Dividend yield strategies in Sweden

(1)

iii

Dividend yield strategies in Sweden

Master’s Thesis 15 credits

Department of Business Studies Uppsala University

Spring Semester of 2018

Date of Submission: 2018-05-29

Erik Chvojka David Lovén

Supervisor: Adri De Ridder

(2)

i

Abstract

This paper examines the dividend yield strategy the "Dogs of the Dow" (DoD) in the Swedish market over different time periods. In particular, we use the period from March 2003 until March 2018 to examine if the strategy beats the Stockholm Stock Exchange Return Index (OMXRI). The DoD strategy outperforms the OMXRI by 765 percent in wealth relative measure over the entire 15-year period of the study and annually by 5.80 percent when comparing geometric mean returns. However, the superior returns lack statistical significance in most time periods suggesting that the outperformance is not strong enough to accept at conventional levels. Interestingly, the outperformance of the strategy is mixed when adjusting for risk using the Sharpe ratio; showing abnormal returns for the entire period, a majority of sub-periods but not for individual years. We find that the returns of the DoD portfolio are mostly captured by the Fama-French three factor model making a case for the efficient market hypothesis. Finally, we examine whether the returns are consistent over different starting points and modifications of the DoD strategy to control for data mining. The different starting points did not show statistically significant abnormal returns for the DoD and confirmed the initial results for the market, value and size premium. The Low 5 modified strategy showed the most promising results illustrating the ease of data mining and setting doubt on the effectiveness of rule-based investment strategies that have little theoretical reasoning.

Keywords: Dogs of the dow, dividend yield investment strategy, portfolio management

(3)

ii

Figure

Figure 1. The cumulative returns of the Swedish DoD and the OMXRI, 2003-2018... 17

Tables

Table I Previous Research ... 9

Table II Constituents in the DoD Portfolios ... 12

Table III The Returns ... 19

Table IV Sharpe Ratio ... 21

Table V Treynor Index ... 23

Table VI Fama-French Three Factor Model ... 25

Table VII Different Starting Months of the Strategy ... 27

Table VIII Modifications of the DoD Strategy ... 28

(4)

1

I. Introduction

Dividend yield strategies have become widely popular in Sweden where headlines like

"The best dividend stocks in the market" (Lundberg, 2018) and books like "Become Rich on Dividends" (Hernhagen, 2016) have been published to the common audience. One of the most popularized dividend yield strategies called "The Dogs of the Dow" (DoD) was first presented by Wall-street analyst John Slatter in 1988 (Dorfman, 1988). By purchasing the 10 highest dividend yielding shares on the Dow Jones Industrial Average index (DJIA) and rebalancing the portfolio on an annual-basis, Slatter showed that this strategy outperformed the DJIA index by 7.59 percent annually for the period 1972-1987. Several studies have since then found empirical evidence confirming and rejecting the strategy (O‟Higgins and Downs, 1991;

Knowles and Petty, 1992; McQueen, Shields, and Thorley, 1997; Domian, Louton, and Mossman, 1998; Hirschey, 2000; Da Silva, 2001; Filbeck and Visscher, 1997; Visscher and Filbeck, 2003; Ap Gwilym, Seaton, and Thomas, 2005; Rinne and Vähämaa, 2011; Qiu, Song, and Hasama, 2013) leading to mixed results.

Several alternative explanations have been proposed to explain these findings. Since the dividend yield serves as a proxy for value (Cochrane, 2001), some authors find that the abnormal returns can be explained by the premium on value stocks. In line with the efficient market theory (Fama, 1970) and the tendency for value stocks to represent firms in some form of financial distress (Fama and French, 1995), the abnormal returns would simply be a compensation for the higher risk inherent in value stocks. Others argue that the abnormal return is due to the market‟s tendency to overreact to both good and bad information as proposed by De Bondt and Thaler (1985). Thus, the temporary underperformance of the stocks would naturally regress to its true value explaining the abnormal returns of the strategy. According to this view the efficient market hypothesis does not always hold. The relation between price and intrinsic value is then a convergence process rather than a static relation (Hirscher, 2000). Alternatively, this anomaly could just be noise in the models that are used to calculate the abnormal expected returns since previous research on the DoD strategy struggles to find statistical significance or at worst a result of data mining (Black, 1993; McQueen and Thorley, 1999; Sullivan, Timmermann and White, 1999).

This study aims to test the DoD strategy in an international context and over a different time period. More precisely, we use data from the Stockholm Stock Exchange (OMXS) in the period ranging from March 2003 until March 2018 to examine how the strategy performs in a different setting and market conditions. Because the methodology of the OMXS30 index

(5)

2 differs from the DJIA, our DoD strategy consist of picking the Top 10 highest dividend yielding stocks from a basket of the Top 100 largest firms by market capitalization on the OMXS and rebalancing every year.

Besides examining the strategy in the Swedish market and adjusting for risk, we investigate whether any extraordinary returns are due to the three factors in the Fama-French (1993) model. It can be argued that the observed market beating returns from the DoD strategy are a result of picking value stocks. Also, due to the broader stock selection criteria in this study, the market beating returns might be explained by a small firm effect (Banz, 1981).

Value would be captured by the Fama-French factor high-minus-low (HML) and size by the factor small minus big (SMB) thus adding a layer of robustness to the results. Also, robustness tests are performed to control for data mining by dividing the data in sub-periods, starting the portfolio in different months and examining extensions of the DoD strategy to check whether the abnormal results are consistent following the methods in McQueen, Shields and Thorley (1997).

This study also updates the fee debate for current times. Trading fees and bid-ask spreads have become significantly lower since their 1976 level of one percent used in previous studies (McQueen, Shields and Thorley, 1997; Hirschey, 2000; Rinne and Vähämaa, 2011). Jones (2002) argues that total costs have consistently halved in price every eight years since 1976 making the round-trip cost less than 0.18 percent in 2002. Additionally, Sweden has introduced an Investment Savings Account (ISK) in 2012 that charges an annual tax on the total value of the portfolio. In such an account, capital gains or dividends are not taxed meaning that investors avoid yearly tax fees from rebalancing the DoD portfolio. Altogether, this makes the DoD strategy cheaper to implement and abnormal returns more pertinent.

The remainder of the paper proceeds as follows. Section II reviews the previous literature.

Section III presents the data and methodology. Section IV discusses and shows the empirical evidence. Section V concludes with final remarks.

(6)

3

II. Literature Review

A. Evidence of the Efficient Market

The efficient market hypothesis (EMH) predicts that capital markets are efficient. This means that the price of a security at any time completely reflects all available information (Fama, 1970). The EMH can be categorized into three forms of efficiency. First, the weak form assumes that all available information consists of historical prices and therefore past prices cannot be used to predict future value. Secondly, the semi-strong form predicts that all public information is incorporated in the price and thus fundamental analysis cannot be used to gain superior returns. Lastly, the strong-form assumes that both public and non-public information is reflected in the security's price and hence, insider information will not lead to abnormal returns (Fama, 1970).

Empirical evidence has been found to support the EMH. Patell and Wolfson (1984) find that the adjustments of share prices occur within 10 minutes of the publication of dividend or earnings, illustrating the speed of the adaptation to new information. Fama and French (1989) provide evidence that the yield-spread between high- and low-grade bonds has greater explanatory power for returns on low-grade bonds than returns on high grade bonds, and greater predictive power for stock returns than for bond returns. This implies that the explanatory power in returns is in fact a risk premium rather than evidence of inefficiencies in the market. Similarly, Fama and French (1988) document that the aggregated return on the market is higher when the dividend yield in the market is high which is concluded to be evidence that a high dividend yield indicates a higher demanded risk premium (because of low expectations of returns) in the market rather than misalignments in the security pricing.

Banz (1981) was the first to document the small-firm effect. The interpretation of the size effect is that smaller firms are traded with a higher risk-premium and thus yield higher returns. The theoretical arguments are as follows: (i) Since fewer analyst monitor small firms, information is less available and makes the investment riskier (Arbel and Strebel, 1983) and (ii) because the shares have less liquidity they tend to be underpriced (Amihud and Mendelson, 1986).

Fama and French (1992) identify the book-to-market as a powerful predictor of security returns. The interpretation is that a high dividend yield or a high book-to-market ratio indicates poor prior performance, a depressed share price, and that the higher return is simply the manifestation of higher associated risk premiums. Further, Fama and French (1992) use

(7)

4 the interpretation of risk premiums associated with size and high book-to-market ratios to extend the original Capital Asset Pricing Model (CAPM) derived by Sharpe (1964), Lintner (1965) and Mossin (1966). The CAPM assumes that the expected return of the portfolio is a positive linear function of the market risk premium. By including the two measured variables size (SMB) and book-to-market equity (HML), Fama and French (1992) suggest that if markets are efficient, the risk of securities are multidimensional. When controlling for size and book-to-market effects, the CAPM beta is deemed to lack explanatory power over projecting average security returns. The Fama-French (1993) three factor model is documented to provide stronger explanatory power in projecting returns for securities, and even though size and the book-to-market ratio are not direct projections of risks they could be seen as proxy for more fundamental determinants of risk. The power of the dividend yield and other value ratios to predict and explain excess return above the risk-free rate has been well documented (Fama and French, 1988; Fama and French 1989; Lamont 1998). Indeed, the regression coefficient of the dividend yield and the coefficient of determination of the regression rise as the time horizon to predict excess returns gets longer.

B. Evidence of Market Inefficiencies

The field of behavioral economics has provided the main arguments and evidence for market inefficiency. Rather than assuming rational behavior, the arguments assume that market prices display an impulsive behavior moving in a random and unpredictable fashion.

These inefficiencies caused by the biases inherent in human behavior are expressed through heuristics used in situations of uncertainty such as predicting future stock returns. For instance, Lakonishok, Shleifer and Vishny (1993) offer an alternative explanation to the size and the value anomalies. They argue that the phenomena of size and the book-to-market ratio are in fact due to systematic errors in forecasts by stock analysts, and hence reflect inefficiencies in the market. The idea is that analysts use past performance to extrapolate earnings too far into the future. Kahneman and Tversky (1972) confirm this claim and add that people tend to put too much weight to recent impressions. Thus, poor prior performance yields underpriced securities and vice versa. La Porta (1996) provides more evidence for the misjudgment of stock analysts and documents that firms with low growth projections perform better than firms with high growth projections.

These findings are consistent with the overreaction hypothesis, or the winner-loser effect, derived by De Bondt and Thaler (1985) suggesting that investors overreact to unexpected

(8)

5 events and thus when market participants recognize their errors, the price of the security reverses. This means that current losers ought to outperform and current winners ought to underperform in the following period as an effect of this overreaction.

C. Dividend Yield Strategies

Dividend yield investment strategies are a widely researched area in the field of finance.

They are often categorized into the broader class of value investment strategies. Some of the common variables that have been found to serve as good proxies for value investments are: a low earnings-price ratio (Basu, 1977), a high dividend yield (Visscher and Filbeck, 2003;

Cochrane, 2001; Hodrick, 1992; Grant, 1995), a low book-to-market ratio (Fama and French, 1992) and a high cash-flow to price-ratio (Chan, Hamao, and Lakonishok, 1991). Stocks that look cheap on the basis of one or more of these ratios have been shown to earn higher returns than "growth" stocks which exhibit the opposite characteristics (Basu, 1977; Estep, Hanson and Johnston, 1983; Fama and French, 1998; Chan, Jegadeesh, and Lakonishok 1995).

Moreover, Fama and French (1988), Hodrick (1992), Grant, (1995) and Litzenberger and Ramaswamy (1979) find a positive relation between expected returns and dividends yields.

Keim (1985) documents that zero dividend yielding firms with a large market capitalization value are the laggers in his dataset. Christie (1990) goes even further and states that zero dividend yielding firms are generating negative excess returns compared to dividend paying firms with similar market capitalization value.

Some evidence has also shown contrary results putting in question the value premium. A weak or no relation between expected return and dividend yield is suggested by Black and Scholes (1974) and Goetzmann and Jorion (1993). Gombola and Liu (1993a) find that high dividend yielding stocks behave differently over time periods. They find a positive relation between the dividend yield and the return during bull markets and a negative relation during bear markets. Moreover, it's also documented that high-yield firms are not homogeneous, and that stocks with a high yield and stable dividend behave differently compared to firms that only offer a high-yield (Gombola and Liu,1993b).

Several studies suggest a U-shaped relationship between the dividend yield and stock return, meaning that both high dividend yielding stocks and stocks without dividend tend to outperform shares that offer a moderate dividend yield (Blume, 1980; Litzenberger and Ramaswamy 1979; Elton, Gruber, and Rentzler, 1983; Keim, 1985). However, the zero dividend yielding firms are identified as being firms with a small market capitalization

(9)

6 suggesting that the excess return is due to the small-firm effect (Elton, Gruber, and Rentzler, 1983; Keim, 1985).

D. Data Mining

Since the field of finance commonly uses ex post data, data mining will be frequently occurring and hard to adjust for. MacKinlay (1995) argues that when one looks at ex post data, one will always be able to identify deviations. And when considering these deviations in a group, they may appear statistically significant even when they are not (Arnott and Asness, 2003; MacKinlay 1995). Arnott and Asness (2003) also state that this problem might arise due to lack of prior economic intuition, or through misinterpreting relationships that are in fact explained by a more intrinsic relationship. Both which will damage the causality of the documented results. However, when trying to adjust for data-snooping, the problem of finding any real deviations arises (MacKinlay, 1995). Iyengar and Greenhouse (1988) warn about the file drawer problem where studies with insignificant results are left unpublished creating an unrepresentative sample when making a meta-analysis leading to wrong conclusions.

Moreover, Black (1993) criticizes the validity of a study where the researcher fails to document everything that has been tested before or puts emphasis on a few significant results.

In other words, a 5 percent significance level means that five out of one hundred cases will prove to be significant only by pure chance and thus, not surprisingly, significance ought to be found if numerous experiments are undertaken. Lastly, Black (1993) points at the problem associated with replication where a researcher chooses to conduct a study the same way others have done using similar data with the knowledge of the patterns prior researchers have found.

Not surprisingly similar results ought to be found confirming the error.

E. Empirical Evidence of the Dogs of the Dow Strategy

The Dogs of the Dow strategy first reached the eye of the public through an article in The Wall Street Journal by Dorfman (1988). It describes how stock analyst John Slatter found a superior investment strategy based on the purchase of the Top 10 highest dividend yielding shares on the DJIA, rebalancing and repeating every year. Slatter provided data for the strategy´s performance from 1972 through 1987 and concluded that it outperformed the DJIA by 7.59 percent on an annual basis. O'Higgins and Downes (1991) and Knowles and Petty (1992) published books confirming these results, with data covering a longer time-period, popularizing the strategy to an even wider audience. Slatter (Dorfman, 1988), O'Higgins and

(10)

7 Downes (1991) and Knowles and Petty (1992) all provide compelling evidence of consistent abnormal returns for the strategy. However, these studies elucidate the absolute return and ignore to incorporate the effects of risk, transaction costs and taxes which are highly relevant when evaluating the performance of an investment strategy.

McQueen, Shields, and Thorley (1997) performed the first academic study of the DoD strategy over the period from 1946 through 1995. In terms of absolute returns, they found that the strategy outperformed the DJIA-index over the 50-year period at a statistically significant level. However, after making the adjustments for risk and transaction costs, they conclude that the superior performance of the strategy is reduced to 0.95 percent yearly period. An amount not large enough to compensate for the additional cost from capital gains and dividend taxes of the DoD strategy.

McQueen and Thorley (1999) follow up on the results by considering the data mining problem (see Black, 1993; Iyengar and Greenhouse, 1988) as a possibility for the abnormal return of the strategy. They examine a self-made variation of the DoD strategy for the period from 1946 until 1972 to find better results than the original DoD, illustrating the fact that data mining is easy and common. They conclude that investors should be particularly skeptical of rule-based strategies that have weak theoretical background because such anomalies ought to disappear sooner or later in efficient markets.

Domian, Louton and Mossman (1998) compared the DoD against the S&P 500 index and finds that the DoD outperforms by 4.8 percent per year over the period from 1965 through 1997. They investigate whether the abnormal return is due to the winner-loser effect by looking at the prior 12-months performance of the DoD firms. The DoD strategy is documented to underperform the S&P 500 index during the prior 12-months aligning the results with the market overreaction hypothesis (De Bondt and Thaler, 1995).

Hirschey (2000) uses a data set from 1961 through 1998, and documents that the strategy outperforms only in absolute terms before adjusting for risk, taxation and transaction costs.

Hirschey (2000) provides a similar explanation as McQueen and Thorley (1999), that the superior returns of the DoD are a consequence of data snooping. This by showing that the DoD strategy outperforms the index during certain time-periods (1970‟s) but underperforms during others (1990‟s) and that over time the strategy has been decreasing in abnormal returns exhibiting the behavior of a dying anomaly,

Several studies have also been conducted outside the U.S. market. Filbeck and Visscher (1997) highlight that the DoD strategy on the British market underperforms the FTSE-100

(11)

8 index in both absolute and risk-adjusted returns between 1985 to 1994 but lack to find statistical significance.

Da Silva (2001) examines the DoD strategy in multiple countries in South America between 1994 and 1999 and finds some inconsistencies among the different countries. The DoD strategy outperforms the benchmark in Argentina, Chile, Colombia, Mexico and Venezuela, but none of the results are statistically significant. On the other hand, the strategy underperformed on the Brazilian market leading to the conclusion that the strategy may indeed add some value in some countries albeit with a degree of caution.

In the Canadian context, Visscher and Filbeck (2003) perform the DoD strategy on the Toronto-35 index from 1988 until 1997 and document significantly higher risk-adjusted returns for the DoD strategy over both the Toronto-35 and TSE-300 indices. Moreover, the abnormal return is also sufficient to survive after adjusting for taxes and transaction costs.

Ap Gwilym, Seaton, and Thomas (2005) test the strategy against the FT-30, FTSE-100, FTSE-250 and the FTSE-300 indices for the period from 1980 through 2001 in the UK. Their results indicate an overperformance in absolute returns, but that the superior return is not enough to compensate for risk, taxes and transaction costs. This confirms the results from Filbeck and Visscher (1997) and concludes that the strategy is not effective in the British context.

Rinne and Vähämaa (2011) test the DoD strategy in the Finnish market from 1988 through 2009 using the OMXH25 index. They conclude that the strategy outperforms after adjusting for risk with statistical significance using bootstrapping methods. However, the abnormal returns vanish after adjusting for taxation and transaction costs. They also find evidence in line with Domian, Louton, and Mossman (1998) that the outperformance aligns with the overreaction hypothesis (De Bondt and Thaler, 1995).

Finally, Qiu, Song, and Hasama (2013), test the DoD strategy in the Japanese context using the NIKKEI 225 index in the period from 1981 to 2010. The DoD strategy is proved to generate high abnormal returns beating the benchmark by 9.46 percent annually. The DoD strategy keeps its advantage after adjusting for risk and transaction costs and the results show statistical significance. Qiu, Song and Hasama (2013) use lower transaction costs than previous studies, assuming transaction costs equivalent to purchasing a NIKKEI 255 tracking index fund. Table I summarizes the previous literature. Most of the studies show that the DoD strategy beats the benchmark in absolute measures. The results are mixed once adjustments for risk, transaction costs and taxes are introduced.

(12)

9

Table I Previous Research Adjustments for Author (s)Market Sample periodDoD (%) Benchmark (%) Difference (%) Measure usedRiskTransaction costsTaxExcess return Slatter (Dorfman 1988) US1972-198818.3910.807.59arithmeticNNN- O‟Higgins and Downes (1991) US1973-199116.6110.436.18arithmeticNNN- Knowles and Petty (1992) US1957-199017.0711.765.31arithmeticNN- McQueen, Shields and Thorley (1997) US1946-199516.7713.713.06arithmeticS.D. & SharpeYYN Filbeck and Visscher (1997) UK1985-19949.4811.58-2.10geometricS.D., Sharpe & Treynor NNN Domian, Louton and Mossman (1998) US1964-1997- - 4.76arithmeticNNN- Hirschey (2000) US1961-199814.1612.391.77arithmeticS.D. YYN Da Silva (2001) Argentina1994-19993.650.483.17geometricS.D. & SharpeYYY Brazil1994-20008.0819.84-11.76geometricS.D. & SharpeYYN Chile1994-200112.281.3310.95geometricS.D. & SharpeYYY Colombia1994-2002-6.40-8.852.45geometricS.D. & SharpeYYY Mexico1994-20023.04-1.784.82geometricS.D. & SharpeYYY Peru1994-20035.543.412.13geometricS.D. & SharpeYYY Venezuela1994-20047.19-2.379.56geometricS.D. & SharpeYYY Visscher and Filbeck (2003) Canada1988-199716.579.956.62arithmeticS.D., Sharpe & Treynor YYY Ap Gwilym, Seaton and Thomas (2005) UK1980-200120.6418.532.11arithmeticS.D.YNN Rinne and Vähämää (2011) Finland1988-200815.5011.004.50arithmeticFama-French, S.D. & SharpeYYN Qui, Son and Hasama (2013) Japan1981-201313.613.979.64arithmeticSharpeYNY

(13)

10

III. Data and Methodology

A. The Data

This study uses Swedish data for the period from 2003 until 2018 to investigate the performance of the DoD strategy. The Swedish market index, the OMXS30, has a different methodology than the DJIA which makes the strategy difficult to translate directly. The OMXS30 is based on the 30 most traded stocks, an index based on liquidity, while the DJIA is based on sector representation and reputation. Hence, the OMXS30 can be overly represented by a sector and entail more risk than the DJIA. In order to have a fairer representation of the total Swedish stock market, this study uses the yearly Top 100 largest firms trading on the Stockholm Stock Exchange ranked by market capitalization, a methodology closer to the S&P 500 index. This list was created from the dataset of all the firms listed on the Stockholm stock exchange for the period of 2003 to 2017 using the Datastream EIKON database. Companies not traded in Swedish krona and firms traded on Xterna-listan, a list consisting of foreign-shares that are cross-listed in Sweden, were removed since the objective is to look at firms that origin from Sweden only.

Moreover, we removed firms with dividend yields above 25 percent under the assumption that they represented firms in high financial distress that would potentially distort the portfolio performance of the Swedish Dogs¹.

Institutional investors who constitute the largest part of the market are more likely than retail investors to treat the special dividend as part of the regular dividend as opposed to a signal of future outperformance (DeAngelo, DeAngelo and Skinner, 2000). As such, no adjustments were made for special dividends. Rather they were counted as part of the regular dividend. The sample in this research shows five occasions where such a dividend made a difference in the portfolio composition².

Sweden allows for dual-class shares which creates duplicates in the Top 100 largest firms list and the Swedish Dogs portfolios since a firm with two different share classes can be included. In the dual-class shares system, shares with higher voting power are used by owners to keep control of the company and are generally less liquid. Since the DoD requires yearly

1Tornet AB excluded for the years 2004 and 2005. (DY of 109 percent and 35 percent). Logica OME excluded 2008 (DY of 344.7 percent)

2Skanska B in 2006, Oriflame Holding in 2012 and Thule Group, Swedish Match and Hexpol B in 2017

(14)

11 rebalancing, liquidity is an important factor to the implementation of the strategy. For this reason, the share class with superior voting rights in each company was removed from the Top 100 list for the entire period.

The OMXRI is used as a benchmark due to the fact that the Top 100 list from which the Swedish Dogs are selected represents a large portion of the Swedish stock market and hence a more relevant benchmark than the OMXS30. The particular starting time-period is chosen because the benchmark OMXRI was created in December 2002. By starting in March 2003, the study gives the index time to stabilize and avoids possible tax effects that can occur by rebalancing in December in line with Visscher and Filbeck (2003). Also, dividends tend to be announced in January in Sweden thus avoiding the price volatility that can distort the returns around announcement day (Lakonishok and Vermaelen, 1986). Moreover, the time-series includes various phases in the business cycle notably the 2003-2007 recovery from the internet bubble burst, the 2007-2009 financial crisis and the following bull market. Monthly data is selected in order to avoid market noise while extracting enough observations to get relevant results.

B. Construction of the Swedish DoD

Firstly, on the last trading day of March each year, we select the Top 100 Swedish firms based on market value. The Top 100 firms are then sorted high-to-low based on the current dividend yield. Thereafter, an equally weighted portfolio of the Top 10 stocks with the highest dividend yield is created. Secondly, the portfolio is held for one year. On the last trading day of March the following year, the portfolio is rebalanced by switching to the new Top 10 dividend yielding stocks. Stocks which are no longer in the Top 10 are sold and replaced, and stocks that are still within the Top 10 are rebalanced so that the portfolio is equally weighted.

Thirdly, the process is repeated on the last trading day of March for the entire period.

All firms included in the Top 10 portfolios are illustrated in Table II. Also shown are the yearly turnover rates of the portfolio compared to the prior year and the number of times an individual share is included in the portfolio during the sample-period. On average the Top 10 portfolio rebalances 5.3 of the firms on an annual basis. This is higher than Filbeck and Visscher (1997) of 5.0, Visscher and Filbeck (2003) of 2.5, and to Rinne and Vähämaa (2011) of 4.9. The average dividend yield of 6.2 percent is near the median of 5.9 percent. Hence, the high dividend yield is the result of the portfolio consisting of firms with high dividend yields as opposed to a few extraordinary outliers. Real Estate and Financial firms represent 30

(15)

12 percent of all constituents over the measured period making the portfolio sensitive to the performance of the housing market, implying a higher risk.

Table II

Constituents in the DoD Portfolios

The table shows the yearly constituents of the DoD portfolio and the total number of times each constituent is included in the portfolio. Moreover, the yearly and average turnovers together with the yearly, average and median dividend yields are displayed.

Company 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Total

D CARNEGIE & CO X X X X X 5

JM X 1

KUNGSLEDEN X X X X X X X 7

BROSTROM X X X 3

RATOS X X X X X X X X X 9

PANDOX X 1

ORESUND INVESTMENT X X X 3

CASTELLUM X X 2

SECO TOOLS X X 2

BILLERUD KORSNAS X X X X 4

LINDEX X X X 3

FABEGE X X X X 4

HOME PROPERTIES X X 2

NCC X X X X X X X X 8

PEAB X X X X X 5

SCANIA X X 2

SKANSKA X X X X 4

HOLMEN X X X X 4

SSAB X 1

AXFOOD X X X X X X 6

VOLVO X X X 3

ICA GRUPPEN X X X X 4

KAPPAHL X X 2

ENIRO X X 2

NORDEA BANK X X X X 4

BOLIDEN X 1

STORA ENSO X X X 3

NOBIA X 1

INDUSTRIVARDEN X X 2

INDUTRADE X 1

ORC GROUP X 1

ASTRAZENECA X X X X X X 6

KLOVERN A X X 2

TELIA COMPANY X X X X X X X X 8

TELE2 X X X X X X X 7

TIETO CORPORATION X X X X X 5

BEIJER ALMA X X X 3

ORIFLAME HOLDING X X X 3

DUNI X X X 3

DIOS FASTIGHETER X 1

ELECTROLUX X 1

MYCRONIC X X 2

KINDRED GROUP X 1

MTG X X 2

ERICSSON X 1

THULE GROUP X 1

RESURS HOLDING X 1

SWEDISH MATCH X 1

HEMFOSA FASTIGHETER X 1

HEXPOL X 1

Total 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10

Turnover 5 6 7 5 4 9 9 5 3 4 4 5 3 5

Average Turnover 5,3

Yearly dividend yield 8,7% 5,1% 4,4% 5,2% 7,4% 9,9% 5,1% 6,5% 6,5% 5,9% 5,3% 5,3% 6,1% 6,0% 5,7%

Average dividend yield 6,2%

Median dividend yield 5,9%

(16)

13

C. Total Returns

To analyze the portfolios, the monthly total return index (RI) of each individual stock is extracted from Datastream. The RI includes capital gains and dividends reinvested in the stock. The monthly returns of the Swedish DoD portfolio are calculated by taking the equally weighted return of the stocks. To calculate the yearly return of the portfolio and the market we use the cumulative return of the monthly data. The geometric mean is used for computing yearly returns in multi-year periods and the entire period since the volatility in returns and the effect of compounding overestimate the arithmetic average.

D. Risk Adjusted Returns

In evaluating the performance of an investment strategy, it is necessary to adjust the returns for the risks associated with the strategy and its benchmark. The DoD strategy might be considered riskier than the comparative benchmark due to lower diversification since only ten securities are held adding individual risk to the portfolio. Also, because the DoD strategy solely looks at the dividend yield, there is a possibility to be overweight in certain sectors during specific periods. Hence, the Sharpe ratio, Treynor index and the Fama-French three factor model are used to test the performance of the Swedish DoD.

D.1. The Sharpe Ratio

The Sharpe ratio computes the excess return per unit of risk measured by the standard deviation (Sharpe, 1966). It is an appropriate measure of performance under the assumption that the Swedish DoD portfolio is not well diversified which is a reasonable assumption if the portfolio is the entire investment. This study uses the average excess monthly returns divided by the standard deviation for the portfolio for the same period. The excess return is computed by subtracting the average risk-free rate defined as the monthly return on the Swedish 1- month treasury bill from the portfolio monthly average returns. The ratios are then multiplied by the square root of 12 to compute the excess yearly return per unit of risk. The Sharpe ratios of the Swedish DoD and the OMXRI are calculated according to the following formula:

̅ ̅ ⁄ √ , (1)

where ̅ represent the monthly return on the portfolio, ̅ is the monthly risk-free rate and is standard deviation of the returns for the portfolio.

(17)

14

D.2. Treynor Index

The Treynor index uses the same analogy as the Sharpe ratio but instead of computing the excess return in units of standard deviation of the portfolio, the Treynor index shows excess return in units of the portfolio beta. The Treynor values relate the portfolio returns to systematic risk under the assumption that the portfolio has diversified away unsystematic risk.

The return is calculated as the average return of the portfolio minus the average risk-free rate divided by the beta of the portfolio (Treynor, 1966). The Treynor index is calculated according to the following formula:

̅ ̅ ⁄ , (2)

where ̅ is the average monthly return on the portfolio, ̅ is the average monthly return on the Swedish 1-month treasury bill, and is equal to the covariance between the portfolio return and the market return divided by the variance of the market return.

D.3. The Fama-French Three Factor Model

We test the performance of the portfolio using the Fama-French three factor model (1993) to see if the excess returns from the investing strategy will be mostly captured by the premiums. The coefficients SMB and HML for the regression were obtained from the Department of Economics at Uppsala University³. The model is calculated using the following formula:

, (3)

where is the expected return on portfolio, is the monthly return on the Swedish 1- month treasury bill, is the abnormal return and is the risk premium on the market. M, SMB and HML are coefficients for the independent variables. The SMB variable is the difference between the returns on small and big firm portfolios after considering the book- to-market ratio. High minus low (HML) uses the same logic as the SML variable but instead of focusing on size, the HML variable capture differences in returns based on book-to-market after controlling for size (Fama and French, 1993).

3 The data of the SML and HML is for the years 2003 through 2016.

(18)

15

E. Transaction Costs and Taxation

Earlier studies (see McQueen, Shields, and Thorley, 1997; Rinne and Vähämaa, 2011) assume a one percent yearly transaction costs including brokerage fees and bid-ask spreads.

According to Jones (2002) the transaction costs for securities have declined from an average of one percent in 1976 to under 0.18 percent in 2002 in the U.S context. Moreover, Jones (2002) argues that total costs have consistently halved in price every eight years since 1976.

This study assumes a similar development in the Swedish context. In the portfolio, the average turnover rate is 5.29 on a yearly basis. This results in an annual turnover rate of 52.9 percent. This generates a yearly transaction cost of 0.095 percent (0.529*0.0018) of the portfolio‟s value. Moreover, the stocks that are retained need to be rebalanced most years. The position in the stocks that generated negative returns are increased and vice versa for the stocks yielding positive returns during each year. The annualized average return for the period is 19.54 percent and the average dividend yield is 6.21 percent giving an average capital gain of 13.33 percent (0.1954-0.0621) per year. If all firms appreciate by the same amount, no rebalancing adjustment is required. On the contrary, the worst-case scenario would lead to an annual rebalancing cost of 13.33 percent of the retained stocks market value. Assuming that capital appreciation is randomly distributed around the long-term mean, half of the remaining stocks would have to be sold to reach equal weight adding an additional 3.1 percent turnover (0.47/2*0.1333), following the method of Rinne and Vähämää (2011). The total cost reaching 0.10 percent for the DoD strategy. In comparison, the costs of holding and retaining an index fund ranges between 0.0 percent and 0.20 percent in Sweden. This suggest that the rebalancing cost of the portfolio and the annual fee for holding an index fund is approximately the same and requires no adjustments on the returns. This is consistent with Qiu, Song, and Hasama (2013) who report that the same level of transaction costs for the DoD strategy and investing in an index fund in the Japanese context.

The taxation of dividends and capital gains have changed in the past decade in Sweden.

Today, investors face the possibility to choose from three different types of accounts which have different tax setups. First, the traditional account offers a tax set up where all realized capital gains and dividends are taxed with 30 percent. In contrast, the other two types of accounts, the Investment Savings Account (ISK) and the Endowment Insurance Account (KF), offer an investor a taxation scheme based on the absolute value of the account. In other words, there are no capital gains tax obligations when rebalancing or when receiving dividend payments. The only difference in taxation between the ISK and KF is that the former

(19)

16 calculates the absolute value on a quarterly basis and the latter on a semi-annual⁴. The ISK account was introduced in 2012 and has since the introduction been the superior account to minimize taxation for a retail investor. We assume that DoD investment strategy and the index benchmark use an ISK account. As follows, adjusting for taxation costs is not necessary.

4The taxation on the ISK is calculated quarterly by multiplying the total account value with the Swedish government borrowing rate (GBR).

The account value for each quarter is then multiplied with the GBR and from this amount 30 percent is taxed. The taxation is calculated using the following formula: = ( ⁄ ) , where is the account value at a particular quarter and is the deposits made to the account during the quarter.

(20)

17

IV. Results

A. Absolute Returns of the Swedish DoD strategy

We compare the cumulative returns from investing SEK1 in the Swedish DoD strategy with an OMXRI buy-and-hold strategy during the period March 2003 until March 2018. The results are illustrated in Figure 1. The terminal value for the Swedish DoD is SEK14.55 while the OMXRI ends with a value of SEK6.90. The holding-period return is 1355 percent for the Swedish DoD and 590 percent for the OMXRI, a difference of 765 percent in wealth relative return. Further, the Swedish DoD value never drops below the OMXRI showing positive preliminary results for the Swedish DoD. The Swedish DoD has a higher standard deviation of 19.65 percent compared with 16.39 percent for the OMXRI for the entire period.

Figure 1. The cumulative returns of the Swedish DoD and the OMXRI, 2003-2018.

We also report the yearly returns, entire period geometric average annual return and 5-year period geometric average annual returns for the DoD versus the benchmark in Table III. The DoD outperforms the OMXRI benchmark consistently during the 15-year period with an annual geometric return of 19.54 percent against 13.75 percent. Measuring the results during individual years yields a more dispersed result where the DoD portfolio outperforms the OMXRI eleven out of fifteen years. For the 5-year holding period the DoD outperforms the OMXRI nine out of eleven times. On absolute returns, the DoD strategy seems to show abnormal returns consistent with previous studies (Dorfman, 1989; O‟Higgins and Downs,

0 2 4 6 8 10 12 14 16

SEK DoD Market

(21)

18 1991; Knowles and Petty, 1992; McQueen, Shields, and Thorley, 1997; Domian, Louton, and Mossman, 1998; Hirschey, 2000; Da Silva, 2001; Visscher and Filbeck, 2003; Ap Gwilym, Seaton, and Thomas, 2005; Rinne and Vähämaa, 2011; Qiu, Song, and Hasama, 2013).

The differences in returns are merely statistically significant for one of the one-year periods (2004-2005) and one of the multi-year holding period (2003-2008) at the two-tailed 5 percent level. Thus, the abnormal results of the Swedish DoD should be interpreted with caution. Interestingly, the 5-year period before the financial crisis has a significant return which might indicate that the global financial crisis (2007-2009) influenced the results.

However, the significance does not return in periods after the financial crisis counter-arguing this point.

The findings also show consistency with Gombola and Liu (1993a) that a dividend yielding strategy seems to outperform the market during bull markets but underperform during bear markets. During the years 2007 to 2009, the Swedish DoD had a geometric average return of -23.29 percent compared to the market‟s -18.10 percent. In the year 2008- 2009, the DoD portfolio underperforms the market by -15.32 percent making it the weakest individual year. All firms in the DoD portfolio have negative returns during this individual year but most notable is the bankruptcy of D. Carnegie & Co. This stands out as the only firm that went bankrupt in the portfolio during the entire investigated time period.

Dividend yield strategies in Sweden