Returns of high dividend yield stocks in the Dutch market

Transcription

1 Master Thesis (20 ects) Returns of high dividend yield stocks in the Dutch market Author: R.C.M. Mosch Student number: Master s in Business Economics, Faculty of Economics and Business, University of Amsterdam Completion date: 16 July 2010 Abstract: The relationship between stock returns and dividend yield is investigated for the Dutch market. We perform conventional tests and regression analyses. We find evidence that high dividend yield stocks achieve statistically excessive returns. However, we would like to control these returns for risk, taxes, and transaction costs. Risk is measured by three different risk measures: Sharpe measure, Treynor measure, and Jensen s alpha. These controlled returns do not achieve excessive returns. Supervisor: Second examiner: dr. J.W.T. Bogers dr. J.E. Ligterink

2 Table of contents 1. Introduction 2 2. Literature review Payout policy The stocks with the highest dividend yield strategy The Efficient Market Hypothesis (EMH) Validity of the high dividend yield strategy 7 3. Methodology Sample period The conventional tests Regression analysis The level of the dividend yield The existence of dividend yield Data Empirical results Statistical results on a single year basis Statistical results on a multiple year basis Economical results The Sharpe ratio The Treynor ratio Jensen s alpha The final economical results corrected for risk, taxes, and transaction costs Economical results for the top-one portfolio and the AEX-index Economical results for the top-five portfolio Economical results for the top-ten portfolio Regression analysis Influence of the level of dividend yield Influence of the existence of dividend yield Summary and conclusions 36 References 39 Appendix 41 1

3 1. Introduction In this thesis we investigate the relationship between dividend yield and stock returns. We are interested in strategies that can beat the market. A general assumption made by other researchers is that the higher the dividend yield, the higher the stock return. In the United States, a strategy called The Dogs of the Dow was developed. This popular strategy is probably one of the most successful strategies ever. The basic variant of this strategy picks, each year, the 10 stocks with the highest dividend yield. The performance of these 10 stocks is compared to a portfolio of all 30 stocks in the Dow Jones Industrial Average (Dow-30). There are several variants of this strategy. For example, you can pick only 5 stocks with the highest dividend yield. Another interesting point is splitting the results into several time periods, for example in bull-market periods and bear-market periods. McQueen et al. (1997) investigated the performance of the 10 stocks with the highest dividend yield over a time period of 50 years. Only stocks from the Dow-30 were chosen. Consequently, their survey is focused on the United States. McQueen et al. (1997) showed that these 10 stocks had a statistically higher mean return than the index. However, this strategy results in higher risk, higher transaction costs, and higher tax payments. After adjusting for risk and transaction costs and allowing for tax disadvantages, the 10 stocks with the highest dividend yield do not outperform the index anymore. The strategy of picking the 10 stocks with the highest dividend yield is applied on several countries throughout the world with different results. Filbeck and Visscher (1997) investigated this strategy for the British market from 1984 till They cannot show significant results of this strategy. However, McQueen et al. (1997) showed that the 10 stocks with the highest dividend yield statistically out perform the index. An explanation of these different conclusions can be found in the way researchers set up their models and the methodologies being used. However, it could also be true that one country is different from the other country. Knowing all these results, we think it is interesting to do research on the Dutch market. The purpose of this thesis is to analyze the stock returns of the stocks with the highest dividend yield in the Dutch stock market. These returns will be compared with the returns of the index. Firstly, we have chosen the Dutch market because there are no articles written on this topic, as far as we know. Furthermore, the AEX-index (which we will use as the benchmark for the Dutch market) contains not only industrial companies. There are many 2

4 financial companies incorporated in this index. This makes it more interesting to investigate. The central question in our thesis is as follows: Do high dividend yield stocks achieve excessive returns in the Dutch market? In Section 2 we present a literature overview. The main articles regarding this topic will be discussed. The methodology will be discussed in Section 3. In this section we explain the setup of the several models that we will use to compare the returns of the stocks with the highest dividend yield and the market. Furthermore, we present an econometric view of these models. Section 4 specifies the data that we want to use and explains how to collect it. Section 5 deals with the empirical results. Finally, a summary including the main conclusions is given in Section Literature review This section contains the relevant literature for our research. Section 2.1 describes the payout policy, focusing on paying dividends. In Section 2.2 several high dividend yield strategies are treated. The main findings from the authors are presented. The critiques on the Efficient Market Hypothesis are explained in Section 2.3. Finally, in Section 2.4, the validity of the high dividend yield strategy is described. 2.1 Payout policy The first question to consider is, why do firms pay dividends? The payout policy is different for every firm. Berk and DeMarzo (2008) describe in their book Corporate Finance the concept of payout policy. When a firm s investments generate free-cash flow, the firm must decide how to use that cash. If the firm has new Net Present Value investment opportunities, it can reinvest the cash and increase the value of the firm. Many young and rapidly growing firms reinvest all of their cash flows in this way. However, mature and profitable firms often find that they generate more cash than they need to fund all of their attractive investment opportunities. If the firm decides to follow the latter approach, it has two choices. According to Berk and DeMarzo (2008), a firm can pay dividend or it can repurchase shares from current owners. By paying dividends, the board authorizes the dividend on the declaration date. The firm will pay the dividend to all shareholders on a specific date, called the record date. The date two days before the record date is called the ex-dividend date; anyone who purchases the stock on 3

5 or after the ex-dividend date will not receive the dividend. Finally, on the payable date the firm mails dividend checks to the registered shareholders. With a share repurchase, the firm uses cash to buy shares of its own outstanding stock. These shares are generally held in the corporate treasury, and they can be resold if the company needs to raise money in the future. The three possible transaction types for a share repurchase are summarized as follows. The first one is an open market repurchase. A firm announces its intention to buy its own shares in the open market, and then proceeds to do so over time like any other investor. The firm may take a year or more to buy the shares, and it is not obligated to repurchase the full amount it originally stated. About 95% of all share repurchases is represented by an open market repurchase. The second one is a tender offer. The firm offers to buy shares at a pre-specified price during a short time period, generally within 20 days. The price is usually set at a premium (typically 10%-20%) to the current market price. The offer often depends on shareholders tendering a sufficient number of shares. If shareholders do not tender enough shares, the firm may cancel the offer and no buyback occurs. The last example of share repurchase is called a targeted repurchase. The purchase price is negotiated directly with the seller. A targeted repurchase may occur if a major shareholder desires to sell a large number of shares but the market for the shares is not sufficiently liquid to sustain such a large sale without severely affecting the prices. Under these circumstances, the shareholder may be willing to sell shares back to the firm at a discount to the current market price. In a perfect capital market the method of payment does not matter. However, paying dividend is more common than share repurchases. Therefore, in this thesis we will focus on the dividend payments by firms. According to Chay and Suh (2009), the cash-flow uncertainty is an important cross-sectional determinant of corporate payout policy. Cash-flow uncertainty, measured by stock return volatility, has a negative impact on the amount of dividends as well as the probability of paying dividends. This could be interesting because stock investors usually do not like high cash-flow uncertainty. Furthermore, Chay and Suh (2009) stated that the impact of cash-flow uncertainty on dividends is generally stronger than the impact of other potential determinants of payout policy, such as the earned/contributed capital mix, agency conflicts, and investment opportunities. 2.2 The stocks with the highest dividend yield strategy McQueen et al. (1997) investigate in their article a strategy that is called the Dow-10 investment strategy, which we explained in the introduction. In their article, the performance 4

6 of this strategy is measured in two ways: statistically and economically. If the performance of the 10 stocks with the highest dividend yield is significantly better than the performance of all the 30 stocks in the Dow Jones Industrial Average (DJIA) you can conclude that the 10 stocks with the highest dividend yield are statistically outperforming. McQueen et al. (1997) have also measured the economic performance in their article by adjusting for higher risk, additional transaction costs, and tax payments. By adjusting for these variables, no significant differences remains. Hence, the 10 stocks with the highest dividend yield do not outperform the market economically. Visscher and Filbeck (2003) apply the strategy described above on the Canadian stock market. Their conclusions are almost similar to the conclusions made by McQueen et al. (1997). The Dow-10 portfolios higher compounded returns were sufficient to compensate for taxes and transaction costs. Furthermore, they showed that the strategy produced higher riskadjusted returns. The Sharpe ratio, which measures total risk, indicates that the 10 stocks with the highest dividend yield results in excessive returns. However, if an investor has other market investments and this strategy is only part of an overall portfolio, the Treynor index, which measures systematic risk, is the appropriate indicator of risk. The 10 stocks with the highest dividend yield show superior performance when adjusted for systematic risk in the Treynor test. Hence, for the Canadian stock market this is a profitable strategy. These two articles thus prove that the dividend yield strategy can achieve excessive returns compared to an index. However, some researchers cannot prove this effect for other stock markets. For example, Filbeck and Visscher (1997) applied this strategy on the British stock market. The Dow-10 portfolio returns exceeded the market returns, on both unadjusted and risk adjusted bases, in only four out of ten years. A possible explanation for the difference in performance between the FTSE-100 (benchmark for the British market) and the DJIA could be the differences between these two indices. Where the DJIA contains only 30 stocks, the FTSE-100 contains 100 stocks. Theoretically, the DJIA contains no financial stocks, but J.P. Morgan and American Express were included in However, the FTSE-100 contains 18 financial stocks, including banks, insurances companies, and life assurance companies. The difference in strategy performance may be partly explained by the extent to which British financial stocks behave differently from the DJIA industrial stocks. Filbeck and Visscher (1997) come up with another difference between the two indices, which may explain the difference in performance. The FTSE-100 is a value-weighted index, whereas the DJIA is a price-weighted index. According to the Dow-10 strategy, the high dividend yield stocks would tend to be 5

7 underpriced. Hence, these stocks are also relatively low priced. As a result, their price movements would have a small effect on the index value. If the high dividend yield stocks in the FTSE-100 include high market value companies, the Dow-10 portfolio stock price movements would also drive the index price movements and no differential return would be evident. 2.3 The Efficient Market Hypothesis (EMH) The Efficient Market Hypothesis states that the expected return of any security should equal its cost of capital, and thus the Net Present Value of trading a security is zero (Berk and DeMarzo, 2008). The EMH is linked with the idea of a random walk, which is defined as a price series where all subsequent price changes represent random departures from previous prices. The logic of the random walk idea is that if the flow of information is unimpeded and information is immediately reflected in stock prices, then tomorrow s price change will reflect only tomorrow s news and will be independent of the price changes today. However, news is unpredictable and thus stock prices must be unpredictable and random. Uninformed investors buying a diversified portfolio should, in theory, obtain the same rate of return as achieved by the experts. A couple of decades ago, the EMH was widely accepted by academic financial economists, for example see the article written by Fama (1970). It was generally assumed that the markets were extremely efficient in reflecting information about individual stocks and about the market as a whole (Malkiel, 2003). It was taken as an assumption that when information arises, the news spreads very quickly and is incorporated into the prices of securities without delay. Thus, neither technical analysis nor fundamental analysis would enable an investor to achieve excess rates of return adjusted for risk. If the EMH is completely true, a high dividend yield strategy, as described in Section 2.2, cannot achieve excessive risk adjusted returns. However, the intellectual dominance of the EMH has become far less accepted nowadays. Many statisticians and financial economists began to believe that stock prices are at least partially predictable. They believe that future stock prices are somewhat predictable on the basis of past stock price patterns as well as certain fundamental valuation methods. Moreover, some economists state that these predictable patterns enable investors to earn excessive risk adjusted returns. Malkiel (2003) concludes that as long as stock markets exist, the collective judgement of investors will sometimes make mistakes. He found that some market participants are demonstrably less than rational. As a result, pricing irregularities and even predictable 6

8 patterns in stock returns can appear over time and even persist for some periods. The market is not always able to correct these mistakes. Moreover, the market cannot be perfectly efficient, or there would be no incentive for professionals to uncover the information that gets so quickly reflected in market prices. Thus, a high dividend yield strategy could theoretically result in excessive risk adjusted returns. 2.4 Validity of the high dividend yield strategy Finally, the key question is why could stocks with a high dividend yield outperform the market? Basically there are two central competing hypotheses. On one hand you have the tax effect hypothesis and on the other hand you have the dividend neutrality hypothesis. Brennan (1970) had originally developed the tax effect hypothesis. He predicts that investors receive higher before-tax, risk adjusted returns on stocks with higher anticipated dividend yields to compensate for the historically higher taxation of dividend income relative to capital gain income. In contrast, Black and Scholes (1974) developed the dividend neutrality hypothesis. This hypothesis states that if investors require higher returns for holding higher yield stocks, corporations would adjust their dividend policy to restrict the quantity of dividends paid, lower their cost of capital, and increase their share price. The theory described in this literature review about high dividend yield strategies will be tested on a sample of Dutch stocks. We describe these tests in the empirical part of the thesis, which will be treated in the next sessions. 3. Methodology In this section the methodology is described. Section 3.1 describes the sample period and why we have chosen for this time frame. Two equations are included to explain exactly how the return of a stock is computed. The research can roughly be split into two parts. Firstly, the conventional tests, like the Student t-test, are used. These tests are described in Section 3.2. The second part contains the regression analysis, which is described in Section 3.3. In both parts we will analyse statistical results as well as economical results. Like we mentioned before, economical results are corrected for additional risk, taxes and transaction costs. 7

9 3.1 Sample period The focus of the research will be on a 10-year time period ranging from 1999 till Monthly data is used for the calculations. Hence, there are 120 observations for each stock. This is a fair number of observations to perform the tests. We have chosen this time period because this is the most recent period and probably the most relevant. Furthermore, this time period captures the dot-com bubble (roughly from 1999 till 2001) and the beginning of the credit crisis (in 2007). For every stock in the AEX-index, the dividend yield is measured using DataStream. The stocks with the highest dividend yield in a given year are picked. Then the return of these stocks in the next year is compared with the return of the whole AEX-index. Hence, at t-1 (a delay of one period) the highest dividend yield stocks are picked and at t the performance of these stocks is measured. All returns are including dividends. The monthly market return and the monthly portfolio return can be calculated as follows: R m,t = IN t + DIV t IN t 1 IN t 1 (1) R p,t = PO t + DIV t PO t 1 PO t 1 Where, in equation (1), R m,t presents the return of the market at time t, IN t presents the index at time t, DIV t presents the received dividend at t, and IN t-1 presents the index at t-1. In equation (2), R p,t presents the return of the portfolio at time t, PO t presents the portfolio of the high dividend yield stocks at t, DIV t presents the received dividend at t, and PO t-1 presents the portfolio of the high dividend yield stocks at t-1. Both equations present the return at time t, measured by the value of the market/portfolio at time t plus the received dividend between t-1 and t minus the value of the market/portfolio at time t-1, dividend by the value of the market/portfolio at time t-1. Hence, the return of the market/portfolio is thus depending on the price movement and the dividends received. The stock prices obtained via DataStream are adjusted prices, which means that the prices are occasionally recomputed by DataStream to take into account capital operations. For example, if a company executed a 2 for 1 stock split, the price of the stock will decrease by 50 percent (restricted to normal price fluctuations). All previous prices are halved by DataStream to make these prices comparable to the new prices. However, the adjusted prices are not adjusted for received dividends. The stock price falls with the amount of dividends received, restricted to normal price fluctuations. For example, if the stock price closes at 100 the day before the ex-dividend date and the declared dividend is 5, the opening price of the stock on the ex- (2) 8

10 dividend date is 95 (restricted to normal price fluctuations). Hence, we have to calculate the stock returns by adding the dividends paid manually, as presented in equation (1) and (2). 3.2 The conventional tests There are several versions of the Dogs of the Dow investment strategy. The most popular one picks the 10 stocks with the highest dividend yield, as we presented in the introduction. However, there are more suggested strategies. For example, you could pick the highest dividend yield stock alone or the five highest dividend yield stocks. In this thesis, we will test the strategy of picking the ten stocks with the highest dividend yield, and these two strategies to check which one has the best results. In the first place, the performance of these strategies will be tested statistically. Consequently, there is no adjustment for higher risk, no tax considerations, and no transaction costs are taken into account. To test for significance, we follow the theory (for example Da Silva (2001)) and thus we will use the Student t-test statistic. This test allows us to compare two different returns; on one hand the return from the high dividend yield portfolio and on the other, the return from the AEX-index. Furthermore, this Student t-test is the most common test used to show significance. The t-statistic is calculated as follows: t = d * n (3) s d Where d is the mean difference between the market and portfolio return each month, s d is the standard deviation of the difference between the returns each month, and n is equal to the number of months (12, 24, 60, or 120). The Student t-test follows a t-distribution denoted as t α,v where α is the significance level and v is the degrees of freedom, which is equal to n-1. Note that the mean difference between the market and portfolio return is the geometric mean. The geometric mean is used because you have to deal with percentages. The following formula expresses the calculation of the geometric mean: n (1+ a 1 ) *(1+ a 2 ) *...* (1+ a n ) 1 (4) Where n is the number of months (12, 24, 60, or 120) and a is the monthly difference between the market s return and the portfolio s return each month. By each monthly difference the value 1 is added, because this formula cannot deal with negative numbers. Each company in the high dividend yield portfolio is given an equal weight. For example, to test for the ten highest dividend yielding stocks each stock is given a weight of 10%. 9

11 On the 1 st of April 1999 the highest dividend yielding stocks were selected for the first portfolio in The dividend yield expresses the dividend per share as a percentage of the share price. The underlying dividend is based on an anticipated annual dividend and excludes special or once-off dividends. Dividend yield is calculated on gross dividends, thus including tax credits. As we mentioned before, we will create three different portfolios. The first portfolio consists of only one stock with the highest dividend yield. The second portfolio consists of five stocks with the highest dividend yield. The third portfolio consists of ten stocks with the highest dividend yield. Each year, on the first trading day of April, this process is repeated in order to create a new high dividend yield portfolio. The portfolios were rebalanced every year by revising the dividend yield indicators for all companies from the AEX-index. We chose to rebalance the portfolio every year in the beginning of April, because the AEX-index is rebalanced in March every year. If new stocks are added to the index, you can almost immediately take these stocks into account for the rebalancing of the new high dividend yield portfolio. Furthermore, this choice also avoids any distortion of the results, if the portfolios were rebalanced at the end of every calendar year. Rebalancing at the end of the year might result in abnormal stock volatility due to year-end stock trading motivated by various tax and/or accounting related reasons. The results obtained by the data and tests above enable me to give a judgement about the significance of the high dividend yield strategy. We will look at one-year time periods as well as multiple-year time periods. If the high dividend yield portfolio obtains significant better results than the market you can conclude that the high dividend yield portfolio outperforms the market statistically. To test the economic performance of this strategy you have to consider risk, taxes, and transaction costs also. The high dividend yield portfolio could be more or less risky than the index. With only ten stocks in the portfolio, some unsystematic risk remains, resulting in a higher standard deviation than the better-diversified AEX-index. We will use the Sharpe ratio (Sharp, 1994), the Treynor indices and Jensen s alphas to adjust for risk. The Sharpe ratio is a measure of the excess return (or risk premium) per unit of risk. This ratio provides the reward to volatility trade-off. The Sharpe measure assumes that investors hold properly diversified portfolios. It also assumes that there are appropriate amounts spent on administration and analysis. This means that the expected rate of return and the variability of each different portfolio in general lie along a straight line. 10

12 The Sharpe ratio is calculated as follows: S = d 1 s d1 * n (5) Where d 1 is the monthly difference between the portfolio, or market, return and the risk-free rate, s d1 is the sample standard deviation of the monthly return differences, and n equals the number of months (12, 24, 60, or 120). The Treynor measure, also know as the reward-to-volatility ratio, provides the reward to CAPM beta risk. This measure is dependent on the characteristic-line. The return of the market is set against the risk-free rate of return. Therefore, it indicates the return earned in excess of that which could have been earned on a risk-free investment, per each unit of market risk. When two portfolios with exactly the same slope are plotted on a graph, with one line parallel but slightly higher than the other line, it means that the upper line portfolio outperforms the lower line portfolio. Reasons for large deviations of the characteristic line can be found in portfolios that are not properly diversified. The calculation of the Treynor measure is almost similar to the calculation of the Sharpe measure. The only difference is that the beta is used as the measurement of volatility. The Treynor index is calculated as follows: T = d 1 β Where d 1 is the mean monthly difference between the portfolio, or market, return and the riskfree rate, calculated over 12, 24, 60, or 120 months and β is the portfolio s beta (the market beta is equal to 1) Jensen s Alpha is based on the Sharpe-Lintner capital asset pricing model. The model is based on five assumptions; all investors have the same homogenous expectations regarding investment opportunities and decision horizons. Investors want to maximize the single periodexpected utility of terminal wealth and are risk averse. All assets are infinitely divisible. All investors have the ability to make a choice among portfolios only on the basis of expected returns and variance of returns. There are zero taxes and zero transaction costs (Naranjo, A., Nimalendran, M., Ryngaert, M., 1998). Jensen s alpha represents the average return on a portfolio over and above that predicted by the capital asset pricing model (CAPM), given the portfolio s beta and the average market return. It could be interpreted as follows: the difference between a portfolio s actual return and the one that could have been achieved on a benchmark portfolio with the same risk as measured by the beta. (6) 11

13 Jensen s alpha is calculated as follows: Where R p is the portfolio s return, R f is the risk-free rate, R m is the market s return, and β is the portfolio s beta. Finally, we will use the Sharpe portfolio performance measure to take risk into account and to make the economic results comparable to each other. The following formula, as used by McQueen et al. (1997), will be used to compute this portfolio performance measure: Rc p = ( Ru p r f ) * M sd + r f (8) P sd (7) Where Rc p is the return of the portfolio, corrected for risk, Ru p is the return of the portfolio uncorrected for risk, r f is the annual mean 1-month Euribor rate, M sd is the standard deviation of the market (AEX-index), and P sd is the standard deviation of the high dividend yield portfolio. However, if the average annual return of the high dividend yield portfolio is lower than the annual mean of the 1-month Euribor rate, equation (8) will give wrong answers. If this is the case, the following formula needs to be used: Rc p = 2* Ru p Ru p r f ( ) * M sd P sd + r f (9) McQueen et al. (1997) consider capital gains tax and tax on any dividends received. In the United States, capital gains are taxed only when the gain is realized, and the most common form of realization is selling. However, according to McQueen et al. (1997), a formal analysis of the tax advantages of the index over the high dividend yield portfolio is not possible because the size of the advantages depends on the individual s marginal tax rate and other considerations. The tax system in the Netherlands is different from that of the United States. In the Netherlands, the realized real gain does not matter when paying tax. You have to pay tax on the average amount of money you invested during the year. This is measured by the sum of your amount on the 1 st of January and the 31 st of December, divided by 2. Whether you realized a real gain or not does not matter. Thus, the capital gains tax is not relevant for analyzing the high dividend yield strategy in the Netherlands. The dividend tax rate is fixed and thus independent on your income or other variables. Until 2005 the dividend tax rate was 25%. From 2006 till present the dividend tax rate is 15%. This 12

14 tax rate matters since the high dividend yield portfolio will pay more dividends compared to the index. Hence, this dividend tax is unfavourable for the high dividend yield portfolio. To correct the high dividend yield portfolios for tax disadvantages the following formula will be used: R i = R e DY *TR (10) Where R i is the return of the portfolio or market including tax disadvantages, R e is the return of the portfolio or market excluding tax disadvantages, DY is the dividend yield, and TR is the tax rate on dividends paid (25% until 2005, 15% from 2006) Following the high dividend yield strategy will probably result in higher transaction costs. McQueen et al. (1997) assumed one-way transaction costs to be 1.00 percent. We will also assume 1.00 percent transaction costs in my thesis. The turnover is used to calculate these transaction costs. Transaction costs for the portfolio are calculated as follows: TC = 2 * AT n *1.00% (11) Where TC is the transaction costs, AT is the average turnover, and n is the number of stocks hold in the portfolio. We present a short summary of what to test using the conventional tests. Firstly, the Student t- test is used to test for significant differences in the returns of the AEX-index on one hand and the top-one, top-five, and top-ten portfolios on the other hand. These tests do not include any risk measures, taxes, and/or transaction costs. Subsequently, the returns of the AEX-index and the returns of the high dividend yield portfolios are measured by several risk measures. These risk measures are the Sharpe measure, the Treynor measure, and Jensen s alpha. These measures present a ratio. Using that ratio you can compare the performance of each portfolio as corrected for risk. McQueen et al. (1997) developed a formula to compute the return corrected for risk, instead of calculating a ratio. We will use this formula. Thereafter, we will correct the returns for taxes and transaction costs as well. All tests are performed on a single year basis as well as on a multiple year basis. 3.3 Regression analysis In addition, we will perform several regression tests to analyse the performance of the high dividend yield strategy. Firstly, we will test the relation between the dividend yield and stock returns. If the coefficient of the dividend yield is positive and significant, we can conclude that a higher dividend yield results in higher stock returns. The influence of the level of the dividend yield is thus tested. We will describe this part in Section Secondly, we will 13

15 test the influence of the existence of dividend. A dummy variable is created that is equal to 1 if the stock has a dividend yield higher than zero and is equal to 0 otherwise. This part is described in Section The level of the dividend yield We will run the regression on the full sample. The econometric program Eviews helps us to create a simple regression of the following form: R i,t = α i, + d i DY t 1 + ε t (12) Where R i,t is the return at t, DY is the dividend yield at t-1, and ε t is the disturbance term. However, we would like to control risk like Naranjo et al. (1998). They emphasize the importance of risk control. In their opinion, the best way to control risk is using the three Fama-French factors. Fama and French (1996) form portfolios on the basis of various stock attributes (book-to-market, price-earnings ratio, sales growth, and size) that have historically produced abnormal returns in a CAPM framework. We will control for risk using the three Fama-French factors too. Naranjo et al. (1998) controlled for taxes too. They infer the implied tax rate from the ratio of the one-year prime grade municipal yield to the one-year T-bill yield. However, Naranjo et al. (1998) found that the size of the yield effect appears to be unrelated to the level of the implied tax rate, and hence the potential tax penalty from receiving taxable dividend income. They also examine shocks to the implied tax rate series. To the extent that it is costly for high dividend yield firms to adjust their dividend policy, they would expect that an unanticipated increase in the implied tax rate would lead to worse performance for higher dividend yield stocks. However, they cannot find such result. Consequently, it is difficult to attribute our documented yield effects to tax effects. Knowing these results, we decide to not take taxes into account on the regression model. Hence, the following regression will be tested: R i,t = α i + s i SMB t + h i HML t + d i DY t 1 + ε i,t (13) Where R i,t is the return on a portfolio at t, SMB t is the difference between the return on a portfolio of small stocks and the return on a portfolio of large stocks (market value is the measure for the size of stocks) at t, HML t is the difference between the return on a portfolio of high book-to-market stocks and the return on a portfolio of low book-to-market stocks at t, DY t-1 is the dividend Yield at t-1, and ε i,t is the disturbance term. 14

16 Note that we will use the same definition for market value as used in DataStream, which is as follows; the market value is the share price multiplied by the number of ordinary shares in issue. The amount in issue is updated whenever new tranches of stock are issued or after a capital change. The market value is displayed in millions of Euros. The book-to-market ratio is calculated manually by DataStream and provides only the market value to book value ratio. This is a contra ratio compared to the book-to-market ratio. Another characteristic that is tested in several other studies is whether there is a seasonal factor in the timing of the stocks with the highest dividend yield strategy. Investing at the start of the year could be critical to the success of the strategy (Da Silva, 2001). However, we decided not to pay attention to this phenomenon The existence of dividend yield To test for the influence of the existence of dividend yield a dummy variable will be created, which results to the following regression: R i,t = α i, + d i DYdummy t 1 + ε t (14) Where R i,t is the return at t, DYdummy is the dummy variable for the dividend yield at t-1, and ε t is the disturbance term This regression will also be controlled for risk, as described above. Hence, the following regression will be tested: R i,t = α i + s i SMB t + h i HML t + d i DYdummy t 1 + ε i,t (15) Where R i,t is the return on a portfolio at t, SMB t is the difference between the return on a portfolio of small stocks and the return on a portfolio of large stocks (market value is the measure for the size of stocks) at t, HML t is the difference between the return on a portfolio of high book-to-market stocks and the return on a portfolio of low book-to-market stocks at t, DYdummy t-1 is the dummy variable for the dividend yield at t-1, and ε i,t is the disturbance term. These regressions will help us to come to a comprehensive conclusion about the effects of dividend yield on stock returns. 15

17 4. Data As we mentioned before, the thesis will be about the dividend yield strategy in the Dutch market. The Amsterdam Exchange Index (AEX-index) is used as the benchmark. The AEXindex consists of the 25 most traded stocks in the Netherlands. This index is chosen as a benchmark for the Netherlands because it is the most important index in this country. For every stock in the index the price and the dividend yield is obtained using DataStream timeseries request. Furthermore, we have to collect the monthly returns of the portfolio and of the index in order to calculate the Sharpe ratio. The beta is also needed to calculate the Treynor index. This data is available at DataStream. The monthly returns will be used to measure the performance of the strategy too. To control for risk using the Fama-French model (1996) we also have to obtain the risk-free interest rate, the market value, and the market to book value. DataStream provides the market value and the market to book value for the stocks. In the United States a short-term government bond is used in practice to indicate the risk-free rate. In EUR countries, it is most common to use the 1-month EURIBOR to indicate the risk-free rate. This interest rate can be found at the website of the European Central Bank (ECB). Note that the 1-month Euribor that is given by the ECB is annualized. However, to make it comparable to the other data we have calculated this rate per month, by the following formula: rf m,t = 121+ rf y,t 1 (16) Where rf m,t is the 1-month Euribor rate per month at t and rf y,t is the 1-month Euribor rate per year at t. 5. Empirical results In this section we present the empirical results. The theory described in the literature review is tested on a sample of stocks from the AEX-index during the years Section 5.1 starts with the statistical results from the conventional tests on a single year basis. In Section 5.2 the same statistical results are presented. However, these results are on a multiple year basis. The economical results are presented in Section 5.3. Several subsections treat the different risk measures. Finally, Section 5.4 comes with the regression results. 16

18 5.1 Statistical results on a single year basis We have examined the stocks with the highest dividend yield. In the first place we tested for a significant difference between this portfolio and the market on a single year basis. In Table 1 we find the yearly performance of the AEX-index and the portfolio of the one stock with the highest dividend yield (Top-one). Discussion Table 1. The yearly performance of the one stock with the highest dividend yield is better in six out of ten years. The yearly performance of the AEX-index is better in four out of ten years. The performance of the one stock with the highest dividend yield is statistically only better in 2006, at 1% level. This is an extremely significant result. However, no other significant results are obtained. The explanation for this can be found in the relatively high standard deviation of the top-one portfolio versus the market. Worth noting is that the average annual return of the top-one portfolio is 3.79% lower than the AEX-index during the years The bad performance of the top-one portfolio is mainly caused by the yearly returns in 2002 and 2008, as you can see in Table 1. Table 1. Top-one dividend yield strategy comparison of compounded returns in years Year AEXindex (yearly return) Top-one (yearly return) Winner Mean difference of market versus top-one a Standard deviation of market versus topone T-test % 58,61% Top-one Top-one AEX-index AEX-index Top-one AEX-index Top-one Top-one **** Top-one ,38 AEX-index **** Statistically differing results at 1% level In Table 2 we find the yearly performance of the AEX-index and the portfolio of the five stocks with the highest dividend yield (Top-five). Discussion Table 2. The yearly performance of the five stocks with the highest dividend yield is better in seven out of the ten years. The yearly performance of the AEX-index is better in only three out of the ten years. The performance of the five stocks with the highest dividend yield is statistically better in the years 2000, 2005, and 2006 at 10% level, 10% level, and 5% level respectively. It is worth noting that the performance of the AEX-index is statistically better in 2007, at 5% level. 17

19 The average annual return of the top-five portfolio is equal to 0.91% and the average annual return of the AEX-index is -4.33%. Hence, the top-five portfolio has a 5.24% higher average annual return than the AEX-index. However, the standard deviation of the top-five portfolio is higher than the standard deviation of the AEX-index, with 9.40% and 6.67% respectively. Table 2. Top-five dividend yield strategy comparison of compounded returns in years on a single year basis. Year AEXindex (yearly return) Top-five (yearly return) Winner Mean difference of market versus top-five Standard deviation of market versus top-five T-test % 57.63% Top-five Top-five Top-five AEX-index Top-five Top-five Top-five * Top-five ** AEX-index ** AEX-index * Statistically differing results at 10% level ** Statistically differing results at 5% level Discussion Table 3. In Table 3 we find the yearly performance of the AEX-index and the portfolio of the ten stocks with the highest dividend yield (Top-ten). The yearly performance of the ten stocks with the highest dividend yield is better in eight out of the ten years. The yearly performance of the AEX-index is better in only two out of the ten years. The performance of the ten stocks with the highest dividend yield is statistically better in the years 1999, 2000, and 2003 at 5% level, 2.5% level, and 10% level respectively. It is worth noting that the performance of the AEX-index is statistically better in 2002, at 10% level. The average annual return of the top-ten portfolio is 6.14% higher than the AEX-index. However, the standard deviation of the top-ten portfolio is higher than the standard deviation of the AEX-index, with 8.29% and 6.67% respectively. If we compare these figures with the figures in Table 5, it suggests a better performance for a portfolio including ten stocks with the highest dividend yield compared to a portfolio including five stocks with the highest dividend yield. 18

20 Table 3. Top-ten dividend yield strategy comparison of compounded returns in years on a single year basis. Year AEXindex (yearly return) Top-ten (yearly return) Winner Mean difference of market versus top-ten Standard deviation of market versus top-ten T-test % 45.74% Top-ten ** Top-ten *** Top-ten AEX-index * Top-ten * Top-ten Top-ten Top-ten AEX-index Top-ten * Statistically differing results at 10% level ** Statistically differing results at 5% level *** Statistically differing results at 2.5% level From 1999 till 2008 the market has gone up over five years and has gone down over five years. Hence, we have five years of a bull market (market has gone up) and five years of a bear market (market has gone down). In Table 3 you can see that if the ten stocks with the highest dividend yield are outperforming, five out of the eight years are bull markets. Furthermore, the two years that the AEX-index was outperforming were both years of a bear market. These results may suggest that the high dividend yield strategy is better performing in a bull market. However, no statistical evidence is provided in this thesis. The portfolio of one stock with the highest dividend yield outperforms the market in six years. The portfolio of 5 stocks with the highest dividend yields outperforms the market in seven years. Finally, the portfolio of ten stocks with the highest dividend yield outperforms the market in eight years. The top-one portfolio has an average annual return of -8.12%, the topfive portfolio has an average annual return of 0.91%, and the top-ten portfolio has an average annual return of 1.81%. The top-one portfolio has only one year reporting a positive statistically difference. The top-five portfolio and the top-ten portfolio have three years reporting a positive statistically difference. Hence, using these figures, the portfolio with ten stocks with the highest dividend yield is performing best on a single year basis. 5.2 Statistical results on a multiple year basis This section provides the statistical results on a multiple year basis. A 10-year, 5-year, and 2- year time period is created. 19

21 Discussion Table 4. In Table 4 we find the yearly performance of the AEX-index and the portfolio of the one stock with the highest dividend yield (Top-one). During the 10-year time period (from 1999 till 2008) the AEX-index has a better performance than the top-one portfolio. However, this result is not significant. From this table we see that the standard deviation of the difference of the market minus the top-one portfolio is relatively high in almost all the years, except in This means that the volatility of the top-one portfolio is relatively high compared to the market. During the 5-year time periods the top-one portfolio outperforms the market in three time periods and underperforms the market in three time periods as well. Again, none of the results are significant. Table 4. Top-one dividend yield strategy comparison of compounded returns in years on a multiple year basis. Year AEXindex (yearly return) Top-one (yearly return) Winner Mean difference of market versus top-five Standard deviation of market versus top-one T-test 10-year time period % % AEX-index year time period % -4.30% Top-one % % AEX-index % -8.16% AEX-index % 8.72% Top-one % Top-one % AEX-index year time period % 35.79% Top-one % % AEX-index % % AEX-index % % Top-one % 73.21% Top-one % 22.50% Top-one % 33.34% Top-one ** % 11.31% Top-one % % AEX-index ** Statistically differing results at 5% level During the 2-year time periods the top-one portfolio outperforms the market in six time periods and underperforms the market in three time periods. During the time period the top-one portfolio performance is statistically better than the market performance. The low standard deviation in the time period is remarkable. During the other time periods no statistical differences are obtained. 20

22 Discussion Table 5. In Table 5 we find the yearly performance of the AEX-index and the portfolio of five stocks with the highest dividend yield (Top-five). During the 10-year time period the top-five portfolio has a better performance than the AEX-index. However, the critical t-value at 10% level is 1.289, which is slightly higher than Hence, this result is not significant. During the 5-year time periods the top-five portfolio outperforms the market in five time periods and underperforms the market in one time period. During the time period the top-five portfolio performs statistically better than the market at 5% level. During the other time periods no significant results can be found. Table 5. Top-five dividend yield strategy comparison of compounded returns in years on a multiple year basis. Year AEXindex (yearly return) Top-five (yearly return) Winner Mean difference of market versus top-five Standard deviation of market versus top-five T-test 10-year time period % 0.88% Top-five year time period % 8.11% Top-five * % 0.96% Top-five % 5.61% Top-five % 9.51% Top-five % 24.50% Top-five % -5.88% AEX-index year time period % 35.70% Top-five ** % 10.16% Top-five * % % AEX-index % % Top-five % 43.55% Top-five % 27.97% Top-five % 34.96% Top-five *** % -0.39% AEX-index % % AEX-index * Statistically differing results at 10% level ** Statistically differing results at 5% level *** Statistically differing results at 2.5% level During the 2-year time periods the top-five portfolio outperforms the market in six time periods and underperforms the market in three time periods. During the time periods , , and the top-five portfolio performs statistically better than the market at respectively 5%, 10%, and 2.5% level. During the other time periods no statistical differences are obtained. 21

23 Discussion Table 6. In Table 6 the yearly performance of the AEX-index and the portfolio of ten stocks with the highest dividend yield (Top-ten) is presented. During the 10-year time period the top-ten portfolio has a better performance than the AEX-index. The performance is statistically better at 2.5% level During the 5-year time periods the top-ten portfolio outperforms the market in all the periods. During the time periods , , and the top-ten portfolio performs statistically better than the market at respectively 5%, 10%, and 5% level. Table 6. Top-ten dividend yield strategy comparison of compounded returns in years on a multiple year basis. Year AEXindex (yearly return) Top-ten (yearly return) Winner Mean difference of market versus top-five Standard deviation of market versus top-ten T-test 10-year time period % 1.67% Top-ten *** 5-year time period % 5.18% Top-ten ** % 0.24% Top-ten * % 2.97% Top-ten % 5.79% Top-ten % 25.97% Top-ten ** % -1.71% Top-ten year time period % 34.24% Top-ten ***** % 10.49% Top-ten ** % % AEX-index % % Top-ten % 48.85% Top-ten * % 27.32% Top-ten * % 26.40% Top-ten ** % 1.71% Top-ten % % Top-ten * Statistically differing results at 10% level ** Statistically differing results at 5% level *** Statistically differing results at 2.5% level ***** Statistically differing results at 0.5% level During the 2-year time periods the top-ten portfolio outperforms the market in eight time periods and underperforms the market in only one time period. During the time periods , , , , and the top-ten portfolio performs statistically better than the market at respectively 0.5%, 5%, 10%, 10%, and 5% level. During the other time periods no statistical differences are obtained. 22

24 The yearly return based on the whole sample is the best for the top-ten portfolio. The top-five portfolio comes on the second place and the top-one portfolio is last. Picking the ten stocks with the highest dividend yield is most common. Generally speaking, the larger the number of stocks in portfolio, the lower the standard deviation of the market returns minus the portfolio return. Hence, providing better results. Focusing on the top-one portfolio, there is only one time period that the high dividend yield strategy achieves statistically excessive returns. During the 2-year time period from 2005 till 2006 a statistically different result at 5% level is obtained in favour of the top-one portfolio. No other significant results are obtained, due to the fact that the standard deviation is relatively high. Hence, we cannot show that the top-one portfolio achieves statistically excessive returns. Focusing on the top-five portfolio, we demonstrated that the high dividend yield strategy achieves statistically excessive returns in one 5-year time period from 1999 till From the nine 2-year time periods, three 2-year time periods show statistically excessive returns for the high dividend yield strategy. Lastly, from the ten 1-year time periods, two 1-year time periods show statistically excessive returns for the high dividend yield strategy and one 1-year time period shows statistically excessive returns for the market. Focusing on the top-ten portfolio, we demonstrated that the high dividend yield strategy achieves statistically excessive returns in a 10-year time period from 1999 till From the six 5-year time periods, three 5-year time periods show statistically excessive returns for the high dividend yield strategy. From the nine 2-year time periods, five 2-year time periods show statistically excessive returns for the high dividend yield strategy. Finally, from the ten 1-year time periods, three 1-year time periods show statistically excessive returns for the high dividend yield strategy and one 1-year time period shows statistically excessive returns for the market. 5.3 Economical results This section treats the economical results. Section starts with the first risk measure; the Sharpe ratio. Section treats the Treynor ratio. Thereafter Jensen s alpha will be calculated in Section Finally, in Section the economic results will be presented in a way that allows it to compare the figures to each other. 23

25 5.3.1 The Sharpe ratio As we explained in Section 3.2, the Sharpe ratio is a measure of the excess return (or risk premium) per unit of risk. Note that a negative Sharpe ratio indicates that a risk-free asset (1- month Euribor) would perform better than the portfolio being analyzed. However, negative Sharpe ratios are not suitable to compare to each other. Hence, a ratio far below zero (for example -1.80) is not necessarily worse than a ratio that is not as far below zero (for example -1.10). For this reason, during several time periods no winner is presented in the table. Discussion Table 7. Table 7 presents the Sharpe ratios for the AEX-index and for the top-one portfolio for several time periods. During the 10-year time period both Sharpe ratios are negative and thus no comparison between both portfolios can be made during this time period. Both the AEX-index and the top-one portfolio win one 5-year time period during the six 5-year time periods. Table 7. Sharpe ratios for the AEX-index and the top-one portfolio Year AEX-index Top-one Winner 10-year time period Inapplicable 5-year time period Inapplicable Inapplicable Inapplicable Top-one AEX-index year time period Top-one Inapplicable Inapplicable Inapplicable AEX-index AEX-index Top-one Top-one year time period AEX-index Top-one Inapplicable Inapplicable AEX-index AEX-index Top-one Top-one Inapplicable Inapplicable 24

26 When we look at the 2-year time periods and the 1-year time periods we see a comparable view; the AEX-index wins as often as the top-one portfolio. Thus, measuring the risk by the Sharpe ratio in this way, you cannot say anything about the performance of the top-one portfolio compared to the AEX-index. Furthermore, there are several negative Sharpe ratios, which make the comparison even worse. Discussion Table 8. The Sharpe ratios for the AEX-index and the top-five portfolio for several time periods are presented in Table 8. Again, during the 10-year time period both Sharpe ratios are negative and thus no comparison between both portfolios can be made during this time period. During the 5-year time periods, the top-five portfolio is presented as the winner, in all four circumstances. The AEX-index wins only one time during the 2-year time periods. During the 1-year time periods, the AEX-index wins two times and the top-five portfolio wins four times. These results suggest a better performance of the top-five portfolio compared to the AEX-index. However, no statistical proof is obtained. Table 8. Sharpe ratios for the AEX-index and the top-five portfolio Year AEX-index Top-five Winner 10-year time period Inapplicable 5-year time period Top-five Inapplicable Top-five Top-five Top-five Inapplicable 2-year time period Top-five Top-five Inapplicable Inapplicable AEX-index Top-five Top-five Inapplicable Inapplicable 1-year time period Top-five Top-five Inapplicable Inapplicable AEX-index AEX-index Top-five Top-five Inapplicable Inapplicable 25

27 Discussion Table 9. In Table 9 the Sharpe ratios for the AEX-index and the top-ten portfolio for several time periods are presented. Again, during the 10-year time period both Sharpe ratios are negative and thus no comparison between both portfolios can be made during this time period. During the 5-year time periods, the 2-year time periods, and the 1-year time periods the winner is always the top-ten portfolio. However, there is not always a winner presented, for the reasons described above. Hence, these figures only suggest a better performance for the top-ten portfolio when compared to the AEX-index. Table 9. Sharpe ratios for the AEX-index and the top-ten portfolio Year AEX-index Top-ten Winner 10-year time period Inapplicable 5-year time period Top-ten Inapplicable Top-ten Top-ten Top-ten Inapplicable 2-year time period Top-ten Top-ten Inapplicable Inapplicable Top-ten Top-ten Top-ten Inapplicable Inapplicable 1-year time period Top-ten Top-ten Inapplicable Inapplicable Top-ten Top-ten Top-ten Top-ten Inapplicable Inapplicable The Treynor ratio This section treats the Treynor ratio. Tables 10, 11, and 12 show the ratios for the top-one, top-five, and top-ten portfolio respectively. As you can see in these tables, negative ratios appear. There are two ways the Treynor ratio might take on a negative value. The first one is 26

28 if the portfolio return is less than the risk-free ratio, and the beta is positive. The second one is if the portfolio return exceeds the risk-free rate, but the beta is negative. Without additional information, a negative Treynor ratio is impossible to interpret. Of the above two circumstances that yield a negative value, the first portfolio performs poorly taking systematic risk but not even matching the risk-free rate. In the second, the portfolio performed well outperforming the risk-free rate while actually shorting systematic risk. There are actually three situations when comparing the Treynor ratio of the AEX-index and the high dividend yield portfolio; both have a positive ratio, one of the two has a positive ratio and the other has a negative ratio, and both have a negative ratio. The interpretation of the ratios in the above situations is as follows: - By two positive ratios: The highest ratio is the best. - By a negative ratio and a positive ratio: The negative ratio is the best if the beta of this ratio is negative (and thus the portfolio return minus the risk-free rate is positive). If the beta of the negative ratio is positive (and thus the portfolio return minus the risk-free rate is negative), the positive ratio is the best. - By two negative ratios: The one with a portfolio return that exceeds that risk-free rate, but has a negative beta is better than the one with a portfolio return less than the risk-free rate, and a positive beta. If both ratios have a portfolio return that is less than the risk-free ratio, and thus a positive beta the following is true; if the high dividend yield portfolio beta is larger than 1 and the ratio is equal to or below the market ratio, the market ratio is the best. Discussion Table 10. Table 10 makes clear that the performance of the AEX-index is better than the top-one portfolio in the 10-year time period, controlled for risk by the Treynor measure. When you look at the smaller time periods, you cannot see a clear pattern. In approximately halve of the cases the AEX-index has a better ratio and in halve of the cases the top-one portfolio has a better ratio. As we showed in Table 4, the top-one portfolio statistically underperforms the AEX-index. However, this result was not significant due to the relatively large standard deviation. 27

29 Table 10. Treynor ratios for the AEX-index and the top-one portfolio Year AEX-index Top-one Winner Portfolio beta 10-year time period AEX-index year time period Inapplicable AEX-index AEX-index Top-one Top-one AEX-index year time period Top-one AEX-index AEX-index Inapplicable Top-one AEX-index Top-one Top-one AEX-index year time period Top-one Top-one AEX-index AEX-index Top-one AEX-index Top-one Top-one AEX-index AEX-index The portfolio return is less than the risk-free ratio, and the beta is positive. 2 The portfolio return exceeds the risk-free rate, but the beta is negative. Discussion Table 11. In Table 11 the Treynor ratios for the AEX-index and top-five portfolio are presented. During the 10-years time period no judgement can be made about which one performs best. Looking to the smaller time periods, the top-five portfolio has a definite advantage. During the 5-year time periods the top-five portfolio wins three times, and the AEX-index wins never. During the 2-year time periods the top-five portfolio wins five times, and the AEX-index one time. At last, during the 1-year time periods, the top-five portfolio wins five times, and the AEX-index wins two times. Hence, no clear verdict can be made of the performance of the AEX-index compared to the performance of the top-five portfolio. However, the smaller time periods suggest a better performance of the top-five portfolio. 28

30 Table 11. Treynor ratios for the AEX-index and the top-five portfolio Year AEX-index Top-five Winner Portfolio beta 10-year time period Inapplicable year time period Top-five Inapplicable Inapplicable Top-five Top-five Inapplicable year time period Top-five Top-five Inapplicable Inapplicable Top-five Top-five Top-five AEX-index Inapplicable year time period Top-five Top-five Inapplicable AEX-index Top-five AEX-index Top-five Top-five AEX-index Inapplicable The portfolio return is less than the risk-free ratio, and the beta is positive. Discussion Table 12. In this table we presented the Treynor ratios of the AEX-index and the top-ten portfolio. Again, during the 10-years time period no judgement can be made about which one performs best. The smaller time periods show a clear advantage for the top-ten portfolio. During all the smaller time periods, the AEX-index wins only two times, while the top-ten portfolio wins fifteen times. However, during seven time periods no winner can be presented due to negative ratios. Hence, like the top-five portfolio, no clear verdict can be made of the performance of the AEX-index compared to the performance of the top-ten portfolio. However, the smaller time periods do suggest a better performance of the top-ten portfolio. 29

31 Table 12. Treynor ratios for the AEX-index and the top-ten portfolio Year AEX-index Top-ten Winner Portfolio beta 10-year time period Inapplicable year time period Top-ten Inapplicable Top-ten Top-ten Top-ten Inapplicable year time period Top-ten Top-ten AEX-index Inapplicable Top-ten Top-ten Top-ten Inapplicable Inapplicable year time period Top-ten Top-ten Inapplicable AEX-index Top-ten Top-ten Top-ten Top-ten Inapplicable Inapplicable The portfolio return is less than the risk-free ratio, and the beta is positive Jensen s Alpha This section treats Jensen s Alpha. Table 13 presents a summary of the results from the Jensen s alpha model. For details, see Table 23 till Table 25 in the appendix. Discussion Table 13. You can see that this model explains 23.80%, 68.43%, and 83.81% for respectively the top-one portfolio, the top-five portfolio, and the top-ten portfolio. Hence, the R-squared of the top-ten portfolio is significantly higher than the top-five portfolio and the top-one portfolio. The R-squared value is dependent on the degree of diversification. As you can expect, the degree of diversification is higher for a larger portfolio. The table below shows that the mean value of α is negative for all three high dividend yield portfolios. The alphas of the top-one portfolio and the top-five portfolio are significantly below zero, at 10% level. The alpha of the top-ten portfolio is significantly lower than zero at 30

32 the 5% level. Hence, when accounting for market risk, as measured by beta, all three high dividend yield portfolios underperform the market. Table 13. Summary of the Jensen s alpha model for all three portfolios. R p R f = α + β( R m R f ) + ε p 10-years Item 10-years Extreme values (yearly) Standard value Minimum Maximum deviation R-squared Top-one R-squared Top-five R-squared Top-ten α Top-one * α Top-five * α Top-ten ** β Top-one β Top-five β Top-ten * Statistically differing results at 10% level ** Statistically differing results at 5% level The final economical results corrected for risk, taxes, and transaction costs This section deals with the final results. The statistical results obtained in Section 5.1 and 5.2 will be corrected for risk, taxes, and transaction costs. We have used the Sharpe portfolio performance measure to correct for risk as presented in equation (8) and (9). Thereafter, the tax disadvantages for the high dividend yield portfolio will be calculated as explained in Section 3.2. Finally the returns will be corrected for transaction costs using the formula presented in equation (10). All three high dividend yield portfolios will be treated separately, starting by the top-one portfolio and the AEX-index, then the top-five portfolio, and lastly the top-ten portfolio Economical results for the top-one portfolio and the AEX-index The average annual return of the top-one portfolio is equal to -8.12% and the standard deviation is equal to 20.10%. The standard deviation of the market is equal to 6.67% and the average annual return of the 1-month Euribor is 3.73%, during the 10-year time period from 1999 till Hence, the risk-adjusted return for the top-one portfolio is equal to % 1, with a standard deviation of 6.67%, equal to the AEX-index. This adjustment is akin to investing 33 percent of the wealth in the top-one portfolio and 67 percent of the wealth in the 1-month Euribor. Adding this high proportion of 1-month Euribor to overly the risky top-one portfolio, results in a risk-adjusted portfolio with a standard deviation exactly equal to that of 31

33 the AEX-index, which facilitates comparisons. Without adjustments for risk, the difference between the top-one portfolio mean return and the AEX-index mean return is 3.79% in favour of the AEX-index. After adjustments for risk, this difference increases to 11.71% ( ) in favour of the AEX-index. Thus, risk alone explains nearly 8% of the top-one portfolio premium. The average dividend yield for the top-one portfolio is 7.90%. The average dividend yield for the AEX-index is 2.75%. By correcting for taxes, the return of the top-one portfolio shrinks from to % 2. The return of the AEX-index shrinks too: from -4.33% to -4.94%. After adjustments for risk and tax disadvantages, the difference increases to 12.84% ( ) in favour of the AEX-index. Finally, the transaction costs are treated. The firm with the highest dividend yield changed nearly every year. Only in the years 2006 and 2007 the same firm stays in the portfolio. Hence, on average 0.89 out of 1 firm changed each year. The AEX-index is, of course, more stable than the top-one portfolio. On average 2.22 out of the 25 firms changed each year. We will assume 1.00 percent one-way transaction costs, like McQueen et al. (1997). Using the formula in equation (11), 1.78% 3 of the top-one portfolio is lost to transaction costs in a typical year, whereas 0.18% of the AEX-index s value is lost in dissipative trading. The average annual return for the top-one portfolio corrected for risk, taxes, and transaction costs is %. This is far below the average annual return of the AEX-index; -5.12% Economical results for the top-five portfolio The average annual return of the top-five portfolio is equal to 0.91% and the standard deviation is equal to 9.40%. The risk-adjusted return of the top-five portfolio is calculated in the same way as in Section and is equal to 0.09% 4. This adjustment is akin to investing 71 percent of the wealth in the top-five portfolio and 29 percent of the wealth in the 1-month Euribor. Without adjustments for risk, the difference between the top-five portfolio mean return and the AEX-index mean return is 5.24% in favour of the top-five portfolio. After adjustments for risk, this difference shrinks to 4.42% ( ) in favour of the top-five portfolio. Thus, risk alone explains less than 1% of the top-five portfolio premium. The dividend yield for the top-five portfolio is 5.89% on average. By correcting for tax disadvantages, the return of the top-five portfolio shrinks from 0.09% to -1.21% 5. After adjustments for risk and taxes, the difference shrinks to 3.73% ( ) in favour of the top-five portfolio. 32

34 And lastly, the transaction costs are treated. The average turnover for the top-five portfolio is Using the same formula as above, you can calculate that the top-five portfolio loses 0.89% 6 to transaction costs. The average annual return for the top-five portfolio corrected for risk, taxes, and transaction costs is -2.10%. This is better than the average annual return of the AEX-index which is -5.12%. However, this average annual return is calculated using the simple arithmetic return, and thus not suitable for compounding. The yearly returns for the top-five portfolio are more volatile than the yearly returns of the AEX-index. These volatile returns are in disadvantage of the total return during the 10-year time period Economical results for the top-ten portfolio The average annual return of the top-ten portfolio is equal to 1.81% and the standard deviation is equal to 8.29%. The risk-adjusted return of the top-five portfolio is calculated in the same way as in Section and is equal to 1.43% 7. This adjustment is akin to investing 80 percent of the wealth in the top-ten portfolio and 20 percent of the wealth in the 1-month Euribor. Without adjustments for risk, the difference between the top-ten portfolio mean return and the AEX-index mean return is 6.14% in favour of the top-ten portfolio. After adjustments for risk, this difference is 5.76% ( ) is favour of the top-ten portfolio. Thus, risk alone explains only 0.40% of the top-ten portfolio premium. The average dividend yield for the top-ten portfolio is 4.76%. The return of the top-ten portfolio shrinks from 1.43% to 0.38% 8, by correcting for taxes. After adjustments for risk and taxes, the difference shrinks to 5.32% ( ) in favour of the top-ten portfolio. Finally, the return will be corrected for transaction costs. The average turnover for the top-ten portfolio is Using the same formula as in Section , you can calculate that the topten portfolio loses 0.56% 9 to transaction costs. The average annual return for the top-ten portfolio corrected for risk, taxes, and transaction costs is -0.18%. This is better than the average annual return of the AEX-index which is -5.12%. However, like we mentioned in Section , this average annual return is calculated using the simple arithmetic return, and thus not suitable for compounding. The yearly returns for the top-ten portfolio are more volatile than the yearly returns of the AEX-index. These volatile returns are in disadvantage of the total return during the 10-year time period. 5.4 Regression analysis As described in Section 3.3, we will also use a regression to test the relationship between dividend yield and stock return. The regressions are split up in two parts: one part will test the 33

35 influence of the level of the dividend yield, which is presented in Section The other part will test the influence of the existence of dividend yield, which is presented in Section Influence of the level of dividend yield This section treats the influence of the level of dividend yield. The regression being used is presented in equation (12). The main results are presented in Table 14. Discussion Table 14. During the 10-year time period, the dividend yield has an average coefficient of per year. You can interpret this result as follows: When the dividend yield increases by 1, the yearly return of the stock increases by 4.72% on average. The result is (extremely) significant at 0.5% level. The results become less clear when we look at the 1- year time periods. During the years 2000, 2001, and 2006 a significant positive relationship between the dividend yield and stock return is obtained. However, during the years 2007 and 2008 a significant negative relationship between the dividend yield and stock return is obtained. Note that all these results are not controlled for risk, taxes, and transaction costs. Table 14. The coefficient, standard error, and the t-statistic for the dividend yield in the following regression: R i,t = α i, + d i DY t 1 + ε Year Coefficient Standard Error T-statistic 10-year time period ***** 1-year time period ** * * * * * Statistically differing results at 10% level ** Statistically differing results at 5% level ***** Statistically differing results at 0.5% level We will control for risk like Naranjo et al. (1998). In Section 3.3 you can find how to deal with risk. Taxes and transaction cost are not taken into account. The reasons are also described in Section 3.3. We will test the regression presented in equation (13). Table 15 provides the main findings. Discussion Table 15. The average coefficient of the dividend yield shrinks from 4.72% to 2.08%. Hence, 2.64% of the additional return can be attributed to increased risk. Nevertheless, there is still a significant positive relationship between the dividend yield and 34

36 stock return at 10% level. Note that the results are only presented during the 10-year time period. The variables SMB and HML are calculated on a yearly basis. Hence, a 1-year time period regression cannot be performed. Table 15. The coefficient, standard error, and the t-statistic for the dividend yield, SMB, and HML in the following regression: R i,t = α i + s i SMB t + h i HML t + d i DY t 1 + ε i Year Variable Coefficient Standard Error T-statistic 10-year time period Dividend Yield * SMB HML ***** * Statistically differing results at 10% level ***** Statistically differing results at 0.5% level Influence of the existence of dividend yield This section treats the influence of the existence of dividend yield. The regression being used is presented in equation (14). The main results are presented in Table 16. Discussion Table 16. During the 10-year time period you can see a remarkable negative coefficient of You can interpret this result as follows: Stocks that have a dividend yield of 0, perform on average 8.79% better than stocks that have a dividend yield higher than 0. However, due to the relatively large standard error, the result is not significant. During the 1-year time periods we found two times a positive significant result. In the years 2001 and 2004 the coefficient of the dividend yield dummy is statistically positive at 0.5% and 10% level. No negatively significant result is found during the 1-year time periods. Note that none of these results are controlled for risk, taxes, and transaction costs. Table 16. The coefficient, standard error, and the t-statistic for the dividend yield dummy in the following regression: R i,t = α i, + d i DYdummy t 1 + ε Year Coefficient Standard Error T-statistic 10-year time period year time period ***** * * Statistically differing results at 10% level ***** Statistically differing results at 0.5% level 35

37 Discussion Table 17. As we found in Table 17, the average coefficient of the dividend yield dummy shrinks from -8.79% to -2.22%. However, this is still a negative amount. Thus, controlled for risk, a stock that has a dividend yield higher than 0 results, on average, in a 2.22% lower stock return than a stock that has a dividend yield equal to 0. However, this result is not significant. The results are only presented during the 10-year time period for the same reason as mentioned above. Table 17. The coefficient, standard error, and the t-statistic for the dividend yield dummy, SMB, and HML in the following regression: R i,t = α i + s i SMB t + h i HML t + d i DYdummy t 1 + ε i Year Variable Coefficient Standard Error T-statistic 10-year time period Dividend Yield dummy SMB HML ***** ***** Statistically differing results at 0.5% level 6. Summary and conclusions We have investigated the relationship between stock returns and dividend yield in the Dutch market. The sample contained all the 25 stock from the AEX-index in the relevant year. The sample period was from April 1999 until March 2008, thus a total of 120 months. The central question in my thesis is as follows: Do high dividend yield stocks achieve excessive returns in the Dutch market? To give an answer to this question, we made a distinction between statistically excessive returns and economically excessive returns. Both returns are measured by conventional tests and regression analysis. The conventional tests were performed on three high dividend yield portfolios and the market (AEX-index). The high dividend yield portfolios were composed as follows: a portfolio consisting of one stock with the highest dividend yield; another portfolio consisting of five stocks with the highest dividend yield; and the third portfolio consisting of ten stocks with the highest dividend yield. The regression tests takes all the stock listed in the AEX-index into account. High dividend yield stocks should achieve statistically excessive return, according to the theory (see for example Filbeck and Visscher (1997)). However, not all researchers can 36

38 demonstrate these statistically excessive returns. For many researchers it is difficult to prove economically excessive returns for high dividend yield stocks. The answer to the central question is that high dividend yield stocks achieve statistically excessive returns in the Dutch market. However, we cannot prove that high dividend yield stocks achieve economically excessive returns in the Dutch market. This thesis makes clear that a high dividend yield portfolio consisting of ten stocks with the highest dividend yield performs better than a high dividend yield portfolio consisting of one stock with the highest dividend yield. This is because of the large volatility and risk of only one stock. The top-ten portfolio achieves statistically excessive returns during the 10-year time period measured by the t-test. The top-five portfolio achieves excessive returns as well. However, these excessive returns were not significant due to a relatively large standard deviation. The top-one portfolio does not achieve excessive returns. These returns were not significant. The economical performance is measured by three different risk measures, taxes and transaction costs. The Sharpe ratios for the portfolios and the AEX-index were negative during the 10-year time period. Negative Sharpe ratios are not suitable for comparison. However, if we look at smaller time periods we find several positive Sharpe ratios suitable for comparison. The top-one portfolio performs worse than the AEX-index, the performance of the top-five portfolio looks the same as the AEX-index, and the top-ten portfolio performs better than the AEX-index. Risk measured by the Treynor ratio gives the following results; the top-one portfolio performs worse than the AEX-index, and we cannot make conclusions about the performance of the top-five and top-ten portfolio during the 10-year time period. However, if we look at smaller time periods the top-five portfolio as well as the top-ten portfolio looks like to be in favour of the AEX-index. The last performance measure is Jensen s alpha. All three alphas of the high dividend yield portfolios are statistically negative. Therefore, when controlling for risk with Jensen s alpha, all three high dividend yield portfolios underperform the market. Finally we keep the tax disadvantages and transaction costs into account. The annual average return of the AEX-index corrected for risk, taxes and transaction costs is equal to -5.12% The annual average returns of the top-one, top-five, and top-ten portfolios corrected for risk, taxes and transaction costs are %, -2.10%, and -0.18% respectively. The top-one portfolio performs clearly worse than the AEX-index. The top-five and top-ten portfolio 37

39 performs slightly better. However, due to relatively large standard deviations, we cannot prove a significant difference. The regression analysis comes with a coefficient of the dividend yield equal to 4.72%. This is statistically different from 0. Hence, dividend yield has a positive significant influence on stock returns. If we control risk by the Fama-French factors, the coefficient of the dividend yield shrinks to 2.08%, which is still statistically different from 0 at 10% level. However, no tax disadvantages and transaction costs are included. From all these figures it becomes clear that high dividend yield stocks achieve statistically excessive returns in the Dutch market. However, you have to create a portfolio including at least 5 high dividend yield stocks. Otherwise, the portfolio is too volatile. If we take risk, taxes, and transaction costs into account, no significant excessive returns can be found. Thus, we cannot find evidence that high dividend yield stock achieve economically excessive returns in the Dutch market. Previous investigations made by other researcher (for example Da Silva (2001) and McQueen et al. (1997)) confirm these results. They also found that high dividend yield stocks achieve statistically excessive returns. However no economically excessive returns are found for high dividend yield stocks. Notes 1. Risk-adjusted return top-one = 2 * ( ) * = Tax corrected return top-one = * ( 0.7* *0.15) = Transaction costs top-one = 2* 0.89 *1.00% =1.78% 1 4. Risk-adjusted return top-five = 2* ( ) * = Tax corrected return top-five = * ( 0.7 * *0.15) = Transaction costs top-five = 2* 2.22 *1.00% = 0.89% 5 7. Risk adjusted return top-ten = 2* ( ) * = Tax corrected return top-ten = * ( 0.7* *0.15) = Transaction costs top-ten = 2* 2.78 *1.00% = 0.56% 10 38

40 References Berk, J., DeMarzo, P. (2008). Book Corporate Finance, Publisher: Pearson Custom Publishing Black, F., Scholes, M. (1974). The effects of dividend yield and dividend policy on common stock prices and returns, Journal of Financial Economics, Vol. 1 pp Blume, M. (1980). Stock returns and dividend yields: Some more evidence, Review of Economics and Statistics, Vol.61 pp Brennan, M. (1970). Taxes, market valuation and corporate financial policy, National Tax Journal, Vol. 23 pp Brzeszczynski, J., Gajdka, J. (2007). Dividend-Driven trading strategies: Evidence from the Warsaw stock exchange, International Advances in Economic Research, Vol. 13 pp Campbell, J.Y., Yogo, M. (2006). Efficient tests of stock return predictability, Journal of Financial Economics, Vol. 81 pp Chay, J.B., Suh, J. (2009). Payout policy and cash-flow uncertainty, Journal of Financial Economics, Vol. 93 pp Cochrane, J.H. (2007). The dog that did not bark: A defense of return predictability, Review of Financial Studies, Vol.21 pp Da Silva, A.L.C. (2001). Empirical tests of the Dogs of the Dow strategy in Latin American stock markets, International Review of Financial Analysis, Vol. 10 pp Domian, D.L., Louton, D.A., Mossman, C.E. (1998). The rise and fall of the dogs of the Dow, Financial Services Review, Vol. 7 pp Fama, E.F. (1970). Efficient Capital Markets: A Review of Theory and Empirical Work, Journal of Finance, Vol. 25 pp Fama, E.F., French, K.R. (1996). Multifactor explanations of asset pricing anomalies, Journal of Finance, Vol. 51 pp Filbeck, G., Visscher, S. (1997). Dividend yield strategies in the British stock market, The European Journal of Finance, Vol. 3 pp Jobson, J., Korkie, B. (1981). Performance Hypothesis Testing with the Sharpe and Treynor Measures, Journal of Finance, Vol. 36 pp Malkiel, B.G. (2003). The efficient market hypothesis and its critics, Journal of Economic Perspectives, Vol. 17 pp McQueen, G., Shields, K., Thorley, S.R. (1997). Does the Dow-10 Investment Strategy 39

41 beat the Dow statistically and economically?, Financial Analysts Journal, Vol. 53, pp Naranjo, A., Nimalendran, M., Ryngaert, M. (1998). Stock returns, dividend yields, and taxes, The Journal of Finance, Vol. 53 pp Sharpe, W.F. (1994). The Sharpe ratio, The Journal of Portfolio Management, Vol. 21 pp Visscher, S., Filbeck, G. (2003). Dividend-Yield strategies in the Canadian stock market, Financial Analysts Journal, Vol. 59 pp Webpage: (composition of the AEX) Webpage: 40

42 Appendix Table 18. Companies that are in the AEX-index in years Stock Total Sector Aegon 10 Financial Services Ahold 10 Consumer Services ABN Amro 9 Financial Services AKZO Nobel 10 Industrial Materials Arcelor Mittal 2 Industrial Materials ASML 10 Hardware Baan 2 Software Corio 1 Financial Services Corporate Express 9 Industrial Materials Corus 1 Industrial Metals DSM 10 Industrial Materials Fortis 10 Financial Services Getronics 7 Business Services Gucci 5 Consumer Goods Hagemeyer 10 Business Services Heineken 10 Consumer Goods Hoogovens 1 Industrial Metals ING 10 Financial Services KLM 1 Business Services KPN 10 Telecommunications KPNQwest 2 Telecommunications Logica 2 Telecommunications Numico 9 Consumer Goods Océ 1 Consumer Goods P&O Nedlloyd 1 Business Services Philips 10 Hardware Randstad 2 Business Services Reed Elsevier 10 Media Rodamco 2 Financial Services Royal Dutch Shell 10 Energy SBM Offshore 6 Energy Tele Atlas 1 Hardware TNT 10 Business Services TomTom 3 Hardware Unibail-Rodamco 1 Financial Services Unilever 10 Consumer Goods UPC 2 Telecommunications Van der Moolen 3 Financial Services Vedior 4 Business Services Vendex KBB 2 Consumer Goods Versatel 3 Telecommunications VNU 8 Media Wolters Kluwer 10 Media Number of stocks staying on the list Turnover (average turnover = 2.22)

43 Table 19. Companies from the AEX-index in the Top-one portfolios in years Stock Total Sector Ahold 1 Consumer Services ABN Amro 1 Financial Services Corus 1 Industrial Metals DSM 1 Industrial Materials Fortis 1 Financial Services Hagemeyer 1 Business Services KPN 3 Telecommunications Van der Moolen 1 Financial Services Number of stocks staying on list Turnover (average turnover = 0.89) Table 20. Companies from the AEX-index in the Top-five portfolios in years Stock Total Sector Aegon 5 Financial Services Ahold 1 Consumer Services ABN Amro 7 Financial Services Corporate Express 2 Industrial Materials Corus 1 Industrial Metals DSM 4 Industrial Materials Fortis 6 Financial Services Hagemeyer 4 Business Services Hoogovens 1 Industrial Metals ING 5 Financial Services KLM 1 Business Services KPN 5 Telecommunications Numico 1 Consumer Goods Rodamco 1 Financial Services Royal Dutch Shell 2 Energy TNT 1 Business Services Van der Moolen 3 Financial Services Number of stocks staying on list Turnover (average turnover =2.22 )

44 Table 21. Companies from the AEX-index in the Top-ten portfolios in years Stock Total Sector Aegon 8 Financial Services Ahold 1 Consumer Services ABN Amro 9 Financial Services AKZO Nobel 5 Industrial Materials Corio 1 Financial Services Corporate Express 3 Industrial Materials Corus 1 Industrial Metals DSM 9 Industrial Materials Fortis 9 Financial Services Hagemeyer 6 Business Services Hoogovens 1 Industrial Metals ING 8 Financial Services KLM 1 Business Services KPN 6 Telecommunications Logica 1 Telecommunications Numico 2 Consumer Goods Randstad 1 Business Services Reed Elsevier 4 Media Rodamco 2 Financial Services Royal Dutch Shell 8 Energy TNT 1 Business Services Unibail-Rodamco 1 Financial Services Unilever 6 Consumer Goods Van der Moolen 3 Financial Services Wolters Kluwer 3 Media Number of stocks staying on list Turnover (average turnover = 2.78) Table 22. Annual returns and standard deviations of the top-one portfolio, top-five portfolio, top-ten portfolio, and AEX-index. Portfolio Average annual return b Standard deviation Geometric mean annual return c Top-one -8.12% 20.10% % Top-five 0.91% 9.40% 0.88% Top-ten 1.81% 8.29% 1.67% AEX-index -4.33% 6.67% -5.51% a All differences are calculated on a monthly basis. b Simple arithmetic mean of the ten annual returns. c Allows for compounding. 43

45 Table 23. The Jensen s alpha model for the top-one portfolio. Year R-squared α β 10-year time period year time period year time period year time period

46 Table 24. The Jensen s alpha model for the top-five portfolio. Year R-squared α β 10-year time period year time period year time period year time period

47 Table 25. The Jensen s alpha model for the top-ten portfolio. Year R-squared α β 10-year time period year time period year time period year time period