• No results found

The effect of dividend payout policy on the volatility of stock : research on the New York Stock Exchange

N/A
N/A
Protected

Academic year: 2021

Share "The effect of dividend payout policy on the volatility of stock : research on the New York Stock Exchange"

Copied!
35
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

The effect of dividend payout policy on the volatility

of stock:

Research on the New York Stock Exchange.

Ralph Willem Frederik van Tulder

Student number: 10434666

Bachelor thesis

Economics and finance

Faculty of Economics and Business

Supervisor: dr. P.J.P.M. Versijp

(2)

1

Abstract

In this thesis the relationship between the dividend payout policies of companies and the volatility of their common stock is researched. Data from 1997 till 2012 was used of the companies listed on the New York Stock Exchange.

The dividend payout policy is divided into two variables: the dividend yield and the payout ratio. Based on dividend related mechanisms a negative relationship is predicted and control variables are defined. Furthermore dummies for the sector the firm is active in were added.

By using a panel data regression on data from 1998 till 2012 on all the companies listed on the NYSE a negative relationship is shown between both the dividend yield and the dividend payout ratio and the volatility of stock. Furthermore the company size, the leverage of a company, the earnings volatility and the sector that a company operates in show significant relationships with the volatility.

(3)

2

Table of content

Abstract ... 1 Chapter 1: Introduction ... 3 1.1 Dividend policy ... 3 1.2 Volatility of stock ... 5

Chapter 2: Literature review ... 6

2.1 Relationship between dividend and stock characteristics ... 6

2.2 Baskin’s Research ... 7

2.2.1 The arbitrage realization effect ... 7

2.2.2 The information effect ... 9

2.2.3 The duration effect. ... 10

2.2.4 The rate of return effects ... 11

2.2.5 Baskin’s regression ... 11

2.3 Earlier findings on other developed and undeveloped markets ... 12

Chapter 3: Methodology ... 14

3.1 Data selection ... 14

3.2 Regression ... 15

3.3 Variable definition ... 16

Chapter 4: Results of the regressions ... 19

4.1 Summary of variables ... 19

4.2 Correlations ... 19

4.3 Regression with only dividend yield and dividend pay-out ratio ... 20

4.4 Regression with control variables ... 21

4.5 Regression with control variables and dummy variables ... 23

4.5 Regression with control variables, dummy variables and dummies for dividend and leverage . 25 Chapter 5: Conclusions ... 27

Appendix A ... 28

Appendix B ... 31

(4)

3

Chapter 1: Introduction

In this thesis the relationship between dividend payout policy and the volatility of common stock is researched. This relationship is important for directors of companies as well as for investors. The understanding of this relationship could give directors and investors advantages and could give directors an instrument to influence the volatility of their stock. This thesis could also shed some light on the “dividend puzzle”: even though 85% of the companies of the S&P 500 pay out dividend according to report from Andrew Birstingl (2015), there is no common sense on what the ideal dividend payout policy is or the exact effects that dividend has.

In this thesis the following research question is answered:

Is there a significant negative relationship between the dividend payout policy and the volatility of common stock?

The dividend payout policy is defined as two variables: the dividend yield and the dividend payout ratio. This thesis focuses on the New York Stock Exchange and thus the American stock market.

1.1 Dividend policy

Dividend policy is a widely researched subject, but there is still no consensus on both the view from the company’s side and from the consumer’s side. Before 1960 the common consensus was that the higher the dividend the company pays, the more attractive the stock was. The “bird in the hand fallacy” was the common theory that described the added value of dividend in that time. This theory states that investors would rather have two dollars paid out in dividends than have a two-dollar stock price increase and thus will dividend be more worth to investors.

Miller & Modigliani (1961) found proof in their research that this theory does not hold in perfect capital markets. Their theory shows that in perfect capital markets, keeping the investment policy of the firm constant, dividend payout policy does not influence the stock price and the return on stock for the investors. They show that consumers could always create their own homemade dividend by selling their stocks and thus is the choice of dividend policy

(5)

4

irrelevant. They also state that the present value of future cash flow is the only factor determining the stock price.

“The dividend puzzle” (Black, 1976) however states that the theory of Miller & Modigliani does not hold in the real economic market. They show that dividend can influence the irrational investor and thus influence the stock price and the return on stock. Black describes how companies use their dividend policy to indirectly influence their shareholders, who act irrational in the real economic market.

Miller & Rock (1985) show in their research that under imperfect information, dividend payout policy can be used as a signaling device to investors. Directors use the dividend to sign information they have, but investors don’t have access to, to the investors. Berk & DeMarzo (2007) state this effect as the dividend signalling hypothesis. Grullon, Michaely, and Swaminathan (2002) found results in line with this hypothesis. They show that an announcement of a 10% increase in dividend will be followed by a 1,34% increase in stockprice. An announcement of a decrease in dividend, and thus giving out a bad signal, will be followed by a 3,71% decrease in stock value.

A theoretical solution for the dividend puzzle, which states that there is no ideal dividend policy, was found in a research paper by Rozef (1982). He found the optimal dividend payout policy by finding the equilibrium of agency costs of exteranal financing and the transaction costs associated with issuing external finance. He states that increasing the dividends will lower the agency costs, but raises the cost of external finance. By increasing the dividend pay untill the two factors reach a minimal total costs, companies can find their optimal dividend ratio.

(6)

5

1.2 Volatility of stock

The volatility of stock is a rate at which the price of a stock increases or decreases. It is calculated by taking the standard deviation from the return on stock for a given time period.

French, Schwert, & Stambaugh (1987) report a positive relationship between the expected market risk premium and the predictable volatility of stock returns. They also give emperical evidence that shows that an unexpected increase in the volatility of stock return increases the expected market premium and thus directly lowers the stock price.

Schwert (1989) tries to explain volatility from a historical macro-economical perspective, but only finds weak evidence that macro-economic volatility can explain stock price volatility. He does show evidence for a positive relation between leverage and stock price volatility.

Cutler, Poterba, & Summers (1989) try to explain the changes in the volatility of stock returns with macro-economic news releases that affect the fundamental company values. They are able to explain slightly less than 50% of the volatility with the news releases. They state the hypothesis that the volatility is also influenced by a re-assessment of excisting fundamental values. This hypothesis is backed up by results from earlier research (Schwert, 1987 & French & Roll, 1986)

(7)

6

Chapter 2: Literature review

2.1 Relationship between dividend and stock characteristics

A lot of research has been done on the relationship between dividend payout policy and fundamental firm values as earnings, stock prices and volatility of stock. Black and Scholes (1974) test their theory that firms with a higher dividend yield will have higher stock returns. They try to show that managers can increase the return on their stock by increasing their dividend pay-out policy. A significant relationship between the dividend yield and the return is not found in the research. Black and Scholes do however find a relationship between dividend yield and dividend pay-out policy; companies with high dividend yield tend to have a high pay-out ratio.

Black and Scholes (1974) also state that after a dividend raise the stock price will temporarily increase, but will return to its old value when investors find that the dividend raise was not because of increased future cash flow. They suggest that companies in need of cash can decrease their dividend, because after the temporary decrease the value will return to its old value.

Fama and French (1988) find that a company’s dividend yield explains more than 25% of the 2 to 4 year stock returns. They found that the longer the horizon, the more the dividend yield explains.

The relation between stock price volatility and future real dividend is researched by Shiller (1981). In his research he finds that the measure of stock price volatility is way to high to be attributed to new information about future dividend. In addition to his findings he states that the volatility has to be explained by more than the future dividend.

(8)

7

2.2 Baskin’s Research

The base for this research paper was laid by Baskin (1989). He was the first to research the relation between dividend payout policy and the volatility of stock. He aimed to find empirical proof for a relationship between these two variables, which would help investors to predict the risk of a company’s stock and give directors an instrument that can change the volatility. He based his hypothesis on four mechanisms that describe the relation between dividend and volatility: the arbitrage realization effect, the information effect, the duration effect and the rate of return effect.

2.2.1 The arbitrage realization effect

The arbitrage realization effect makes use of the main assumption that markets aren’t perfect. This contradicts the theory of Miller & Modigliani (1961). Baskin’s (1989) theory states that although they might be difficult to exploit, arbitrage opportunities are present in real life (imperfect) markets. This is due to the fact that not every investor knows the true value of a stock, which can cause some stocks to be underpriced. Baskin argues that investors will take advantage of this underpricing and thus increase the stock price. He also shows mathematically that when a company does not pay dividends, there is no arbitrage opportunity. This is explained by the formula’s below. This will cause a lower volatility for non- or low dividend paying stock.

This theory is based on the following mathematical functions and definitions: P = the value of the stock.

P* = the present value of future dividend (anticipated with full information). A = the discount from the intrinsic value of the stock.

This gives us the formula by definition:

(9)

8

Where (1-A) denotes the underpricing of the stock. Note that one should have more information about the company’s value than the market has in order to know that a stock is underpriced.

The rate of return for someone exploiting this arbitrage opportunity is described by: 2 𝑅𝑎𝑡𝑒 𝑜𝑓 𝑟𝑒𝑡𝑢𝑟𝑛 = 𝐷/𝑃 + 𝑔𝑟𝑜𝑤𝑡ℎ 𝑜𝑓 𝑠𝑡𝑜𝑐𝑘 𝑣𝑎𝑙𝑢𝑒.

Where D/P is the dividend divided by the current stock price, thus the dividend yield. The growth of the value of the stock is the amount gained from the stock returning from P to its intrinsic value P*. These two add up to the total rate of return.

The discount rate formula used in this thesis is denoted as:

3 𝐷𝑖𝑠𝑐𝑜𝑢𝑛𝑡 𝑅𝑎𝑡𝑒 = 𝐷/𝑃 ∗ + 𝑔𝑟𝑜𝑤𝑡ℎ 𝑜𝑓 𝑠𝑡𝑜𝑐𝑘 𝑣𝑎𝑙𝑢𝑒. Using (1) to rewrite (3) and substituting in (2) the following is computed:

4 𝑅𝑎𝑡𝑒 𝑜𝑓 𝑅𝑒𝑡𝑢𝑟𝑛 = 𝐷𝑖𝑠𝑐𝑜𝑢𝑛𝑡 𝑟𝑎𝑡𝑒 + 𝐴 (𝐷𝑃)

Baskin (1989) shows that the return is the underpricing multiplied by the dividend yield. The higher the dividend yield, the higher the rate of return is for investors who exploit this underpricing. When there is no dividend and the stock shows no tendency to return to its true value, there is no rate of return gained from arbitrage and thus there is no arbitrage opportunity. Dividend yield thus increase the possible return from arbitrage on a stock.

Investors who exploit these arbitrage opportunities will cause the stock to rise and thus increase the volatility.

(10)

9

2.2.2 The information effect

The information effect is also known as the signing effect as described by Berk & DeMarzo (2014). They state that a high dividend might signal to the investors that the company has enough money to keep growing and to return money to the investors. As dividends tend to be sticky they signal to the investors that the directors trust that the company can keep up with paying these dividends in the future (Berk & DeMarzo, 2007).

This theory is also described by Allen and Rachim (1996) and researched by Miller and Rock (1985). Miller and Rock state that a company can signal that the directors trust that the company is doing well by issuing (high) dividend and simultaneously releasing the earnings announcement. Investors will believe in the stability of the company. When investors believe the company is able to sustain their income and continue to create value, the expected future cash flow will be more stable and thus helps sustain the company’s value. When the company’s value remains stable, keeping all else equal their stock value will remain stable and thus the volatility will decrease.

As Berk & DeMarzo (2014) showed the value of a stock (read company) is calculated as followed:

5 𝑃! =𝐷𝑖𝑣!+ 𝑃! 1 + 𝑟!

Assuming that the dividend and 𝑟! are stable, the price will change when 𝑃!changes. 𝑃! is the future stock price and is unknown in the real markets, so it can be replaced by E(𝑃!) , the expected future stock price.

So it is shown that ceteris paribus when the expected stock price is stable, thus when the investors trust the company will sustain its performance, the stock price itself will be stable and volatility will be low.

(11)

10

2.2.3 The duration effect.

The duration effect states that high dividend yield stock is less sensitive to fluctuations in the discount ratio. Baskin argues in his paper (1989) that high-dividend paying stock implies more short-term cash flow. As the short-term cash flows can be compared to a short-term debt, he argues that the high dividend paying stock will, as does the short-term debt, always stay close to par.

Baskin (1989) supports this theory by the use of the Gordon growth model. The Gordon growth model is a model designed by Gordon and Shapiro published in 1956. It calculates the intrinsic value of a stock by using the present value of growing future dividends.

Baskin uses the Gordon growth model to establish a relationship between dividend and volatility as followes.

The stock price as calculated by the Gordon growth model:

6 𝑃!= 𝐷!!! (𝐾𝑒 − 𝑔)

Assuming that growth (g) and the equity discount rate (Ke) are constant. Taking the first derivative to Ke to get the change in 𝑃!:

7 𝑑𝑃! 𝑑𝐾𝑒=

−(𝐷!!!) (𝐾𝑒 − 𝑔)!

If the formula is rewritten to a form where the equity discount rate is shown as a ratio of the dividend yield the following formula is received:

8 𝑑𝑃! 𝑑𝐾𝑒 𝑃!/𝐾𝑒 = 𝐾𝑒 𝐷!!!/𝑃!

(12)

11

As can be concluded from formula 8, a stock with a high dividend yield is less sensitive to changes in Ke than a stock with low dividend yield. So a higher dividend yield reduces the volatility of a stock.

2.2.4 The rate of return effects

The rate of return effect makes use of the assumption that companies with high growth opportunities retain their earnings for reinvestment. It is thus likely that companies that pay out less of their income (lower pay-out ratio) will create value due to new investment opportunities. Baskin argues that creating value trough existing assets will be less vulnerable to prediction errors.

As can be seen in equation (5), when P1 is predicted correct P0 will be stable. If the

expected value of P1 contains more errors, P0 will change when those errors are discovered

(and thus the expected P1 changes). A change in P0 will create volatility in the stock (P0).

Baskin states that high dividend paying stock has a better predictable future value, which helps the stock price to stay stable.

2.2.5 Baskin’s regression

Based on the four mechanisms mentioned above Baskin designed a formula with the two main components, dividend yield and pay-out ratio, and added some control variables. The control variables that he used are the earnings volatility, the firm size, the yearly asset growth of the company and the leverage ratio. Furthermore he added dummy variables for the industries that the companies are active in: Mining/Oil, Utility/Railroad, Wholesale/Retail, Financial and Service. The findings of his research are reported in chapter 2.3.

(13)

12

2.3 Earlier findings on other developed and undeveloped markets

Baskin was the first to test the relationship between dividend pay-out policy and the volatility of stock. In 1989 he did his research on 2344 American companies that were available through Compustat files of 1986. He defined the dividend payout policy as two variables: the dividend yield and the dividend payout ratio. The dividend yield is the ratio of the dividend per stock divided by the stock price. This ratio shows how much of the investment is returned to the shareholder as dividend per year. The dividend payout ratio is the total amount of cash that is paid out as dividend to the stockholders divided by the total earnings before income taxes of the company. These ratios are further discussed in chapter 3.

By applying a regression on the data mentioned above, Baskin came to a significant relationship between dividend yield and stock price volatility. He also shows that there is no evidence for a relationship between the dividend payout ratio and the volatility of stock. He did find a significant relationship for size, earnings volatility and leverage. These findings provide proof for the theories mentioned in 2.2.

After Baskin’s research this method was repeated several times. The first notable researchers to reproduce this investigation in a different area were Allen & Rachim in 1996. They repeated Baskin’s research on companies listed in Australia from 1972 till 1985. They found a significant negative relationship between the pay-out ratio and the volatility, whereas Baskin had found that this relationship was not significant. They also found that the firm size, leverage and earnings volatility have significant influence on the volatility. Allen and Rachim reject the duration and the arbitrage effect because these theories use the dividend yield as a crucial factor. They also reject the rate of return effect because there is no significant effect from the variable growth on the dividend yield and the pay-out ratio. If the rate of return effect works in real life, they say, than this would be the case.

Hussainey, Oscar Mgbame, & Chijoke-Mgbame reproduced Baskin’s research in 2011 regarding the UK market. They did not include the financial sector in their research. In their research they found emperical evidence for both the dividend yield and the dividend pay-out ratio. They further found as expected a negative relationship between size and volatility and a positive relationship between leverage and volatility.

(14)

13

Lashgari & Ahmadi (2014) researched the relationship between dividend on the Tehran stock market in Iran. They found a significant effect of dividend pay-out ratio, but no significant effect for size, leverage and earnings volatility.

Furthermore there are some more bachelor theses and master theses on this subject, all with different outcomes with regards to the relationship between dividend payout policy and the volatility of stock.

The later researches thus show different results compared to Baskin. Some of the theses find a strong correlation between dividend yield and dividend pay-out ratio and decide to drop one of those variables. The choice of which variable they drop will have influenced their outcome. Baskin has kept both and found that only the dividend yield was significant.

This research uses more data than the research of Baskin and uses a different regression tactic than the older researches. All the old researches used Ordinary Least Squares on averaged data per company. In this research a regression on panel data will be used, as will be explained in chapter 3.

(15)

14

Chapter 3: Methodology

3.1 Data selection

The research area of this thesis is limited to the New York Stock Exchange. All the equity data was collected from the Compustat database by Wharton WRDS. The yearly data of all NYSE listed companies was retrieved. However, not all the data was complete. When companies missed essential data, their data was omitted and not taken into account in the regression.

A second dataset with quarterly equity data was received from Compustat in order to calculate the yearly earnings volatility (see paragraph 3.2). This dataset was merged with the previous mentioned yearly data after the calculation of the earnings volatility.

For all the NYSE companies the following yearly data was gathered from 1997 till 2012: dividends per share, earnings before income taxes, total dividends, total market value, total asset value and long term debt. Furthermore the companies were divided into different industrial sectors denoted by the official Global Industry Classification Standard (GIC) sector codes. These were also received from Compustat. The GIC codes were used to create dummy variables.

The data on the stocks of the companies was received from the CRSP database by Wharton WRDS. Again the stock data for all active NYSE companies was retrieved. The data was received monthly in order to calculate the yearly volatility and the average yearly stock price. After calculating these two variables the data was converted to yearly data and merged with the equity data. This created the final dataset that was used for this thesis.

(16)

15

3.2 Regression

Based on the earlier mentioned mechanisms described by Baskin (1989) the following regression was designed in order to research the relationship between dividend pay-out policy and the volatility of stock:

𝑉𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦 = 𝛽!+ 𝛽!𝐷𝑌𝑖𝑒𝑙𝑑 + 𝛽!𝑃𝑟𝑎𝑡𝑖𝑜 + 𝛽!𝐸𝑣𝑜𝑙 + 𝛽!𝑆𝑖𝑧𝑒 + 𝛽!𝐺𝑟𝑜𝑤𝑡ℎ + 𝛽!𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 Where:

Volatility The volatility of the stock DYield The dividend yield Pratio The pay-out ratio

Evol The quarterly earnings volatility Size The firm size (total market value) Growth The yearly asset growth

Leverage The leverage ratio

In addition dummy variables regarding the following industrial sectors were added:

Energy, Industrials, Materials, Consumer discretionary, Consumer staples, Healthcare, Financials, Information technology, Telecommunication services and Sector Unknown. In the regression the dividend yield and the payout ratio are the two variables this research is about, the other variables are control variables.

The data is a panel data set divided per company and per year. By using panel data instead of averaging all the values, like Baskin and previous researchers did, the amount of observations will be significantly higher. Also panel data analysis allows us to control for changing characteristics of a company over the years. The GLS random effects model will be used to analyze the data. Because the variables used tend to be sticky, we will use robust standard errors.

(17)

16

3.3 Variable definition

The regression mentioned in 3.2 contains a series of variables, from which the dividend yield and the dividend payout ratio are the main components, whereas the rest are control variables. These variables are chosen based on the mechanisms described in chapter 2. To get each variable in the right form to perform the regression, some transformations were done. How each variable is computed and defined is described below:

Volatility

To calculate the volatility of the stock monthly closing prices were gathered. The monthly return on these stock prices was computed and used to calculate the standard deviation of the return on the stocks, which is by definition the volatility. The volatility was calculated per year. 𝑀𝑜𝑛𝑡ℎ𝑙𝑦 𝑟𝑒𝑡𝑢𝑟𝑛 = 𝑝! − 𝑝!!! 𝑝! 𝑉𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦 = 1 𝑁 (𝑀𝑜𝑛𝑡ℎ𝑙𝑦 𝑟𝑒𝑡𝑢𝑟𝑛 − 𝑀𝑒𝑎𝑛 𝑚𝑜𝑛𝑡ℎ𝑙𝑦 𝑟𝑒𝑡𝑢𝑟𝑛 )! ! !!! Dividend Yield

The dividend yield is defined as the ratio of dividend paid out to the average stock price. To compute the dividend yield the total dividend per share is used as received from Compustat. The average share price, calculated as yearly average of the monthly stock closing prices, is used as share price in this ratio. By dividing the dividend per share by the share price the dividend yield was computed. The dividend yield is used as a ratio in the regression.

𝐷𝑖𝑣𝑖𝑑𝑒𝑛𝑑 𝑦𝑖𝑒𝑙𝑑 = 𝐷𝑖𝑣𝑖𝑑𝑒𝑛𝑑 𝑝𝑒𝑟 𝑠ℎ𝑎𝑟𝑒 𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑠ℎ𝑎𝑟𝑒 𝑝𝑟𝑖𝑐𝑒

Payout ratio

The payout ratio of a company is the percentage of its yearly income it returns to the investors in the form of dividend. A higher payout ratio means the company pays out a larger part of its profit to their owners/shareholders. To calculate this value the total dividend paid out that

(18)

17

year was divided by the total earnings before taxes. Both values were received from the Compustat database.

𝑃𝑎𝑦𝑜𝑢𝑡 𝑟𝑎𝑡𝑖𝑜 =𝑇𝑜𝑡𝑎𝑙 𝑑𝑖𝑣𝑖𝑑𝑒𝑛𝑑 𝑝𝑎𝑖𝑑 𝐸𝐵𝐼𝑇

Earnings volatility

The volatility of Earnings is calculated by taking the quarterly EBIT to total asset ratio. This ratio is used to calculate the volatility of the earnings per year. The return on the earnings was calculated per quarter, which was used to measure the yearly standard deviation. This standard deviation is the earnings volatility.

𝐸𝑞𝑢𝑖𝑡𝑦 𝑡𝑜 𝐴𝑠𝑠𝑒𝑡 𝑟𝑎𝑡𝑖𝑜 = 𝑃𝑟𝑒𝑡𝑎𝑥 𝑖𝑛𝑐𝑜𝑚𝑒 𝑇𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠 𝑅𝑒𝑡𝑢𝑟𝑛 𝑜𝑛 𝑒𝑞𝑢𝑖𝑡𝑦 = 𝐸 𝑡𝑜 𝐴 𝑟𝑎𝑡𝑖𝑜! 𝐸 𝑡𝑜 𝐴 𝑟𝑎𝑡𝑖𝑜!!!− 1 𝐸𝑎𝑟𝑛𝑖𝑛𝑔𝑠 𝑣𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦 = 𝑁1 (𝑅𝑒𝑡𝑢𝑟𝑛 𝑜𝑛 𝑒𝑞𝑢𝑖𝑡𝑦 − 𝑀𝑒𝑎𝑛 𝑅𝑒𝑡𝑢𝑟𝑛 𝑜𝑛 𝑒𝑞𝑢𝑖𝑡𝑦 )! ! !!! Size

The size of the firm is measured by the total market value in millions of dollars. On this total market value a base 10 log transformation is applied to receive an order of magnitude.

𝐹𝑖𝑟𝑚𝑠𝑖𝑧𝑒 = log!"(𝑀𝑎𝑟𝑘𝑒𝑡 𝑣𝑎𝑙𝑢𝑒)

Growth

The variable growth is the yearly growth of the assets. It is calculated by subtracting the value of the total assets at the beginning of the year from the value at the end of the year and dividing that by the beginning value. The growth is thus defined as a percentage.

(19)

18

𝐺𝑟𝑜𝑤𝑡ℎ = 𝑇𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠! – 𝑇𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠!!! 𝑇𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠!

Leverage

For leverage the ratio of long-term debt to total assets is used in order to compare the companies.

𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 = 𝑇𝑜𝑡𝑎𝑙 𝑙𝑜𝑛𝑔 𝑡𝑒𝑟𝑚 𝑑𝑒𝑏𝑡 𝑇𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠

Restrictions were set on some of these variables in order to avoid irregular outliers that could bias the regression. These limits were carefully selected in order to only omit the outliers and to not bias the regression. The limitations that were set with their corresponding histograms are included in appendix B.

(20)

19

Chapter 4: Results of the regressions

4.1 Summary of variables

In table 1 the summary of the used variables is given. This gives a clear view of the variables with their means, standard deviations, minimum and maximum. The data denoted below are after adjustments for outliers as explained in paragraph 3.3.

Variable Obs Mean Std, Dev, Min Max

Volatility 6542 0,106 0,060 0,018 0,499 Dyield 6542 0,019 0,023 0,000 0,149 Dpayout 6542 0,182 0,218 0,000 1,000 Evol 6542 0,011 0,014 0,000 0,100 Firmsize 5536 3,308 0,694 1,016 5,706 Leverage 6542 0,235 0,184 0,000 0,997 Growth 6542 0,105 0,183 -0,493 0,999 Table 1

As can be seen the dataset consists of 6542 observations. These observations come from 650 different companies. Firmsize has only 5536 observations due to missing data from the Compustat database.

4.2 Correlations

To test for multicollinearity and other correlation errors a correlation test was ran. The results are displayed in table 2.

Volatility Dyield Dpayout Evol Firmsize Leverage Growth

Volatility 1 Dyield -0,29 1 Dpayout -0,27 0,79 1 Evol 0,20 -0,08 -0,04 1 Firmsize -0,27 0,03 0,06 -0,14 1 Leverage -0,01 0,19 0,16 -0,03 -0,12 1 Growth 0,03 -0,08 -0,08 -0,05 0,056 -0,04 1 Table 2

As can be seen the dividend payout ratio and the dividend yield show the highest correlation. This is no surprise, because of their shared dependence on the value of the dividend paid out.

(21)

20

When a company raises its dividend, keeping all else equal both the dividend yield and the dividend payout ratio will rise. The same high correlation was found in the paper of Allen & Rachim (1996). This correlation however will not cause a multicollinearity problem due to the high amount of observations used in this research.

The dividend yield and dividend payout ratio both show a negative relationship with the volatility. This concludes with the expectations and the mechanics described earlier.

The negative correlation between firmsize and volatility can be explained by the fact that larger companies have more divers investments, which can be seen as risk spreading. The more spread of the investments and income, the less the stock will react to a shock in a certain market. Because of this reduced risk the volatility will be lower as the size of the company gets bigger.

The relationship between leverage and both dividend variables is somehow remarkable. It shows that the higher the leverage is, the higher the dividend of a company is. Also the negative relationship between leverage and firm size is remarkable. One would expect bigger companies with more diversified activities to be able to bear more debt and thus have a higher leverage. However, when one looks at the calculations of both variables, it can be seen that the variable firmsize is computed using market value and leverage is computed using total assets. Running a correlation test on these variables shows a correlation of 0.49. So if the firm size is higher, market value must be higher, which will go hand in hand with a higher total asset value. Keeping the debt level equal, the leverage will thus drop. If the leverage would rise with the firm size, the company has to issue relatively more debt than the total assets rise.

4.3 Regression with only dividend yield and dividend pay-out ratio

At first the core regression containing only the variables from the research question was executed. As explained in chapter 3 the regression will be a panel data regression with the random effects model. The random effects model is applied in order to allow for differences between companies. One of these differences is for instance the sector that the company operates in. Assuming the effects of dividend increase are constant for all the companies, and thus implicating the fixed effects model, is not valid. Therefore the effects are assumed to be random and the random effects model is used.

(22)

21

There is no reason to assume the standard errors are homoscedastic, so this thesis uses robust error terms in its regressions. The regression of the random effects method using only the core variables is displayed below in table 3.

Volatility Coef,

Robust Std,

Err, z P>z [95% Conf, Interval]

Dyield -0,298 0,072 -4,160 0,000 -0,438 -0,158 Dpayout -0,027 0,007 -3,800 0,000 -0,042 -0,013 _cons 0,119 0,002 72,490 0,000 0,116 0,122

Table 3

As can be seen both the dividend yield and the dividend payout have a significant effect on the volatility of stock. A one-point increase of the dividend yield will lead to a 0.298 decrease in volatility. Regarding the value of both means (see table 1) the dividend yield is able to influence the volatility with only small amounts. The coefficient of the dividend payout ratio is 10 times as low as that from the dividend yield. Its mean however is 10 times larger. This means, assuming both variables are able to change in the same way and with the same percentages, the influence of the dividend payout ratio is about the same as that of the dividend yield. For the complete regression output including R2 , the amount of observations and the chi2 test see appendix A.

4.4 Regression with control variables

The second regression consists of the core variables as described in 4.3 with the variables added as stated in chapter 3. These variables are all based on economic theories and the results will therefore show the validity of these theories. As with the base regression a random effects panel regression was performed. The results of this regression are given in table 4.

(23)

22

Volatility Coef,

Robust Std,

Err, z P>z [95%Conf, Interval]

Dyield -0,456 0,079 -5,750 0,000 -0,611 -0,300 Dpayout -0,021 0,008 -2,760 0,006 -0,037 -0,006 Evol 0,567 0,079 7,160 0,000 0,412 0,723 Firmsize -0,029 0,002 -14,750 0,000 -0,032 -0,025 Leverage 0,010 0,007 1,550 0,121 -0,003 0,024 Growth 0,007 0,005 1,440 0,150 -0,002 0,016 _cons 0,204 0,007 29,460 0,000 0,190 0,217 Table 4

The core variables show a very different relationship with the volatility in comparison with the core regression under paragraph 4.3. Dividend yield has a coefficient of -0.46, which is significant at the 95% level. This means that when the dividend yield goes up with 1 point, the volatility will decrease with -0.46 points. Increasing the dividend yield by increasing the dividend policy of a firm seems to be an effective way to decrease the volatility of a stock.

The biggest difference is the coefficient of the dividend yield, which is 1,5 times as high as in the first regression without the control variables. It is likely to be caused by an omitted variable bias in the first regression.

The dividend payout policy decreases with 0.006 from 0.027 to 0.021. This decrease might also be caused by the omitted variable bias in the first regression.

The control variables differ from each other in results. The earnings volatility shows a highly positive relationship with the volatility as expected. This relationship is significant at the 99% level. The variable for firm size seems to have a smaller influence with the volatility, but due to the way this variable is computed, with a mean value of 3.3 its influence on the volatility will even be bigger than that of the other variables.

The last two variables, growth and leverage show no significant relationship with the volatility, even not at the 90% level.

Because of potential multicollinearity errors this test is also run with the dividend yield and the dividend payout ratio omitted.

The two separate regressions do not differ much from the original regression. Both the dividend yield and the dividend payout ratio show higher effects on the volatility. This is easily explained by the omission of the other core variable. The dividend yield seems to take over the effects of the dividend payout ratio and vice versa. The other variables stay the same in influence and significance.

(24)

23

4.5 Regression with control variables and dummy variables

In order to efficiently research the relationship between dividend payout policy and the volatility of stock in the best way possible dummy variables were added regarding the type of business the company is in. Each type of business has its own risk characteristics and own type of investments. Because of this there could be a significant difference between types of business. As explained in chapter 3 the Global Industry Classification Standard is used to determine in which industry the company operates. The utilities sector was omitted to avoid the dummy trap. All the results are thus compared to the base dummy: the utilities sector. This sector was chosen as base dummy it is expected to show a low volatility due to a stable supply and demand and due to strict regulation. Thus was tested if the utilities sector shows a significant different volatility than the other sectors.

As above the random effects regression was used with robust standard errors. The results are displayed below.

Volatility Coef,

Robust Std,

Err, z P>z [95%Conf. Interval]

Dyield -0,384 0,0793 -4,840 0,000 -0,540 -0,229 Dpayout -0,022 0,0078 -2,780 0,005 -0,037 -0,006 Evol 0,529 0,0814 6,490 0,000 0,369 0,688 Firmsize -0,027 0,0019 -14,290 0,000 -0,031 -0,024 Leverage 0,013 0,0067 1,960 0,050 0,000 0,026 Growth 0,006 0,0048 1,180 0,239 -0,004 0,015 Cenergy 0,042 0,0060 6,990 0,000 0,030 0,053 Cmaterials 0,034 0,0059 5,770 0,000 0,022 0,045 Cindustrials 0,025 0,0055 4,430 0,000 0,014 0,035 Cconsdis 0,030 0,0059 5,160 0,000 0,019 0,042 Cconsstap 0,023 0,0066 3,500 0,000 0,010 0,036 Chealth 0,026 0,0062 4,280 0,000 0,014 0,039 Cfinancials 0,019 0,0054 3,550 0,000 0,009 0,030 Cinfo 0,037 0,0064 5,820 0,000 0,025 0,050 Ctele 0,032 0,0076 4,240 0,000 0,017 0,047 _cons 0,172 0,0091 18,870 0,000 0,154 0,190 Table 5

(25)

24

In table 5 the results of the regressions are displayed. As can be seen some small changes in the variables occur compared to the regression without dummy variables. The influence of the dividend yield drops, but this is corrected by all the positive dummy variables. Dividend payout ratio, the firm size and the earnings volatility show no large changes.

What is remarkable is the significance of leverage. Where leverage wasn’t significant at 90% at the regressions without the dummy variables for sectors, it now shows significance at the 95% level. This could be a sign that there was an omitted variable bias before the dummies were added.

As expected, all the dummy variables show a significantly higher volatility compared to the utilities sector. The energy sector and the information technology sector show the highest volatility. The relatively high volatility of the energy sector might be caused by the high political importance, which might cause quickly changing expectations due to political changes. Further research on the characteristics of these sectors might be necessary to explain the outcome of these results.

To check if without the possible multicollinearity problem these findings still hold, again two regressions were performed in which the dividend yield or the dividend payout ratio was omitted. Both regressions show that with omitting one of the variables both the dividend yield and the dividend payout ratio stay significant at 99%. The same as before can be seen; the coefficients of both variables rise slightly to adjust for the missing variable. Some of the dummy variables go from insignificant to significant, but this can be attributed to their higher values due to the missing negative variable. This again shows us that multicollinearity is not a problem in this regression.

(26)

25

4.5 Regression with control variables, dummy variables and dummies for dividend and leverage

The last and most complete regression was performed with inclusion of dummies for companies that pay no dividend and companies that are unleveraged. The no dividend dummy was added because companies that pay no dividend might have different characteristics than companies that do pay dividends. Also this might give a more detailed image about how issuing dividend might decrease the volatility of a company.

The dummy for no leverage was added in order to determine if investors have more trust in unleveraged companies. Through the earlier mentioned information effect companies could signal to investors that they have enough capital to keep on growing without the need of loans. Thus a negative value for the dummy for no leverage would be expected. The summary of these dummies and the results of the regression are displayed below.

Variable Obs Mean Std, Dev, Min Max

Cnodiv 6542 0,327 0,469 0 1

Cnolev 6542 0,069 0,254 0 1

Table 6

As can be seen, in 33% of the observations the company doesn’t pay out any dividend at all. Also, in 7% of the observations the company is unleveraged.

(27)

26 Table 7

Both dummy variables are added in this regression as Cnodiv and Cnolev. The dummy for no dividend shows a significant relationship with the volatility. This means that the volatility of companies that pay no dividend is significantly different than those who do pay dividend. This was also concluded in the earlier regressions, where filling in zero for dividend yield and dividend payout ratio would give a higher volatility, but now can be seen that by issuing dividend a company might instantly lower its volatility. The regression shows us that in this dataset the companies that pay no dividend at all have a volatility of 0.012 higher than companies that do pay a small dividend.

The dummy for no leverage is very insignificant and thus shows that having no leverage at all does not influence the volatility of the stock. This could mean that companies being totally equity financed do not influence expectations of investors.

Furthermore the dividend yield and dividend payout ratio show a lower impact than before, which is simply explained by the fact that the constant is lower by introducing the dummy for no dividend paying firms.

(28)

27

Chapter 5: Conclusions

The results of the regression give a clear answer on the research question. From the regressions can be seen that the dividend yield and the dividend payout ratio have a significant influence on the volatility of the stock. Compared to their both means they influence the volatility comparably. Evidence is found that increasing the dividend yield with 1 percent point decreases the volatility with 0.3 percent point. A percent point increase in the dividend payout ratio will decrease the volatility with 0.016 percent point. These results do not prove any of the mechanisms described in chapter 3 wrong.

This research has also shown that companies that do not pay out dividend have a significantly higher volatility than companies that do pay out dividend. Issuing dividend will thus lead to a lower volatility.

Furthermore, the size of a company shows a significant and negative effect on the volatility. The higher the market value of a company, the lower the volatility. In contrast, the earnings volatility shows a strong positive relationship with the volatility. Companies that have a more volatile income will have a more volatile stock price.

When the dummy for no dividend was not taken into account, the leverage of a company also showed a significant positive relationship with the volatility. The more debt to assets a company has, the higher the volatility.

Also, as expected, large differences were found between the sectors that the companies operate in: the utilities sector displaying the lowest volatility and the energy sector displaying the highest.

(29)

28

Appendix A

Regression 1

Regression 2

rho .19766568 (fraction of variance due to u_i) sigma_e .04957751 sigma_u .02460779 _cons .2037717 .0069165 29.46 0.000 .1902157 .2173278 Growth .0068595 .0047692 1.44 0.150 -.0024879 .0162069 leverage .0103869 .0066969 1.55 0.121 -.0027388 .0235125 Firmsize -.0285715 .0019365 -14.75 0.000 -.0323669 -.0247761 Evol .5673666 .0792866 7.16 0.000 .4119677 .7227656 dpayout -.0214741 .0077808 -2.76 0.006 -.0367242 -.0062239 dyield -.4558509 .0792822 -5.75 0.000 -.6112411 -.3004607 volatility Coef. Std. Err. z P>|z| [95% Conf. Interval] Robust

(Std. Err. adjusted for 572 clusters in CompanyID) corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000

Wald chi2(6) = 481.61

overall = 0.1662 max = 15 between = 0.3117 avg = 9.7 R-sq: within = 0.0755 Obs per group: min = 1 Group variable: CompanyID Number of groups = 572 Random-effects GLS regression Number of obs = 5536

(30)

29

Regression 3 .

rho .17320906 (fraction of variance due to u_i) sigma_e .04957751 sigma_u .02269197 _cons .171752 .0091 18.87 0.000 .1539165 .1895876 Ctele .0320305 .0075623 4.24 0.000 .0172086 .0468524 Cinfo .0373532 .0064146 5.82 0.000 .0247809 .0499256 Cfinancials .0191033 .0053851 3.55 0.000 .0085488 .0296578 Chealth .0264386 .0061772 4.28 0.000 .0143315 .0385458 Cconsstap .0232851 .0066482 3.50 0.000 .0102549 .0363153 Cconsdis .0302493 .005857 5.16 0.000 .0187699 .0417287 Cmaterials .0339361 .0058842 5.77 0.000 .0224034 .0454689 Cindustrials .0245001 .0055362 4.43 0.000 .0136493 .0353509 Cenergy .0416636 .0059587 6.99 0.000 .0299848 .0533425 Growth .0056884 .0048277 1.18 0.239 -.0037737 .0151505 Leverage .0131904 .0067157 1.96 0.050 .0000279 .0263529 Firmsize -.0273398 .001913 -14.29 0.000 -.0310892 -.0235904 Evol .5286481 .0813981 6.49 0.000 .3691107 .6881854 Dpayout -.0217532 .0078216 -2.78 0.005 -.0370832 -.0064232 Dyield -.384385 .0793378 -4.84 0.000 -.5398843 -.2288857 Volatility Coef. Std. Err. z P>|z| [95% Conf. Interval] Robust

(Std. Err. adjusted for 572 clusters in CompanyID) corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000

Wald chi2(15) = 678.22

overall = 0.1922 max = 15 between = 0.3778 avg = 9.7 R-sq: within = 0.0770 Obs per group: min = 1 Group variable: CompanyID Number of groups = 572 Random-effects GLS regression Number of obs = 5536

(31)

30

(32)

31 0 2 4 6 8 10 D e n si ty 0 .1 .2 .3 .4 .5 Volatility 0 20 40 60 80 100 D e n si ty 0 .05 .1 .15

dividend yield - dividend per share/shareprice average yearly

0 5 10 15 D e n si ty 0 .2 .4 .6 .8 1

dividend payout - total dividend to earnings ratio Appendix B Volatility with restriction Volatility < 0,5 Dividend yield with restriction Dyield < 0,15 Dividend payout ratio with restriction 0 < Dpayout < 1 Earnings volatility with restrictions Evol < 0,1 Firmsize with restrictions 0 > Firmsize Leverage with restrictions Leverage < 1 0 20 40 60 80 D e n si ty 0 .02 .04 .06 .08 .1 Earnings volatility 0 .2 .4 .6 D e n si ty 1 2 3 4 5 6

Size - log10 market value

0 1 2 3 4 5 D e n si ty 0 .2 .4 .6 .8 1

(33)

32 Histogram of Growth with restrictions -0,5 > Growth >1 0 1 2 3 4 D e n si ty -.5 0 .5 1

(34)

33

Bibliography

Allen, D., & Rachim, V. (1996). Dividend policy and stock price volatility: Australian evidence. Applied Financial Economics , 6 (2), 175 - 188.

Baker, H. K., Powell, G. E., & Veit, E. T. (2002). Revisiting the dividend puzzle: Do all of the pieces now fit? Review of Financial Economics , 11 (4), 241-261.

Baskin, J. (1989). Dividend policy and the volatility of common stock. The journal of

portfolio management , 15 (3), 19-25.

Berk, J., & DeMarzo, P. (2007). Corporate Finance. Pearson Education.

Birstingl, A. (2015). Dividend quarterly Q3 2015. Retreived from Factset website: http://www.factset.com.

Black, F. (1986). Noise. The journal of finance , 41 (3), 529-543.

Black, F. (1976). The dividend puzzle. The Journal of Portfolio Management , 2 (2), 5-8. Black, F., & Scholes, M. (1974). The effects of dividend yield and dividend policy on common stock prices and returns. Journal of financial economics , 1 (1), 1-22.

Brav, A., Graham, J., Harvey, C., & Michaely, R. (2005). Payout policy in the 21st century.

Journal of financial economics , 77 (3), 483-527.

Cutler, D. M., Poterba, J. M., & Summers, L. H. (1989). What moves stock prices?. The

Journal of Portfolio Management , 15 (3), 4-12.

Fama, E., & French, K. (1988). Dividend yields and expected stock returns. Journal of

financial economics , 22 (1), 3-25.

Frankfurter, G. M., & Wood, B. G. (2002). Dividend policy theories and their empirical tests.

International Review of Financial Analysis , 11 (2), 111-138.

French, K. R., & Roll, R. (1986). Stock return variances: The arrival of information and the reaction of traders. Journal of financial economics , 17(1), 5 - 26.

French, K., Schwert, W., & Stambaugh, R. (1987). Expected stock returns and volatility.

Journal of financial Economics , 19 (1), 3-29.

Gordon, M., & Shapiro, E. (1956). Capital Equipment Analysis: The Required Rate of Profit.

Management science , 3 (1), 102 -110.

Grullon, G., Michaely, R., & Swaminathan, B. (2002). Are Dividend Changes a Sign of Firm Maturity? Journal of Business , 75 (3), 387–424.

Hussainey, K., Oscar Mgbame, C., & Chijoke-Mgbame, A. M. (2011). Dividend policy and share price volatility: UK evidence. The journal of risk finance , 12 (1), 57-68.

(35)

34

Lashgari, Z., & Ahmadi, M. (2014). The impact of dividend policy on stock price volatility in the Tehran stock exchange. Kuwait Chapter of the Arabian Journal of Business and

Management Review , 3 (10), 273 -283.

Miller, M., & Modigliani, F. (1961). Dividend policy, growth, and the valuation of shares.

The journal of Business , 34 (4), 411 -433.

Schwert. (1989). Why does stock market volatility change over time? The journal of finance ,

44 (5), 1115-1153.

Shiller, R. J. (1981). Do stock prices move too much to be justified by subsequent changes in dividends?. The American economic review , 71 (3), 421-436.

Stock, J., & Watson, M. (2003). Introduction in Econometrics. Boston: Addison Wesley.

Referenties

GERELATEERDE DOCUMENTEN

Europol, the AIVD and the Key Figure Programme (or the municipality of Amsterdam), presented several different focal points within their strategies. This is not very

When the American doctrine is applied to the case of InnovaThor v Generix, InnovaThor’s Swiss-claim could be considered a patented medical treatment method. This method can be

Monetary policy arrangements and asset purchase programs Firstly, in this section is presented how monetary policy is arranged in the United States, the Euro Area, the United

Next to this, we can conclude that in all cases, except for the stock exchange in Shanghai, the effect of negative news has a larger impact on volatility than positive.. All

This paper has investigated the short-run and long-run impact of the short-term interest rate, long-term interest rate, money supply, inflation rate, industrial production,

To answer the first and second hypothesis (did the official Brexit referendum announcement resulted in an increase in volatility of the individual European Stock Indexes?

The main contribution and crucial question of this paper was if the VAEX is an unbiased and efficient predictor of the future volatility on the Dutch stock market index

Furthermore, Schwert (1990) argued in an early paper that trading volume and stock market volatility moved together. As more investors participate in the market, stock