4% 13.3% 14.4% 15.5% 16.2% 14.4% 14.7% 1.4%
1984 - 2011 9.2% 11.5% 12.7% 13.0% 14.2% 12.8% 12.2% 1.9%

15

Electronic copy available at: https://ssrn.com/abstract=2056317

Widely used global equity indices

This section will review the total return of a global dividend index and compare its return and risk to global

‘market’ and ‘value’ indices. The data is based on the MSCI World Index in USD and provided by Bloomberg.

Data is total return data and the total return definition is net of dividend withholding taxes. The MSCI world

High Dividend Index is designed to reflect the performance of equities (excl. REIT’s) with above average

dividend yields, that are perceived both sustainable and persistent.

In the period from June 1995 to March 2012, the MSCI World High Dividend Yield Index has outperformed

both the broad MSCI World Index and the MSCI World Value Index. This outperformance of the MSCI World

High Dividend Yield Index has occurred with a lower volatility and lower beta over the period, suggesting

that dividend investing is a low-risk investment strategy.

Exhibit 9: Return and Volatility since June 1995, MSCI World indices, Bloomberg data
MSCI World High MSCI World Index MSCI World Value
Dividend Yield Index Index
Full period
Start date 30/06/1995 30/06/1995 30/06/1995
End date 30/03/2012 30/03/2012 30/03/2012
# of months data 201 201 201
Geometric Return, % p.a. 8.1% 5.9% 6.0%
Beta 0.90 1.00 0.98
Jensens's Alpha 2.8% 0.0% 0.3%
Volatility, % 15.7% 16.1% 16.3%
Return / Volatility 0.52 0.36 0.37
First sub-period
Start date 30/06/1995 30/06/1995 30/06/1995
End date 31/10/2003 31/10/2003 31/10/2003
# of months data 100 100 100
Geometric Return, % p.a. 9.7% 5.9% 6.6%
Second sub-period
Start date 28/11/2003 28/11/2003 28/11/2003
End date 30/03/2012 30/03/2012 30/03/2012
# of months data 100 100 100
Geometric Return, % p.a. 6.2% 5.6% 5.3%
Note: One month is skipped between the two sub-periods in order to make them equally long.

16

Electronic copy available at: https://ssrn.com/abstract=2056317

To test the robustness of the results, the data set is divided into two sub-periods. In both sub-periods, the

dividend index outperformed, although more so in the first sub-period than in the second. The

outperformance of the dividend index, even net of withholding taxes is even more remarkable considering

that for some investors in some countries in some periods, capital gains may have been treated more tax

favorably than dividends. If this indeed has been the case historically, it only reinforces the argument that

dividends destroys less value on average than M&A’s and share buy-backs.

Other risk and utility considerations

Portfolios based on dividend investing principles have a lower risk of experiencing years with negative total

returns. The theoretical argument comes easy: A higher starting point (historical outperformance) and a

lower volatility imply a lower risk of an observation below zero.

Exhibit 10: Total Return Probability Functions

Probability

Dividend Index

Market

<-- Loss Gain --> Total Return

Note: Illustrative only.

This pattern is also found in the empirical data. In the 16 years from 1996 to 2011 (both inclusive), the MSCI

World High Dividend Index has shown positive total return in 13 years (81%), as compared to 11 years

(69%) for the broad market index (MSCI World) and 12 years (75%) for the value index.

17

Electronic copy available at: https://ssrn.com/abstract=2056317

From prospect theory, we know that a loss is perceived with around twice the negative utility as a similar

gain. Assuming a utility of +1 in years with a positive total return and -2 in years with a negative total

return, the long-term utility for investors is up to seven times higher when investing in dividend portfolios

than investing in a broad ‘market portfolio’. Traditional value portfolios are found somewhere in between.

Exhibit 11: Total Investor Utility from Different Investment Strategies

Investor Utility MSCI World 'Dividend' 'Value' MSCI World 'Dividend' 'Value'
Prospect Theory Utility Weights Traditional' Utility Weights
% time with Total Return > 0 69% 81% 75% 69% 81% 75%
% time with Total Return < 0 31% 19% 25% 31% 19% 25%
Gain Utility 1 1 1 1 1 1
Loss Utility -2 -2 -2 -1 -1 -1

Total Utility 0.06 0.44 0.25 0.38 0.63 0.50

Total Utility Index 100% 700% 400% 100% 167% 133%

Assuming instead traditional symmetric utility weights of +1 for gains and -1 for losses, dividend investing

again scores best in terms of relative investor utility.

In the F&F D/P data set, the fourth quintile had 80% of years with positive total return vs. 75% for the

‘average’ quintile (value weighted portfolios to mirror the value weighted MSCI indices). Figures pretty

much in line with the shorter history of the MSCI World indices. Hence, the higher investor utility claim is

not without long-term merit. Indeed, John D. Rockefeller has been quoted of saying:”Do you know the only

thing that gives me pleasure? It’s to see my dividends coming in”. In this aspect, he was also ahead of his

time.

Conclusion

Based on the analysis in this paper, the following general conclusion may be drawn on dividend investing:

 Being unique in it’s own right, dividend investing overlap with value investing and low-risk investing.

 High dividend yield stocks tend to outperform low dividend yield stocks over the long-run.

18

Electronic copy available at: https://ssrn.com/abstract=2056317

 Within the F&F D/P portfolios, the best trade-off between risk and return is not found among the stocks

with the highest dividend yield (fifth quintile) but rather in stocks with high, but not highest dividend

yield (fourth quintile). This suggests that extreme dividend yield stocks entail some risks that may be

avoided by focusing instead not only on dividend yield but also on dividend persistence and dividend

sustainability.

 The MSCI World High Dividend Yield has outperformed both the MSCI World Index and the MSCI World

Value Index in the period since 1995, and exhibited lower risk.

 Two dividend specific explanations appear plausible to explain the outperformance of dividend investing

strategies: Lower agency costs of free cash flows and a probable systematic mis-pricing (undervaluation)

of high dividend paying stocks.

 Compared to the broad market and the value index, The MSCI World High Dividend Yield Index has seen

fewer years with negative total returns, adding significantly to investor perceived utility.

 The results seem to be robust to various investing horizons, starting dates and ending dates.

While academics have called the dividends decision both irrelevant or a puzzle, the empirical data speaks

for itself: It works. Given that most of the dividend irrelevance arguments are based on un-realistic

assumptions and given that many of the arguments in favor of a generous, but not too generous dividend

pay-out are behavioral in nature, chances are that it will continue to work.

19

Electronic copy available at: https://ssrn.com/abstract=2056317

References
Alles, Lakshman and Y. T. Shen.”Dogs of the Dow’ Down Under”. The Finsia Journal of Applied Finance,
Issue 3 (2008), pp. 30-38.

Al-Malkawi, Husam-Aldin, M. Rafferty and R. Pillai. ”Dividend Policy: A Review of Theories and Empirical
Evidence”. International Bulletin of Business Administration, Issue 9 (2010), pp. 171-200.

Andersson ,Magnus and E. Hall, P. Örn. “The Dividend Effect: A Study of the OMXS30”. Bachelor Thesis,
School of Economics and Management, Lund University (2010)

Andrikopoulos, Panagiotis and A. Daynes. ”An Investigation of the Value Premium in the UK Stock Market
1988-2002: Some New Evidence”. University of Porthmouth, Working Paper, 2004.

Ang, Andrew and G. Bekaert. “Stock Return Predictability: Is It There?”. Columbia University and NBER,
Working Paper (24 February 2006).

Baker, Malcolm, B. Bradley and J. Wurgler. ”Benchmarks as Limits to Arbitrage: Understanding the Low-
Volatility Anomaly”. Financial Analyst Journal, Vol. 67, No. 1 (2011), pp. 40-54.

Basu, Sanjoy. ”Investment Performance of Common Stocks in Relation to Their Price-Earnings Ratios: A Test
of the Efficient Market Hypothesis”. The Journal of Finance, Vol. 32, No. 3 (June 1977), pp. 663-682.

Black, Fischer, M. C. Jensen and M. Scholes. ”The Capital Asset Pricing Model: Some Empirical Tests”. In
Michael C. Jensen’s (ed.) “Studies in the Theory of Capital Markets”, Praeger Publishers Inc., 1972.

Black, Fischer. ”Beta and Return – Announcement of the ‘death’ of beta seem premature”. The Journal of
Portfolio Management, Fall 1993, Vol. 20, No. 1 (1993), pp. 8-18.

Blitz, David C., P. van Vliet. ”The Volatility Effect: Lower Risk without Lower Return”. Erasmus University,
Working Paper, 2007.

Brzeszczynski, Janusz and J. Gajdka.”Performance of high dividend yield investment strategy on the Polish
Stock Market 1997-2007”. Investment Management and Financial Innovations, Vol. 5, Issue 2 (2008), pp.
86-92.

Brzeszczynski, Janusz and K. Archibald, J. Gadjka, J. Brzeszczynski. “Dividend yield strategy in the British
stock market 1994-2007. Heriot-Watt University, Edinburgh. Working Paper (February 2008).

Chan, Louis K. C. and Josef Lakonishok. ”Value and Growth Investing: Review and Update”. Financial Analyst
Journal, Vol. 60, No. 1 (January/February 2004), pp. 71-86.

Cowan, David and Sam Wilderman (2011):”Re-Thinking Risk: What the Beta Puzzle Tells Us about
Investing”; GMO White Paper.

Crutchley, Claire E. and R. S. Hansen. ”A Test of Agency Theory of Managerial Ownership, Corporate
Leverage, and Corporate Dividends”. Financial Management, Vol. 18, No. 4 (Winter 1989), pp. 36-46.

20

Electronic copy available at: https://ssrn.com/abstract=2056317

Da Silva, André L.C. ”Empirical Tests of the Dogs of the Dow Strategy in Latin American Stock Markets”.
International Review of Financial Analysis, Vol. 10, Issue 2 (Summer 2001), pp. 187-199.

Dahlstedt, Jakob and O. Engellau. ”High Dividend Yield as Investment Strategy”. Stockholm School of
Economics, Mater Thesis in Finance (May 2006).

De Bondt, Werner F. M. and Richard Thaler (1985):”Does the Stock Market Overreact?”; The Journal of
Finance, Vol. 40, Issue3, July 1985, pp. 793-805

Domain, Dale L., D. A. Louton and C. E. Mossman. “The rise and fall of the “Dogs of the Dow””. Financial
Services Review, Vol. 7 (1998), pp. 145-159.

Easterbrook, Frank H. ”Two Agency Costs Explanations of Dividends”. American Economic Review, Vol. 74,
No. 4 (1984), pp. 650-659.

Fama, Eugene F. and K. R. French. ”Dividend Yields and Expected Stock Returns”. Journal of Financial
Economics, Vol. 22, Issue 1 (October 1988), pp. 3-25.

Fama, Eugene F. and K. R. French. ”The Cross-Section of Expected Returns”. Journal of Finance, Vol. XLVII,
No. 2, (June 1992), pp. 427-465.

Fama, Eugene F. and K. R. French. ”Common Risk Factors in the Returns of Stocks and Bonds”. Journal of
Financial Economics, Vol. 33, No. 1 (February 1993), pp. 3-56.

Filbeck, Gret and S. Visscher. “Dividend Yield Strategies in the British Stock Market”. European Journal of
Finance, Vol. 3, Issue 4 (1997), pp. 277-289.

Frazzini, Andrea and L. Pedersen. ”Betting against Beta”. New York University, Working Paper, 2011.

Fuller, Russell J., L. C. Huberts and M. J. Levinson. ”Return to E/P Strategies, Higgledy-Piggledy Growth,
Analysts’ Forecast Errors, and Omitted Risk Factors”. The Journal of Portfolio Management, Vol. 19 (Winter
1993), pp. 13-24.

Grantham, Jeremy. GMO Quarterly Letter (April 2012).

Grice, Dylan. “Popular Delusions: Why we overpay for excitement (and the secret pleasure of being
boring)”. Societe Generale Cross Asset Research, 4 April 2012.

Hirschey, Mark. “The “Dogs of the Dow” Myth”. School of Business, University of Kansas, Working Paper
(December 1999).

Jensen, Michael C. ”Agency Costs of Free Cash Flow, Corporate Finances and Takeovers”. American
Economic Review, Vol. 76, No. 2 (1986), pp. 323-329.

Kottkamp, Stefan and M. Otte. ”Die langfristige Performance von DAX-Dividendenstrategien”. Kredit und
Kapital (Heft 3/2001).

Lakonishok, Josef, A. Schleifer and R. W. Vishny. “Contrarian Investment, Extrapolation, and Risk”. The
Journal of Finance, Vol. 49, No. 5 (1994), pp. 1541-1578.

21

Electronic copy available at: https://ssrn.com/abstract=2056317

McQueen, Grant and K. Shields, S. R. Thorley. “Does the “Dow-10 Investment Strategy” Beat the Dow
Statistically and Economically?”. Financial Analyst Journal, Vol. 53, No. 4 (July/August 1997), pp. 66-72.

Miller, Merton and F. Modigliani. ”Dividend Policy, Growth, and the Valuation of Shares”. Journal of
Business, Vol. 34 (October 1961), pp. 411-433.

Nicholson, Francis S. ”Price/Earnings Ratios”. Financial Analyst Journal, Vol. 16, No. 4 (1960), pp. 43-45.

Nilsson, Elina. “Empirisk studie av investeringsstrategin “Dogs of the Dow” på den tyska aktiemarknaden”.
Lund University, Master Thesis in Economics (November 2011).

O’Higgins, Michael and J. Downes.”Beating the Dow”. HarperBusiness, HarperCollinsPublishers (1992).

O’Higgins, Michael and J. Downes.”Beating the Dow”. HarperBusiness, HarperCollinsPublishers (2000).

Patel, Pankaj N., S. Yao, R. Carlson, A. Banerij, J. Handelman. “Dividends and Taxes”. Quantitative Research,
Credit Suisse, 01 March 2012.

Rinne, Eemeli and S. Vähämaa. “The ‘Dogs of the Dow’ Strategy Revisited: Finnish Evidence”. The European
Journal of Finance, Vol. 17, Issue 5-6 (2001), pp. 451-469.

Risager, Ole. ”The Value Premium on the Danish Stock Market: 1950-2004”. Working Paper 20-2005,
Department of Economics, Copenhagen Business School, 2005.

Robertson, Donald and S. Wright.”Dividends, Total Cash Flow to Shareholders, and Predictive Return
Regression”. The Review of Economic Statistics, Vol. 88 (February 2006), pp. 91-99.

Sahu, Chinmoy. “Effectivenes of ‘Dogs of the Dow’ Investment Strategy in the Indian Context”. Vikalpa, Vol.
26, No. 1 (January-March 2001), pp. 65-71.

Sefton, James et al.”Why is low-risk investing successful”; UBS Investment Research, Q-Series, 23
September 2011.

Shiller, Robert J. “Do Stock Prices Move Too Much to be Justified by Subsequent Changes in Dividends?”.
American Economic Review, Vol. 71, No. 3 (June 1981), pp. 421-436.

Visscher, Sue and G. Filbeck. ”Dividend-yield strategies in the Canadian stock market”. Financial Analyst
Journal, Vol. 59, Issue 1 (2003), pp. 99-106.

Wang, Carol, J. E. Larsen, M. F. Ainina, M. L. Akhbari and N. Gressis. “The Dogs of the Dow in China”.
International Journal of Business and Social Science, Vol. 2, No 18 (October 2011), pp. 70-80.

Data sources:

The Kenneth R. French data library at Dartmouth College may be fond via the following internet-link:

http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html

22

Electronic copy available at: https://ssrn.com/abstract=2056317

Bloomberg indices used:

MSCI World High Dividend Yield USD Net TR, Ticker:“MHDYWOUN Index”
MSCI Daily TR Net World USD, Ticker:”NDDUWI Index”
MSCI Daily TR Net Value World USD, Ticker: “NDUVWI Index”

23

Electronic copy available at: https://ssrn.com/abstract=2056317