Predicting future stock returns of the German stock market : comparing predictability by dividend, total payout and net payout yield from 1989 to 2014

(1)

1

Predicting future

stock returns of the

German stock market

Comparing predictability by dividend, total payout

and net payout yield from 1989 to 2014

Bachelor’s Thesis Author: Daniel Goudswaard

Economie en Bedrijfskunde Studentnumber:10528709 Specialization: Financiering en Organisatie Date of final draft version: Supervisor: Mr. Guilherme V.E.P. De Oliveira 19-06-2016

I’d like to thank several people for helping me in the last couple of months to complete this bachelor’s thesis. I would like to thank my thesis supervisor Mr. Guilherme V.E.P. De Oliveira for supporting me with my work for all these months and Dr. Philippe J.P.M. Versijp for his guidance in writing this thesis. I would also like to thank my family and friends who supported me throughout my study and while writing this thesis. To all of you, thank you for your help!

Daniel Goudswaard

The dividend yield has lost some it allure as a predictor or stock market returns early in this century. In this paper this general consensus is analyzed for the German Stock market in the period from 1989 till 2014. Furthermore to see if other payout yield predict stock market returns a total of three types of payouts have been tested by using the Fama-Macbeth model. This is the first paper that tests all three payouts outside the US and the first that tests net payout yield in Germany. During the writing of this paper evidence has been found on the relation between stock market returns and dividend yield and no evidence of the predictability of stock market returns by total payouts or net payouts has been found.

(2)

2

1. Introduction

Investors always try to find effective methods to predict future stock returns. This has always been an important topic of interest, since there are many different views on predicting stock returns. The basic premise on which this paper builds is that stock returns can be predicted by how much investors are currently being paid out. Payouts are directly distributed from firm’s earnings to investors and are the direct component of stock market returns. The premise is one of the oldest but has recently come under growing scrutiny, however almost all of the research comes from the US. The aim of this paper is to help find evidence on whether this the premise holds or not in Germany. The cultural differences between the US and Germany make it interesting to study this because of the differences in how companies see and treat shareholders and stakeholders. To do this I use the Fama-Macbeth model on three different explanatory variables to test this theory are examined; dividend yield, total payout yield and net payout yield. The dividend yield is the foundation of these three measures with total payout yield and net payout yield being extensions of this by integrating more data. The goal of this paper is to analyze and compare the predictability of stock returns by dividend yield, total payout yield and net payout yield for the German stock market. I found some evidence on the relation between stock market returns and the dividend yield and no evidence of a relation between the stock market returns and total payout yield or net payout yield. This indicates that the relation between returns and the various payout measures is affected by the differences between the US and German stock market. This warrant for more research on payout measures as predictors of stock market returns but also makes a case for more country specific research on factors that influence stock market returns.

Historically German firms do not focus directly as much on shareholders and financial markets as their American counterparts; this comes from a different kind of capitalism. In the US they have the more individualistic Anglo-Saxon model of capitalism; here the focus is more singularly and directly on the shareholder (Becht & Röell, 1999). In Germany they have the Rheinlandisch model of capitalism, a model in which companies focus on more generally on all stakeholders e.g. customers, employees and shareholders. These differences are not just culturally but are also embedded in laws and regulations. In the US firms have a lot of freedom, but in the end they are obligated by law to serve the shareholders of the firm. In Germany there are more laws and regulations to protect all stakeholders. For example a share repurchase program in the US can be started by the board of directors of a firm. In Germanynot only all repurchase programs need to be approved by the annual shareholders’ meeting, but they have to publicly announced, they have to take place within 18 months of the announcement and cannot be more than 10% of share capital per period. (Hackethal & Zdantchouk, 2004). While Germany has the largest economy by GDP and the highest median total payout yield of Europe I find very little to no research about this subject has been done on Germany (Von Eije & Megginson, 2008). It is interesting to see if the results from research on German firms are similar to the results from research on US firms.

In 2001 Fama & French find that the dividend yield is losing its allure as a predictor of future stock returns. They find the cause in the declining propensity to pay dividends among US firms. In 2008 Skinner further confirms this and finds that dividends are being replaced by share repurchases, this is theorized to shift the explanatory power from dividends to share repurchases. Boudoukh, Michaely,

(4)

4

Richardson, and Roberts (2007) suggest dividends are just a component of the return investors are paid out, just like share repurchases. Therefore they argue that payouts as a whole should be used instead of dividends. They define three measures for payouts: the first is the dividend yield and is based on dividends. The second measure is the total payout yield and is based on dividends in combination with share repurchases. The third measure is net payout yield and incorporates dividends, share repurchases and share issuances. They argue that dividend yield is the least complete measure and net payout yield the most complete measure of payouts thus dividend yield must be the weakest and net payout yield the strongest predictor of stock returns, with total payout yield being in between. They test their hypotheses in their US sample and find as they predicted; dividend yield has no predictive value and net payout yield has the strongest predictive value for stock returns.

Research from Von Eije and Megginson (2008) shows that the dividend and share repurchase policies are similar in Europe to those in the US. In Europe share repurchases have also become an important way to pay out cash to shareholders. Similar to the US they find a decline in dividend payouts, although less steep, from approximately 80% in 1989 to 60% in 2005. Because of the similar dividend and share repurchase policies, the same kind of relationships between payouts and stock market returns are expected as found by Boudoukh et al. (2007). However, because of the differences between Germany and the US the relationship between returns and the three payout yields are expected to be less pronounced. Concretely this leads to four hypotheses; the first hypothesis is that like in the US dividend yield has no predictive value in Germany. The second and third hypotheses are that the total payout yield and net payout yield do predict future returns significantly in Germany. The fourth hypothesis is that as a predictor of future stock returns net payout yield is more significant than total payout yield.

To test the predictability of the three payout measures the two-stage regression model by Fama and Macbeth (1973) in combination with factors identified by Fama and French (1993) that are used. These are chosen because they are two of the most established and widely used papers on this subject. They form the foundation to regress returns on beta, size, book-to-market and our predictive yield variables; dividend yield, total payout yield and net payout yield. To my knowledge this is the first study to examine the predictive content of all three payout yields outside of the US: dividend yield, total payout yield and net payout yield. The results of this paper show there is evidence that the dividend yield is a predictor of stock returns while there is none for total payout yield and net payout yield in Germany. This indicates that the differences between the US and Germany influence what payout measure can predict stock market returns best, but also leads to questioning whether research from the US has applicability in the rest of the world. The results are not in line with those of the US, on which this paper was based, but were more similar to research on other European countries. This paper indicates that there should be more research on the different payout measures, but in a broader sense more country specific research on which factors influence stock market returns.

In the next chapter first the key concepts and literature leading to the writing of this paper will be discussed. After that a chapter about the methodology and data used in this paper follows. Next the results of the paper are shown and discussed. Finally, this paper ends with the conclusion and remarks about where to focus on in future research.

(5)

5

2. Literature Review

In this chapter we will first cover some basic concepts fundamental tot this paper about how investors and firms interact. After that the history of predicting stock returns by firm payout yields will be examined. This starts in the last century with the dividend yield, and evolves to the more complete payout measures of total payout yield and net payout yield. In this paper we will look at three predictors of stock returns and see how they compare for the German stock market.

First some basic finance concepts will be explained through Berk and DeMarzo’s (2014) guiding. Firms at some point usually need external funding. A very important option to obtain external funds is to attract investors and sell them part of the firm by issuing shares. The investors who buy these shares then become shareholders of the firm and hope to profit from their investment in the firm, they expect returns. With returns is meant the theoretical growth in value of a stock holding over a specific period assuming dividends are reinvested in the stock. According to the risk-return tradeoff high potential returns require higher risk taking. But investors do not simply wish to take high risks on their capital; they usually want to maximize the return with a certain amount of risk they are willing to take. Two important related concepts are the risk free rate and the risk premium, the risk-free rate is the return of (near) riskless deemed assets. Usually the return of long term government bonds is taken to proxy for this. The risk premium is the return an asset should offer above the risk-free rate for the amount of risk the asset has and can be seen as a refinement of the risk return trade-off.

The return an investor receives consists of two components; the capital gain or loss they make on the change in share price and the dividends they receive. Capital gain or loss is dependent on if the share price at that moment is higher or lower than what the share was bought for. Since share prices are driven by expectations of future firm performance prices can go can go up and down with changing investors’ expectations. The capital gain or loss therefore does not materialize until a share is sold. So, if an investor needs cash from his investment he normally needs to sell his shares, but it is uncertain whether he makes a gain or a loss on this until he actually sells them. To take some of the uncertainty away from investors firms can payout dividends. A dividend is a portion of the firm’s earnings distributed directly to shareholders. Dividends are the direct form of return for investors, they are determined by the firm but once paid out the firm can’t change the value or reclaim it for other uses. To receive dividends investors don’t have to sell any shares and can usually choose if they want their dividend in the form of cash or extra shares.

The return from dividend payouts is determined by the firm, but firms can also influence the other component of return, capital gain. Firms can do this by issuing or repurchasing new shares. This decreases or increases the percentage of the firm a share represents thus if the firm is valued the same an increase in shares through an issuance will decrease the value per share while a decrease of shares through repurchases will increase the value per share. Because of this increase in share prices through share repurchases these can be seen as an alternative means to indirectly pay out investors, while issuances do the opposite. However, firm value and therefore share prices are ultimately driven by expected future firm performance. Share issuances and repurchases only change the percentage each share represents, but not the actual value of the firm.

(6)

6

way for firms to pay out investors in the early in the 20th century, the dividend yield, dividends as a percentage of the firm’s market value was then identified as a key variable in explaining stock returns (Dow, 1920). The rationale for predictability is twofold: one is that from firms that have high payouts it is expected that business is going well and two is that firms that have had high payouts in the past are expected to have high payouts in the future (Dow, 1920). Because firm management is expected to have the best information on the firm’s future profitability and the firms’ management decides if and how much dividends are paid out, the “information content hypothesis of dividends” states that dividend changes are likely to be interpreted by investors as a change in the firm’s expected future profitability (Miller & Modigliani, 1961). Because the expected future profitability is the driver of stock market prices this could mean that a change in dividends could also signal a change in expected capital gains.

However, DeAngelo, DeAngelo, and Skinner (1996) find in their study of 145 US firms with suddenly declining earnings that there is virtually no support for Miller and Modigliani’s theory. They say this is mainly due to over-optimism of management about the future and the limited magnitude of dividends on the company’s cash position making it a less important aspect. Research by Benartzi, Michaely, and Thaler (1997) among 1025 US firms that pay quarterly dividends shows that if a firm increases dividends it does not predict growth in future earnings but says something about past earnings growth. However they do find that these firms have returns higher than the market so conclude there is limited evidence to support the information content hypothesis of dividends. Fama and French find in 2001 that among 5113 US firms the number of dividend paying firms in has declined from 66.5% in 1978 to 20.8% in 1999. There are two explanations for this decline; the first is that that firm characteristics have changed to smaller firms with larger investments per earnings, these firms often never pay dividends. The second explanation for the decline is that also given their characteristics fewer firms pay dividends than before. Goyal and Welch (2003) find that dividend’s predictive ability has disappeared from the 1990’s, but even before that period it’s explanatory ability was only due to two years, 1973 & 1974 and it had no out-of-sample performance.

While the evidence against dividend’s as a predictor for stock returns was building, Grullon and Michaely (2002) find in their sample of 15843 US firms that from 1983 onwards firms gradually shifted from paying out dividends to using the same funds for the repurchase of shares. This suggests that dividends and repurchases may be substitutes. This can also be deducted from research on agency and signaling models where dividends and repurchases are seen as similar instruments. For instance research by Brav, Graham, Harvey, and Michaely (2005) confirm this relationship. From a US survey amongst 384 financial executives they find that CFO’s view dividends as less important than they used to and that share repurchases are now an important means of pay out. They also say that Miller and Modigliani’s information content hypothesis is more applicable to share repurchase than dividends. This is because managers clearly state dividend changes adapt slower to firm performance compared to decisions about share repurchases as dividends are seen as a more stable factor. This difference in perception suggests that while repurchases can be seen as substitutes for dividends they are not perfect substitutes. (Von Eije & Megginson, 2008).

Despite not being perfect substitutes Boudoukh et al. (2007) say that if dividends are (partially) substituted by share repurchases both should be taken into account when testing Miller and Modigliani’s hypothesis. Previously, research

(7)

7

focused primarily on dividends, but Boudoukh et al. (2007) say that dividends and share repurchases can both be seen as payouts to investors and thus both need to be included. They further argue that the most complete measure for payouts should be the best predictor of stock market returns and define three measures to test their hypothesis: the first is the standard dividend yield; this is expected to be the worst predictor as it is the least complete measure. The second measure is the total payout yield and is based on dividends in combination with share repurchases. The third and most complete measure is net payout yield, this incorporates dividends, share repurchases and share issuances. They find their hypothesis confirmed with net payout yield being the strongest predictor of stock market returns, total payout yield the second strongest predictor and with the dividend yield they find no predictive ability for their US sample.

Von Eije and Megginson (2008) show that in Europe, including in Germany, the same dividend and share repurchase policy changes take place as in the US. Share repurchases have also become an important way to pay out shareholders in Europe and dividend payouts are similarly declining from approximately 80% in 1989 to 60% in 2005. This raises the question if the payout measures constructed by Boudoukh et al. (2007) also are predictors of stock market returns for the Europe. Ang and Bekaert (2007) find that in Germany from 1953 to 2001 dividend yield has short term predictability if monthly data are used, as in this paper. This is contrary research from the US that indicates dividends are losing their predictive power (DeAngelo et al. 1996, Michaely and Thaler, 1997, Fama and French 2001, Goyal and Welch 2003, Boudouk et al. 2007). Ang and Bekaert (2007) further find no predictive ability if quarterly data is used when corrected non-constant errors. Koren and Valentincic (2013) find in their study among 3117 UK firms that dividends are much more important for firms’ payout policies than share repurchases and issuances. This is because even though repurchases are almost as common as dividend payouts their effect is much smaller. This could explain some the findings of Andriosopoulos, Chronopoulos, and Papadimitriou (2014); contrary to the US they find that both dividends and total payout yield have predictive power in the UK and in France. Total payout yield however does not outperform dividend yield in UK and French stock markets. This indicates that share repurchases are independent from dividend policies in UK and French stock markets and that these are not substitutes as Boudoukh et al. (2007) suggested.

As shown by Von Eije and Megginson (2008) in Germany similar changes in payout policies take place as in the US. Therefore the same kinds of relationships are expected to be found as in previous research done in the US. However, as other European research from Ang and Bekaert (2007) and Andriosopoulos et al. (2014) has shown this does not have to be the case. Following the combined results of previous literature there are four hypotheses to be investigated in this paper. The first hypothesis is the dividend yield has predictive value, in the US this is not the case as was shown by Boudoukh et al. (2007), but Ang and Bekaert (2007) show that there is predictability in Germany. They did note that the time period examined was from 1953 to 2001 and that they found the relation only when monthly data was used, as is used in this paper. The second and third hypotheses are that the total payout yield and net payout yield do predict future returns significantly in Germany. This is in line with previous research on total payout yield in the UK from Andriosopoulos et al. (2007) and on total payout yield and net payout yield in the US by Boudoukh et al. (2007). The fourth hypothesis is that as a predictor of future stock returns net

(8)

8

payout yield is more significant than total payout yield. This is because Boudoukh et al. (2007) argues that net payout yield is a more complete metric of payouts compared to the other payout yields. They also find evidence to support this theory for firms in the US. Taking all this into account in the next chapter the methodology and data used in this paper will be discussed and analyzed.

3. Methodology & Data

In the first section of this chapter the methodology used in this paper is discussed. In the second section all the necessary variables and data are described and and the third section provides some descriptive statistics to gain a bit more insight on the sample.

3.1 Methodology and model

In this section the methodology and model used in this paper will be discussed. To test the predictability of the three payout measures the main model used in this paper is the Fama-Macbeth (1973) regression model. This is done while controlling for risk factors identified by Fama and French (1993). They identified three common risk factors for publicly traded firms; size, book-to-market ratio and beta, the relation between the firms return and overall book-to-market return. These three are the factors that will be controlled for since they are these are widely accepted and used as the foundation of many papers on this subject.

The Fama-Macbeth model is used to estimate risk premiums in linear factor models. The methods were first introduced by Black, Jensen and Scholes (1972), and refined by Fama and Macbeth (1973). It involves two-stages; the first stage is a time series regression of returns on the risk factors for each firm. This produces estimates of factor loadings which will be used as generated regressors. The second pass regresses firm returns cross-sectionally on the generated regressors for each month from 1989 to 2014 and provides the standard errors corrected for cross-sectional correlation.

Black, Jensen and Scholes (1972) pointed out that the risk premium estimates from the second pass cross-sectional regression contain an inherent errors-in-variables bias because of estimation errors in the generated regressors from the first stage. In this paper this bias is addressed by creating diversified portfolios organized by particular characteristics. Black, Jensen and Scholes (1972) and Fama and Macbeth (1973) show that this portfolio approach reduces estimation errors in the generated regressors for large samples. This mitigates the errors-in-variables bias because the generated regressors of from the first stage are less affected by firm specific risk due to the diversification.

The procedure starts with creating 10 portfolios for every month. Each of these portfolios represents a decile in firm market value for that month, a size decile. The smallest firms in market value go in portfolio 1 up to the largest in portfolio 10. After all firms having been put into size deciles all size decile are divided in again ten deciles, creating a total of a hundred portfolios. This time each of the size deciles is ordered on firm beta through the same methods as size. After creating each of the 100 size/beta portfolios the equal-weighted

return for each portfolio each month are calculated.

The method of Dimson (1979) is employed to correct for biases due to choices in data window. Ang and Bekaert (2007) found that changes in data window can lead to different outcomes. To provide more robust results the method of Dimson is used to mitigate biases due to choice of data window,

(9)

9

usually caused by infrequent trading. The method is employed by doing the first stage Fama-Macbeth regressions on both the market returns and lagged market returns. The estimated generated regressors are then summed and assigned to each stock in the corresponding portfolio before starting with the second stage regressions.

Then in the second stage the factor premiums for each factor is calculated by running the cross-sectional regressions. This is done for all three payout yields and control variables beta, firm size and book to market ratio. The logarithm of the predictive payout yields, size and book-to-market ratio are used to see the effects on a percentage change of the factors.

𝑅𝑖,𝑡= 𝛼𝑖+ 𝑏1,𝑡ln⁡(𝐷𝑖𝑣𝑖𝑑𝑒𝑛𝑑⁡𝑌𝑖𝑒𝑙𝑑𝑖,𝑡)+𝑏2,𝑡𝛽𝑖,𝑡+ 𝑏3,𝑡ln⁡(𝑆𝑖𝑧𝑒𝑖,𝑡) + 𝑏4,𝑡ln⁡(𝐵𝑡𝑀𝑖,𝑡) + +𝑒𝑖,𝑡 (1)

𝑅𝑖,𝑡= 𝛼𝑖+ 𝑏1,𝑡ln⁡(𝑇𝑜𝑡𝑎𝑙⁡𝑃𝑎𝑦𝑜𝑢𝑡⁡𝑌𝑖𝑒𝑙𝑑𝑖,𝑡)+𝑏2,𝑡𝛽𝑖,𝑡+ 𝑏3,𝑡ln⁡(𝑆𝑖𝑧𝑒𝑖,𝑡) + 𝑏4,𝑡ln⁡(𝐵𝑡𝑀𝑖,𝑡) + +𝑒𝑖,𝑡 (2)

𝑅𝑖,𝑡= 𝛼𝑖+ 𝑏1,𝑡ln⁡(𝑁𝑒𝑡⁡𝑃𝑎𝑦𝑜𝑢𝑡⁡𝑌𝑖𝑒𝑙𝑑𝑖,𝑡)+𝑏2,𝑡𝛽𝑖,𝑡+ 𝑏3,𝑡ln⁡(𝑆𝑖𝑧𝑒𝑖,𝑡) + 𝑏4,𝑡ln⁡(𝐵𝑡𝑀𝑖,𝑡) + +𝑒𝑖,𝑡 (3)

3.1 Data description

The data used will be described following the equations above. They will be discussed from left to right, first return, then the explanatory payout variables and the finally the control variables.

The return is the dependent variable and is actually the risk premium, the return provided by an asset in excess to the return of a no-risk investment. The return consists of the monthly return minus the risk-free rate. With the return variable is meant the theoretical growth in value of a stock by holding it for a month assuming dividends are reinvested in the stock. Monthly instead of for instance yearly data is chosen for comparability with other research. For the risk free rate variable the 10-year German Government Bond yield is taken, this is chosen because it is regarded as the return on an investment with near zero risk and the safe alternative to investing in German stocks. All monthly returns for July of year t to June of year t+1 are merged with market value, book value, dividend, share repurchase and share issuance data in year t-1 to ensure all factors are known before the returns they are assumed to predict.

Next are the three explanatory payout variables, these are the same as the payout measures used by Boudoukh et al. (2007); dividend yield, total payout yield and net payout yield.

The dividend yield is seen as the most basic payout measure and consists of the variables fiscal year-end dividend and the year-end market value of that year. The dividend is divided by the market value to normalize it into a yield so that it can easily be compared to other yields.

Dividend yield Dividends𝑖,𝑡

MV𝑖,𝑡 (4)

The second payout measure, total payout, already contains more information about a firm and is defined as the sum of dividends and share repurchases. This is again normalized into a yield by dividing dividends and share repurchases by the year-end market value.

(10)

10

Total Payout yield (Dividends+Share⁡Repurchases)𝑖,𝑡

MV𝑖,𝑡 (5)

The third payout measure is considered most complete. It contains information about the firm’s dividends, share repurchases and share issuances. Net payouts are the sum of dividend and share repurchases minus share issuances and is again normalized into a yield by dividing them by the year-end market value of that year.

Net Payout yield (Dividends+Share⁡Repurchases−Share⁡Issuance)𝑖,𝑡

MV𝑖,𝑡 (6)

The explanatory payout yield from t-1 are merged with monthly returns from July of year t to June of year t+1 to ensure the explanatory payout yields are known before the returns they are to predict.

The control variables beta, size and book-to-market ratio have been shown to have a relation with stock market returns by Fama & French (1993) and are widely accepted and used and form the foundation for many papers on this subject. The firm beta is the relationship between firm returns and market returns, widely called beta in finance literature.

Beta Cov(r𝑖−r𝑓,r𝑚−r𝑓)

Var(r𝑚−r𝑓) (7)

Beta is estimated by regressing firm risk premium on the market risk premium. However, to correct for biases due to choice of windows Dimson’s method (1979) is used to get a better estimate. This done by doing a second regression of the firm risk premium on the lagged market risk premium and adding the two estimates resulting in the corrected firm beta. These estimates are updated annually and cover the previous 24 months of historical data. As a proxy for the market CDAX performance index is chosen, this is a composite index of all stocks traded on the Frankfurt Stock Exchange in the General Standard or Prime Standard market segments. As with stock returns all dividends are assumed to be reinvested in the index.

For the factor size the market value in June of year t is taken for each of the firms. The book-to-market ratio factor is defined as the fiscal year-end book value plus the balance sheet deferred taxes divided by the firms’ market value of the 31st_{of December of that year. So this means that the book-to-market for} year t-1 and the firm size in June of year t are being used with the returns from July of year t to June of year t+1 to ensure that all the factors are known before the returns they are assumed to predict.

All German stocks listed the Frankfurt Stock Exchange have been covered in the time period 1989-2014. All data for this paper was acquired from the Worldscope database through DataStream.

For the analysis in this paper the sample selection and variable construction methods of Fama and French (1993) are followed as is standard in this field to ensure comparability. The procedures entail only including non-financial firms are included to ensure their different firm characteristics don’t cloud the analysis. Then all firms with insufficient or incorrect data are removed and firms with negative book equity, because negative book equity would mean the firm has negative value. Attrition in the sample is mainly due to delisting firms; therefore both listed and delisted firms are used to ensure no survivorship

(11)

11

bias takes place and firms need to be listed for at least two consequent years. This to make sure that ‘unusual’ firms with irregularities are kept out of the analysis and conclusions can be drawn from regression estimates. These proceedings give a sample 728 firms with a total of 231790 observations.

After that the data will be trimmed to avoid that extreme observations, outliers, are given excessive weight in regressions and to ensure consistency across analyses. The trimming procedure will be the same as used by Boudoukh et al. (2007) for comparability and consists of trimming four variants; the upper and lower 0,5% observations of Book to market, the upper 5% of dividend and payout yield observations and lastly the upper and lower 2,5% of observations for net payout yield.

After these procedures the data is ready to be used for analysis. To be thorough, all analysis have been done with correcting for non-constant errors terms both with and without the final trimming procedure, the whole selected sample, for only positive dividend yield sample and for the sample from 2004, to see if there was any difference. The answer to this is no, the analysis does give different numbers but the results lead to the same conclusions. In the next section the descriptive results will be covered to gain better insight into the data sample.

3.3 Descriptive Statistics

For the descriptive results in this section first some basic information on the variables and their interaction with each other will be covered. Afterwards some of the variables will be analyzed a bit further to investigate if the trends seen in other literature also apply for Germany. I find the same kind of trends and results as described by earlier literature. The descriptive statistics confirm expectations from literature and show no irregularities.

First the base variables that form the core of the research are looked at. As can be seen in Table A1 of the Appendix the amount of observations for each variable is not equal. This is because DataStream could not provide all data, for each company every year. To see how the core variables interact the correlation matrix of all the core variables is presented in Table A3.

As can be observed in Table A3 return only has a significant relation with share issuances, as expected this is a negative correlation. The other relations with return are highly insignificant, dividends being the least significant with a correlation that is strangely negative. What is notable is how significantly correlated all the other variables are. To further investigate in Table A4 and A5 the correlation matrix of our factors are presented including zero-yield firms and excluding zero-yield firms.

Tables A4 and A5 mostly support our assumptions, though none of the correlations is particularly strong. Because of this, multicollinearity can also be ruled out. In Table A4 the risk premium shows a positive correlation with all three payout yields and book-to-market ratio. Of the three payout yield net payout yield shows the strongest correlation with the risk premium, but interestingly enough dividend yield has a stronger correlation than total payout yield. This indicates that Germany has comparable results as in the UK and France where total payout yield did not outperform dividend yield as a predictor of stock market returns. Beta and size have a negative correlation with the risk premium, this is as expected for size, but not for beta. This implies that the return of the goes down if the market as a whole rises. However, beta is highly insignificant and barely negative making this less relevant.

(12)

12

In table A5, without zero yield firms in the sample, beta is positive but still highly insignificant. All signs are as expected and are, except for beta, highly significant. Here again net payout yield has the strongest relation to risk premium, but also again dividend has a stronger relation than total payout yield.

Next is investigated if the trends about the disappearance of dividends and the substitution for share repurchases also take place in Germany. Figure A1 of the Appendix shows dividends still are the dominant means to payout investors in the German stock market. This makes the distinction between Total Payouts and Dividends less relevant. Figure A1 of the Appendix shows that the percentage of dividend paying firms has declined, but is rising again since 2004. Furthermore the percentage of firms that has been using repurchases has risen from 0% in 1996 to approximately 15% in 2014. Figure A2 of the Appendix shows that aggregate dividend payouts have become less aligned with total payouts, but that net payouts have become more aligned to both over time. This implies that firms use share repurchases more often than before, but also that there share issuance are more aligned with the dividend and share repurchases policies than before. The scatter plots of figures A8 through A10 show a negative relationship between the payout yields and the logarithm of the risk premium. And distributions in A11 through A18 show no irregularities in the sample.

Our assumptions made based on literature mostly are confirmed. The still dominant position of dividends in Germany makes the distinction between dividends and total payouts less relevant. However, though less pronounced, the same kind of trend can be spotted in Germany as shown in the US with respect to the substitutability between dividend and share repurchases.

4. Results and Discussion

In this chapter the results of the methodology will be presented and discussed. The results from the Fama-Macbeth regressions can be found from Table A6 through Table A14 for each regression the estimate and standard errors are shown.

In Table A6 monthly returns have been regressed on; dividend yield; portfolio beta; size, book-to-market ratio and beta of the firm for the trimmed sample. The table shows that the dividend yield is positively significant at the 10% level for all combinations with or without the control variables. It is even significant at the 1% level in combination with the size factor and on its own, indicating a very strong relationship.

Table A7 and Table A8 show the same kind of regression with the dividend yield being replaced by the total payout yield and net payout yield, respectively. Both total payout yield and net payout yield show no significance with any of the control variables. Total payout yield is positively insignificant, while net payout yield is unexpectedly negatively insignificant. Control variables size and book-to-market ratio sometimes show significance, while neither beta does.

Tables A9 through A11 show similar results for the sample if firms that do not payout dividends are excluded. However the results show are less pronounced as before; dividend yield is no longer significant at the 10% level for all combinations and only is significant at the 1% level in combination with size and the R-squared of that combination is higher than that of the same combination in the sample with the zero-yield firms.

In table A12 through A14 the same regression and sample as in tables A9 trough A11 is used, but now starting in 2004 instead of 1989. Now the dividend

(13)

13

yield is significant at the 10% level for all combinations, but none at the 1% level. However seven of the nine combinations are significant at the 5% level, indicating that there is a strong relationship between returns and dividend yield from 2004 onwards.

Overall, contrary to the hypotheses of this paper these tables indicate a relationship between the dividend yield and returns and no relationship between returns and total payout yield or net payout yield. Only the dividend yield coefficient is regularly positive and significant, the total payout yield and net payout yield are insignificant in all regressions, sometimes even negatively. The control variables size and book-to-market ratio sometimes show significance, while neither beta does in any of the regressions. The dividend yield shows a significantly stronger relation with returns compared to both the total payout yield and the net payout yield. The results are similar in the non-trimmed sample and the trimmed sample, but as expected less pronounced due to the outliers still in the sample. Similar results arise with and without zero-yield firms and both in the period starting in 1989 as in the period starting in 2004. There are some differences in significance levels in the various settings but overall the conclusions are the same.

The results show there is no evidence for any of the four hypotheses; dividend yield does have predictive ability in Germany, while total payout yield and net payout yield do not have this. Net pay out yield is also not more significant than total payout yield.

5. Conclusion

In the US the dividend yield has lost some it allure as a predictor or stock market returns early in this century. In this paper this general consensus is analyzed for the German Stock market in the period 1989 to 2014. Three types of payouts have been tested by using the Fama-Macbeth model. In this paper evidence is found of the dividend yield being a predictor of stock market returns for the German stock market. This is contrary to research from the US, but in agreement with earlier research from Ang and Bekaert (2007) for Germany. However, the main goal of this paper, to find a significant relation between stock market returns and total payouts and net payouts respectively has not been reached. During the writing of this paper there was no evidence of the predictability of stock market returns by total payouts or net payouts.

This paper shows that of the three payout measures the dividend yield is the best payout yield predictor of stock market returns in Germany. That the dividend yield has predictive value is in line with earlier research in Germany by Ang and Bekaert (2007). In the UK and France Koren and Valentincic (2013) and Andriosopoulos et al. (2014) also find that the dividend yield has predictive value and that the total payout yield does not outperform the dividend yield. This is the first study that finds that both the total payout yield and net payout yield have no predictive value for the German stock market.

Finally some remarks about the writing of this paper and suggestions for future research; data was one of the main concerns during the writing of this paper. I think increasing the scope to other countries similar to Germany such as the Netherlands, Austria and Switzerland could have given more robust results while still being able to research the predictability of stock market returns for countries that have Germany’s Rheinlandisch model of capitalism.

(14)

14

differences in stock market predictability between countries. Almost all research now is focused on the US, but it becomes more and more clear that despite the fact that most local stock markets can be entered globally through the internet there are still differences we don’t comprehend. It would be interesting to examine the different dynamics of country specific stock markets and what determines these differing dynamics. Perhaps a relation can be found between legal and cultural factors that can indicate how well specific stock market predictors work for certain country. This meta study could then be used to use these predictors in the international financial markets of today.

(15)

15

References

Andriosopoulos, D., Chronopoulos, D. K., & Papadimitriou, F. I. 2014. “Can the information content of share repurchases improve the accuracy of equity

premium predictions?” Journal of Empirical Finance (26): 96–111.

Ang, A. & Bekaert, G. 2006. “Stock Return Predictability: Is it There?” The Review of Financial Studies. Vol. 20 (3).

Becht, M., & Röell, A. 1999. “Blockholdings in Europe: an international comparison.” European Economic Review, Vol. 6: 289-295.

Benartzi, S., Michaely, R., & Thaler, R. H. 1997. “Do changes in dividends signal the future or the past?” Journal of Finance (52): 1007–34.

Berk, J., & DeMarzo, P. (3rd_{global ed.) 2014 “Corporate Finance” Edinburgh Gate} Harlow Essex: Pearson education Limited.

Black, F., Jensen, M., & Scholes, M. 1972. “The Capital Asset Pricing Model: Some Empirical Results.” Studies in the Theory of Capital Markets. New York: Praeger.

Boudoukh, J., Michaely, R., Richardson, M., & Roberts, M. R. 2007. “On the Importance of Measuring Payout Yield: Implications for Empirical Asset Pricing.” The Journal of Finance. Vol. 62 (2).

Brav, A., Graham, J. R., Harvey, C. R., & Michaely, R. 2005. “Payout policy in the 21st century.” Journal of Financial Economics. Vol. 77: 483–527.

DeAngelo, H., DeAngelo, L., & Skinnder, D. J. 1996. “Reversal of fortune

Dividend signaling and the disappearance of sustained earnings growth.” Journal of Financial Economics (40): 341-371.

Dimson, E. 1979. “Risk Measurement when shares are subject to infrequent trading.” Journal of Financial Economics (7):197-226.

Dow, C. H. 1920. “Scientific stock speculation.” The Magazine of Wall Street. Fama, E.F., & French, K. R. 1993. “Common risk factors in the returns on stocks and bonds.” Journal of Financial Economics. Vol. 33: 3–56.

(16)

16

Fama, E.F., & French, K. R. 2001. “Disappearing dividends: Changing firm characteristics or lower propensity to pay?” Journal of Financial Economics. Vol. 60: 3–43.

Fama, E.F., & Macbeth, J. 1973. “Risk, Return, and Equilibrium: Empirical Tests. Journal of Political Economy.” Vol. 81: 607-636.

Goyal, A., & Welch, I. 2003. “Predicting the equity premium with dividend ratios. Management Science.” Vol. 49: 639–654.

Grullon, G., & Michaely, R. 2002. “Dividends, share repurchases and the substitution hypothesis.” Journal of Finance. Vol. 57: 1649–1684.

Hackethal, A., & Zdantchouk, A. 2004. “Share Buy-Backs in Germany Overreaction to Weak Signals?” Frankfurt am Main: Univ. Fachbereich Wirtschaftswiss.

Koren, J., & Valentincic, A. 2013. “Shareholders’ pay-out-related thresholds and earnings management.” Economic and Business Review. Vol. 15 (2): 151–73. Miller, M. H., & Modigliani, F. 1961. “Dividend Policy, Growth, and the Valuation of Shares.” The Journal of Business. Vol. 34. (4): 411-433.

Skinner, D. J. 2008 "The Evolving Relation between Earnings, Dividends, and Stock Repurchases." Journal of Financial Economics. Vol. 87. (3): 582-609. Valkanov, R. 2003. “Long-horizon regressions: Theoretical results and applications.” The Journal of Financial Economics. Vol. 68: 201–232.

Von Eije, H., & Megginson, W.L. 2008. “Dividends and share repurchases in the European Union.” Journal of Financial Economics. Vol. 89: 347–374

(17)

17

Appendix

Table A1: Summary of all base variables Variable Worldscope

code Description # observations Mean

Standard deviation

Return RI

TOTAL RETURN INDEX: A return index is available for individual equities and unit trusts. This shows a theoretical growth in value of a share holding over a specified period, assuming that dividends are re-invested to purchase

additional units of an equity or unit trust at the closing price applicable on the ex-dividend date.

151,077 0.7178974 24.46358

Market Value MV

MARKET VALUE represents the Market Price * Common Shares Outstanding. For companies with more than one type of common/ordinary share, market capitalization represents the total market value of the company. In

millions (€).

13,125 1,482.56 6,618.784

Book Value 5491 BOOK VALUE – OUTSTANDING SHARES – FISCAL represents the book value (proportioned common equity divided

by outstanding shares) at the company’s fiscal year end. In thousands (€). 10,520 821,193 3,812,703 Deferred

Taxes 3263

DEFERRED TAXES represent the accumulation of taxes which are deferred as a result of timing differences

between reporting sales and expenses for tax and financial reporting purposes. In thousands (€). 8,253 10,471.88 335,009.4 Dividend 5376 COMMON DIVIDENDS (CASH) represent the total cash common dividends paid on the company's common stock

during the fiscal year, including extra and special dividends. In thousand (€). 10,644 39,628.21 226,996.8 Share

Repurchase WC04751

COMMON & PREFERRED REDEEMED, RETIRED, CONVERTED, ETC. represents funds used to decrease the

outstanding shares of common and/or preferred stock. In thousands (€). 6,426 10,951.13 137,827.7 Share

Issuance WC04251

NET PROCEEDS FROM SALE/ISSUE OF COMMON & PREFERRED represents the amount a company received from the sale of common and/or preferred stock. It includes amounts received from the conversion of debentures or preferred stock into common stock, exchange of common stock for debentures, sale of treasury shares, shares

issued for acquisitions and proceeds from stock options. In thousands (€).

7,923 24,685.91 232,239.8

Table A2: Summary of all composed variables

Variable Description # observations Mean Standard deviation

Risk premium ReturnStock/Portolio –Return10-year German Bonds 230,048 -4.092245 19.98856

Beta Systemic risk of a stock/portfolio compared to the market 126,916 .5428876 1.482737

Portfolio Beta Beta of portfolio 95,196 .6305196 .6538233

Size Market Value in June 12,666 1,483.976 6,420.966

Book-to-Market Book Value/Market Value 86,567 .7921797 .8484011

Dividend Yield Dividends/Market Value 87,499 .0225598 .0173437

Total Payout Yield (Dividend+Share Repurchase)/Market Value 57,224 .0256253 .0211577

(18)

18 Table A3: Correlation matrix of the base variables

Return Market Value Book Value Deferred Taxes Dividend Share

Repurchases Share Issuances Return 1 Market Value -0.0085 1 0.5482 Book Value 0.0117 0.8306 1 0.4083 0.0000 Deferred Taxes 0.0108 0.0734 0.1779 1 0.4453 0.0000 0.0000 Dividend -0.0002 0.7759 0.7934 0.1506 1 0.9905 0.0000 0.0000 0.0000 Share Repurchases 0.0188 0.3411 0.3105 0.0376 0.3222 1 0.1857 0.0000 0.0000 0.0081 0.0000 Share Issuances -0.0327 0.4113 0.3474 -0.0512 0.2433 0.0630 1 0.0212 0.0000 0.0000 0.0003 0.0000 0.0000

Table A4: Correlation matrix of factor variables including zero yield firms

Risk premium Beta Size

Book-to-Market Dividend Yield

Total Payout Yield Net Payout Yield Risk premium 1 1 1 Beta -0.0021 1 1 0.6790 1 Size -0.0108 0.0939 1 1 0.0371 0.0000 1 Book-to-Market 0.0308 0.0071 -0.0925 1 1 0.0000 0.1710 0.0000 1 Dividend Yield 0.0157 -0.0799 -0.0035 0.1584 1 1 0.0024 0.0000 0.4959 0.0000 1

Total Payout Yield

0.0137 -0.0577 0.0299 0.1401 0.8971 1

0.0080 0.0000 0.0000 0.0000 0.0000

1

Net Payout Yield 0.0247 -0.0755 0.0530 0.0835 0.4967 0.5333 1

(19)

19

Table A5: Correlation matrix of factor variables excluding zero yield firms

Figure A1: Percentage of firms that pays out dividends and share repurchases

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 19 89 19 90 19 91 19 92 19 93 19 94 19 95 19 96 19 97 19 98 19 99 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 20 12 20 13 20 14 Dividend paying firms (%) Share repurchasing firms (%)

Non-zero Risk premium Beta Size

Book-to-Market Dividend Yield

Total Payout Yield Net Payout Yield Risk premium 1 1 1 Beta 0.0026 1 0.6438 1 Size -0.0114 0.1172 1 0.0422 0.0000 1 Book-to-Market 0.0187 -0.0220 -0.0944 1 0.0009 0.0001 0.0000 1 Dividend Yield 0.0246 -0.0879 -0.0524 0.2625 1 0.0000 0.0000 0.0000 0.0000 1

Total Payout Yield

0.0228 -0.0599 -0.0160 0.2382 0.8854 1

0.0000 0.0000 0.0044 0.0000 0.0000

1

Net Payout Yield 0.0283 -0.0857 0.0341 0.1223 0.4882 0.5276 1

(20)

20

Figure A2: Aggregate Payouts of Dividend, Total Payout and Net Payout in € bln *1000

Figure A3: Mean Dividend, Total Payout and Net Payout incl. zero-yield firms in €*1000

1 40 3 30 2 20 1 10 0 0 --10

(21)

21

Figure A4: Mean Dividend, Total Payout and Net Payout excl. zero-yield firms in €*1000

Figure A5: Mean Return of sample

(22)

22 Figure A6: Mean risk premium of the sample

Figure A7: 10-year German Government Bond yield Rp

Bond yield

(23)

23

Figure A8: The logarithm of the risk premium on Dividend yield

Figure A9: The logarithm of the risk premium on Total Payout yield

Figure A10: The logarithm of the risk premium on Net Payout yield Ln (Rp)

Ln (Rp)

(24)

24 Figure A11: Distribution of the logarithm of Returns

Figure A12: Distribution of the logarithm of the risk premium

(25)

25

Figure A13: Distribution of the logarithm of Dividend Yield excl. zero-yield firms

(26)

26

Figure A15: Distribution of the logarithm of Net Payout Yield excl. zero-yield firms

(27)

27

Figure A17: Distribution of the logarithm of Book-to-Market ratio

(28)

28

Table A6: Fama-Macbeth regression model on trimmed sample including zero-yield firms

(1) (2) (3) (5) (6) (7) (8) (9) (10)

VARIABLES Risk premium Risk premium Risk premium Risk premium Risk premium Risk premium Risk premium Risk premium Risk premium

lnDivT 0.00203*** 0.00122* 0.00228*** 0.00164* 0.00113* 0.00151** 0.00159* 0.00187** 0.00149* (0.000762) (0.000707) (0.000752) (0.000899) (0.000592) (0.000676) (0.000929) (0.000889) (0.000853) PortfolioBeta 0.000510 0.000349 0.00124 (0.00134) (0.00139) (0.00163) lnSize 0.000185 0.000175 0.000619 0.00135*** 0.00117*** (0.000382) (0.000426) (0.000495) (0.000479) (0.000426) lnBtMT 0.00214* 0.00266** 0.00310** 0.00248** (0.00119) (0.00121) (0.00127) (0.00109) Beta -0.000443 -0.00123 (0.00130) (0.00150) Constant -0.0310*** -0.0296*** -0.0312*** -0.0319*** -0.0296*** -0.0297*** -0.0315*** -0.0383*** -0.0340*** (0.00346) (0.00334) (0.00374) (0.00398) (0.00309) (0.00376) (0.00463) (0.00446) (0.00429) Observations 87,499 66,062 86,938 63,703 75,915 65,759 49,736 63,417 56,642 R-squared 0.008 0.021 0.023 0.029 0.024 0.035 0.064 0.052 0.066 Number of groups 305 287 305 305 287 287 287 305 287

(1) (2) (3) (5) (6) (7) (8) (9) (10)

lnTotT 0.000750 0.000908 0.000920 0.000606 0.000802 0.000971 0.000673 0.000596 0.000684 (0.000891) (0.000895) (0.000880) (0.00132) (0.000742) (0.000890) (0.00155) (0.00163) (0.00135) PortfolioBeta 0.00243 0.00250 0.00241 (0.00154) (0.00161) (0.00185) lnSize 0.000548 0.000152 0.000386 0.00132** 0.000959* (0.000436) (0.000477) (0.000593) (0.000637) (0.000547) lnBtMT 0.00189 0.00348** 0.00345 0.00255* (0.00152) (0.00160) (0.00219) (0.00146) Beta 0.000855 0.000674 (0.00144) (0.00168) Constant -0.0349*** -0.0315*** -0.0378*** -0.0345*** -0.0311*** -0.0327*** -0.0329*** -0.0419*** -0.0363*** (0.00413) (0.00413) (0.00489) (0.00482) (0.00380) (0.00523) (0.00607) (0.00552) (0.00580) Observations 57,224 45,594 56,949 46,351 52,351 45,456 37,120 46,175 42,468 R-squared 0.014 0.031 0.036 0.049 0.034 0.050 0.098 0.079 0.095 Number of groups 305 287 305 305 287 287 287 305 287

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

(29)

29

(1) (2) (3) (4) (5) (6) (7) (8) (9)

lnNetT 0.000460 0.00103 0.000984 -0.000245 0.000699 0.00118 -0.000592 0.00155 -0.000695 (0.000854) (0.000828) (0.000854) (0.00166) (0.000734) (0.000839) (0.00159) (0.00256) (0.00154) PortfolioBeta 0.00199 0.00205 0.00188 (0.00156) (0.00161) (0.00206) lnSize 0.000392 0.000230 0.000436 0.00158 0.000834 (0.000427) (0.000470) (0.000597) (0.00102) (0.000553) lnBtMT 0.00124 0.00380** 0.00378 0.00351** (0.00465) (0.00153) (0.00616) (0.00147) Beta 0.000856 1.24e-05 (0.00168) (0.00199) Constant -0.0348*** -0.0303*** -0.0353*** -0.0363*** -0.0313*** -0.0315*** -0.0366*** -0.0379*** -0.0391*** (0.00423) (0.00397) (0.00483) (0.00665) (0.00365) (0.00485) (0.00619) (0.00901) (0.00593) Observations 37,763 31,304 37,620 30,974 36,193 31,222 25,776 30,886 29,826 R-squared 0.018 0.040 0.045 0.067 0.042 0.063 0.118 0.105 0.118 Number of groups 305 287 305 305 287 287 287 305 287

Table A9: Fama-Macbeth regression model on trimmed sample without zero-yield stocks

(1) (2) (3) (4) (5) (6) (7) (8) (9)

lnDiv 0.00151** 0.00127** 0.00183*** 0.00109 0.000891 0.00156** 0.00148 0.00127 0.00123 (0.000656) (0.000643) (0.000665) (0.000908) (0.000589) (0.000649) (0.000971) (0.00103) (0.000902) PortfolioBeta 0.00128 0.000913 0.00185 (0.00137) (0.00137) (0.00166) lnSize 0.000594 0.000437 0.000844* 0.00140*** 0.00137*** (0.000369) (0.000395) (0.000487) (0.000467) (0.000431) lnBtMT 0.00185* 0.00254** 0.00344*** 0.00251** (0.00111) (0.00123) (0.00122) (0.00115) Beta -2.92e-05 -0.000780 (0.00152) (0.00170) Constant -0.0321*** -0.0299*** -0.0342*** -0.0330*** -0.0308*** -0.0311*** -0.0335*** -0.0397*** -0.0363*** (0.00345) (0.00331) (0.00365) (0.00408) (0.00311) (0.00381) (0.00472) (0.00448) (0.00439) Observations 56,726 43,916 56,451 41,220 50,885 43,767 33,124 41,099 38,258 R-squared 0.012 0.027 0.030 0.032 0.033 0.043 0.076 0.058 0.081 Number of groups 305 287 305 305 287 287 287 305 287

(30)

30

(1) (2) (3) (4) (5) (6) (7) (8) (9)

lnTotT 0.000483 0.00116 0.000827 -0.00106 0.000669 0.00125 -0.000757 -0.000733 -0.000922 (0.000884) (0.000960) (0.000887) (0.00133) (0.000823) (0.000968) (0.00159) (0.00164) (0.00144) PortfolioBeta 0.00223 0.00180 0.00177 (0.00156) (0.00157) (0.00192) lnSize 0.000710* 0.000419 0.000275 0.00120** 0.000816 (0.000407) (0.000443) (0.000578) (0.000599) (0.000544) lnBtMT 0.00280* 0.00424*** 0.00431** 0.00392*** (0.00147) (0.00158) (0.00216) (0.00150) Beta 0.00110 0.000666 (0.00159) (0.00191) Constant -0.0353*** -0.0305*** -0.0386*** -0.0391*** -0.0319*** -0.0327*** -0.0360*** -0.0448*** -0.0403*** (0.00420) (0.00434) (0.00487) (0.00496) (0.00396) (0.00521) (0.00615) (0.00549) (0.00582) Observations 41,299 33,520 41,156 33,766 38,675 33,438 27,486 33,678 31,756 R-squared 0.019 0.040 0.043 0.051 0.043 0.061 0.106 0.082 0.106 Number of groups 305 287 305 305 287 287 287 305 287

(1) (2) (3) (4) (5) (6) (7) (8) (9)

lnNetT 0.000507 0.00114 0.00106 -0.000234 0.000770 0.00132 -0.000514 0.00159 -0.000647 (0.000856) (0.000829) (0.000857) (0.00166) (0.000733) (0.000840) (0.00159) (0.00256) (0.00154) PortfolioBeta 0.00185 0.00182 0.00170 (0.00156) (0.00160) (0.00206) lnSize 0.000406 0.000278 0.000434 0.00155 0.000816 (0.000427) (0.000469) (0.000597) (0.00102) (0.000553) lnBtMT 0.00121 0.00375** 0.00372 0.00346** (0.00465) (0.00153) (0.00616) (0.00147) Beta 0.000803 -6.15e-05 (0.00168) (0.00199) Constant -0.0346*** -0.0299*** -0.0351*** -0.0363*** -0.0311*** -0.0312*** -0.0362*** -0.0376*** -0.0388*** (0.00425) (0.00398) (0.00484) (0.00666) (0.00365) (0.00486) (0.00620) (0.00902) (0.00593) Observations 37,594 31,162 37,451 30,827 36,041 31,080 25,653 30,739 29,694 R-squared 0.018 0.040 0.045 0.066 0.042 0.063 0.118 0.105 0.118 Number of groups 305 287 305 305 287 287 287 305 287

(31)

31

Table A12: Fama-Macbeth regression model on trimmed sample without zero-yield stocks from 2004

(1) (2) (3) (4) (5) (6) (7) (8) (9)

lnDivT 0.00176** 0.00198** 0.00182** 0.00177** 0.00137* 0.00196** 0.00184** 0.00186** 0.00143* (0.000771) (0.000795) (0.000769) (0.000732) (0.000767) (0.000788) (0.000748) (0.000727) (0.000731) PortfolioBeta 0.000919 0.00167 0.00225 (0.00199) (0.00175) (0.00174) lnSize -0.000153 -0.000456 -0.000130 0.000270 0.000247 (0.000511) (0.000496) (0.000480) (0.000501) (0.000491) lnBtMT 0.00190* 0.00184 0.00215* 0.00229** (0.00112) (0.00119) (0.00113) (0.00113) Beta -0.000398 -1.27e-06 (0.00208) (0.00194) Constant -0.00900* -0.00898* -0.00793 -0.00817* -0.0104** -0.00692 -0.00875* -0.00934* -0.0109** (0.00479) (0.00461) (0.00500) (0.00486) (0.00461) (0.00530) (0.00510) (0.00495) (0.00487) Observations 28,124 23,626 28,069 26,997 27,487 23,604 22,654 26,942 26,329 R-squared 0.009 0.024 0.025 0.018 0.027 0.037 0.046 0.035 0.051 Number of groups 136 136 136 136 136 136 136 136 136

(1) (2) (3) (4) (5) (6) (7) (8) (9)

lnTotT 0.00133** 0.000963 0.00133** 0.00110* 0.000811 0.000922 0.000619 0.00105* 0.000546 (0.000610) (0.000649) (0.000614) (0.000602) (0.000636) (0.000662) (0.000690) (0.000603) (0.000654) PortfolioBeta 0.000538 0.000771 0.000814 (0.00193) (0.00176) (0.00182) lnSize 0.000304 -0.000109 -0.000112 0.000382 0.000279 (0.000481) (0.000488) (0.000510) (0.000502) (0.000493) lnBtMT 0.00120 0.00105 0.00154 0.00149 (0.00112) (0.00122) (0.00117) (0.00113) Beta -0.000102 3.42e-05 (0.00199) (0.00189) Constant -0.0111** -0.0129*** -0.0131*** -0.0111** -0.0131*** -0.0127** -0.0129** -0.0136*** -0.0150*** (0.00459) (0.00436) (0.00502) (0.00474) (0.00448) (0.00510) (0.00525) (0.00510) (0.00503) Observations 24,045 20,154 24,012 23,103 23,419 20,132 19,329 23,070 22,460 R-squared 0.008 0.022 0.024 0.019 0.026 0.036 0.048 0.035 0.051 Number of groups 136 136 136 136 136 136 136 136 136

(32)

32

(1) (2) (3) (4) (5) (6) (7) (8) (9)

lnNetT 0.00126* 0.00124* 0.00133* 0.000917 0.00100 0.00123 0.000891 0.000894 0.000662

(0.000707) (0.000728) (0.000729) (0.000718) (0.000710) (0.000752) (0.000775) (0.000743) (0.000747)

PortfolioBeta 0.000935 0.00177 0.00161

(0.00204) (0.00190) (0.00194)

lnSize 6.16e-05 -0.000375 -0.000347 0.000101 9.76e-05

(0.000497) (0.000503) (0.000505) (0.000506) (0.000482) lnBtMT 0.00123 0.00106 0.00129 0.00125 (0.00111) (0.00114) (0.00108) (0.00104) Beta -9.47e-05 0.000119 (0.00224) (0.00213) Constant -0.0111** -0.0117** -0.0114** -0.0117** -0.0123*** -0.0102* -0.0107* -0.0124** -0.0136*** (0.00499) (0.00461) (0.00527) (0.00516) (0.00461) (0.00539) (0.00549) (0.00534) (0.00518) Observations 22,067 18,700 22,034 21,303 21,753 18,678 18,011 21,270 20,956 R-squared 0.009 0.025 0.028 0.020 0.028 0.040 0.051 0.038 0.054 Number of groups 136 136 136 136 136 136 136 136 136

Predicting future stock returns of the German stock market : comparing predictability by dividend, total payout and net payout yield from 1989 to 2014